key: cord-297198-dneycnyr authors: khan, t.; lopez, t.; khan, t.; ali, a.; syed, s.; patil, p.; hatoum, a. title: re: a british society of thoracic imaging statement: considerations in designing local imaging diagnostic algorithms for the covid-19 pandemic date: 2020-05-27 journal: clin radiol doi: 10.1016/j.crad.2020.05.009 sha: doc_id: 297198 cord_uid: dneycnyr nan re: a british society of thoracic imaging statement: considerations in designing local imaging diagnostic algorithms for the covid-19 pandemic q3 q2 sirdwe read with great interest the statement from nair and colleagues of the british society of thoracic imaging (bsti), which provides a thorough insight into the role of imaging in combating the covid-19 pandemic; however, we noticed a potentially consequential error within the text. in answering question 2, the authors comment that "cxr may be abnormal in the majority of covid-19 cases", incorrectly inferring that the study of huang et al. found "bilateral radiographic abnormalities in 40/41 (98%) of cases". 1 the study of huang et al. in the lancet does mistakenly state that they found "bilateral involvement of chest radiographs" in 40/41 patients in table 2 of their results; however, from reading the main text, it is clear that the imaging method they are referring to is actually chest computed tomography (ct), not chest radiography (cxr): "on admission, abnormalities in chest ct images were detected among all patients. of the 41 patients, 40 (98%) had bilateral involvement ( table 2 )". 2 the mismatch between the results table and the main text in the lancet may have contributed to this error. in fact, it is evident from current literature that the sensitivity of cxr is limited for covid-19. the study of guan et al., also cited by nair et al., reported cxr abnormalities in only 162/274 (59.1%) of covid-19 patients. 3 in addition, a more recent study by wong et al. reported abnormal cxr in 44/64 (68.8%) of covid-19 patients on presentation 4 ; however, it is important to note that both of these studies included hospitalised patients, representing individuals with more severe illness. a recently published study from new york city presents a different picture of the sensitivity of cxr in ambulatory care. weinstock et al. reported cxr findings of 636 covid-19 patients presenting to urgent-care centres and found that only 168 (26.4%) were reported originally as abnormal. 5 subsequently, they had 11 board-certified radiologists re-read these radiographs with prior knowledge of the patients' covid-19 diagnosis. in spite of this, the panel classified only 41.7% of cxr findings as abnormal. 5 therefore, in light of the above, it is possible that cxr may not be abnormal in the majority of covid-19 cases presenting to emergency departments. the authors declare no conflict of interest. a british society of thoracic imaging statement: considerations in designing local imaging diagnostic algorithms for the covid-19 pandemic clinical features of patients infected with 2019 novel coronavirus in wuhan, china clinical characteristics of coronavirus disease 2019 in china frequency and distribution of chest radiographic findings in covid-19 positive patients chest x-ray findings in 636 ambulatory patients with covid-19 presenting to an urgent care center: a normal chest x-ray is no guarantee clinical radiology j o ur n a l ho m e p a g e : w w w . c l i ni c a l ra d i o l og y o nl i n e . n e t 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 key: cord-028786-400vglzm authors: oloko-oba, mustapha; viriri, serestina title: diagnosing tuberculosis using deep convolutional neural network date: 2020-06-05 journal: image and signal processing doi: 10.1007/978-3-030-51935-3_16 sha: doc_id: 28786 cord_uid: 400vglzm one of the global topmost causes of death is tuberculosis (tb) which is caused by mycobacterium bacillus. the increase rate of infected people and the recorded deaths from tb disease is as a result of its transmissibility, lack of early diagnosis, and inadequate professional radiologist in developing regions where tb is more prevalent. tuberculosis is unquestionably curable but needs to be detected early for necessary treatment to be effective. many screening techniques are available, but chest radiograph has proven to be valuable for screening pulmonary diseases but hugely dependent on the interpretational skill of an expert radiologist. we propose a computer-aided detection model using deep convolutional neural networks to automatically detect tb from montgomery county (mc) tuberculosis radiographs. our proposed model performed at 87.1% validation accuracy and evaluated using confusion matrix and accuracy as metrics. tuberculosis (tb) is a contagious disease that is considered worldwide as a significant source of death from a single transmittable agent as well as among the top 10 sources of death [1, 2] . the tb disease is caused by the mycobacterium tuberculosis bacillus which is easily contractible by having close contact with infected individuals. this disease mostly affects the lungs but can as well affects other parts of the body [3, 4] . world health organization (who) estimated about 10 million individuals fell sick as a result of tuberculosis disease in 2018 which resulted in about 1.2 million deaths from the previous 1.3 million deaths recorded in 2017 [2] . tuberculosis disease is more prevalent in developing regions and can affect both males and females but more prominent in males. among the total number of individuals infected with tuberculosis in 2017, 1 million cases were reported in children aged less than 14, 5.8 million cases in males, and 3.2 million cases in females [1] . tuberculosis disease is certainly curable but needs to be detected early for appropriate treatment. several lung examination techniques are available but the chest radiographs conversationally known as chest x-ray or cxr for short is a prominent screening tool for detecting abnormalities in the lungs [5, 6, 14] . basically tb manifestation can be detected on cxr, however, quality cxr imaging equipment along with skilled radiologists to accurately interpret the cxr is either limited or not available in tb prevailing regions [7, 8] . a geographical report by the world health organization of most tuberculosis cases for 2018 is shown in table 1 . due to the deadly nature of tb disease and the rate at which it can easily be spread, who has laid emphasis on more proactive measures for continuous reduction of tb cases and deaths [9] . also, the decision to embark on a mission to put an end the universal tb epidemic by the year 2030 is underway as contained in the 2019 global tuberculosis report [2] . the lack of skilled radiologist, high number of patients waiting in line to be screen and mostly outdated equipment which results in a high rate of errors in properly screening the cxr remain a major problem that requires prompt attention. as a result, to profer solution to the issue of limited or lack of expert radiologist and misdiagnosis of cxr, we propose a deep convolutional neural networks (cnn) model that will automatically diagnose large numbers of cxr at a time for tb manifestation in developing regions where tb is most prevalent. the proposed model will eliminate the hassle of patients waiting in line for days to get screened, guarantee better diagnosis, performance accuracy and ultimately minimize cost of screening as opposed to the process of manual examination of the cxr which is costly, time-consuming, and prone to errors due to lack of professional radiologist and huge number of the cxr pilled up to be diagnosed. the evolution of computer-aided detection and investigative systems has offered a new boost attributed to the emerging digital cxr and the ability of computer vision for screening varieties of health diseases and conditions. although much impressive research has been carried out in the last few years regarding computeraided detection, nevertheless lots more finding is required in the field medical imaging to improve the existing methods and find convenient lasting solutions to deadly medical conditions as the case of tuberculosis and many more. a processing method that combines the local binary pattern with laplacian of gaussian was employed in [10] for the manual detection of tuberculosis nodules in cxr. this research centers on accentuating nodules by the laplacian of gaussian filter, lung segmentation, ribs suppression and the use of local binary pattern operators for texture classification. computer-aided diagnosis system presented by [11] for screening tuberculosis using two different convolutional neural networks architectures (alexnet and vggnet) to classify cxr into positive and negative classes. their experiment which is based on montgomery and shenzhen cxr datasets found vggnet outperformed alexnet as a result of a deeper network of vggnet. the performance accuracy of 80.4% was obtained for alexnet while vggnet reached 81.6% accuracy. the authors conclude that improved performance accuracy can be achieved by increasing the dataset size used for the experiment. one of the first research papers that utilized deep learning techniques on medical images is shown in [12] . the work was based on popular alexnet architecture and transfer learning for screening the system performance on different datasets. the cross dataset performance analysis carried out shows the system accuracy of 67.4% on the montgomery dataset and 83.7% accuracy on the shenzhen dataset. a convnet model involving classifications of different manifestations of tuberculosis was presented in [13] . this work looked at unbalanced, less categorized x-ray scans and incorporate cross-validation with sample shuffling in training their model. the peruvian tuberculosis datasets comprising of a total of 4701 image samples with about 4248 samples marked as abnormal containing six manifestation of tuberculosis and 453 samples marked as normal were used for the experiment to obtain 85.6% performance accuracy. cnn has also been applied by the authors of [15] for extracting discriminative and representative features from x-ray radiographs for the purpose of classifying different parts of the body. this research has exhibited the capabilities of cnn models surpassing traditional hand-crafted method of feature extraction. an approach based on deep learning for the classification of chest radiographs into positive and negative classes is depicted in [16] . the cnn structure employed in this work consists of seven convolutional layers and three fully connected layers to perform classification experiments. the authors compared three variety optimizers in their experiments and found the adam optimizer to perform better with validation accuracy of 0.82% and loss of 0.4013. other methods that have been utilized for tb detection and classification includes: support vector machine [17, 18] , k-nearest neighbor [19] , adaptive thresholding, active contour model, and bayesian classifier [20] , linear discriminant analysis [21] . it is evident from the related work that more effort is required in dealing with the tuberculosis epidemic that has continued as one of the topmost causes of death. in view of this, we have presented an improved performance validation accuracy concerning to detecting and classifying cxr for tb manifestation. the montgomery county (mc) cxr dataset was employed in this research. the mc dataset is a tb specific dataset made available by the national library of medicine in conjunction with the department of health services maryland, u.s.a for research intent. this dataset composed of 58 abnormal samples labeled as "1" and 80 normal samples labeled as "0". all samples are of the size 4020 by 4892 pixels saved as portable network graphic (png) file format as shown in fig. 1 . this dataset is accompanied by clinical readings that give details about each of the samples with respect to sex, age, and manifestations. the dataset can be accessed at https://lhncbc.nlm.nih.gov/publication/pub9931 [23] . since deep neural networks are hugely dependent on large data size to avoid overfitting and achieve high accuracy [22], we performed data augmentation on the mc dataset as a way of increasing the size from 138 samples to 5000 samples. the following types of augmentation were applied: horizontal left and right flip with a probability = 0.3, random zoom = 0.3 with an area = 0.8, top and bottom flip = 0.3, left and right rotation = 0.5. other preprocessing task employed here includes image resizing, noise removal and histogram equalization. the data augmentation procedure used in this work is not such that gives room for data redundancy. a model based on deep convolutional neural network (cnn) structure has been proposed in this work for the detection and classification of tuberculosis. cnn models are based upon feed-forward neural network structures for automatic features selection and extraction as a result of taking advantage of the inherent properties of images. the depth of a cnn model has an impact on the performance of the features extracted from an image. cnn models have many layers but the convolutional layer, maxpooling, and the fully connected layer are regarded as the main layers [15] . at the time of model training, diverse parameters are optimized in the convolution layers for extracting meaning features before is been pass on to the fully connected layer where the extracted features are then classified into the target classes which in this case is "normal and abnormal" classes. our proposed cnn structure is composed of feature extraction and features classification stages. the feature extraction stage consists of convolution layers, batch normalization, relu activation function, dropout, and max pooling while the classification stage contains the fully connected layer, flatten, dense and a softmax activation function. there are 4 convolution layers in the network for extracting distinct features from the input image with shape 224 × 224 × 3 that is passed to the first convolutional layer learning 64, 3 × 3 filters, the same as the second convolutional layer. both the third and fourth convolutional layers learn 128, 3 × 3 filters. relu activation function and batch normalization were employed in all the convolutional layers but only the second and fourth layer uses max pooling with a 2 × 2 pooling size and 25% dropout. the fifth layer which is the fully connected layers output 512 feature that is mapped densely to 2 neurons required by the softmax classifier for classifying our images into normal and abnormal classes. the detail representation of our proposed tb detection model is presented in table 2 . at each convolution layer, the feature maps from the preceding layer convolute with kernels which are then fed through the relu activation function to configure the feature output maps. also, each output map can be formulated with respect to several input maps. this can be mathematically written as: where y i j depicts the j th output feature of the l th layer, f (.) is a nonlinear function, n j is the input map selection, y l−1 i refer to the i th input map of l − 1 th layer, m l ij is the kernel for the input i and output map m in the j th layer, and a l j is the addictive bias associated with the j th map output. maxpooling layer carryout a downsampling operation of the input map by calculating the maximum activation value in each feature map. the downsampling of maxpooling is done to partly control overfitting and is formally written as: where α l j represent the multiplicative bias of every feature output map j that scale the output back to its initial range, down(.) can be substituted for either avg(.) or max(.) over an n × n window effectively scaling the input map by n times in every dimension. our model is trained using the stochastic gradients descents (sgd) optimizer with an initial learning rate set to 0.001, batch size of 22 samples, momentum equals 0.9, regularization l2 equals 0.0005 to control overfitting and training loss. the sgd optimizer is given below as: where α j is the gradient of the loss w.r.t α, n is the defined learning rate, α is the weight vector while x and y are the respective training sample and label. the softmax classifier is used to process and classify the features that have been extracted from the convolutional stage. softmax determine the probability of the extracted features and classify them into normal and abnormal classes defined as: where q mean the input vector to the output layer i that is depicted from the exponential element. the structure of our model is presented in fig. 2 . presented in this paper is a model that aids early detection of tuberculosis using cnn structure to automatically extract distinctive features from chest radiographs and classify them into normal and abnormal categories. we did not test other architectures in this research; instead, their performance accuracy is reported in table 3 . the histogram equalizer we applied to enhance the visibility of the data samples, which makes the extracted features more evident, is one of the contributing factors responsible for the improved performance. we will consider the shenzhen, jsrt, kit, and indiana datasets in our future work while we continue to aim for optimal performance accuracy. world health organization: global tuberculosis report tuberculosis: a global health problem causes of stigma and discrimination associated with tuberculosis in nepal: a qualitative study world health organization: tuberculosis prevalence surveys: a handbook world health organization: chest radiography in tuberculosis detection: summary of current who recommendations and guidance on programmatic approaches deep learning at chest radiography: automated classification of pulmonary tuberculosis by using convolutional neural networks can tuberculosis patients in resource-constrained settings afford chest radiography? towards tuberculosis elimination: an action framework for low-incidence countries detecting tuberculosis in chest radiographs using image processing techniques tb detection in chest radiograph using deep learning architecture a novel approach for tuberculosis screening based on deep convolutional neural networks. in: computer-aided diagnosis tx-cnn: detecting tuberculosis in chest x-ray images using convolutional neural network testing for tuberculosis discriminative feature extraction from x-ray images using deep convolutional neural networks deep-learning: a potential method for tuberculosis detection using chest radiography detecting tuberculosis in radiographs using combined lung masks combination of texture and shape features to detect pulmonary abnormalities in digital chest x-rays automatic detection of abnormalities in chest radiographs using local texture analysis a hybrid knowledgeguided detection technique for screening of infectious pulmonary tuberculosis from chest radiographs fusion of local and global detection systems to detect tuberculosis in chest radiographs a survey on image data augmentation for deep learning key: cord-350636-ufwfitue authors: shumilov, evgenii; hosseini, ali seif amir; petzold, golo; treiber, hannes; lotz, joachim; ellenrieder, volker; kunsch, steffen; neesse, albrecht title: comparison of chest ultrasound and standard x-ray imaging in covid-19 patients date: 2020-09-02 journal: ultrasound int open doi: 10.1055/a-1217-1603 sha: doc_id: 350636 cord_uid: ufwfitue purpose: the covid-19 pandemic poses new challenges for the medical community due to its large number of patients presenting with varying symptoms. chest ultrasound (chus) may be particularly useful in the early clinical management in suspected covid-19 patients due to its broad availability and rapid application. we aimed to investigate patterns of chus in covid-19 patients and compare the findings with results from chest x-ray (crx). materials and methods: 24 patients (18 symptomatic, 6 asymptomatic) with confirmed sars-cov-2 by polymerase chain reaction underwent bedside chus in addition to crx following admission. subsequently, the results of chus and crx were compared. results: 94% (n=17/18) of patients with respiratory symptoms demonstrated lung abnormalities on chus. chus was especially useful to detect interstitial syndrome compared to cxr in covid-19 patients (17/18 vs. 11/18; p<0.02). of note, chus also detected lung consolidations very effectively (14/18 for chus vs. 7/18 cases for cxr; p<0.02). besides pathological b-lines and subpleural consolidations, pleural line abnormality (89%; n=16/18) was the third most common feature in patients with respiratory manifestations of covid-19 detected by chus. conclusion: our findings support the high value of chus in the management of covid-19 patients. purpose the covid-19 pandemic poses new challenges for the medical community due to its large number of patients presenting with varying symptoms. chest ultrasound (chus) may be particularly useful in the early clinical management in suspected covid-19 patients due to its broad availability and rapid application. we aimed to investigate patterns of chus in covid-19 patients and compare the findings with results from chest x-ray (crx). materials and methods 24 patients (18 symptomatic, 6 asymptomatic) with confirmed sars-cov-2 by polymerase chain reaction underwent bedside chus in addition to crx following admission. subsequently, the results of chus and crx were compared. results 94 % (n = 17/18) of patients with respiratory symptoms demonstrated lung abnormalities on chus. chus was especially useful to detect interstitial syndrome compared to cxr in covid-19 patients (17/18 vs. 11/18; p < 0.02). of note, chus also detected lung consolidations very effectively (14/18 for chus vs. 7/18 cases for cxr; p < 0.02). besides pathological b-lines and subpleural consolidations, pleural line abnormality (89 %; n = 16/18) was the third most common feature in patients with respiratory manifestations of covid-19 detected by chus. conclusion our findings support the high value of chus in the management of covid-19 patients. at present, national health systems throughout the world are being overwhelmed by the amount of respiratory tract infections associated with the novel coronavirus 2019 (sars-cov-2) that can cause coronavirus disease 2019 (covid-19). sars-cov-2 has proved to be highly contagious, spreading globally within a very short time and prompting the who to declare it a pandemic on march 11, 2020. referring to positive-stranded rna viruses of the coronaviridae family, sars-cov-2 demonstrates a high sensitivity to the human airway epithelial cells, resulting in a variety of respiratory symptoms including acute respiratory distress syndrome occurring in up to 5 % of cases due to its cytopathic effects [1] . given the explosive spread of the virus as well as the fact that an estimated 5 % of patients with sars-cov-2 infections have severe or critical symptoms that require hospitalization [2] , clinicians are facing an enormous logistical and medical challenge, including the appropriate choice of diagnostic imaging methods. thoracic imaging has turned out to be an essential part for the diagnostic workup and clinical management of covid-19 patients. in particular, computed tomography (ct) of the chest has been shown to be a highly efficient tool and is the gold standard for the early detection of covid-19 pneumonia according to several studies [3] [4] [5] . however, routine chest ct upon admission to the emergency room may not be available in most medical centers around the world and may expose patients with other upper respiratory tract infections or a potential mild course of covid-19 to unnecessary radiation. therefore, bedside diagnostic imaging is a desirable and rapid solution that may have great potential to be implemented in algorithms for the early clinical management of covid-19 patients. interestingly, conventional chest x-ray (cxr) may often fail to capture early signs of covid-19 pneumonia such as ground-glass opacity [3] . in contrast, initial data from china and italy may indicate that point-of-care chest ultrasonography (chus) might be more appropriate to diagnose patterns of interstitial syndrome and alveolar consolidations, and may even reach similar diagnostic accuracy as ct scans [6] . therefore, point-of-care chus might constitute a rapid, cost-effective and safe imaging tool that may be positioned at the interface between cxr and chest ct. here, we aim to describe patterns of chus in covid-19 patients and systematically compare our findings with results from cxr. patients 24 patients with confirmed sars-cov-2 by rt-pcr were admitted to our university hospital from march to april 2020. cxr was performed as a standard radiologic investigation for the assessment of lung abnormalities. furthermore, all patients underwent routine bedside chus following admission. in addition, clinical as well as laboratory data were recorded. statistical tests were performed using the chi-squared test, and p-values < 0.05 were considered significant. the retrospective data analysis was approved by a decision of the local ethics committee (№ 38/4/20). bedside chus was performed on venue 50 and logiq e9, ge medical systems, usa. a standard chus protocol was used. in detail, patients were investigated by both linear (8.4-13.0 mhz) and convex probes (4.0-5.0 mhz) in supine and sitting position at six predetermined examination points (ventral, lateral and dorsal chest wall in apical and basal position, respectively). the chus assessment parameters included the amount (pathologic ≥ 3/field of view) and distribution of b-lines (unilateral, bilateral, focal, multifocal, confluent), pleural line abnormalities (unilateral, bilateral), consolidations (unilateral, bilateral, focal, multifocal, confluent), abnormal lung sliding and pleural effusions [6] . depending on the clinical setting, cxr was performed either in the posterior-anterior and lateral position or in the antero-posterior position with findings described according to the glossary of the fleischner society [7] . cxr evaluation focused on the presence and distribution of hazy increased opacities, consolidations, and pleural effusions. the assessment of chus and cxr images was performed by two blinded investigators. the median age of patients was 65 years (range: 25-95 years) with 16 males and 8 females. the most common symptom was dyspnea (n = 14/24) followed by cough (n = 10/24) and fever (n = 7/24). 10 patients (42 %) required at least 2 liters of oxygen per minute and fig. 1b) , and an equal uni-and bilateral distribution (7/14 each). again, no confluent pattern for consolidations was present as well. along this line, abnormal lung sliding was observed only in 3 of 18 patients (17 %) and occurred only in cases with multifocal consolidations. 10 cases (56 %) showed small pleural effusions with a mostly bilateral manifestation (7/10) (▶ table 2 ) (▶ fig. 1c) . regarding the anatomical distribution of lung abnormalities by chus, the lower lobe (9/18) or both the lower and the upper lobe of note, pathological findings were not seen on chus in 7 out of 24 patients (29 %). however, only one of these patients had covid-19 whereas the remaining six cases were asymptomatic sars-cov-2 carriers. comparing the chus results to standard cxr (available for 23 out of 24 patients), the most common sign in covid-19 patients (n = 18) was hazy increased opacity (▶fig. 1e) in 11/18 (61 %, p < 0.02 compared to b-lines on chus) cases followed by consolidations (▶ fig. 1f ) (7/18; 38 %; p < 0.02 compared to consolidations on chus), and pleural effusion (5/18; 28 %; p = 0.09 compared to chus) (▶table 3). the lesions in cxr were predominantly present in the lower (8/18; 44 %) or in both the lower and the upper lobes (7/18; 39 %) and tended to be distributed bilaterally (13/18; 72 %). of note, only 5/9 lesions (55 %) in the lower lobe and 4/7 lesions (57 %) in the upper and the lower lobe were detected by both chus and cxr, suggesting poor agreement between chus and cxr. regarding asymptomatic sars-cov-2 carriers (cxr for 5/6 available), only 1/5 demonstrated local hazy increased opacity and none of them had consolidations. currently, multiple challenges are associated with the management of the covid-19 pandemic. regarding thoracic imaging, rapid and cost-effective diagnostic tools are urgently needed to cope with the large number of patients. in our study, we investigated patterns in fact, the vast majority of covid-19 patients demonstrated lung abnormalities on chus. notably, these results are in line with two other recent studies that investigated covid-19 patients by chus [8, 9] . in particular, chus was especially informative for revealing different manifestations of interstitial syndrome. the same refers to the detection of lung consolidations by chus. in particular, consolidations in covid-19 cases were characterized by a rather focal and mostly subpleural appearance presenting frequently not only in basal but also in apical parts of the lung. in conjunction with the frequent presence of alveolar consolidations, pleural line abnormalities were the third most common sign among patients with symptomatic covid-19. although it is difficult to directly correlate anatomical locations of lung abnormalities between chus and cxr without having chest ct imaging as an anatomical reference, our results suggest relatively poor agreement between chus and cxr for the anatomical locations of lung pathologies. recently, peng et al. reported the rare presence of pleural effusion in 20 covid-19 patients investigated by chus [6] . in contrast, our data provide evidence that small pleural effusions were present in almost half of covid-19 patients and detected more often by chus compared to cxr. regarding the anatomical distribution of involved lung lobes, chus predominantly showed affection of the lower lobes, but simultaneous lower and upper lobe involvement was also recorded. simultaneous affection of the upper lobe was associated with a more severe clinical course as evidenced by frequent intermediate care unit admission, more severe dyspnea, and a higher rate of systemic inflammation. thus, we conclude that us examinations should always involve apical parts of the lung independent of basal findings. interestingly, all six asymptomatic sars-cov-2 carriers showed no abnormalities on chus which was in accordance with cxr results. this aspect may be of interest especially in emergency departments where clinicians have to make decisions as to whether additional imaging modalities such as cxr or ct scans need to be employed, and ultimately whether the patient can be dismissed to ambulatory care or has to be admitted. following the ongoing active exploration of the diagnostic role of chus in the covid-19 pandemic, clinical and sonographic classification of covid-19 pneumonia was recently suggested. [10] . finally, the frequent finding of bilateral and multilobar lesions on chus in our study confirmed earlier observations by ct scans that peripheral subpleural distribution of lung lesions is frequently found in covid-19 patients [3, 4, 11] . to that end, several studies could show the association of chus findings with ct abnormalities in direct comparison with each other [6, 8, 9] . using chest ct as the reference standard, lu et al. reported on the successful application of lung ultrasound score in covid-19 patients with a diagnostic accuracy of 76.7 %, 76.7 % and 93.3 % for mild, moderate and severe lung lesions, respectively [12] . despite all of the advantages of chus, deep lung lesions cannot be evaluated by ultrasound, and there are limitations for several patient groups, such as patients with high body mass index or restricted mobility. our study has several limitations: first, we only compare chus and cxr and do not provide a gold standard with chest ct imaging. second, the described lung pathologies on chus and cxr are ▶table 3 x-ray findings of n = 18 covid-19 patients. by no means diagnostic for covid-19 and could also be found in several other pulmonary conditions such as viral pneumonia, lung embolisms or congestive heart failure. furthermore, chus is highly observer-dependent and can only capture peripheral lung pathologies. in summary, chus represents a useful tool for rapid and informative lung assessment in covid-19 patients at first clinical presentation and is convenient as a follow-up investigation that could potentially reduce radiation exposure and support clinical decision making. although chus may not be as accurate and sensitive as chest ct scans, it seems to be highly sensitive with respect to detecting peripheral pulmonary pathologies. further multicenter studies should evaluate the diagnostic power and clinical value of chus in the initial assessment and follow-up examinations of covid-19 patients as well as define criteria regarding whether and when chus may replace cxr and/or ct. a novel coronavirus from patients with pneumonia in china characteristics of and important lessons from the coronavirus disease 2019 (covid-19) outbreak in china: summary of a report of 72 314 cases from the chinese center for disease control and prevention imaging profile of the covid-19 infection: radiologic findings and literature review ct imaging features of 2019 novel coronavirus (2019-ncov) time course of lung changes on chest ct during recovery from 2019 novel coronavirus (covid-19) pneumonia chinese critical care ultrasound study g. findings of lung ultrasonography of novel corona virus pneumonia during the 2019-2020 epidemic fleischner society: glossary of terms for thoracic imaging covid-19 pneumonia manifestations at the admission on chest ultrasound, radiographs, and ct: single-center study and comprehensive radiologic literature review can lung us help critical care clinicians in the early diagnosis of novel coronavirus (covid-19) pneumonia? thoracic ultrasound and sars-covid-19: a pictorial essay initial ct findings and temporal changes in patients with the novel coronavirus pneumonia (2019-ncov): a study of 63 patients in wuhan, china a clinical study of noninvasive assessment of lung lesions in patients with coronavirus disease-19 (covid-19) by bedside ultrasound the authors declare that this study did not receive any funding. the retrospective data analysis was approved by a decision of the local ethics committee № 38/4/20. the authors report no relevant conflicts of interest. key: cord-266672-t85wd0xq authors: bagnera, silvia; bisanti, francesca; tibaldi, claudia; pasquino, massimo; berrino, giulia; ferraro, roberta; patania, sebastiano title: performance of radiologists in the evaluation of the chest radiography with the use of a “new software score” in coronavirus disease 2019 pneumonia suspected patients date: 2020-07-20 journal: j clin imaging sci doi: 10.25259/jcis_76_2020 sha: doc_id: 266672 cord_uid: t85wd0xq objectives: the purpose of this study is to assess the performance of radiologists using a new software called “covid-19 score” when performing chest radiography on patients potentially infected by coronavirus disease 2019 (covid-19) pneumonia. chest radiography (or chest x-ray, cxr) and ct are important for the imaging diagnosis of the coronavirus pneumonia (covid-19). cxr mobile devices are efficient during epidemies, because allow to reduce the risk of contagion and are easy to sanitize. material and methods: from february–april 2020, 14 radiologists retrospectively evaluated a pool of 312 chest x-ray exams to test a new software function for lung imaging analysis based on radiological features and graded on a three-point scale. this tool automatically generates a cumulative score (0–18). the intrarater agreement (evaluated with fleiss’s method) and the average time for the compilation of the banner were calculated. results: fourteen radiologists evaluated 312 chest radiographs of covid-19 pneumonia suspected patients (80 males and 38 females) with an average age of 64, 47 years. the inter-rater agreement showed a fleiss’ kappa value of 0.53 and the intra-group agreement varied from fleiss’ kappa value between 0.49 and 0.59, indicating a moderate agreement (considering as “moderate” ranges 0.4–0.6). the years of work experience were irrelevant. the average time for obtaining the result with the automatic software was between 7 s (e.g., zero covid-19 score) and 21 s (e.g., with covid-19 score from 6 to 12). conclusion: the use of automatic software for the generation of a cxr “covid-19 score” has proven to be simple, fast, and replicable. implementing this tool with scores weighed on the number of lung pathological areas, a useful parameter for clinical monitoring could be available. e severe acute respiratory syndrome coronavirus 2 is a newly emerging zoonotic agent appeared in december 2019 and causing the coronavirus disease 2019 (covid-19). [1] e virus rapidly infected all provinces of china and spread to the rest of the world. [2] covid-19-infected patients develop pneumonia with associated symptoms of fever (98%), cough (76%), and myalgia or fatigue (44%). [3] e infection results in a syndrome leading, in some cases, to a critical care respiratory condition, requiring specialized management at intensive care units. [4, 5] e current diagnostic criterion for covid-19 is the positive result of a nucleic acid test by real-time reverse transcription-polymerase chain reaction (rt-pcr). [2] oracic imaging with chest radiography (or chest x-ray, cxr) and computed tomography (ct) plays an important role for diagnosis, management, and follow-up of covid-19 pneumonia patients. [6] e typical radiological imaging demonstrated clear destruction of the pulmonary parenchyma, including interstitial inflammation and extensive consolidation. [7] many groups describe lung opacities in more than one lobe and bilateral lower lobe consolidations (as opposed to community acquired bacterial pneumonia which tends to be unilateral and involving a single lobe). identifying multifocal air-space disease on cxr can be a significant clue to covid-19 pneumonia; [8] some investigators have noticed that the air-space disease tends to have a lower bilateral lung distribution (in the 33-85% of cases). [8, 9] high frequency of peripheral lung involvement (in the 77-88% of cases) is another feature readily identifiable on cxr. ground glass opacities emerges in other cases evaluated (ggo, in the 55-79% of cases), and it is extremely difficult to detect on cxr. cxr is a less sensitive modality in the detection of covid-19 lung disease compared to ct (with a reported baseline cxr sensitivity of 69%). [9] ct is more sensitive than cxr for early disease and it is also important to exclude alternative diagnoses (pulmonary thromboembolism, and heart failure), [10, 11] but the use of cxr could offer some advantages. however, due to infection control issues related to patient transport to ct suites, the inefficiencies introduced in ct room decontamination, and lack of ct availability in parts of the world, portable chest radiography (cxr) will likely be the most commonly utilized modality for identification and follow-up of lung abnormalities. in fact, the american college of radiology notes that ct decontamination required after scanning covid-19 patients may disrupt radiological service availability and suggests that portable chest radiography may be considered to minimize the risk of cross-infection. [12] furthermore, in cases of high clinical suspicion for covid-19, a positive cxr may obviate the need for ct. many groups have tried to schematize radiologist's work providing various "diagnostic scores" with the aim to make the report more standardized; actually, all of these researches are "ct score. " e use of a specific score for cxr report has not been well investigated; the purpose of this study is to clarify the use of a specific cxr digital tool and to investigate its validity in terms of inter-reader concordance. fourteen radiologists participated in this study, divided into four groups based on work experience (group 1: < to 5 years; group 2: from 5 to 15 years; group 3: 15-20 years; and group 4: > to 20 years), as indicated in table 1 . from february 24, 2020, to april 3, 2020, more than 1560 patients with suspected diagnosis of covid-19 viral pneumonia underwent chest radiography (cxr) in our institution. to evaluate a new tool called "covid-19 score" made available to radiologists for lung imaging analysis, we retrospectively included in the study patients who underwent at least two consecutive chest x-rays for a total of 312 exams. is instrument is a software function available within the radiological examination reporting platform of our institute; table 1 : asl to4 radiologists participated on this study and divided into four groups (based on the number of years of work experience). for each zone radiologists assigned the score based on predominant features, graded on a 3-point scale as follow: (0) for negative exam; (1) for predominant interstitial changes (as reticular and/or interlobular septal thickening); (2) in the presence of both interstitial changes and alveolar consolidation (as interstitial changes predominance); and (3) in the presence of both interstitial changes and alveolar consolidation (as alveolar consolidation predominance). points from the values for each zone (from a to f) were added to obtain a final total cumulative score (from 0 to 18). e software can automatically generate the final "covid-19 score;" the data can be recorded and filed for display in subsequent reports and it can be used as a comparison parameter in sequential radiological examinations (for a temporal evaluation of the trend). e radiologists retrospectively and blindly assessed the pool of selected chest x-ray exams and reported the "covid-19 score" values in an excel database. when filing, the radiologists had no clinical information and did not know the outcome of the rt-pcr covid test (actually considered the gold standard for the diagnosis). chest x-ray images were viewed on barco workstation. e time required to complete the banner of each exam for the achievement of the overall score was reported and the insertion of the "covid-19 score" within the structured reports of cxr was assessed, as a prognostic evaluation parameter. e practical implications of using this new tool in the radiological cxr evaluation were considered. all the exams were performed with patients in the supine position with a single ap projection. e portable chest x-ray device used for investigations was: gmm mecall easy slide 30 (for the ivrea hospital), philips practice 33 plus (for the ciriè hospital), and technix tms 300 x-ray (for the chivasso hospital). all these mobile equipment have been used to perform chest examinations at the patient's bed. after each procedure, the equipment was sanitized (with chlorine-containing disinfectant solutions). additional precautions were taken by radiology technicians (tsrm), to preserve themselves and prevent the transmission of the virus in the health-care environment. ese precautions include the correct use of personal protective equipment (ppe; such as visor and mask, disposable gloves, and over-gown with headgear) and adequate awareness and training on the modalities relating to their use, dressing/undressing, and disposal. to investigate the potential role of the "covid-19 score" in the report of cxr, the concordance analysis among the 14 radiologists and the four groups were performed using the fleiss' kappa statistic. e final scores automatically generated by the software was grouped in three score classes (0-4, 5-13, and 14-18) relating to a low, moderate, and high degree of covid-19 prognostic value. fourteen radiologists (three of the group 1; five of the group 2; three of the group 3; and three of the group 4) evaluated 312 chest x-ray images of covid-19 pneumonia suspected patients (80 males and 38 females) with an average age of 64, 47 years. e radiologists belong to the three hospitals of our company (one from ivrea, nine from ciriè, and four from chivasso). no radiologist has reported difficulties in accessing the software function or in filling out the banner. no issues with the automatic registration of the "covid-19 score" were found. e results of the inter-rater and intra-group concordance, obtained with the fleiss' kappa statistic, showed a moderate agreement (0.53) among the radiologists (considering ranges: k <0.4 for a concordance definable as "poor;" 0.4 < k < 0.6 definable as "moderate;" 0.6 < k < 0.8 definable as "good;" and k > 0.8 definable as "excellent"). e intra-group analysis leads to the same conclusion, with fleiss' kappa values between 0.49 for group 1 and 0.59 for group 3, indicating that the years of work experience was irrelevant in the use of this new software application. e average time for filling in the banner of a chest x-ray for obtaining the "covid-19 score" with automatic software was: seven seconds for negative exams (with score equal to zero); 13 s for tests with score between one and six; 21 s for exams with intermediate score (from 7 to 12, where the presence of more pathological findings made it more difficult to identify and interpret the image), and 15 s for images strongly indicative of pulmonary impairment (with score values from 13 to 18). rapid and accurate diagnosis of covid-19 is critical during the epidemic. oracic imaging with chest radiography (cxr) and ct is key tools for pulmonary disease diagnosis and management, but their role in the management of covid-19 has not been considered within the multivariable context of the severity of respiratory disease, pre-test probability, risk factors for disease progression, and critical resource constraints. [13, 14] to solve these problems, we hypothesize a diagnostic model based on radiological imaging and clinical manifestations alone, independent of the nucleic acid test. analysis of specific cases, where radiologists were wrong reveals that the mistakes were made when the covid-19 chest imaging findings are either subtle (likely reflecting early time in the disease process) or when covid-19 has atypical signs. it is worth noting that non-covid-19 pneumonia can also have typical appearance of covid-19. is poses a dilemma because mandated quarantine for all suspected cases can put significant strain on medical infrastructure, health-care providers, and the lives of patients, but may need to be followed as a necessary precaution due to variation in presentation with timing of disease. [15] multiple studies demonstrated that the common covid-19 chest imaging findings are represented by bilateral involvement, peripheral distribution, and lower zone dominance; these can also be appreciated on cxr. [8] in the scenario, where there is high clinical suspicion of covid-19 it is conceivable that a positive cxr may obviate the need for a ct, thus reducing burden on ct units in this pandemic. although chest ct has high sensitivity, it has low specificity. [10, 16] is low specificity may stem from the fact that it is difficult to distinguish covid-19 findings from findings of other disease on chest ct. another critical issues should be considered: e availability of the ct scanner and the time required to clean and disinfect the equipment after performing imaging on a suspected patient. is problem can be solved using cxr portable equipment. e use of these mobile devices allows the exam to be performed in the same room the patient is isolated into, therefore reducing the risk of covid-19 transmission along the transport route to a ct scanner. in hospitalized patients, cxr can be used as a first step to assess the progression of the disease or to make alternative diagnoses, such as lobar pneumonia (suggestive of bacterial superinfection), pleural effusion, or pneumothorax. e role of imaging in covid-19 patients' management should be considered in a multivariable context; the physicians should try to choose the best tools in a specific context considering different practice settings, different phases of the epidemic outbreak, and specific resources availability. in this study, we tested a new software application called "covid-19 score" that can be used in the reporting of chest x-ray imaging in patients suspected covid-19, based on radiological semantic features. using this tool, we found some critical issues: e range (proposed by the design company) with values from 0 to 18 is too wide (it does not allow a precise statistical analysis and generates multiple interpretative variables in radiologists). in addition, the "covid-19 score" obtained with the cumulative method (by automatic addition and without a specific attribution of the area/s concerned) is difficult to correlate with the true clinical status of the patient. by stratifying the scores into three classes (0-4, 5-13, and 14-18), we saw that the extreme values can only be representative of subject's symptomatology. in fact, the class with "covid-19 score" from 0 to 4 correlates with negative imaging and pauci-symptomatic patient. e class with "covid-19 score" from 15 to 18 correlates with pathological imaging and with severely symptomatic patient is shown in figure 2 . on the other hand, the intermediate class with "covid-19 score" from 5 to 13, poses major interpretative problems, due to the copresence of different situations even with identical score values, as shown in figures 3 and 4 . for example, a score of six could derive both from the sum of a score of one in all anatomical areas (from a to f) and from the sum of a maximum score attributed in two ipsilateral anatomical areas (equivalent to a 2/3 lung pathological involvement) or contralateral (equivalent to a bilateral pathological involvement process). to overcome this correlation problem between "score imaging" and "clinic status, " we propose that the "covid-19 score" in cxr be weighed on the number of anatomical lung sectors (a-b) involved; this could offer clinicians a quantifiable data that can be used for patient monitoring over time. e use of automatic software for the generation of a "covid-19 score" within the reports of chest x-rays in covid-19 pneumonia suspected patients has proven to be simple, fast, and replicable in different working contexts with a moderate agreement among the radiologists and without significant differences related to the work experience. implementing this tool with scores weighed on the number of lung pathological areas can be a future parameter of clinical interest for the prognostic temporal evaluation of the improvement or worsening trend of a patient's health. patient's consent not required as patients identity is not disclosed or compromised. nil. ere are no conflicts of interest. figure 2 : (a) 69-year-old man with pneumonia disease who presented with fever, cough, and dyspnea. portable cxr image shows lungs with bilateral pulmonary involvement with widespread interstitial disease and areas of consolidation in different zones of each lung with cxr "covid-19 score" value of 18; (b) 64-yearold man with respiratory distress (intubated and subjected to mechanical ventilation). portable cxr image shows lungs with multiple small areas of lung consolidation affecting the whole lung parenchyma bilaterally with cxr "covid-19 score" value of 18. b a figure 3 : 65-year-old man arrived in the emergency room with sudden high fever (39.8°c) and intense cough. (a) e first portable cxr examination shows lungs with minimal disventilative bibasal areas associated with reticular and hazy medium and the lower left lobe opacities (with interstitial changes predominance) with cxr "covid-19 score" value of 3. (b) after a week, the second portable cxr exam shows lungs with progressive worsening with bilateral medium and lower lobe consolidations (with alveolar consolidation predominance) with cxr "covid-19 score" value of 8. b a figure 4 : (a) 67-year-old man with respiratory distress (intubated and subjected to mechanical ventilation). portable cxr image shows lungs with multiple small areas of lung consolidation (with alveolar consolidation predominance) in the whole lung parenchyma bilaterally with cxr "covid-19 score" value of 16. (b) 3-years-old man with respiratory distress (intubated and subjected to mechanical ventilation). portable cxr image shows lungs with reticular and hazy medium and lower bilateral lobe opacities (with interstitial changes predominance) with cxr "covid-19 score" value of 12. a novel coronavirus from patients with pneumonia in china a diagnostic model for coronavirus disease 2019 (covid-19) based on radiological semantic and clinical features: a multi-center study clinical features of patients infected with 2019 novel coronavirus in wuhan, china e species severe acute respiratory syndrome-related coronavirus: classifying 2019-ncov and naming it sars-cov-2 sars-cov-2 viral load in upper respiratory specimens of infected patients imaging changes in patients with 2019-ncov imaging profile of the covid-19 infection: radiologic findings and literature review frequency and distribution of chest radiographic findings in covid-19 positive patients portable chest x-ray in coronavirus disease-19 (covid-19): a pictorial review chest ct findings in coronavirus disease-19 (covid-19): relationship to duration of infection ct features of coronavirus disease 2019 (covid-19) pneumonia in 62 patients in wuhan, china acr recommendations for the use of chest radiography and computed tomography (ct) for suspected covid-19 infection. united states: american college of radiology e role of chest imaging in patient management during the covid-19 pandemic: a multinational consensus statement from the fleischner society e many faces of covid-19: spectrum of imaging manifestations performance of radiologists in differentiating covid-19 from viral pneumonia on chest ct sensitivity of chest ct for covid-19: comparison to rt-pcr key: cord-337507-cqbbrnku authors: cozzi, andrea; schiaffino, simone; arpaia, francesco; pepa, gianmarco della; tritella, stefania; bertolotti, pietro; menicagli, laura; monaco, cristian giuseppe; carbonaro, luca alessandro; spairani, riccardo; paskeh, bijan babaei; sardanelli, francesco title: chest x-ray in the covid-19 pandemic: radiologists’ real-world reader performance date: 2020-09-10 journal: eur j radiol doi: 10.1016/j.ejrad.2020.109272 sha: doc_id: 337507 cord_uid: cqbbrnku purpose: to report real-world diagnostic performance of chest x-ray (cxr) readings during the covid-19 pandemic. methods: in this retrospective observational study we enrolled all patients presenting to the emergency department of a milan-based university hospital from february 24th to april 8th 2020 who underwent nasopharyngeal swab for reverse transcriptase-polymerase chain reaction (rt-pcr) and anteroposterior bedside cxr within 12 h. a composite reference standard combining rt-pcr results with phone-call-based anamnesis was obtained. radiologists were grouped by cxr reading experience (group-1, >10 years; group-2, <10 years), diagnostic performance indexes were calculated for each radiologist and for the two groups. results: group-1 read 435 cxrs (77.0 % disease prevalence): sensitivity was 89.0 %, specificity 66.0 %, accuracy 83.7 %. group-2 read 100 cxrs (73.0 % prevalence): sensitivity was 89.0 %, specificity 40.7 %, accuracy 76.0 %. during the first half of the outbreak (195 cxrs, 66.7 % disease prevalence), overall sensitivity was 80.8 %, specificity 67.7 %, accuracy 76.4 %, group-1 sensitivity being similar to group-2 (80.6 % versus 81.5 %, respectively) but higher specificity (74.0 % versus 46.7 %) and accuracy (78.4 % versus 69.0 %). during the second half (340 cxrs, 81.8 % prevalence), overall sensitivity increased to 92.8 %, specificity dropped to 53.2 %, accuracy increased to 85.6 %, this pattern mirrored in both groups, with decreased specificity (group-1, 58.0 %; group-2, 33.3 %) but increased sensitivity (92.7 % and 93.5 %) and accuracy (86.5 % and 81.0 %, respectively). conclusions: real-world cxr diagnostic performance during the covid-19 pandemic showed overall high sensitivity with higher specificity for more experienced radiologists. the increase in accuracy over time strengthens cxr role as a first line examination in suspected covid-19 patients. since the start of the covid-19 pandemic, international recommendations [1, 2] stated that the diagnosis of sars-cov-2 infection should primarily rely on viral testing rather than on chest imaging. this endorsed reference standard, i.e. reverse transcriptase-polymerase chain reaction (rt-pcr) on nasal or throat swabs, has become essential in the triage and monitoring phases of patients with suspected sars-cov-2 infection performance [3] , but is encumbered by a sensitivity oscillating between 38% and 89% [4] [5] [6] . moreover, during the pandemic peak, rt-pcr response times became often incompatible with appropriate triaging and management of the high number of suspect covid-19 cases simultaneously presenting to emergency departments [7] [8] [9] , forcing the incorporation of imaging in the diagnostic pathway to compensate both rt-pcr aforementioned shortcomings [2, 10, 11] . while the use of chest cteven as a triaging testwas almost ubiquitous [11] [12] [13] , both initial reports from china and a recent meta-analysis highlighted its low specificity [14] . therefore, an ever-growing number of institutions have come to prefer chest x-ray (cxr), also taking into account that it can be performed with portable equipment in isolation rooms [15] or even in external settings [16] . such choice also minimizes potential contact between patients and operators, as well as other patients [15] [16] [17] [18] . this has been the case in our hospital, located less than 25 miles from the first pandemic hotspot in lombardy, italy. apart from small-scale case series [19, 20] , three major retrospective studies have so far evaluated the diagnostic performance of cxr performed as a triaging test on emergency department admission [21] [22] [23] . the two largest by far are a retrospective review by a single radiologist of 518 cxrs acquired during the first phase of the pandemic peak (from march 1 st to march 15 th )with a resulting overall sensitivity of 57% [22] and a study coming from our group and performed on 535 patients [23] . in our analysis we instead considered the this retrospective observational study was approved by the ethics committee of blinded and performed between february 24 th and april 8 th , 2020, at blinded, a university hospital mainly focusing on cardiovascular diseases but promptly converted to a primarily covid-19-dedicated hospital during the pandemic peak. we included in this study all patients presenting to our emergency department for suspected sars-cov-2 infection who underwent both a nasopharyngeal swab for rt-pcr and an anteroposterior bedside cxr within 12 hours from admission. at our hospital, cxrs are reported by the on-duty radiologist within about 60-90 minutes if performed during the day shift (07:00-20:00), and at the beginning of the following working day if performed during the night shift (20:00-07:00). considering the delay in the availability of rt-pcr results caused by the high number of patients incessantly presenting to the emergency department during the pandemic peak in our region, all cxrs in the study period were reported by radiologists forcedly blinded to rt-pcr results. for this study's purposes, as previously described [23] , we then built a composite reference standard to improve rt-pcr sensitivity, by combining rt-pcr results with phone-call-based complete anamnesis in rt-pcr-negative patients who had not repeated the swab during hospitalization. considering the rather unspecific nature of cxr findings in patients with covid-19 pneumonia, a radiologist with 5 years of experience in cxr interpretation (blinded) reviewed all routine cxr reportsbeing blinded for the original radiologists' signaturesin order to classify them dichotomously as positives or negatives for covid-19. the absence of pulmonary abnormalities on a cxr determined its classification as a negative one, while the presence of interstitial infiltratesassociated or not with alveolar infiltrateswith predominantly bilateral and basal distribution on a cxr implied its classification as a positive examination [1, 2, 11] . conversely, cxr findings j o u r n a l p r e -p r o o f unrelated to covid-19, such as lobar alveolar infiltrates, (typically associated with bacterial pneumonia), pleural effusion, pneumothorax, were considered as non-covid-19-related finding for the purpose of this dichotomization. we grouped the seven radiologists from our department by their cxr reading experience: group 1 included 4 radiologists (r1, r2, r3, and r4) with 10 or more years of experience in cxr reading; group 2 included 3 (r5, r6, and r7) radiologists with less than 10 years of experience in cxr reading. all radiologists were board-certified: if a resident was in charge of drafting a first version of the report, the report was always checked by a board-certified radiologist and the final version was signed by the same board-certified radiologist. only one of the seven radiologists (in group 1) has a particular dedication to breast imaging but practices at least half of his time as a general radiologist. overall and patient-sex-specific diagnostic performance indexes were calculated for each radiologist and for the two groups over the 6-week timeframe and according to the first and second half of all cxrs read for each radiologist. data were presented as sensitivity, specificity, positive predictive value, negative predictive value, accuracy, positive likelihood ratio, negative likelihood ratio, and their 95% confidence interval (ci). statistical analyses were performed using microsoft excel 2019 (microsoft corporation, redmond, wa, usa). in the six-week study period, r1 read 180 cxrs, with a 79% disease prevalence, r2 read figure 1 shows an example of a true positive and of a false negative case both for group 1 and group 2, while table 1 details overall performance indexes of all readers and table 2 shows the results of readers performance evaluation according to patients subgroups and different timeframes (i.e. the first and second three-week periods). considering the first half and the second half of all cxrs read by each radiologist, we observed an increase in disease prevalence for 5 out of 7 readers: disease prevalence in the cxr subset read by r1 increased from 77% to 81%, from 64% to 77% for r2, from 86% to 90% for r4, from 70% to 86% for r5, from 64% to 77% for r6, while decreasing from 85% to 75% for r3 and from 71% to 43% for r7. group table 3 the role of cxr in covid-19 imaging could be paramount in settings with temporarily-or permanently-limited rt-pcr availability, as anticipated by murphy et al. [24] , who also warned against potential low diagnostic performance of cxr when reported by nondedicated chest radiologists. real-world data from this study, albeit conducted in a highprevalence region and during a sars-cov-2 pandemic peak, seem to provide a better scenario, in which radiologists with less than 10 years of experience matched the 89.0% sensitivity attained by radiologists with more than 10 years of experience, with similar disease prevalence in the cxr subsets read by each group (73% versus 77%, respectively). a non-negligible cost for group 2 to attain such a sensitivity was a consistently lower specificity (41%, 95% ci 25%-59%)a value similar to the pooled specificity reported for chest ct by a meta-analysis of 3 studies from non-high-epidemic areas and 2 studies from high-epidemic areas (37%, 95% ci 26%-50%) [14] while group 1 showed a smaller difference between sensitivity and specificity, with a constantly higher accuracy ( table 2) . such pattern was also observed comparing different timepoints or the total number of cxrs read by each radiologist: between the first and second half of the six-week study period overall accuracy increased from 76% to 86%, with corresponding increases both in group 1 and group 2; between the first and second half of cxrs read by each reader, overall accuracy increased from 81% to 84%, again with corresponding increases in both groups, albeit more pronounced in the less experienced group 2 (1% difference for group 1, 11% difference for group 2). this trend was most likely driven in both groups by adaptation to the escalation of examined cases (from 195 in the first three weeks to 340 in the following three) with an increase in sensitivity and accuracy mirrored by a specificity decrease. of note, we can observe how in both groups there was a comparable number of readers exhibiting an j o u r n a l p r e -p r o o f inverse tendency towards a decrease in accuracy ( figure 1 ) and sensitivity (figure 2 ), reinforced by a decrease in specificity in all but one less-experienced reader (figure 3 ). limitations of this study include its retrospective and monocentric nature, the fact that each radiologist read a different subset of images, and the imbalance in the number of cxrs read by group 1 and group 2, with the lesser-experienced group 2 reading 18.6% of all cxrs. however, the closely proportionate disease prevalence between the two groups substantiates the comparability of subsequent findings and seems to suggest a more pronounced influence of overall radiological experience on the diagnostic performance of each group. such an hypothesis should be verified with a conventional multi-reader study, to ascertain if these differences in diagnostic performance are also influenced by the number of covid-19-positive cxrs read by each radiologist, or indeed result from a combination of these factors. however, we should also consider that any multi-reader study performed after a pandemic outbreak would not reproduce the condition of the first outbreak, when the new disease first spread in a country. other than a conventional multi-reader study, further evaluations of real-world diagnostic performance should also target the potential impact on diagnostic performance of various types of subspecialty radiological training and of centrespecific contingencies, such as presence and employment of residents, different radiologists workloads, and disparities in cxr reporting conducted during day or night shifts. in addition, the result herein reported should be considered in light of the pandemic peakwith very high disease prevalenceand could be not reproducible in low prevalence settings [25, 26] . being this a real-world data study, our results rely on a practical dichotomization of cxr reports: their potential generalizability must be therefore very carefully considered, especially when, in case of suspected covid-19, we have a non-typical cxr for sars-cov-2 pneumonia. clinical translation of our findings would still result in at least two different scenarios, also taking into account the unspecific nature of cxr findings in covid-19 pneumonia and j o u r n a l p r e -p r o o f other viral pneumonias. first, when a patient displays suspicious symptoms for covid-19 that can however be justified by alternative pathological cxr findings pointing to another disease (such as pleural effusion, pneumothorax, bacterial pneumonia), the management of the patient would remain the one that would have normally been followed in the detected condition. otherwise, if in a general situation of increased patient influx to emergency departments a patient presents with suspicious symptoms for covid-19 but no suggestive cxr findings or other findings that can justify a covid-19 diagnosis, the use of chest ct could be considered [2, 14] . however, taking into account the suboptimal diagnostic performance of chest ctin particular the potentially low specificity and positive predictive value [14] -if the patient's clinical conditions are stable and it is therefore possible to wait for rt-pcr confirmation of sars-cov-2 infection, preventive isolation would remain the safest approach. to summarize, the real-world diagnostic performance of cxr during the covid-19 pandemic peak reached a relatively well-balanced overall accuracy (76%-86%), with an 89% sensitivity and a higher specificity for the more experienced radiologists (66%), lower for the less experienced radiologists (41%). such data play in favour of the use of cxr as first line examination when chest imaging is required in suspected covid-19 patients during a pandemic peak. the role of chest imaging in patient management during the covid-19 pandemic: a multinational consensus statement from the fleischner society use of chest imaging in the diagnosis and management of covid-19: a who rapid advice guide should rt-pcr be considered a gold standard in the diagnosis of covid-19? positive rate of rt-pcr detection of sars-cov-2 infection in 4880 cases from one hospital in diagnosis of the coronavirus disease (covid-19): rrt-pcr or ct? laboratory diagnosis of emerging human coronavirus infections -the state of the art detection of 2019 novel coronavirus (2019-ncov) by realtime rt-pcr detection profile of sars-cov-2 using rt-pcr in different types of clinical specimens: a systematic review and meta-analysis real-time rt-pcr in covid-19 detection: issues affecting the results radiology department preparedness for covid-19: radiology scientific expert review panel integrated radiologic algorithm for covid-19 pandemic chinese expert consensus statement a systematic review of chest imaging findings in covid-19, quant diagnostic performance of ct and reverse transcriptase-polymerase chain reaction for coronavirus disease 2019: a meta-analysis radiology department strategies to protect radiologic technologists against covid19: experience from wuhan coronavirus (covid-19) outbreak: what the department of radiology should know frequency and distribution of chest radiographic findings in patients positive for covid-19 imaging evaluation of covid-19 in the emergency department the role of initial chest x-ray in triaging patients with suspected covid-19 during the pandemic diagnostic impact of bedside chest x-ray features of 2019 novel coronavirus in the routine admission at the emergency department: case series from lombardy region covid-19 on the chest radiograph: a multi-reader evaluation of an ai system variation of a test's sensitivity and specificity with disease prevalence this study was partially supported by ricerca corrente funding from italian ministry of health to irccs policlinico san donato j o u r n a l p r e -p r o o f key: cord-331891-a6b1xanm authors: cozzi, diletta; albanesi, marco; cavigli, edoardo; moroni, chiara; bindi, alessandra; luvarà, silvia; lucarini, silvia; busoni, simone; mazzoni, lorenzo nicola; miele, vittorio title: chest x-ray in new coronavirus disease 2019 (covid-19) infection: findings and correlation with clinical outcome date: 2020-06-09 journal: radiol med doi: 10.1007/s11547-020-01232-9 sha: doc_id: 331891 cord_uid: a6b1xanm aim: the purpose of this study is to describe the main chest radiological features (cxr) of covid-19 and correlate them with clinical outcome. materials and methods: this is a retrospective study involving patients with clinical-epidemiological suspect of covid-19 infection, who performed cxrs at the emergency department (ed) of our university hospital from march 1 to march 31, 2020. all patients performed rt-pcr nasopharyngeal and throat swab, cxr at the ed and clinical-epidemiological data. rt-pcr results were considered the reference standard. the final outcome was expressed as discharged or hospitalized patients into a medicine department or intensive care unit (icu). results: patients that had a rt-pcr positive for covid-19 infection were 234 in total: 153 males (65.4%) and 81 females (34.6%), with a mean age of 66.04 years (range 18–97 years). thirteen cxrs were negative for radiological thoracic involvement (5.6%). the following alterations were more commonly observed: 135 patients with lung consolidations (57.7%), 147 (62.8%) with ggo, 55 (23.5%) with nodules and 156 (66.6%) with reticular–nodular opacities. patients with consolidations and ggo coexistent in the same radiography were 35.5% of total. peripheral (57.7%) and lower zone distribution (58.5%) were the most common predominance. moreover, bilateral involvement (69.2%) was most frequent than unilateral one. baseline cxr sensitivity in our experience is about 67.1%. the most affected patients were especially males in the age group 60–79 years old (45.95%, of which 71.57% males). rale score was slightly higher in male than in female patients. anova with games-howell post hoc showed significant differences of rale scores for group 1 vs 3 (p < 0.001) and 2 vs 3 (p = 0.001). inter-reader agreement in assigning rale score was very good (icc: 0.92—with 95% confidence interval 0.88–0.95). conclusion: in covid-19, cxr shows patchy or diffuse reticular–nodular opacities and consolidation, with basal, peripheral and bilateral predominance. in our experience, baseline cxr had a sensitivity of 68.1%. the rale score can be used in the emergency setting as a quantitative method of the extent of sars-cov-2 pneumonia, correlating with an increased risk of icu admission. at the end of 2019 a novel virus, named sars-cov-2 (severe acute respiratory syndrome coronavirus 2), expanded globally from china with the first italian cases dating back to february 2020 [1] . this new coronavirus causes a highly infectious disease, commonly called coronavirus disease 19 : lung infection can result in severe pneumonia up to more aggressive acute respiratory distress syndrome (ards) [2, 3] . genetic sequencing of sars-cov-2 has permitted the rapid development of real-time reverse transcription polymerase chain reaction (rt-pcr) of viral nucleic acid, and nowadays this is the diagnostic gold standard [2] . however, this serologic examination has several limitations due to the high number of false-negative tests and the delayed results. radiological evaluation of patients with clinical-epidemiological suspect of covid-19 is mandatory, especially in the emergency department (ed) while waiting for rt-pcr results, in order to have a rapid evaluation of thoracic involvement. the recent covid-19 radiological literature focuses primarily on computed tomography (ct) findings, which is more sensitive and specific than chest x-ray (cxr): in particular, in china ct is used as a first-line diagnostic method for covid-19 [4, 5] . nonetheless, it has to be remembered that performing ct scan is not easy during this pandemic, considering not only the excessive radiation exposure especially to younger patients but also the mandatory scanner disinfection procedures that have to take place. the most italian hospitals are employing cxr as the first-line method, with faster results comparing with those of rt-pcr, especially by using portable x-ray units which reduce the movement of patients and so minimizing the risk of cross-infection [6] [7] [8] . therefore, the purpose of our study is to better understand the main radiographic features of covid-19 pneumonia, by describing the main cxr findings in a selected cohort of patients, also correlating the radiological appearance with rt-pcr examination and patients outcome (intended as discharged or hospitalized into a medicine department or intensive care unit). cxrs of patients with clinical-epidemiological suspect of covid-19 infection performed at the ed of our university hospital from march 1 to march 31, 2020, were retrospectively reviewed. inclusion criteria were: patients' age between 18 and 99 years, rt-pcr nasopharyngeal-throat swab and cxr performed immediately at the ed access, clinical-epidemiological data suspect for covid-19 infection and their duration at the time of ed access (fever, cough, dyspnea, respiratory impairment, diarrhea, asthenia, myalgia and dysgeusia). rt-pcr results were considered the reference standard. for the radiological assessment we selected only patients with rt-pcr-positive results. the final outcome was expressed as discharged or hospitalized patients into a medicine department or into an intensive care unit (icu). all cxrs were acquired as digital radiographs with the same portable x-ray unit (fdr go plus-fujifilm, italia) in the isolation wards of our ed. cxrs were performed in the postero-anterior or antero-posterior projection. all images were stored in a picture archiving and communication system (pacs, syngo-siemens). an independent and retrospective review of each cxr was performed by two thoracic radiologists in order to define the number of radiological suspects of covid-19 infection; after this, they defined the predominant pattern of covid-19 pneumonia presentation in patients with a positive rt-pcr. in case of discordance, a consensual agreement was reached. radiographic features including consolidation, ground-glass opacities (ggo), pulmonary nodules and reticular-nodular opacities were diagnosed according to the fleischer society glossary of terms [9] . moreover, cxrs were assessed for the presence of a specific distribution of the disease (mostly peripheral or perihilar predominance), monolateral (right or left lung) or bilateral disease, upper or lower or diffuse predominance. all thoracic images were also assessed for evidence of other associated pulmonary pathology (cardiomegaly, hilar vascular congestion, pleural effusion, pneumothorax). finally, to quantify the extent of covid-19 lung involvement, a severity score was applied (radiographic assessment of lung edema-rale) [10] . following rale indications, each cxr was given a score between 0 and 48, ranging from the absence of any pathological sign (score 0) to the complete pathological involvement of lung parenchyma (score 48). the score was separately assessed by each of the two radiographers. statistical analysis was performed with spss (spss chicago il, usa). descriptive statistics of rale score were calculated for each group of patients. analysis of variance (anova) was performed to detect possible differences among rale score estimated in the following groups: discharged patients, hospitalized patients into a medicine department, hospitalized patients into an icu (respectively, group 1, 2 and 3 in the following). homogeneity of the variance was established between groups by means of levene test, to adequately choose the post hoc test: bonferroni in case of detected significant homogeneity of variance, otherwise games-howell. intraclass correlation coefficient (icc) was calculated to assess inter-reader agreement in assigning rale score. statistical significance threshold was set at p = 0.05. we found 482 patients fulfilling the following selecting criteria: presence of clinical-epidemiological suspect of covid-19 infection and rt-pcr and cxr performed at the ed admission. patients with a rt-pcr-positive results for covid-19 infection were 234: of these, 153 were males (65.4%) and 81 females (34.6%), with a mean age of 66.04 years (range 18-97 years). only 13 cxrs were negative for radiological thoracic involvement (5.6%). the others showed variable features as described in tables 1 and 2 . the following alterations were more commonly observed: 135 patients with lung consolidations (57.7%), 147 (62.8%) with ggo, 55 (23.5%) with nodules and 156 (66.6%) with reticular-nodular opacities. patients with consolidations and ggo coexistent in the same radiography were 35.5% of total. in rt-pcr-positive patients, we found also signs nonspecific for covid-19 pneumonia as hilar or vascular congestion (39.3%), cardiomegaly (29.9%), pleural effusion (16.6%) and pneumothorax (2.4%). peripheral (57.7%) and lower zone distribution (58.5%) were the most common predominance. bilateral involvement (69.2%) was most frequent than unilateral one. given the results, baseline cxr sensitivity in our experience is about 68.1%. in our population the most affected patients were in the age group of 60-79 years old (43.6%, of which 71.57% males); patients older than 80 years (23.1%) often presented more advanced lung involvement (fig. 1) . nine patients (4%) were immediately discharged from ed, and the others were hospitalized in medicine department or icu (table 3) . a total of 34 (15.3%) patients died in the 30 days included in this study (9 in group 2 and 25 in group 3). descriptive statistics of rale score for each group is reported in table 4 . rale score was slightly higher in male than in female patients. levene test showed significant inhomogeneities of variances among groups 1, 2 and 3; thus, games-howell post hoc test was adopted. anova with games-howell post hoc showed significant differences of rale scores for group 1 vs group 3 (p < 0.001) and group 2 vs group 3 (p = 0.001). box and whisker plot of rale score estimated in groups 1-3 is reported in fig. 2 . inter-reader agreement in assigning rale score was very good: icc, with 95% confidence interval in parentheses, was 0.92 (0.88-0.95). in the context of a global pandemic, the radiological approach should be aimed at a rapid classification of the patient with suspected covid-19 infection. all the italian radiology departments are at the forefront in the diagnosis, in the quantification and in the follow-up of covid-19 infection. the italian society of radiology (sirm) recommends using cxr as a first-line imaging tool and reserves to chest ct others additional roles as the identification of covid-19 pneumonia typical features in selected cases [6] [7] [8] 11] . multiple recent studies indicate that cxr may not have the diagnostic power of ct, but it still has a role in managing the pandemic [6, 12, 13] . in fact, although ct has a high sensitivity (around 97-98%), it has a very low specificity in detecting typical features of sars-cov-2 pneumonia [12, 14, 15] . our study reveals a cxr sensitivity substantially in accordance with the most recent literature (68.1%), where a variability between 69 and 90% is described [12, 13] . in our tertiary center, ct examination is usually performed at ed (generally after a cxr) only in specific situations: in case of clinical-radiological discordance (when cxr is negative for infective lung involvement, but there is a high clinical-epidemiological suspect), in case of acute complication (p.e. pulmonary embolism or severe respiratory failure) or after intubation before transporting patient to icu (fig. 3) . it is operationally much 2 box and whisker plot of rale score estimated in each group defined by outcome: discharged patients (group 1), hospitalized patients into a medicine department (group 2), hospitalized patients into an intensive care unit (group 3). rale score showed statistically significant differences between group 1 vs 3 and 2 vs 3 more complex to perform ct scans, especially considering the disinfection procedures that have to take place after each examination. for this reason, portable x-ray machine is very useful, inexpensive and radiographs can be taken to the patient's bedside, reducing radiology department's exposure to infection and minimizing the risk of cross-infection. our study confirms the main radiological characteristics in covid-19 patients described in previous studies: in most cases, cxr shows patchy or diffuse reticular-nodular opacities and consolidation, with basal, peripheral and bilateral predominance (figs. 4, 5) . in case of monolateral involvement, right lung is affected more frequently than the left one (58% vs 42%). our cxr was performed in a period between 2 and 15 days after the onset of symptoms, with cases of more advanced lung involvement in patients around the tenth day of illness. we applied the rale score, used for the quantification of lung involvement in ards [10] , in order to standardize and objectively quantify the radiographic report and to produce a prognostic score at the patient's admission. we found a significant statistical correlation between rale score and patients' outcome, with a rale score higher than 15 points which correlates with an increased risk of being admitted to icu. so, the above data confirmed rale score as a valid standardized prognostic score. in our experience, rale score has demonstrated to be highly reproducible, as we found with an optimal interobserver reproducibility (icc: 0.92 [95% ci 0.88-0.95]). the role of cxr in clinical monitoring of this patients, especially in icu, is still debated: fleischner society do not recommend daily cxr in stable intubated patients, but it is also true that the disease can progress very quickly [16, 17] . our opinion is that radiological imaging should be always associated with clinical and laboratoristic parameters when monitoring the disease course, waiting for readily effective therapies. obviously, every hospital in the world has its own "radiological" organization and management of the patient with sars-cov-2 infection, but it is always necessary to fig. 3 chest x-ray in three patients with severe respiratory failure, immediately intubated at the arrival in emergency department and transferred to icu fig. 4 peripheral consolidation in covid-19 pneumonia. images in a, b and c show three cases of male patients with subpleural consolidations and bilateral involvement maintain a balance between the safety of health professionals and the diagnostic resources that we can use in this pandemic [16] . our work has several limitations: first of all, the retrospective nature of the study and the lack of a non-covid-19 control group in the study of the prognostic score, thus limiting evaluation of sensitivity and specificity of cxr. moreover, also the lack of comparison between rale score and patients' comorbidities is a limitation because some clinical data were available only in a small percentage of subjects. in addition, there is a difference in size between the three prognostic groups and a difference in the period of time between the onset of symptoms and cxr's execution. in future, it will be necessary to validate the method considering also these variables and evaluate how the rale score varies in radiological follow-up. in our study we applied a visual evaluation either in the diagnostic and in the prognostic approach. it would be interesting to continue the study with application of an artificial intelligence (ai) software for a more sophisticated image analysis. although there are some data about ai and chest ct, neural network applied to chest radiographs needs further investigations and it is too early to apply this new technology in the clinical practice [7, 18, 19] . in conclusion, we describe the main features of covid-19 thoracic involvement on cxr in our cohort of patients. in our experience, baseline cxr had a sensitivity of 67.1%. the rale score can be used in the emergency setting as a quantitative method of the extent of sars-cov-2 pneumonia, correlating with an increased risk of icu admission. the results of our study could help radiologists in identifying the highest risk patients, allowing for timely initiation of treatments currently available against sars-cov-2 infection. conflict of interest the authors declare that they have no conflict of interest related to the publication of this article. ethical approval all procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 helsinki declaration and its later amendments or comparable ethical standards. informed consent informed consent was obtained from all individual participants included in the study. the study protocol was submitted to the ethics committee of our referring center. a novel coronavirus from patients with pneumonia in china a novel coronavirus outbreak of global health concern epidemiological and clinical characteristics of 99 cases of 2019 novel coronavirus pneumonia in wuhan, china: a descriptive study the role of imaging in 2019 novel coronavirus pneumonia (covid-19) ct features of coronavirus disease 2019 (covid-19) pneumonia in 62 patients in wuhan, china facing the covid-19 emergency: we can and we do use of ct and artificial intelligence in suspected or covid-19 positive patients statement of the italian society of medical and interventional radiology advoc acyand-econo mics/acr-posit ion-state ments /recom menda tions -for-chest -radio graph y-and-ct-for-suspe cted-covid -19-infec tion fleischner society: glossary of terms for thoracic imaging severity scoring of lung edema on the chest radiograph is associated with clinical outcomes in ards extension of coronavirus disease 2019 (cvid-19) on chest ct and implications for chest radiograph interpretation frequency and distribution of chest radiographic findings in covid-19 positive patients correlation of chest ct and rt-pcr testing in coronavirus disease 2019 (covid-19) in china: a repost of 1014 cases sensitivity of chest ct for covid-19: comparison to rt-pcr the role of chest imaging in patient management during the covid-19 pandemic: a multinational consensus statement from the fleischner society it's not over until it's over: the chameleonic behavior of covid-19 over a six-day period artificial intelligence distinguishes covid-19 from community acquired pneumonia on chest ct ai-driven tools for coronavirus outbreak: need of active learning and cross-population train/test models on multitudinal/multimodal data key: cord-345528-rk16pt0i authors: yasar, y.; karli, b. t.; coteli, c.; coteli, m. b. title: mantiscovid: rapid x-ray chest radiograph and mortality rate evaluation with artificial intelligence for covid-19 date: 2020-05-08 journal: nan doi: 10.1101/2020.05.04.20090779 sha: doc_id: 345528 cord_uid: rk16pt0i the novel coronavirus pandemic has negative impacts over the health, economy and well-being of the global population. this negative effect is growing with the high spreading rate of the virus. the most critical step to prevent the spreading of the virus is pre-screening and early diagnosis of the individuals. this results in quaranteeing the patients not to effect the healthy population. covid-19 is the name of the disease caused by the novel coronavirus. it has a high infection rate and it is urgent to diagnose many patients as we can to prevent the spread of the virus at the early stage. rapid diagnostic tools development is urgent to save lives. mantiscovid is a cloud-based pre-diagnosis tool to be accessed from the internet. this tool delivers a rapid screening test by analyzing the x-ray chest radiograph scans via artificial intelligence (ai) and it also evaluates the mortality rate of patients with the synthesis of the patient history with the machine learning methods. this study reveals the methods used over the platform and evaluation of the algorithms via open datasets. covid-19 is the infectious disease caused by the most recently discovered novel coronavirus. this new virus and the caused disease were unknown before the outbreak began in wuhan, china, in december 2019. covid-19 is now a pandemic (who announced at march 12) affecting many countries globally [1] . covid-19 can appear with many symptoms such as fever, fatigue, dry cough, anorexia, myalgia (muscle ache), dyspnea(shortness of breath), expectoration (the coughing up and spitting out of material from the lungs, bronchi, and trachea, sputum), pharyngalgia, diarrhea (loose, watery stools),nausea, dizziness, headache, vomiting and abdominal pain. respiratory system symptoms are more common, and onset of disease pneumonia appears to be the most frequent serious manifestation of infection, characterized primarily by fever, cough, dyspnea, and bilateral infiltrates on chest imaging [2] [16] . for the diagnosis, researchers define gold standard and lateral methods to overwhelm the pandemic via early diagnosis, but they have disadvantages such as high duration for rt-pcr (reverse transcription polymerase chain reaction), low sensitivity for antigen/antibody tests, high radiation exposure for computer tomography (ct). since these methods have many disadvantages, a rapid diagnosis pre-screening test is crucial to keep the population well-being. physicians' labor is the most important factor to fight with the novel coronavirus. in addition, decision support systems (dss) which are improved with the artificial intelligence (ai) and machine learning methods are timesaving activities for the physicians. a rapid analysis for the chest x-ray (cxr) scans, ct, infection rate or mortality rate with the machine learning methods are some of the helpful tools and researchers are trying to build such tools for pre-screening covid-19. this study defines a deployed environment 1 for rapid evaluation of the mortality rate and cxr scans via machine learning tools. the analysis explained in this study is the submitted work for the #euvsvirus hackathon 2 . the article goes as follows: sec. 2 discusses about the diagnosis methods and their disadvantages. sec. 3 reveals the algorithms in detail. sec. 4 is the evaluation of each algorithm deployed over the platform and sec. 5 concludes the work. rt-pcr is the gold standard to diagnose the covid-19 [3] but it is necessary to keep in mind that there are not enough studies to evaluate the accuracy and predictive values of the tests. sensitivity of testing depends on the precise of rt-pcr assay, type of specimen, duration of illness at testing. in a study at china, 1070 samples collected from 205 patients. bronchoalveolar lavage fluid had the most positive results 14 of 15 patients (93 percent). specimens from upper respiratory tract had fewer positive results. from 326 pharyngeal swabs there were 136 positive results (32 percent) and from 8 nasal swabs there were 5 positive results (63 percent). researchers in this study suggest that testing specimens from multiple sides may improve the sensitivity [4] . according to the statement and the study before it is necessary to do studies for evaluating rt-pcr tests. the second diagnostic tool defined in the literature is ct. according to a study in wuhan, china with 1014 patient's chest ct has 97 percent sensitivity based on positive rt-pcr results. researchers suggest that chest ct could be a primary tool to diagnose covid-19. unfortunately, specificity was 25 percent in other words 75 of 100 patients were false positive [5] . ct scan on pediatric patients should also be avoided due to radiation [6] . besides that, the problem is that resources to diagnose are not sufficient. in the literature, machine learning methods are also applied for the diagnosis through ct images [18] [19] and they are used for supporting the clinician's decisions. however, they can not be applied in large number of samples to exploit the speed of ai interpretation. therefore, lateral and rapid diagnostic tools are required for pre-elimination steps. according to the statement before pneumonia is the most common clinical manifestation and has fatal outcomes. that is the inspiration of solution. in order to diagnose the pneumonia, cxr appears to be a useful tool and to diagnose the community acquired pneumonia. in fact, it is a gold standard with clinical evaluation and microbiologic test results [7] . unfortunately, for covid-19 cxr has some limitations. in a study on 64 patients some patients had no abnormality in their cxr even they had positive initial rt-pcr. cxr may be normal in early or mild disease [8] , but this test can be used more frequently due to low radiation exposure. the health systems of the governments also have enough x-ray scan devices to satisfy the population. in spite of that cxr is a low-cost diagnostic tool and in healthcare chest x-rays are used frequently and we should keep in mind that that study has limitations as researchers stated. that inspires second part of our solution. some researchers also publish their ai architectures according to the detection of covid-19 patients [20] [21] . by evaluating the fact that cxr is not enough to speak about certain diagnosis, a supporting parameter can be provided to the physician for the conditions where ai can not catch the condition. in this study, this parameter is defined as mortality rate that shows the risk of mortality for the person who is infected with covid-19 disease. there are people who are at higher risk for severe illness. those at high-risk from covid-19 are people 65 years and older, people who live in nursing home or long-term care facility. also people of all ages with underlying medical conditions, particularly if not well controlled, including, people with chronic lung disease or moderate to severe asthma, serious heart conditions, immunocompromised (many conditions can cause a person to be immunocompromised, including cancer treatment, smoking, bone marrow or organ transplantation, immune deficiencies, poorly controlled hiv or aids, and prolonged use of corticosteroids and other immune weakening medications), people with severe obesity (body mass index [bmi] of 40 or higher), diabetes ,chronic kidney disease undergoing dialysis, liver disease [9] . there should be a risk evaluation parameter to direct these patients to other diagnostic services and identify these patients who are at the risk. 3 proposed solution for pre-screening covid-19 3 mantiscovid is the hybrid platform proposed via this study. the evaluation platform has two outputs after screening the group of patients as the prediction about the risk in covid-19 via cxr and the mortality rate. two parameters can be used as a decision support system to direct the patient for the other highly sensitive tests. in the case where 1 mantiscovid : https://scan.mantiscope.com/?lang=en 2 https://euvsvirus.org/ 2 . cc-by-nc 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 8, 2020. . https://doi.org/10.1101/2020.05.04.20090779 doi: medrxiv preprint mantiscovid cannot catch covid-19 patient via ai elimination from cxr, the physician can change approaching style to the patient via evaluating the mortality rate. if the patient has no symptoms and negative test results but have a higher mortality rate in that population, it is advised that follow up the patient closely but that advise should be evaluated by clinicians. although forecasting of mortality rate is not easy due to dependence over many conditions, we have tried to estimate the mathematical model as linear regression in terms of 7 parameters via using the open datasets over the patients [10] . the system works over the cloud as a web page and each physician can access to the system for the available computations. the system is designed to be scattered over millions of people to provide scalability via cloud. flow diagram of the platform is seen from fig. 1 . the system has two types of forecasting such as the mortality rate and cxr posteroanterior (pa) chest view via ai elimination. these methods used to implement the processing flow of forecasting the mortality rate is given in the following details. during the development, open source datasets are used and the python code for data synthesis and grouping from these sets is given 3 . the dataset used to compute the mortality rate is from open source covid-19 survival calculator. the mortality rate from dataset were modeled as a linear regression method by connecting it to 7 parameters [10] . the chosen mortality rate factors were age, sex, height, weight, health conditions (asthma, carcinoma, chronic kidney disease compromised immune system, coronary heart disease chronic obstructive lung disease, diabetes, hiv positive, hypertension (high blood pressure), other chronic illness), blood type and smoking type. we have discarded some of mortality rate adjustment parameters to obtain a minimum dataset. the methods used to obtain the linear regression model for the mortality rate are the open source libraries associated with the python scikit-learn [15] . in the presentation of the cxr pa patient elimination ai platform, three types of ai algorithms are available to check whether the person is risky or not. the flow diagram can be seen from the fig. 2 . the sequence has a direct role to detect the risk from x-ray scans. as an aim, this hybrid module is not used for certain diagnosis due to low sensitivity occurring via pa chest view. it is only used as a decision support system for the clinician. the first model is generated to classify whether the input image is x-ray image or not. the model is trained via a dataset containing cxr pa scans with 211 images and images with 210 images [12] [13] . dataset is divided into training (70 percent) and test (30 percent) sets. if an image is detected as besides of x-ray image, the other algorithms are not processed, and it warns the user to check the uploaded image. keras library is used to generate the convolutional neural network (cnn) architecture for the starting point of the classifier [17] . in the input layer, 32 convolutional filters with size 3x3 are inserted with rectified linear unit (relu) activation function. inputs are accepted as 200x200 with red-green-blue (rgb) colors which makes input shape 200x200x3. max pooling layers and 2 hidden layers are inserted with 3x3 filter after convolution. each hidden layer contains 32 convolutional filters with size 3x3 relu activation functions with the addition of max-pooling layer with 2x2 filters. then, flatten operation is applied. fully connected layer is added with 64 units with relu activation function. later, dropout technique is applied with the rate of 0.5 to avoid overfitting. finally, fully connected layer is used with a unit with sigmoid function (1, if the image is x-ray image; 0, otherwise). the classifier ai in the middle revealing whether the patient is risky (covid-19 / pneumonia) or not is the trained model (modified version of alexnet [11] ) according to the open datasets strengthened with the currently obtained covid-19 patients' day by day [12] . the dataset is split into train (0.7), test (0.2) and validation (0.1) to enable the platform for the evaluation. it totally contains 8851 normal, 121 covid-19 and 9579 pneumonia patient data [13] . the ai architecture of the object classifier and training batch sizes are given in fig. 3 . the last part defined as the object detection case is the detection of the lung opacity parts associated with the bacteria type pneumonia. the ai neural network is the trained tiny version of you only look once (yolo) [14] to mark the single objects in cxr pa images. the anomaly is named as lung opacity. the dataset used in this part is the annotated cxr pa obtained through the open source datasets [13] . the system presented in this paper contains 4 different machine learning techniques to provide a rapid analysis of cxr pa view and the mortality rate evaluation. these algorithms are evaluated one by one with the allocated data. each evaluation results can be seen within the following details. 1) linear regression model to forecast the mortality rate: this evaluation contains allocated 10000 patient data gathered from the open source dataset. the mean squared error and the standard deviation between the actual and predicted values are computed as 2.87 and 1.69 respectively. fig. 4 shows the details about the actual and predicted values via testing with 150 conditions. 2) binary classifier to check whether uploaded image is x-ray scan or not: this evaluation contains additional 1000 x-ray and 1030 random images to be used in the evaluation of the binary classifier algorithm. the correctly predicted values for the x-ray scans and random samples are given as 971 and 943 separately. it is observed that the averaged accuracy is about 96 percent. 3) classifier to separate covid-19/pneumonia/normal patients: this part contains randomly selected 150 patient data to be separated via the ai algorithm. the sensitivity of the risky patient detection (covid-19/pneumonia) is 84 percent. the sensitivity for the detection of normal patients is 67 percent. the confusion matrix can be seen in the following fig. 4 . as the main objective is the detection of risky patients in high accuracy, the system conditions can be improved via gathering more patient data which is known as covid-19 certainly. lesion detection over the bacteria type x-ray scans: this part contains randomly selected 30 patient data to be analyzed via the ai algorithm. the evaluation criterion is comparing the number of lesions between the real annotated and predicted x-ray scans. the allocated dataset is divided into three groups as 2 lesions (10 image), 1 lesion (10 image) and no lesion (10 image). the detected and the actual number of lesions over the groups are given as in the 4 . cc-by-nc 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 8, 2020. mantiscovid is a cloud-based pre-screening platform that would be useful for the elimination of non-risky patients in the coronavirus pandemic. since the algorithms are developed and evaluated using the open source data and platforms, this system needs to be evaluated via the clinical perspective by the legal authorities. this platform is an internet-based pre-diagnosis tool to be verified via the universities during the covid-19 diagnosis. however, a physician would also use this tool as an assistant to scan and diagnose the patients due to the rapid diagnosis requirement to prevent spreading of the coronavirus. in fact, the gold standards used to diagnose the covid-19 patients also have high false alarm rates because of the epidemiology of the disease is not fully known. as a future work, the clinical validation of the internet-based platform will start with the known legal authorities and other state of the art ai architectures will be used to increase the sensitivity and specificity of the platform. the integration over the platform will continue with the chest anteroposterior (ap) views, chest ct and ultrasound images by providing novel ai architectures. infection rate for the novel coronavirus would also be evaluated with the usage of tracking the patient's locations. after the coronavirus pandemic, this platform would also be used for scanning the society in terms of the mortality rate and x-ray scans in any other illnesses and symptoms. there should be more studies to identify case fatality rate and infection rate in the case where proper data will be collected to train ai. . cc-by-nc 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted may 8, 2020. . https://doi.org/10.1101/2020.05.04.20090779 doi: medrxiv preprint . cc-by-nc 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted may 8, 2020. . https://doi.org/10.1101/2020.05.04.20090779 doi: medrxiv preprint clinical characteristics of 138 hospitalized patients with 2019 novel coronavirus-infected pneumonia in initial public health response and interim clinical guidance for the 2019 novel coronavirus outbreak-united states detection of sars-cov-2 in different types of clinical specimens correlation of chest ct and rt-pcr testing in coronavirus disease 2019 (covid-19) in china: a report of 1014 cases. radiology the role of ct for covid-19 patient's management remains poorly defined lesion detection algorithm results. from top to bottom, original cxr pa images, annotated images via experts, ai examination results respectively infectious diseases society of america/american thoracic society consensus guidelines on the management of community-acquired pneumonia in adults frequency and distribution of chest radiographic findings in covid-19 positive patients. radiology covid19 survival calculator squeezenet: alexnet-level accuracy with 50x fewer parameters and < 0.5 mb model size covid-19 image data collection chestx-ray8: hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases yolov3: an incremental improvement machine learning for neuroimaging with scikit-learn clinical features of patients infected with 2019 novel coronavirus in wuhan, china. lancet hands-on machine learning with scikit-learn, keras, and tensorflow: concepts, tools, and techniques to build intelligent systems. o'reilly media artificial intelligence distinguishes covid-19 from community acquired pneumonia on chest ct rapid ai development cycle for the coronavirus (covid-19) pandemic: initial results for automated detection patient monitoring using deep learning ct image analysis automatic detection of coronavirus disease (covid-19) using x-ray images and deep convolutional neural networks covid-net: a tailored deep convolutional neural network design for detection of covid-19 cases from chest radiography images key: cord-238881-tupom7fb authors: yeh, chun-fu; cheng, hsien-tzu; wei, andy; chen, hsin-ming; kuo, po-chen; liu, keng-chi; ko, mong-chi; chen, ray-jade; lee, po-chang; chuang, jen-hsiang; chen, chi-mai; chen, yi-chang; lee, wen-jeng; chien, ning; chen, jo-yu; huang, yu-sen; chang, yu-chien; huang, yu-cheng; chou, nai-kuan; chao, kuan-hua; tu, yi-chin; chang, yeun-chung; liu, tyng-luh title: a cascaded learning strategy for robust covid-19 pneumonia chest x-ray screening date: 2020-04-24 journal: nan doi: nan sha: doc_id: 238881 cord_uid: tupom7fb we introduce a comprehensive screening platform for the covid-19 (a.k.a., sars-cov-2) pneumonia. the proposed ai-based system works on chest x-ray (cxr) images to predict whether a patient is infected with the covid-19 disease. although the recent international joint effort on making the availability of all sorts of open data, the public collection of cxr images is still relatively small for reliably training a deep neural network (dnn) to carry out covid-19 prediction. to better address such inefficiency, we design a cascaded learning strategy to improve both the sensitivity and the specificity of the resulting dnn classification model. our approach leverages a large cxr image dataset of non-covid-19 pneumonia to generalize the original well-trained classification model via a cascaded learning scheme. the resulting screening system is shown to achieve good classification performance on the expanded dataset, including those newly added covid-19 cxr images. abstract. we introduce a comprehensive screening platform for the covid-19 (a.k.a., sars-cov-2) pneumonia. the proposed ai-based system works on chest x-ray (cxr) images to predict whether a patient is infected with the covid-19 disease. although the recent international joint effort on making the availability of all sorts of open data, the public collection of cxr images is still relatively small for reliably training a deep neural network (dnn) to carry out covid-19 prediction. to better address such inefficiency, we design a cascaded learning strategy to improve both the sensitivity and the specificity of the resulting dnn classification model. our approach leverages a large cxr image dataset of non-covid-19 pneumonia to generalize the original well-trained classification model via a cascaded learning scheme. the resulting screening system is shown to achieve good classification performance on the expanded dataset, including those newly added covid-19 cxr images. more specifically, the proposed dnn learning proceeds in three stages. in the first stage, the model is trained to predict the mask of lung regions to emphasize the targeted areas of concern and alleviate the effect of irrelevant annotations on a cxr image. incremental learning is then deployed in the subsequent stage so that the pre-trained dnn can learn to classify the additional covid-19 images, while retaining its classification performance on the original data. the design is to first filter out images of the normal category in the second stage and then, in the final stage, conduct fine-grained classification to divide the pneumonia candidates into two specific types, namely, covid-19 and non-covid-19. we report promising results on both the open and clinical covid-19 datasets. the outbreak of covid-19 disease (a.k.a., sars-cov-2) has been affecting the world in an unprecedented way. while intensively global research efforts are being made to seek its effective treatments or vaccines, at the core of the most urgent concern is to prevent the pandemic from further spreading into an uncontrollable and chaotic status. in this work, we are endeavoring to establish an open ai-based platform to seamlessly carry out preliminary large-scale screening of potential covid-19 patients, and provide early detection at the onset of the infection from the viewpoint of radiology imaging. countries such as the united states and japan are now actively practicing social distancing and seem to receive encouraging outcomes to curb further spread of covid-19. still, an affordable and reliable procedure to effectively screen the potentially infected patients is much needed. the reverse-transcription polymerase chain reaction (rt-pcr) testing is the preliminary evaluation to assess whether a subject is at risk of covid-19. however, the less satisfactory sensitivity of rt-pcr has its limitation on reducing the false-negative rate [1] and may overlook a good number of covid-19 patients, especially in their early stage of infection. it is essential to find a complementary testing to boost the screening confidence and chest radiology, especially chest x-ray (cxr), with the advantages of affordability, efficiency and reliability, is promising in this regard. as will be demonstrated in the experiments, the proposed ai-based cxr screening system of covid-19 could effectively detect the infection with high sensitivity, even several days prior to the confirmed rt-pcr results. compared with chest computer tomography (ct), chest x-ray imaging is more suitable for being incorporating into a large-scale ai-based screening platform for covid-19. the supporting evidence is threefold. first, performing ct scanning for comprehensive screening is not practical to contain a pandemic outbreak in that even the well-established healthcare system of a developed country simply does not have such a capacity to do so. second, cleaning the ct scanning equipment takes considerably longer time than the case of using chest x-ray. the deep cleaning is necessary to prevent subsequent patients from being infected of covid-19. third, chest x-rays are already widely adopted as a de facto screening procedure, while ct scanning equipment is not equally popular and is mostly available in primary healthcare institutes. their distinction in public use also reflects in the availability of large-scale open datasets of chest x-ray images over ct. the ease of collecting image data is crucial in training a reliable ai-based screening system, especially under the current circumstance. the proposed chest x-ray screening platform leverages the pneumonia classification system developed by taiwan ai labs to detect covid-19. to extend the model, we collaborate with several medical research centers in taiwan to collect chest x-ray images from covid-19 patients at various stages, and re-train the pneumonia classification system using a three-stage cascaded learning strategy. specifically, in the first stage, the ai model is trained to predict from a given cxr image the regions of interest, i.e., the mask of lung regions. in the second stage, features from the predicted regions are extracted to decide whether the image is a positive case of pneumonia. in the final and the third stage, the platform would further make decision on whether the underlying case is covid-19 or another type of pneumonia. the resulting screening system also yields the stage-wise predicted heatmaps and thus provides explainable image clues leading to the classifications. to demonstrate the effectiveness of the proposed platform for covid-19 screening, we report the inference results from both the open datasets and those from the collected clinical cases. we denote the dataset for learning the classification network as , where x i is a chest x-ray image and y i is the pneumonia class label. in our implementation, we have y i = 0 to reflect that x i is a cxr image of normal case and y i ∈ {1, 2} for the covid-19 and non-covid-19 pneumonia, respectively. the training set can be further decomposed as d = d o ∪ d c and |d o | |d c | to indicate that d o is the original (large) collection for conventional pneumonia classification and d c includes all the covid-19 cxr images. the overall network architecture of the proposed three-stage cascaded learning is illustrated in fig. 1 . we detail the design details as follows. to prevent an ai-based system from learning inconsistent information or noisy annotations, we design the pneumonia classification system to focus on the essential regions of interest, which in our case is the lung areas. we consider a u-net model [2] to first predict a mask of the lung regions in a cxr image, and use the resulting mask to filter out image areas that could distract the ai system from learning the crucial pneumonia-relevant features for the underlying classification problem. simply put, the task of the first stage, as shown in fig. 2(c) , can be thought of as a preprocessing step to segment the mask of the lung regions in each given cxr image. pneumonia classification. having excluded non-informative regions from each cxr image x i , the task of the succeeding second stage of the proposed cascaded learning is to carry out binary classifications: normal versus pneumonia. that is, the densenet-121 [3] pneumonia classifier is trained to predict x i as negative if y i = 0, and otherwise positive if y i ∈ {1, 2}. under this design setting, all cxr images of pneumonia are expected to be classified as positive no matter what their type is. after all, the goal of this stage is to filter out non-pneumonia samples from further considerations. it is constructive to explain the pneumonia classification outcome of a cxr image x i by using the cam [4] interpretation technique. as shown in fig. 1 , performing global average pooling (gap) yields a 1024-d feature vector f 2 i , which can be reshaped into a 32 × 32 tensor h 2 i and then upsampled to h 2 i of the original input resolution, 512 × 512. we express the cam modeling at stage 2 as follows. the heatmap h 2 i derived in (1) can be interpreted as the importance distribution over the lung regions leading to the pneumonia classification outcome of x i . intuitively, the ai-based readout of a cam heatmap can be used to aide physicians to identify anomalous spots in the input cxr image x i . recall that the training dataset d comprises d o and d c . the latter includes the covid-19 images and is expected to expand constantly when new cxr samples are provided by our collaborative medical centers. the frequent changes in the available training data may require re-training the ai model to improve classification performance on covid-19 images, while it also may degrade the overall pneumonia classification in stage 2. to overcome this dilemma, we deploy incremental learning to ensure a robust model learning in stage 2. we first adopt a pneumonia classification model by taiwan ai labs, which is pre-trained on d o , and then perform incremental learning on d c by optimizing the model parameters θ with respect to the following loss function: where ce is the cross-entropy loss, kl is the knowledge distillation loss for learning the prediction output of the pre-trained model and λ ≥ 0 is a parameter to weigh the effect of knowledge distillation. when λ is set to 0, the incremental learning is simply reduced to re-training the model with the updated data d. covid-19 screening. at the stage 3 of the cascaded learning, the task is to distinguish the specific type of pneumonia: covid (y = 1) and non-covid (y = 2). notice that the dataset d o solely comprises non-covid cxr images, which could correspond to normal cases and viral or bacterial pneumonia. we can write the input to the stage 3 classification as where symbolizes pixel-wise product. in fig. 1 , an example ofx i is illustrated as a mask of informative regions relevant to predicting an input cxr image x i as a covid-19 or non-covid-19 pneumonia case at the stage 3. analogous to the derivation of h 2 i in (1), we can explore the heatmap responses to gain insights into how the classification system distinguishes covid-19 from other types of pneumonia. we consider gradcam [5] to generate the heatmap h 3 i in that the resulting gradcam responses tend to be concentrated rather than diffusive. the heatmap h 3 i yields interpretable image cues relevant to how likely the ai system predicts the pneumonia case of x i as covid-19. inspired by [5] , we also generate guided-gradcam (guided activation) as detailed responses to further focus on patterns to distinguish covid-19 pneumonia. in collaborations with ntuh, tmuh and nhia, we evaluate our method on both open and clinical collections of cxr images. details about the datasets are listed in table 1 and table 2 . we learn the model of each stage separately, i.e., all parameters other than those of the current training stage are kept unchanged. lung segmentation masking. the preprocessing stage 1, lung segmentation, is trained on tuberculosis chest x-ray image data sets that consist of two parts, montgomery county x-ray set and shenzhen hospital x-ray set. the total number of samples contains 406 normal case images and 394 tuberculosis case images. we randomly sample 80% as our training/validation sets and 20% as our test set. our stage 1 model achieves 0.88 dice similarity coefficient (dsc) on the test set. fig. 3 shows some segmentation outcomes from our stage 1 model. we further investigate the effect of stage 1 considering the whole screening pipeline. table 3 compares the pneumonia classification (stage 2) results between models with lung segmentation masking and models without lung segmentation masking. table 1 . we separate the data into three groups: noraml (n), pneumonia (p, non-covid-19), and covid-19 (c). normal and non-covid-19 pneumonia cxr images are collected from padchest [7] and rsna [8] , while covid-19 samples are selected from [6] . for the data from padchest and rsna, we randomly divide it into training/validation/testing in the ratio of 80%/10%/10%. as to the data of covid-19, we split it into training/validation/testing in the ratio of 50%/25%/25%, due to its small sample size. during stage 2, binary labels are considered, where non-covid-19 pneumonia (p) and covid-19 (c) samples are labeled as 1 (positive) and normal samples are labeled as 0 (negative). for each batch of training, we evenly sample each group to account for imbalanced distributions among normal, non-covid-19 pneumonia, and covid-19 samples. the masked images by stage 1 model (lung segmentation, as in fig. 3 ) are used as the input of stage 2 model for both training and inference. after achieving results of pneumonia classification of stage 2 as in table 4 , we then train the stage 3 model to distinguish covid-19 from non-covid-19 pneumonia, using only non-covid-19 pneumonia (p) and covid-19 (c) training samples. all parameters but those of stage 3 are fixed during training. in this stage, marked as "3: covid-19" in table 4 , covid-19 (c) samples are labeled as 1 (positive) while non-covid-19 pneumonia (p) samples are labeled as 0 (negative). table 4 shows fine-tune stage 2 and 3 on clinical datasets (ntuh and tmuh). from table 4 , we observe the performance gaps in both validation and testing between clinical data and open data. (we will further elaborate this discovery in the next section.) to further improve the ai-based classification performance and achieve clinical applicability, we collaborate with ntuh, tmuh, and nhia to collect more clinical covid-19 cxr samples, as summarized in table 2 . starting from a pre-trained model using only open datasets, we fine-tune our model using clinical data. instead of using only clinical data, we combine open datasets and clinical datasets to prevent from overfitting during our fine-tuning phase, and consider the incremental learning strategy as in (2) . as in the experiments on open datasets, we fine-tune our pneumonia classification (stage 2) and covid-19 screening (stage 3) sequentially. based on the final results shown in table 5 , our method is able to recognize normal, pneumonia (both non-covid-19 and covid-19), non-covid-19, and covid-19 cases respectively with high auc, sensitivity, and specificity. for qualitative result, those cases sampled from the testing set of open dataset in fig. 6 and the clinical dataset in fig. 5 demonstrate reasonable model activation in different stages. fine-tune stage 3 on nhia dataset. to evaluate the clinical applicability of our method, we fine-tune our stage 3 model on a relatively larger dataset (nhia), which includes 306 covid-19 (c) samples from 306 confirmed cases in taiwan and 3,069 non-covid-19 pneumonia (p) samples. of those 3,069 pneumonia (p) samples, 50% of them are diagnosed with bacterial pneumonia and the other half are diagnosed with viral pneumonia. table 6 includes the validation and testing results on nhia dataset, suggesting the potential of using our method for distinguishing covid-19 from both common bacterial and viral pneumonia. early covid-19 detection with our screening system further, to gain insights into the advantages of using the proposed cxr screening platform, we investigate its efficiency of early prediction over rt-pcr so that the needed medical treatment can be performed as early as possible. to this end, we describe the overall performance of our screening platform and the case study on the nhia dataset, particularly the validation (n=54) and the test (n=55) set. out of these 109 samples detected by rt-pcr, there are 47 samples with visible signs of pneumonia, which are double confirmed by our medical specialist. based on this subset, our model correctly detects 42 samples; that is, the sensitivity of our model achieves 89.36%. out of these 109 samples, taken in a period of three months (january 2020 to march 2020), the advantage of using our system for early detection of covid-19 versus using rt-pcr can be observed from in the comparative study. fig. 4 illustrates the details of each case. -case 24 was detected by our system, 17 days prior to rt-pcr confirmed (on january 31, 2020). comparison with relevant studies. to evaluate the efficacy of our multistage method, we further conduct an experiment on the training/testing data split setting of covid-net [9] . the cxr samples are taken entirely from open datasets [6, 8] , which consist of 13,594 cxr images (7966/5476/152 in n/p/c) in the training set and 231 cxr images (100/100/31 in n/p/c) in the testing set. as shown in table 7 , our method is more sensitive (and specific at inverse perspective of binary classification) than covid-net in both normal and covid-19 predictions. note that in the stage 2 of our method, the binary classification is to predict normal versus both covid-19 and non-covid-19 pneumonia. since the outbreak of covid-19, intensive efforts have being made by healthcare experts in hospitals to reach diagnosis result for each patient even with supplemental symptoms. suggested by medical experts, because of the efficiency and availability of cxr, we propose an ai-based screening system to recognize covid-19 pneumonia in cxr images. regarding coarse to fine manners, our cascaded method consists of lung segmentation, pneumonia recognition, and covid-19 recognition as hierarchical screening. the proposed approach outperforms a previous method on open dataset of covid-19 cases and is able to reach clinical-grade performance on ntuh and tmuh clinical data. moreover, our method has been integrated into the internal system of taiwan nhia and cdc, achieving over 80% sensitivity and specificity on nhia clinical test cases. key future research challenges include the sensitivity improvement on cases with mild symptoms, medical studies of covid-19 cases with recognized lesion patterns, and the refinement of lung segmentation to uncover subtle and relevant regions. sensitivity of chest ct for covid-19: comparison to rt-pcr u-net: convolutional networks for biomedical image segmentation densely connected convolutional networks learning deep features for discriminative localization grad-cam: visual explanations from deep networks via gradient-based localization covid-19 image data collection padchest: a large chest x-ray image dataset with multi-label annotated reports rsna pneumonia detection challenge covid-net: a tailored deep convolutional neural network design for detection of covid-19 cases from chest radiography images we would like to thank national taiwan university hospital (ntuh), taipei medical university hospital (tmuh), and taiwan national health insurance administration (nhia) to provide clinical covid-19 data for our studies. the experts from ntuh also provide us insightful feedback about medical findings and results evaluation on cxr images. for the deployment of our covid-19 screening system, we appreciate the support from taiwan centers for disease control and taiwan executive yuan, letting us contribute to the safeguard of taiwan citizens' well-being during this global pandemic. key: cord-282198-ugmv9om1 authors: pare, joseph r.; camelo, ingrid; mayo, kelly c.; leo, megan m.; dugas, julianne n.; nelson, kerrie p.; baker, william e.; shareef, faizah; mitchell, patricia m.; schechter-perkins, elissa m. title: point-of-care lung ultrasound is more sensitive than chest radiograph for evaluation of covid-19 date: 2020-06-19 journal: west j emerg med doi: 10.5811/westjem.2020.5.47743 sha: doc_id: 282198 cord_uid: ugmv9om1 introduction: current recommendations for diagnostic imaging for moderately to severely ill patients with suspected coronavirus disease 2019 (covid-19) include chest radiograph (cxr). our primary objective was to determine whether lung ultrasound (lus) b-lines, when excluding patients with alternative etiologies for b-lines, are more sensitive for the associated diagnosis of covid-19 than cxr. methods: this was a retrospective cohort study of all patients who presented to a single, academic emergency department in the united states between march 20 and april 6, 2020, and received lus, cxr, and viral testing for covid-19 as part of their diagnostic evaluation. the primary objective was to estimate the test characteristics of both lus b-lines and cxr for the associated diagnosis of covid-19. our secondary objective was to evaluate the proportion of patients with covid-19 that have secondary lus findings of pleural abnormalities and subpleural consolidations. results: we identified 43 patients who underwent both lus and cxr and were tested for covid-19. of these, 27/43 (63%) tested positive. lus was more sensitive (88.9%, 95% confidence interval (ci), 71.1–97.0) for the associated diagnosis of covid-19 than cxr (51.9%, 95% ci, 34.0–69.3; p = 0.013). lus and cxr specificity were 56.3% (95% ci, 33.2–76.9) and 75.0% (95% ci, 50.0–90.3), respectively (p = 0.453). secondary lus findings of patients with covid-19 demonstrated 21/27 (77.8%) had pleural abnormalities and 10/27 (37%) had subpleural consolidations. conclusion: among patients who underwent lus and cxr, lus was found to have a higher sensitivity than cxr for the evaluation of covid-19. this data could have important implications as an aid in the diagnostic evaluation of covid-19, particularly where viral testing is not available or restricted. if generalizable, future directions would include defining how to incorporate lus into clinical management and its role in screening lower-risk populations. novel coronavirus, sars-cov-2, is responsible for causing the coronavirus disease 2019 . with an estimated case fatality rate of 1%, covid-19 has resulted in over 305,000 deaths worldwide to date. 1 covid-19's mortality pocus is more sensitive than cxr for evaluation of pare et al. what do we already know about this issue? lung ultrasound (lus) has been shown to outperform chest radiograph (cxr) in its ability to detect abnormalities with non-coronavirus disease 2019 (covid-19) pulmonary infections. what was the research question? to determine if b-lines detected by lus are more sensitive for the associated diagnosis of covid-19 than an abnormal cxr. what was the major finding of the study? b-lines detected by lus were more sensitive for the associated diagnosis of covid-19 than an abnormal cxr. in locations where viral testing is not available or has significant delays, lus may provide important information for the evaluation of suspected covid-19. is primarily due to lung injury resulting in acute respiratory distress syndrome (ards). 2 the definition of ards has changed over time; however, using the 2012 berlin definition it would include acute bilateral lung injury in the absence of fluid overload, causing hypoxemia and respiratory failure. 3 physicians evaluating patients may wish to order radiographic imaging to screen for findings of covid-19, evaluate severity of pulmonary involvement, or assess for alternative etiologies of illness. radiographic results may alter the treating physician's concern for covid-19 thereby guiding patient counseling, or supporting clinical choices such as hospitalization, the need for closer follow-up, or anticipating complications of the disease. the american college of radiology (acr) recommended the use of portable chest radiograph (cxr) when medically necessary for patients with suspected or known covid-19, which does not include screening purposes. 4 however, it is estimated that portable cxr is only 69% sensitive for findings of covid-19. 5 when compared to cxr, lung ultrasound (lus) may offer improved diagnostic accuracy in the evaluation of patients with suspected covid-19 pneumonia. lus has a high sensitivity and often out-performs cxr in the diagnosis of other pulmonary infections. 6 lus findings for covid-19 have been reported in the literature and include b-lines, pleural abnormalities, and subpleural consolidations. [7] [8] [9] evaluation of b-lines is already within the scope of practice for emergency physicians (ep), and instruction in interpreting lus is part of current residency education standards. 10 lus is a safe, readily available tool that can be employed by eps to provide real-time clinical assessment for covid-19. lab testing utility is hampered by delays in results, accuracy, and availability. cxr may miss pulmonary disease, and the acr has cautioned against routine screening with chest computed tomography (ct), citing concerns of poor specificity of ground-glass opacities for covid-19 as well as infection control procedures necessary to decontaminate the ct scanner. 4 regarding infection control procedures, we expect that portable (or hand-held) ultrasounds would be easier to decontaminate than portable cxr machines or ct suites. our primary aim was to determine whether detection of b-lines on lus, among patients without alternative etiologies for their presence, is more sensitive for the diagnosis of covid-19 than cxr. our secondary aim was to evaluate the proportion of patients with covid-19 that have secondary lus findings of pleural abnormalities and subpleural consolidations. this was a retrospective, observational, cohort study of patients undergoing covid-19 testing (based on real-time reverse transcriptase-polymerase chain reaction [rt-pcr] of nasopharyngeal sampling performed on an assay developed by the center for regenerative medicine at boston university, operating under an emergency use authorization], who also had both diagnostic lus and cxr for the evaluation of covid-19 in the emergency department (ed). this study had institutional review board approval and was conducted based on standards for reporting of diagnostic accuracy studies (stard) guidelines and best practices for retrospective reviews. 11 this investigation was performed at a large urban academic ed in the united states with >140,000 visits per year. the ed is associated with an emergency medicine residency and clinical ultrasound fellowship, and has six dedicated portable ultrasound machines (philips sparq, wayne, pa; and mindray te7, arnold, md). all ultrasound studies are transferred wirelessly and stored in qpath (telexy, blaine, wa). there was no formal education for lus specific to covid-19; however, all physicians have had structured training in lus. all physicians were provided literature from a small study of 20 patients with covid-19 that had 12 lung zones evaluated with ultrasound, which found 75% of patients had abnormal lus findings at the posterior lung bases. 9 when performing point-of-care ultrasound in the clinical setting, all eps at our institution are required to archive at least one image that is representative of their findings. pocus is more sensitive than cxr for evaluation of covid-19 all ultrasound studies completed in the ed between march 20, 2020-april 6, 2020, were reviewed for lus imaging. we reviewed the electronic health record (ehr), epic (verona, wi) to determine whether covid-19 testing was performed. subjects were included for evaluation if they had a covid-19 test performed during the index hospitalization or within two weeks of the lus examination. at the hospital during this time period, covid-19 testing was performed only on people with symptoms concerning for disease, and no routine screening practices were in place. however, performance of viral testing was at physician discretion, and those without viral testing were excluded from analysis. we also excluded subjects if they did not have a cxr. lastly, based on ehr review from patient history or physician documentation, patients were excluded if they had reasons for alternative causes of b-lines (congestive heart failure, renal disease leading to volume overload, or underlying lung disease), as it would not be possible to determine the etiology of the abnormal ultrasound results. all lung ultrasounds were reviewed by two expert eps, both with clinical ultrasound fellowship training (jrp and kcm), who were blinded to covid-19 results. when disagreements occurred, a third ultrasound fellowship-trained, blinded independent expert reviewer adjudicated (mml). lus were scored as positive or negative after review of all images. subjects were considered to have a positive lus if any b-lines were detected. the reviewers further graded positive ultrasounds as having 1-2 b-lines or ≥3 b-lines. 12 if b-lines coalesced, the score was graded as ≥3 b-lines if the area of b-lines took up ≥30% of the intercostal space. although ground-glass opacities can manifest as thinner b-lines <3mm apart, we allowed for percentage grading to account for coalescing in addition to "light beam" artifact, which is a broader, band-shaped artifact described in covid-19. 13 because covid-19 is reported to cause focal and diffuse lung disease, we chose the image with the most b-lines detected at one intercostal space to score each patient. the images were subsequently evaluated for subpleural consolidations and pleural abnormalities ( figure 1 and online supplemental videos a-e). we defined subpleural consolidations as an area of hypoechoic focus at the pleural line. these areas may be associated with increased b-lines originating from this area of hypoechoic focus. for pleural abnormalities we defined this as a) loss of pleural line echogenicity; b) irregular contour of the pleural line; or c) areas that appeared >3 millimeters in thickness by visual estimation. 14 secondary lus findings were determined by a consensus of all reviewers. finalized cxr reports were recorded. we classified cxrs as positive if the report included infection in the differential, as defined by words such as opacity, consolidation, or airspace disease. cxrs were classified as negative if no abnormality was noted, an abnormality was noted but attributed to a non-infectious etiology, or was inconclusive for infectious process. after lus scoring and data collection, clinical data including demographics, co-morbidities, vital signs, and laboratory values, was collected from the ehr by two investigators (jrp and fs) using a standardized abstraction technique and entered into redcap. the primary outcome measure was the sensitivity of lus compared to cxr for the detection of covid-19, using the rt-pcr laboratory test as the reference standard. secondary outcome measures were the proportion of additional secondary lus findings (pleural abnormalities or subpleural consolidation) detected. a sample size of 43 patients with an estimated sensitivity of 40% for cxr and 70% for lus yields 81% power with an alpha of 0.05 assuming 70% disease prevalence. we used an estimated sensitivity of 40% based on results of cxr findings in influenza, as the referenced paper of 69% was not available at the time this study was designed. 5, 15 we compared sensitivities of lus and cxr using a two-sided mcnemar's test. patient demographics were evaluated with descriptive statistics, fisher's exact tests, wilcoxon sum-ranked test, chi-squared tests, and welch's t-test. inter-rater reliability for pocus is more sensitive than cxr for evaluation of covid-19 pare et al. the primary outcome between the two primary reviewers was assessed by cohen's kappa. 16 a total of 304 ultrasound studies were completed over the 18-day study period (figure 2 ). of these, 81 had lus performed. among these, 43 met inclusion criteria, and 27/43 tested positive for covid-19 by rt-pcr (63%). four patients admitted with initial negative results were retested, and two were found to be positive. these two subjects were classified in the 27 total patients with covid-19. table 1 describes the demographic and clinical information of the included patients. the sensitivity and specificity of b-lines on lus associated with covid-19 were 88.9% (95% ci, 71.1-97.0) and 56.3% (95% ci, 33.2-76.9), respectively. the association between cxr and covid-19 results had a sensitivity and specificity (appendix) of 51.9% (95% ci, 34.0-69.3) and 75.0% (95% ci, 50.0-90.3). lus was more sensitive than cxr for the association of pulmonary findings of covid-19 (p = 0.013). while there was a trend for cxr to be more specific for the associated diagnosis of covid-19, this was not found to be statistically significant (p = 0.453). additional lus test characteristics are provided in table 2 . cohen's kappa for interrater agreement between the two expert lus reviewers for the primary outcome was strong (κ = 0.83, 95% ci, 0.65-1.00). there were only three cases out of 43 where there was disagreement on the primary outcome between the two reviewers. these involved cases where b-lines were more subtle. b-lines were more frequently detected in patients with covid-19 (24/27 patients with covid-19 and 7/16 patients without, p < 0.001). of the 27 patients with confirmed covid-19 infection, 21 had pleural abnormalities (77.8%) and 10 had subpleural consolidations (37%). of the 16 subjects without covid-19, three had pleural irregularities (18.8%) and two had subpleural consolidations (12.5%). there was a mean of 6.2 lus images recorded per patient, which was not significantly different between covid-19 results, and a median of 6 lus images taken per patient. images were more frequently obtained with a curvilinear probe 37/43, (86%), than the phased array probe, 6/43 (14.0%). of the lus studies, 8/43 (18.6%) were completed by residents or physician assistants, 4/43 (9.3%) by an ultrasound fellow, 17/43 (39.5%) by ultrasound faculty, and 14/43 (32.6%) by non-fellowship trained eps. of the cxrs performed, 42/43 (97.7%) were performed as portable examinations. the one 2-view cxr was a false negative. to our knowledge this is the first study to evaluate the test characteristics of lus for covid-19. we also are the first to compare the diagnostic performance of lus to the more conventional use of cxr. although preliminary, this work provides important results for the application of lus for detection of covid-19. this investigation offers compelling evidence that b-lines detected by lus are more frequently associated with covid-19 than an abnormal cxr. this finding is in line with the performance of lus in other pulmonary disease entities. 6, 10 we used rt-pcr as the reference standard for diagnosis of covid-19. however, it is known that the test characteristics of rt-pcr are dependent on collection technique, timing in disease process, and processing technique. in our population there were two negative rt-pcr tests that were positive on repeat testing. both patients with initially negative rt-pcr tests had positive lus findings; thus, it is possible lus is more sensitive than rt-pcr for covid-19. further research would be necessary to substantiate this theory. our study reports a sensitivity of 52% for cxr, which is lower than the reported 69% for portable cxr. it is unknown whether the radiologists in that previous study were blinded, and it is also unclear how body mass index or other variables may have resulted in our reported lower sensitivity for cxr. it is unknown how two-view cxrs would perform for the detection of lung involvement from covid-19, as it might evidence that lus is more sensitive for the associated diagnosis of covid-19 than cxr has potential global implications. these results may be of particular importance to settings with significant delays in viral rt-pcr testing, settings in which rt-pcr testing is restricted or not available, or where cxr or ct are not accessible. further scientific investigation could determine how lus at the time of initial evaluation may aid the physician in counseling patients with regard to findings suggestive of covid-19. our investigation provides important new data for the role of lus relative to cxr for patients being evaluated for covid-19. conversely, lus did have a lower specificity than cxr. as noted, 1-2 b-lines may be non-pathologic; however, only one patient in this study was found to have 1-2 b-lines that did in fact have covid-19. it is possible that using lus with only one or two b-lines to direct care for patients suspected of having covid-19 could lead to unnecessary isolation or further medical testing. additionally, there are other etiologies for lus b-lines, and our results will likely be most valuable when interpreted in the clinical context of the medical evaluation. physicians should have an estimation of pretest probability when performing and interpreting diagnostic testing, and lus for covid-19 is no exception to this rule. in this population with a high prevalence of disease (as judged by rt-pcr results), a positive lus was a good predictor of disease. further work is necessary to better delineate how to incorporate these findings into screening for asymptomatic patients, diagnostic algorithms, and clinical management strategies. since this was a retrospective study, it is unclear why physicians chose to perform both cxr and lus. it is also unknown whether the result of either diagnostic test affected the physician's choice to perform the other test. additionally, the treating physician was not blinded to the patient's history, exam, or cxr. it is possible that knowledge of these data points would change the extent to which the physician performed their lus. despite this, there were a similar number of images recorded for patients with and without covid-19. over half of the studies performed were performed by non-fellowship trained eps. further work is needed to validate these findings in a population of eps without fellowship training. identification of b-lines is a core skill of eps; therefore, we anticipate the findings would be similar. another limitation was the use of rt-pcr for the diagnosis of covid-19, as it likely misses some cases. some of the tests classified as false positive may have actually been true positives. rt-pcr was chosen as the reference standard since that is what is currently used at our, and most, institutions nationally, and viral culture is not feasible at this time. inconclusive cxrs were scored as negative, which might favor the analysis toward lus. this was done, in accordance with stard guidelines, because inconclusive cxrs do not provide diagnostic guidance in real time. 11 we used b-lines in this study as a reliable marker for covid-19. it is possible a comprehensive evaluation including pleural abnormalities and subpleural consolidations would improve the test characteristics of lus. we chose to only include b-lines for our assessment as b-lines are already familiar to eps and would be easier to implement. we included any number of b-lines (one or more) as abnormal; however, it has been reported 1-2 b-lines may not be pathologic. we selected this approach to maximize the sensitivity of lus at the cost of specificity. this investigation provides evidence that lus is more sensitive for the associated diagnosis of covid-19 than cxr when excluding patients with other expected causes of b-lines. this work could have important implications where viral testing is restricted or alternative diagnostic imaging is not available. further work may find lus for the evaluation and care of covid-19 patients to be of clinical benefit and may also have a role to guide testing as screening and contact tracing are expanded. the many estimates of the covid-19 case fatality rate clinical characteristics of 138 hospitalized patients with 2019 novel coronavirus-infected pneumonia in wuhan, china acute respiratory distress syndrome: the berlin definition american college of radiology. acr recommendations for the use of chest radiography and computed tomography (ct) for available at: https://www. acr.org/advocacy-and-economics/acr-position-statements/ recommendations-for-chest-radiography-and-ct-for-suspected-covid19-infection frequency and distribution of chest radiographic findings in covid-19 positive patients accuracy of lung ultrasonography versus chest radiography for the diagnosis of adult community-acquired pneumonia: review of the literature and meta-analysis can lung us help critical care clinicians in the early diagnosis of novel coronavirus (covid-19) pneumonia? findings of lung ultrasonography of novel corona virus pneumonia during the 2019-2020 epidemic a preliminary study on the ultrasonic manifestations of peripulmonary lesions of non-critical novel coronavirus pneumonia (covid-19) diagnosing acute heart failure in the emergency department: a systematic review and metaanalysis guidelines for reporting diagnostic accuracy studies: explanation and elaboration relevance of lung ultrasound in the diagnosis of acute respiratory failure: the blue protocol what's new in lung ultrasound during the covid-19 pandemic transforming growth factor beta-1 as a predictor of fibrosis in tuberculous pleurisy pulmonary imaging of pandemic influenza h1n1 infection: relationship between clinical presentation and disease burden on chest radiography and ct a coefficient of agreement for nominal scales key: cord-297396-r1p7xn3a authors: ng, ming-yen; wan, eric yuk fai; wong, ho yuen frank; leung, siu ting; lee, jonan chun yin; chin, thomas wing-yan; lo, christine shing yen; lui, macy mei-sze; chan, edward hung tat; fong, ambrose ho-tung; yung, fung sau; ching, on hang; chiu, keith wan-hang; chung, tom wai hin; vardhanbhuti, varut; lam, hiu yin sonia; to, kelvin kai wang; chiu, jeffrey long fung; lam, tina poy wing; khong, pek lan; liu, raymond wai to; man chan, johnny wai; ka lun alan, wu; lung, kwok-cheung; hung, ivan fan ngai; lau, chak sing; kuo, michael d.; ip, mary sau-man title: development and validation of risk prediction models for covid-19 positivity in a hospital setting date: 2020-09-15 journal: int j infect dis doi: 10.1016/j.ijid.2020.09.022 sha: doc_id: 297396 cord_uid: r1p7xn3a objectives: to develop:(1) two validated risk prediction models for covid-19 positivity using readily available parameters in a general hospital setting; (2) nomograms and probabilities to allow clinical utilisation. methods: patients with and without covid-19 were included from 4 hong kong hospitals. database was randomly split 2:1 for model development database (n = 895) and validation database (n = 435). multivariable logistic regression was utilised for model creation and validated with the hosmer-lemeshow (h-l) test and calibration plot. nomograms and probabilities set at 0.1, 0.2, 0.4, 0.6 were calculated to determine sensitivity, specificity, positive predictive value (ppv) and negative predictive value (npv). results: 1330 patients (mean age 58.2 ± 24.5 years; 50.7% males; 296 covid-19 positive) were recruited. first prediction model developed had age, total white blood cell count, chest x-ray appearances and contact history as significant predictors (auc = 0.911 [ci = 0.880-0.941]). second model developed has same variables except contact history (auc = 0.880 [ci = 0.844-0.916]). both were externally validated on h-l test (p = 0.781 and 0.155 respectively) and calibration plot. models were converted to nomograms. lower probabilities give higher sensitivity and npv; higher probabilities give higher specificity and ppv. conclusion: two simple-to-use validated nomograms were developed with excellent aucs based on readily available parameters and can be considered for clinical utilisation. developed two simple-to use nomograms for identifying covid-19 positive patients probabilities are provided to allow healthcare leaders to decide suitable cut-offs variables are age, white cell count, chest x-ray appearances and contact history model variables are easily available in the general hospital setting. objectives: to develop: (1) two validated risk prediction models for covid-19 positivity using readily available parameters in a general hospital setting; (2) nomograms and probabilities to allow clinical utilisation. patients with and without covid-19 were included from 4 hong kong hospitals. database was randomly split 2:1 for model development database (n=895) and validation database j o u r n a l p r e -p r o o f calibration plot. models were converted to nomograms. lower probabilities give higher sensitivity and npv; higher probabilities give higher specificity and ppv. two simple-to-use validated nomograms were developed with excellent aucs based on readily available parameters and can be considered for clinical utilisation. coronavirus disease 2019 has spread rapidly worldwide and as of 6 th september 2020, there are now ~27 million cases worldwide and ~900,000 deaths 1 . respiratory and non-respiratory complications of covid-19 are also becoming increasingly apparent 2, 3 . reverse transcription polymerase chain reaction (rt-pcr) is regarded as a vital tool in identifying the severe acute respiratory syndrome coronavirus 2 (sars-cov-2) and quarantining covid-19 patients to prevent further spread of the disease 4 . furthermore, it is the definitive test in confirming the diagnosis of covid-19. however, availability of rt-pcr kits maybe difficult in various countries 4 and from specimen collection to report generation, the tests could take 48-72 hours to confirm a positive or negative result 5 . therefore, clinical assessment, blood tests and imaging have been recommended to help identify potential covid-19 positive patients 6 . various strategies have been proposed including widespread computed tomography (ct) scanning 7-9 , greater use of chest x-rays (cxr) 10, 11 , identification of low lymphocyte counts 8, 11 to determine patients more likely to have covid-19 12, 13 and thus more suitable for testing. as yet, the data which supports these strategies are predominantly based on data of covid-19 patients 12, 13 but without comparisons to patients with other conditions and symptoms overlapping with covid-19 (eg. fever, shortness of breath, cough). several issues have arisen in trying to determine the likelihood of a covid-19 diagnosis. firstly, in the early stages of the pandemic when the disease was limited to a few countries, travel and contact history may have been helpful to increase suspicion of a covid-19 j o u r n a l p r e -p r o o f diagnosis, but in some countries where there is established community transmission, this has resulted in patients being covid-19 positive but with no knowledge of possible contact. secondly, different countries have adopted different strategies due to socioeconomic factors and healthcare resources. thus, a covid-19 prediction model based on clinical, laboratory and radiological findings which presents the sensitivity, specificity, positive predictive value (ppv) and negative predictive value (npv) would allow public healthcare systems to decide a suitable strategy on prioritizing tests when such rt-pcr availability is constrained. in this study, we aimed to construct a prediction model utilising patient characteristics, commonly available hematological and biochemical blood tests and cxr findings which can identify covid-19 patients within a cohort of patients who presented to hospitals for various disease conditions and underwent testing for covid-19. in addition, we aimed to create a separate model in the event that contact history is not available in order to determine the presence of covid-19. research ethics approval was obtained from the hong kong west cluster cxr images were searched via the electronic patient record system. baseline cxr images were reviewed and interpreted by radiologists blinded to the patient's covid-19 status. assessment was based on identifying the common findings of covid-19 on cxr which were (i) consolidation or ground glass opacity and (ii) absence of pleural effusion 14, 17 . this was done in a binary format (present or absent) to make this more reproducible in the clinical environment for front-line clinicians. image quality was assessed in 702 randomly chosen cxrs (53% of entire cohort of cxrs) by 8 radiologists separately. we ensured that the cxrs were taken from each of the 4 hospitals. image quality was assessed on a scale from 1 to 3. see supplementary table 2 for examples of cxrs graded as 1, 2 and 3. briefly, cxrs which could not be interpreted with any confidence were graded 1. cxrs with suboptimal image quality but lung changes and pleural effusion could be interpreted with some confidence were graded 2. cxrs with good quality such that lung changes and pleural effusions can be diagnosed with high confidence were graded 3. patients positive for covid-19 were compared to those negative for covid-19 patients. continuous variables were compared using student t-tests. categorical variables were compared using chi-squared tests. the database was randomly split on a 2:1 basis for the the selection was finished until the difference in bic of all remaining risk factors <10. 18 to test the nonlinear effect of selected clinical parameters, quadratic term of significant continuous predictors were considered. given that patients can present without knowledge of contact history with an infected person, a further model was developed with one having contact history removed, to represent an event in which contact history is unknown. in order to validate the model, the discrimination and calibration power of models were examined. the area under the receiver operating characteristic curve (auc) were conducted to evaluate the discrimination power, where 0.7 to 0.8 of auc is considered acceptable, 0.8 j o u r n a l p r e -p r o o f to 0.9 is considered excellent, and more than 0.9 is considered outstanding discrimination power. meanwhile, hosmer-lemeshow (h-l) test and calibration plot was used to test how well the percentage of observed covid-19 positive matches the percentage of predicted covid-19 positive over deciles of predicted risk. a p-value >0.05 is needed to conclude that there are insignificant differences between the observed and expected outcomes and therefore the model has good overall calibration. different probabilities were used to evaluate the model performance based on the sensitivity, specificity, ppv and npv. sensitivity analysis was conducted to examine the robustness of the model. multiple imputation was applied to handle missing data. the chained equation method was used to impute each missing value twenty times, adjusted for all baseline covariates and outcomes. moreover, 10-fold cross validation was applied to evaluate the discrimination and calibration power. to facilitate the risk prediction models used for screening in routine busy clinical practice, simple nomograms were developed. the effect of each predictor in the model was converted to a score and summation of all predictors that can be mapped to an estimated risk of covid-19 positive. the nomograms were plotted using nomolog package in stata 19 . sensitivity, specificity, ppv, npv were determined for the following probabilities which were: figure 3 and figure 4 were developed based on the derived risk prediction models. using the overall cohort model nomogram (figure 3 ) as an example, if a patient suspected to have covid-19 is aged 50, has no contact history, wcc of 2x10 9 cells/l and a cxr with no consolidation/ggo and absent pleural effusion (peff) the scoring will be as follows: age has two steps, so for age 50 at step 1, allocate 2 points; for step 2: allocate 7.5 points. for no contact history which is step 3, allocate 0 points. for a total white cell count (wcc) of 2x10 9 cells/l at step 4: allocate 7 points. for a cxr with no consolidation/ggo and absent peff at step 5, allocate 0 points. therefore, they would be allocated a total score of 16.5 points which equates to 0.6-0.7 probability (ie. 60-70% probability) of being covid-19 positive. in our study, we have developed two risk prediction models for determining covid-19 positive patients which have been validated with a separate dataset. both models have an excellent auc with good matching with the validation dataset. the models are based on parameters (ie. total wcc, cxr consolidation/ggo with absent pleural effusions) which are available in general hospitals as well as clinical data (ie. age with or without contact history). we have also provided nomograms to determine probability of covid-19 with several different probabilities illustrated to show the sensitivity, specificity, ppv and npv so that clinicians or healthcare systems can decide which probabilities would make the best cut-offs for rt-pcr testing. the development of these nomograms will hopefully improve frontline clinicians' diagnostic accuracy in identifying patients with covid-19 where rt-pcr may not be available or rapid results cannot be provided. commission. thus our data provides evidence that these initial observations of covid-19 were indeed accurate. cxr consolidation/ggo with absent pleural effusions is the typical appearance of covid-19 radiologically 14 . this model confirms that using cxr in addition to other parameters is j o u r n a l p r e -p r o o f helpful in identifying covid-19 patients. this has already been incorporated into societal recommendations and our models provide evidence to support this approach despite the lower sensitivity of cxr compared to ct 10, 11 . our model did not incorporate ct as ct was not easily available for our covid-19 positive patients and indeed the negative patients. this would likely be the scenario globally during this pandemic. ct with its higher sensitivity 21 will likely improve diagnostic accuracy but this is dependent on the facilities in each health service. not all health services can dedicate ct scanners for covid-19 diagnosis due to either a lack of scanner availability and/ or the extensive cleaning required after each covid-19 scan which reduces the radiology department's productivity 22 . in our study, we wanted to focus on parameters which would be easily accessible to all patients seen in the general hospitals, as some health systems even struggle to make chest x-rays and wcc available 23 . in our cohort, age is a significant predictor for covid-19. in this, study, the covid-19 patients were significantly younger than the negative patients. this can be partly explained by younger patients being more mobile and thus being more susceptible to develop covid-19 compared to the older population who may travel less. review of previous publications have indicated that patients with covid-19 are usually younger. in korea, one paper indicated that >60% of patients 24 were <49 years old whilst in china, 51.2%-55.1% 12, 13 of patients were <49 years old. the two nomograms in this study allocated higher scoring to the younger patients including children. this is possibly due to children having less symptoms and even less radiological changes 25, 26 making the identification of covid-19 more difficult. indeed, this possibly explains the noticeably less children confirmed to have covid-19 and possibly explains the statistical significance of age in the models for determining patients who are positive for sars-cov2. however, age as a predictor is very much representative of this cohort. in a different healthcare system where more elderly patients present, age as a predictor will likely need to be further investigated. the models we have established can set different probabilities in order to allow medical systems to self-determine the pre-test probability required for rt-pcr testing. moreover, the nomograms have been developed to visualize the sophisticated mathematical equation so that it can be adopted in the routine busy clinical practice. however, it should be emphasised that rt-pcr remains the gold standard for diagnosing covid-19 and that focus should be made on making rt-pcr easily available for testing patients as well as increasing the time taken for results to be made available. our study has several limitations. firstly, the covid-19 cases are reflective of practice in hong kong which has been active in screening for covid-19 which has included asymptomatic patients (5.6% in this cohort) with contact history and patients with mild symptoms. this may not be representative in other health systems worldwide so this model needs to be validated in those health systems. secondly, the chest x-rays were assessed by radiologists, so whether these results will be similar with frontline clinicians is uncertain. however, the assessment was simplified in order that frontline clinicians can focus their search on cxr to consolidation/ggo and absence of pleural effusions. furthermore, some health systems have access to radiology support to review cxrs and this model possibly justifies this practice if logistically feasible. thirdly, inflammatory markers like c-reactive protein, creatnine kinase, lactacte deyhydrogenase were not included in the model as a significant proportion of patients did not have these markers measured at time of admission. whether these markers prove useful will require further study. lastly, asymptomatic patients made up a very small proportion of patients and thus further validation with an asymptomatic cohort would be required to validate this model. in conclusion, we present two models which have 3 or 4 readily available parameters to improve the accuracy of identifying covid-19 amongst patients suspected of having covid-19 with or without known contact history. this will help identify patients most likely to benefit from rt-pcr testing and thus help better allocate rt-pcr testing where this resource is limited. table 2 . a total score is calculated from the addition of the scores for the variables chest x-ray (cxr) consolidation/ ground glass opacity (ggo), contact history, white cell count and age. note that age has two steps whilst other variables only have 1 step. the total score can then be marked on the bottom row and compared with the probability scale above. for example, a patient suspected to have covid-19 aged 50 (step 1: allocate 2 points; step 2: allocate 7.5 points), has no contact history (step 3: allocate 0 points), total white cell count (wcc) of 2x10 9 cells/l (step 4: allocate 7 points) and a cxr with no consolidation/ggo and absent pleural effusion (pe) (step 5: allocated 0 points), would receive a total score of 16.5 points which equates to a probability of between 0.6 and 0.7. a clinician then refers to the probability table (table 3) and decides what degree of sensitivity, specificity, positive predictive value or negative predictive value is adequate for their setting. table 4 . a total score is calculated from the addition of the scores for the variables pleural effusion, chest x-ray (cxr) consolidation/ ground glass opacity (ggo), white cell count, age and vomiting symptom. note that age has two steps whilst other variables only have 1 step. the total score can then be marked on the bottom row and compared with the probability scale above. for example, a patient suspected to have covid-19 aged 40 (step 1: allocate 2 points; step 2: allocate 8.5 points), total white cell count (wcc) of 8x10 9 cells/l (step 3: allocate 3 points) and a cxr with consolidation/ggo and absent pleural effusion (pe) (step 4: allocated 1.5 points), would receive a total score of 15 points which equates to a probability of between 0.6 and 0.7. a clinician then refers to the probability table (table 3) and decides what degree of sensitivity, specificity, positive predictive value or negative predictive value is adequate for their setting. world health organization. coronavirus disease (covid-19) weekly epidemiological update recovered covid-19 patients show ongoing subclinical myocarditis as revealed by cardiac magnetic resonance imaging world health organization. laboratory testing strategy recommendations for covid-19 guidance and standard operating procedure covid-19 virus testing in nhs laboratories the role of chest imaging in patient management during the covid-19 pandemic: a multinational consensus statement from the fleischner society covid-19): a perspective from china therapeutic and triage strategies for 2019 novel coronavirus disease in fever clinics covid-19 pneumonia: what has ct taught us? the lancet infectious diseases acr recommendations for the use of chest radiography and computed tomography (ct) for suspected covid-19 infection british society of thoracic imaging. bsti nhse covid-19 radiology decision support tool clinical features of patients infected with 2019 novel coronavirus in wuhan, china. the lancet 2020;0:201160. 15. world health organization. global surveillance for covid-19 caused by human infection with covid-19 virus: interim guidance improved molecular diagnosis of covid-19 by the novel, highly sensitive and specific covid-19-rdrp/hel real-time reverse transcription-polymerase chain reaction assay validated in vitro and with clinical specimens imaging profile of the covid-19 infection: radiologic findings and literature review bayesian model selection in social research design and evaluation of analytical tools for emergency department management based on machine learning techniques index for rating diagnostic tests sensitivity of chest ct for covid-19: comparison to rt-pcr a role for ct in covid-19? what data really tell us so far adoption of covid-19 triage strategies for low-income settings transmission potential and severity of covid-19 in south korea clinical and epidemiological features of 36 children with coronavirus disease 2019 (covid-19) in zhejiang, china: an observational cohort study. the lancet infectious diseases differences in clinical and imaging presentation of pediatric patients with covid-19 in comparison with adults sensitivity, specificity, positive predictive value and negative predictive value are stated in percentage with 95% confidence intervals in brackets key: cord-310228-bqpvykce authors: borkowski, a. a.; viswanadham, n. a.; thomas, l. b.; guzman, r. d.; deland, l. a.; mastorides, s. m. title: using artificial intelligence for covid-19 chest x-ray diagnosis date: 2020-05-26 journal: nan doi: 10.1101/2020.05.21.20106518 sha: doc_id: 310228 cord_uid: bqpvykce coronavirus disease-19 (covid-19), caused by a novel member of the coronavirus family, is a respiratory disease that rapidly reached pandemic proportions with high morbidity and mortality. it has had a dramatic impact on society and world economies in only a few months. covid-19 presents numerous challenges to all aspects of healthcare, including reliable methods for diagnosis, treatment, and prevention. initial efforts to contain the spread of the virus were hampered by the time required to develop reliable diagnostic methods. artificial intelligence (ai) is a rapidly growing field of computer science with many applications to healthcare. machine learning is a subset of ai that employs deep learning with neural network algorithms. it can recognize patterns and achieve complex computational tasks often far quicker and with increased precision than humans. in this manuscript, we explore the potential for a simple and widely available test as a chest x-ray (cxr) to be utilized with ai to diagnose covid-19 reliably. microsoft customvision is an automated image classification and object detection system that is a part of microsoft azure cognitive services. we utilized publicly available cxr images for patients with covid-19 pneumonia, pneumonia from other etiologies, and normal cxrs as a dataset to train microsoft customvision. our trained model overall demonstrated 92.9% sensitivity (recall) and positive predictive value (precision), with results for each label showing sensitivity and positive predictive value at 94.8% and 98.9% for covid-19 pneumonia, 89% and 91.8% for non-covid-19 pneumonia, 95% and 88.8% for normal lung. we then validated the program using cxrs of patients from our institution with confirmed covid-19 diagnoses along with non-covid-19 pneumonia and normal cxrs. our model performed with 100% sensitivity, 95% specificity, 97% accuracy, 91% positive predictive value, and 100% negative predictive value. finally, we developed and described a publicly available website to demonstrate how this technology can be made readily available in the future. the novel coronavirus severe acute respiratory syndrome coronavirus 2 (sars-cov-2), which causes the respiratory disease coronavirus disease-19 (covid19) , was first identified as a cluster of cases of pneumonia in wuhan, hubei province of china on december 31, 2019. 1 within a month, the disease had spread significantly, leading the world health organization (who) to designate covid-19, a public health emergency of international concern (pheic). on march 11, 2020 , the who declared covid-19 a global pandemic. 2 as of may 7, 2020, the virus has infected more than 3.6 million people, with over 250,000 deaths worldwide. 3 the dramatic impact the spread of covid-19 has had on social, economic, and healthcare issues throughout the world has been reviewed. 4 prior to the 21st century, members of the coronavirus family had minimal impact on human health. 5 however, in the past 20 years, outbreaks have highlighted an emerging importance of coronaviruses in morbidity and mortality on a global scale. although less prevalent than covid-19, severe acute respiratory syndrome (sars) in 2002-2003 and middle eastern respiratory syndrome (mers) in 2012 likely had higher mortality rates than the current pandemic. 5 based on this recent history, it is reasonable to assume that we will continue to see novel diseases with similar significant health and societal implications. the challenges presented to health care providers by such novel viral pathogens are numerous, including methods for rapid diagnosis, prevention, and treatment. in the current study, we focus on diagnosis issues, which were evident with covid19 with the time required to develop rapid and effective diagnostic modalities. we have previously reported the utility of using artificial intelligence (ai) in the histopathologic diagnosis of cancer. [6] [7] [8] ai was first described in 1956 and involves the field of computer science in which machines are trained to learn from experience. 9 machine learning (ml) is a subset of ai and is achieved by using mathematical models to compute sample data sets. 10 current ml utilization employs deep learning with neural networks algorithms, which can recognize patterns and achieve complex computational tasks often far quicker and with increased precision than can humans. [11] [12] [13] in addition to applications in pathology, ml algorithms have both prognostic and diagnostic applications in multiple medical specialties such as radiology, dermatology, ophthalmology, and cardiology. 6 it is predicted that ai will impact almost every aspect of health care in the future. 14 in this manuscript, we examine the potential for ai to diagnose patients with covid-19 pneumonia using chest radiographs (cxr) alone. this is done using microsoft customvision, a readily available, automated ml platform. 15 employing ai to both screen and diagnose emerging health emergencies such as covid-19 has the potential to dramatically change how we approach medical care in the future. in addition, we describe the creation of a publicly available website that could augment covid-19 pneumonia cxr diagnosis. one hundred three cxr images of covid-19 were downloaded from github covid-chest-xray dataset. 16 five hundred images of non-covid-19 pneumonia and 500 images of the normal lung were downloaded from the kaggle rsna pneumonia detection challenge dataset. 17 to balance the dataset, we expanded the covid-19 dataset to 500 images by slight rotation (probability=1, max rotation=5) and zooming (probability=0.5, percentage_area=0.9) of the original images using the augmentor python package. 18 thirty random cxr images from the veteran's administration (va) pac system were obtained for the validation dataset. this dataset included ten cxr images from hospitalized covid-19 patients, ten cxr pneumonia images from non-covid-19 patients, and ten normal cxrs. covid-19 diagnoses were confirmed with a positive test result from the xpert® xpress sars-cov-2 pcr platform. 19 microsoft customvision is an automated image classification and object detection system that is a part of microsoft azure cognitive services. 15 it has a pay as you go model with fees depending on your computing needs and usage. it offers a free trial to users for two initial projects. the service is web-based with an easy to follow graphical user interface. no coding skills are necessary. in microsoft customvision, we created a new project with the following setup: project type -classification, classification type -multiclass (single tag per image), domains -compact general for small size and easy export to tensorflow.js model format. with the project created, we proceeded to upload our image dataset. each class was uploaded separately and tagged with the appropriate label (covid_pneumonia, non-covid pneumonia, and normal lung). the system rejected 16 covid-19 images as duplicates. the final customvision training dataset consisted of 484 images of covid-19 pneumonia, 500 images of non-covid-19 pneumonia, and 500 images of normal lung. once uploaded, microsoft customvision self-trains using the dataset upon initiating the program. (figure 1) microsoft azure custom vision was used to train the model. custom vision can be used continuously to execute the model, or the model can be compacted and decoupled from azure. in this case, the model was compacted and decoupled for use in a web app. an angular web app was created with tensorflow.js. tensorflow.js is a javascript library that enables dynamic download and execution of ml models. within a user's web browser, the model is executed when an image of a cxr is submitted. confidence values for each classification are returned. in this design, after the initial web . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 26, 2020. . https://doi.org/10.1101/2020.05.21.20106518 doi: medrxiv preprint page and model is downloaded, the web page no longer needs to access any server components and performs all operations in the browser. although the solution works well on mobile phone browsers and in low bandwidth situations, the quality of predictions may depend on the browser and device one uses. at no time does an image get submitted to the cloud. overall, our trained model showed 92.9% precision and recall. precision and recall results for each individual label were 98.9% and 94.8% for covid-19 pneumonia, 91.8% and 89% for non-covid-19 pneumonia, and 88.8% and 95% for normal lung. (figure 2 ) next, we proceeded to validate the training model on the va data by making individual predictions on 30 images from the va dataset. our model performed very well with 100% sensitivity (recall), 95% specificity, 97% accuracy, 91% positive predictive value (precision), and 100% negative predictive value. we have successfully demonstrated the potential of using ai algorithms in assessing cxrs for covid-19. we first trained the microsoft customvision automated image classification and object detection system to differentiate cases of covid-19 from pneumonia from other etiologies as well as normal lung cxrs. we then tested our model against known patients from our medical center. the program achieved 100% sensitivity (recall), 95% specificity, 97% accuracy, 91% positive predictive value (precision), and 100% negative predictive value in differentiating the three scenarios. using the trained ml model, we proceeded to create a website that could augment covid-19 cxr diagnosis. 20 the website works on mobile as well as desktop platforms. one can take a cxr photo with a mobile phone or upload it from the file. the ml algorithm would provide the probability of covid-19 pneumonia, non-covid-19 pneumonia, or normal lung x-ray diagnosis. (figure 3 ) emerging diseases such as covid-19 present numerous challenges to healthcare providers, governments, and businesses, as well as to individual members of society. as evidenced with covid-19, the time from first recognition of an emerging pathogen to the development of methods for reliable diagnosis and treatment can be months, even with a concerted international effort. the gold standard for diagnosis of covid-19 is by reverse transcriptase polymerase chain reaction (rt-pcr) technologies, but early rt-pcr testing produced less than optimal results. [21] [22] [23] even after the development of reliable tests for detection, making test kits readily available to health care providers on an adequate scale presents an additional challenge as evident with covid-19. the lack of availability of diagnostic rt-pcr with covid-19 initially placed increased reliability on presumptive diagnoses via imaging in some situations. 24 most of the literature evaluating radiographs of covid-19 patients . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted may 26, 2020. . https://doi.org/10.1101/2020.05.21.20106518 doi: medrxiv preprint focuses on chest computed tomography (ct) findings, with initial results suggesting ct was more accurate than early rt-pcr methodologies. 22, 23, 25 the radiological society of north america expert consensus statement on chest ct for covid-19 states that ct findings can even precede positivity on rt-pcr in some cases. 23 however, currently they do not recommend the use of ct scanning as a screening tool. furthermore, the actual sensitivity and specificity of ct interpretation by radiologists for covid-19 are unknown. 23 characteristic ct findings include ground-glass opacities (ggos) and consolidation most commonly in the lung periphery, though a diffuse distribution was found in a minority of patients. 22, 24, [26] [27] [28] lomoro and colleagues recently summarized the ct findings from several reports which described abnormalities as most often bilateral and peripheral, subpleural and affecting the lower lobes. 27 not surprisingly, ct appears more sensitive at detecting changes with covid-19 than cxr, with reports that a minority of patients had ct changes before changes visible on cxr. 24, 27 we focused our study on the potential of ai in the examination of cxrs in patients with covid-19, as there are several limitations to the routine use of ct scans with conditions such as covid-19. aside from the more considerable time required to obtain cts, there are issues with contamination of ct suites, sometimes requiring a dedicated covid-19 ct scanner. 24, 29 the time constraints of decontamination or limited utilization of ct suites can delay or disrupt services for both covid-19 and non-covid-19 patients. because of these factors, cxr may be a better resource to minimize the risk of infection to other patients. besides, accurate assessment of abnormalities on cxr for covid-19 may identify patients in whom the cxr was performed for other purposes. 24 cxr is more readily available than ct, especially in more remote or underdeveloped areas. 29 finally, as with ct, cxr abnormalities are reported to have appeared before rt-pcr tests became positive in a minority of patients. 24 cxr findings described in covid-19 patients are similar to those of ct and include ggos, consolidation, and hazy increased opacities. 24, 26, 27, 29, 30 like ct, the majority of patients demonstrated greater involvement in the lower zones and peripherally 24, 26, 27, 29, 30 most patients showed bilateral involvement. however, while these findings are common in covid-19 patients, they are not specific and can be seen in other conditions such as other viral pneumonia, bacterial pneumonia, injury from drug toxicity, inhalation injury, connective tissue disease, and idiopathic conditions. applications of ai in interpreting radiographs of various types are numerous, and extensive literature has been written on the topic. 31 using deep learning algorithms, ai has multiple possible roles to augment traditional radiograph interpretation. these include the potential for screening, triaging, and increasing the speed to render diagnoses. it also can provide a rapid "second opinion" to the radiologist to support the final interpretation. in areas with critical shortages of radiologists, it potentially can be used to render the definitive diagnosis. with covid-19, imaging studies have been shown to correlate with disease severity and mortality, and ai could assist in monitoring the course of the disease as it progresses and potentially identifies patients at greatest risk. 28 there is excellent potential should a rapid diagnostic test as simple as a cxr be able to reliably impact containment and prevention of the spread of contagions such as covid-19 early in its course. few studies have assessed using ai in the radiologic diagnosis of covid-19, most of which utilize ct scanning. bai and colleagues demonstrated increased accuracy, sensitivity, and specificity in distinguishing chest cts of covid-19 patients from other types of pneumonia. 22, 32 a separate study demonstrated the utility of using ai to differentiate covid-19 from community-acquired pneumonia with ct. 33 however, the effective utility of ai for cxr interpretation has been demonstrated as well. 14, 34 implementation of convolutional neural network layers has allowed for reliable differentiation of viral and bacterial pneumonia with cxr imaging. 35 evidence suggests that there is great potential in the application of ai in the interpretation of radiographs of all types. finally, as mentioned, we have developed a publicly available website based on our studies. 20 it should be stressed that this website is for research use only. to utilize the website, images must have protected health information (phi) removed before uploading. the information on the website, including texts, graphics, images, or other material, is for research purposes and may not be appropriate for all circumstances. the website does not provide medical, professional, or licensed advice and is not a substitute for consultation with a health care professional. medical advice should be sought from a qualified health care professional for any questions, and the website should not be used for medical diagnosis or treatment. we have utilized a readily available, commercial platform to demonstrate the potential of ai to assist in the successful diagnosis of covid-19 pneumonia on cxr images. while this technology has numerous applications in radiology, we have focused on the potential impact on future world health crises such as covid-19. the findings have implications for screening and triage, initial diagnosis, monitoring disease progression, and identifying patients at increased risk of morbidity and mortality. based on the data, a website was created to demonstrate how such technologies could be shared and distributed to others to combat entities such as covid-19 moving forward. our study offers a small window into the potential for how ai will likely dramatically change the practice of medicine in the future. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted may 26, 2020. . https://doi.org/10.1101/2020.05.21.20106518 doi: medrxiv preprint who director-general's opening remarks at the media briefing on covid-19 -11 coronavirus disease (covid-19) journal pre-proof the socio-economic implications of the coronavirus and covid-19 pandemic: a review the emergence of sars, mers and novel sars-2 coronaviruses in the 21st century comparing artificial intelligence platforms for histopathologic cancer diagnosis apple machine learning algorithms successfully detect colon cancer but fail to predict kras mutation status using apple machine learning algorithms to detect and subclassify non-small cell lung cancer the dartmouth college artificial intelligence conference: the next fifty years some studies in machine learning using the game of checkers neural networks and statistical models deep learning in neural networks: an overview deep learning high-performance medicine: the convergence of human and artificial intelligence custom vision -home covid-19 image data collection biomedical image augmentation using augmentor xpress sars-cov-2 has received fda emergency use authorization chest ct findings in coronavirus disease-19 (covid-19): relationship to duration of infection correlation of chest ct and rt-pcr testing in coronavirus disease 2019 (covid-19) in china: a report of 1014 cases radiological society of north america expert consensus statement on reporting chest ct findings related to covid-19. endorsed by the society of thoracic radiology, the american college of radiology, and rsna frequency and distribution of chest radiographic findings in covid-19 positive patients authors sensitivity of chest ct for covid-19: comparison to rt-pcr epidemiological and clinical characteristics of 99 cases of 2019 novel coronavirus pneumonia in wuhan, china: a descriptive study covid-19 pneumonia manifestations at the admission on chest ultrasound, radiographs, and ct: single-center study and comprehensive radiologic literature review coronavirus disease 2019 (covid-19) imaging reporting and data system (covid-rads) and common lexicon: a proposal based on the imaging data of 37 studies cardiothoracic imaging portable chest x-ray in coronavirus disease-19 (covid-19): a pictorial review chest imaging in patients hospitalized with covid-19 infection -a case series a comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis ai augmentation of radiologist performance in distinguishing covid-19 from pneumonia of other etiology on chest ct artificial intelligence distinguishes covid-19 from community acquired pneumonia on chest ct chexpedition: investigating generalization challenges for translation of chest x-ray algorithms to the clinical setting zhang correspondence k. identifying medical diagnoses and treatable diseases by image-based deep learning this material is the result of work supported with resources and the use of facilities at the james a. haley veterans' hospital and computational resources of the interknowlogy, llc. this activity has been approved by the james a. haley veterans' hospital office of research and development. key: cord-013065-oj0wsstz authors: rodríguez-fanjul, javier; guitart, carmina; bobillo-perez, sara; balaguer, mònica; jordan, iolanda title: procalcitonin and lung ultrasound algorithm to diagnose severe pneumonia in critical paediatric patients (prolusp study). a randomised clinical trial date: 2020-10-08 journal: respir res doi: 10.1186/s12931-020-01476-z sha: doc_id: 13065 cord_uid: oj0wsstz background: lung ultrasound (lus) in combination with a biomarker has not yet been studied. we propose a clinical trial where the primary aims are: 1. to assess whether an algorithm with lus and procalcitonin (pct) may be useful for diagnosing bacterial pneumonia; 2. to analyse the sensitivity and specificity of lus vs chest x-ray (cxr). methods/design: a 3-year clinical trial. inclusion criteria: children younger than 18 years old with suspected pneumonia in a paediatric intensive care unit. patients will be randomised into two groups: experimental group: lus will be performed as first lung image. control group: cxr will be performed as first pulmonary image. patients will be classified according to the image and the pct: a) pct < 1 ng/ml and lus/cxr are not suggestive of bacterial pneumonia (bn), no antibiotic will be prescribed; b) lus/cxr are suggestive of bn, regardless of the pct, antibiotic therapy is recommended; c) lus/cxr is not suggestive of bn and pct > 1 ng/ml, antibiotic therapy is recommended. conclusion: this algorithm will help us to diagnose bacterial pneumonia and to prescribe the correct antibiotic treatment. a reduction of antibiotics per patient, of the treatment length, and of the exposure to ionizing radiation and in costs is expected. trial registration: nct04217980. therefore, there is a need to have an examination tool that can be used at the patient's bedside and which is easily reproducible to help detect lung consolidations. lus has recently emerged as a radiation-free technique; it is non-invasive, with a high interobserver [10] agreement for lung pathologies such as consolidation [11] , pleural effusion [12] , interstitial syndrome [13] , and pneumothorax [14] . to determine the etiology, conventional microbiology tests such as blood culture, pleural aspiration, and bronchoalveolar lavage are usual practices, but some of these are invasive, may not detect all etiologies [15] , and the results may not be immediate. besides this, the use of biomarkers such as procalcitonin (pct) has become more widespread during the past 10 years, helping clinicians diagnose bacterial etiology, especially in patients who have only had a fever for a few hours or those admitted to intensive care units [16] [17] [18] [19] [20] . the use of pct has allowed for a decrease in antibiotic prescription [21, 22] , even in nosocomial and community pneumonia [23, 24] . despite its role in the diagnosis of pneumonia, pct values without any other tests may not be a complete diagnostic biomarker for pneumonia. quality of care is defined by the world health organization as medical care in which the patient is diagnosed and treated correctly, according to current medical knowledge (scientific and technical quality), and to their biological factors (optimal state of health to attain), with a minimum cost (efficiency), minimum possible exposure to risk for more harm, and maximum patient satisfaction [25] . at the hospital level, a quality assurance plan should include different levels of action. the first step is level 1 of quality promotion, which requires institutional support and the availability of clinical and quality protocols. the second step is the research level: descriptive studies for detecting and quantifying a specific situation or health issue, and the use of databases or specific studies to evaluate health services, among others. therefore, we propose this clinical trial, based on combining lus and pct in an algorithm with the aim to improve quality of care in children with pneumonia in a picu. we hypothesize that the diagnostic performance of lus and pct will be better than conventional cxr. the study is designed as a randomized, blinded clinical trial of children with severe community or nosocomial pneumonia. it will be conducted at a single picu at sant joan de déu hospital, a tertiary children's hospital in barcelona. period of recruitment and follow-up was from september 2017 to december 2019. our primary goal is to improve the quality of care in children with suspected community or nosocomial pneumonia in a picu. the main objectives are: 1. to assess whether a diagnostic algorithm for pneumonia that combines lus and pct may be useful in indicating and determining the duration of antibiotic treatment. 2. to analyse the sensitivity and specificity of lus, compared to cxr, for severe community or nosocomial pneumonia. the secondary aims are: 1. to quantify the irradiation dose avoided using lus to replace cxr, and to determine if there is an associated decrease in costs. inclusion: children under 18 years old with suspected community pneumonia who require admission to the picu or patients with suspected nosocomial pneumonia during their picu stay. exclusion: patients with underlying pathologies such as cystic fibrosis or who are immunocompromised. patients who develop nosocomial pneumonia after being included in the study due to community pneumonia. patients who have a cxr taken before being admitted to the picu. withdrawal and abandonment criteria: violation of study protocol, withdrawal of parental consent, death. patients withdrawing or with loss of protocol adherence will be excluded from the study. -community pneumonia: patients with compatible clinical suspicion (fever, cough, tachypnoea, shortness of breath, abnormal respiratory auscultation sounds, hypoventilation, tubular breath sounds, thoracic or abdominal pain), compatible cxr (lobar consolidation, airspace opacity, pleural effusion, bullae, etc.) or lus, changes in blood with a c-reactive protein higher than 50 mg/l and/or pct higher than 1 ng/ml. -nosocomial pneumonia: based on the clinical pulmonary infection score (cpis) ( table 1 ). -ventilator-associated pneumonia: defined according to the center for disease control criteria (cdc). lus procedure lus will be performed by any of the 8 intensive care physicians who have received standard training in lus (winfocus pncus bl1p) and who have at least 3 years of experience using it. team sessions focusing on the diagnosis of pneumonia with lus will be repeated every 3 months to ensure quality and consistency in the lus exam. the supplemental data included in the shah et al. article will be used [26] . subjects will be examined while they are in supine position. imaging will be performed using a portable ultrasound device (toshiba® xario 200). a 12-mhz linear or a 5-mhz convex probe will be used, depending on the weight or size of the patient. a scan will be taken systematically in 3 areas for each hemithorax (anterior, lateral, and posterior), according to international recommendations [27] . each area will be examined longitudinally and transversally. in each area the following will be evaluated [27] : alines, b-lines (number and distance between them), lung sliding (m-mode), pleural space, lung consolidations, small subpleural consolidations, dynamic air bronchogram, vascular pattern, presence of lung point, and lung pulse. the determination of a bacterial pneumonia ultrasound pattern will be based on the presence of lung consolidation with air bronchograms, which in initial stages are detected as small subpleural hypoechoic zones of less than 1 cm with bronchogram (not seen using conventional cxr) [28] [29] [30] [31] . the determination of a viral pneumonia ultrasound pattern will be based on the presence of b-lines or coalescent b-lines with small subpleural consolidations of less than 1 cm, without bronchogram [32, 33] . the first thing to do will be to obtain consent from the parent(s) or legal guardian(s). patients who meet the inclusion criteria and sign the inform consent were randomly assigned to two groups using the "random" function in ms-excel xp® program. a binary series of random numbers were generated according to the procedure described by friedman. to procure a similar number of patients in both groups, the procedure created the sequence through a balanced block sampling. the series of numbers were held by the principal investigator and depending on the number of the patient tit was be assigned to on or another group. a total of 8 physicians enrolled participants, and the principal investigator assigned participants to interventions, depending on the randomized list. (fig. 1) -experimental group 1: the paediatrician-researcher (pr) performs the lus at admission/suspicion as the first lung image test. if the paediatrician assistant (pa) requires a cxr, it can be performed, but the pr will not see the cxr. patients will be subdivided into 3 groups: a. if pct is < 1 ng/ml and lus is not suggestive of bacterial pneumonia (normal or viral), the patient will not receive an antibiotic. b. if lus is suggestive of bacterial pneumonia, regardless of pct value, an antibiotic will be recommended. c. if lus is not suggestive of bacterial pneumonia, but pct values are > 1 ng/ml, an antibiotic will be recommended to cover other infectious etiologies. -control group 2: cxr will be performed as a first lung image test. criteria to start an antibiotic will depend on the current unit protocol. a. if pct is < 1 ng/ml and cxr is not suggestive of bacterial pneumonia (normal or viral), the patient will not receive an antibiotic. b. if cxr is suggestive of bacterial pneumonia, regardless of pct value, an antibiotic will be recommended. c. if cxr is not suggestive of bacterial pneumonia, but pct value is > 1 ng/ml, an antibiotic will be recommended to cover other infectious etiologies. radiological and ultrasound patterns will be classified as: pneumonia (viral or bacterial), atelectasis, or parapneumonic pleural effusion (table 2 ). lus will be performed every day following admission and recorded and stored. lus images will be later analysed by a paediatric radiologist who is an expert in lus and who has not seen the initial assessment and cxr, in order to evaluate interobserver agreement. cxr will be also be reported on by a paediatric radiologist consultant who has not seen the other results. the antibiotics protocol guided by lus and pct will be considered an improvement in the quality of care if a reduction in the prescription of antibiotics is observed, and also if there is a reduction in the number of days on antibiotics. another primary outcome will be the increase in the sensitivity and specificity when diagnosing bacterial pneumonia using lus. secondary outcomes will be the reduction of the irradiation dose using the new protocol (with a reduction in economic costs as well), and a high lus interobserver agreement. the study coordinator will register patients, verifying compliance with all the inclusion criteria. an external company will be appointed to monitor the study and ensure compliance with correct clinical practice principles (iche6). once the notebooks are audited, they will be entered into a validated database, one with restricted access by user level, which is equipped with inconsistency detection filters, and which affords data traceability until the database is no longer needed. data access: all the physicians and clinical researchers involved in the study, the ethics committee, and the relevant health authorities will have access to the data. sample size calculations will be performed using the statistical program ene 2.0®. the main variable will be the existence of differences between cxr and lus in patients with community pneumonia and nosocomial pneumonia. h1 will be considered as the existence of differences between lus and cxr. an 80% power will be required to detect differences in the contrast of the null hypothesis h0: p1 = p2, using a bilateral x 2 test for two independent samples. if we consider a significance level of 5%, it will be necessary to include 63 units in the control group and 63 units in the experimental group. a total of 126 patients will be included for community pneumonia. twenty-eight patients per group will be included for nosocomial pneumonia; therefore, there will be 56 patients in total for this kind of pneumonia. after 12 months of recruitment, a preliminary analysis will be carried out to guarantee the safety of the patients. respiratory infection represents around 30-70% of the picu admissions, depending on the season. there will be an estimated 200-300 recruitable patients per year, so it is expected that the calculated sample size will be attainable. using the "random" function of the ms-excel xp® program, a binary series of random numbers will be generated, according to the procedure described by friedman [34] . this procedure creates the sequence by means of balanced block sampling to ensure a similar number of patients in each group. the series of numbers will be in the possession of the picu's head researcher. depending on that number, each patient will be assigned to one group or the other. if a patient is randomized but does not complete the treatment, their data will not be analysed, and their random number will not be reused. the categorical variables will be compared using the chi-square test. the quantitative analysis will be compared using student's t-test or the mann-whitney u test, depending on whether the sample follows a normal distribution or not. a multivariate logistic regression analysis will be performed on those variables with statistical significance or a clear tendency in the univariate analysis to detect which factors represent a protective factor or not, in terms of quality. the analysis of the interobserver agreement will be performed using cronbach's alpha. a p < 0.05 will be considered significant. the statistical program spps® 20.0 will be used. nowadays, care standards are focused on the quality of care. as the world health organization stipulates, the patient must be diagnosed and treated correctly. this clinical trial is focused on improving the quality of care for paediatric patients with suspected bacterial pneumonia. lus has good diagnostic accuracy for pneumonia in children, even if the exam is performed by a nonexpert physician [35] . our algorithm will help us to diagnose bacterial pneumonia accurately, and to prescribe the correct antibiotic treatment. a reduction in patients on antibiotics and in the number of days on antibiotics is expected. secondarily, a reduction in exposure to ionizing radiation and in costs is expected. for many years, lus has been integrated into the management of critically ill paediatric and neonatal patients at our hospital. some articles have been published by our group regarding the use of lus in different pathologies, such as prematurity, pulmonary arterial hypertension, during the postoperative period after cardiopulmonary bypass, etc. [36] [37] [38] . for the application of this clinical trial, all the researchers will be intensivists who are experts in lus, and regular internal training will be essential to guarantee objective results. in addition, our group has extensive experience in the use of pct for the diagnosis of bacterial infection in other medical situations, such as after cardiopulmonary bypass in children and new-borns [19, 39] . furthermore, we have experience in pct-guided antibiotic policy, with a reduction in the number of days on antibiotics without adverse events in children with nosocomial infections [40] and in children after cardiopulmonary bypass [22] . the use procalcitonin and lung ultrasound algorithm will help us diagnose bacterial pneumonia accurately and prescribe the correct antibiotic treatment. a reduction in patients on antibiotics and in a reduction in exposure to ionizing radiation and in costs is expected. this clinical trial is focused on improving the quality of care for paediatric patients with suspected bacterial pneumonia. supplementary information accompanies this paper at https://doi.org/10. 1186/s12931-020-01476-z. additional file 1 supplemental table 1 . clinical pulmonary infection score. table 2 . summary of the main findings using chest x-rays and lung ultrasound for diagnosing pneumonia. additional file 3 supplemental figure 1 . study protocol diagram. we expect that with the estimated number of patients, relatively high for a paediatric study, and the homogeneity of the patients suffering from this pathology, we will have sufficient statistical power to obtain reliable data. another limitation could be that since ultrasound-based assessments are user-dependent, there could be interobserver variability. thanks to the internal training program, we believe that this limitation will be minimized. this protocol is for a study that it is ongoing, authors are still analysing data. no publications containing the results of this study have been already published neither submitted to any journal. authors disclose any potential financial or ethical conflicts of interest regarding the contents. clinical trial, trial registration: nct04217980. what is the key message of your article? we propose an algorithm to use in the diagnosis of pneumonia to improve the quality of care in paediatric critical care. what does it add to the existing literature? there are few papers about the use of lung ultrasound with pneumonia but, to our knowledge, this is the first clinical trial to assess the use of lung ultrasound and laboratory biomarkers to improve the diagnosis of bacterial pneumonia in critically ill children admitted to a paediatric intensive care unit. what is the impact? the algorithm should improve bacterial pneumonia diagnosis and therefore the treatment of either community-acquired or nosocomial pneumonia in patients with respiratory infection suspicion. the impact should be seen in reduction of antibiotic indication in patients with non-bacterial pneumonia, reduction of radiation given to paediatric patients and reduction costs of its technique. all the authors contributed to the conception and design of the protocol, acquisition of data, analysis and interpretation of data; drafting the article and revising it critically for important intellectual content; and they approval the final version to be published. this study has been financially supported by official grant from spain sanitary ministerium. no other institutions have financed this research. this study has been funded by instituto de salud carlos iii through the project " pi16/01040" (co-funded by european regional development fund/european social fund "a way to make europe"/"investing in your future"). project " pi16/01040", funded by instituto de salud carlos iii and co-funded by european union (erdf/esf, "a way to make europe"/"investing in your future"). funding: isciii ("pi16/01040"), co-funded by erdf/esf, "a way to make europe"/"investing in your future"). availability of data and materials not applicable. the study will be conducted in accordance with the ich cp (ich e6 and ich e11) and the declaration of helsinki. the protocol study was approved by the local health care ethics committee and the institutional review board of the sant joan de déu hospital. the identities of the patients will be kept confidential throughout the entire study. complete filiation data and written consent will be kept in the researcher's file. the data obtained will be treated according to the organic law 3/2018 on protection of personal data. according to this law, the personal data collected from the subjects will be only those necessary to fulfil the study's objectives. study participants will have the right to access their personal data and to request its rectification or cancellation. given that the patients included in the study will be minors or will be in critical condition, in all cases informed consent will be requested from the legal representatives of these patients. not applicable. the authors declare that they have no competing interests. commiting to child survival: a promise renewed surviving sepsis campaign the spectrum of viral pathogens in children with severe acute lower respiratory tract infection: a 3-year prospective study in the pediatric intensive care unit prospective multicenter study of viral etiology and hospital length of stay in children with severe bronchiolitis surviving sepsis campaign: international guidelines for management of severe sepsis and septic shock viral-bacterial coinfection affects the presentation and alters the prognosis of severe communityacquired pneumonia challenges in severe communityacquired pneumonia: a point-of-view review childhood community-acquired pneumonia: a review of etiology-and antimicrobial treatment studies high discordance of chest xray and computed tomography for detection of pulmonary opacities in ed patients: implications for diagnosing pneumonia interobserver agreement in the evaluation of b-lines using bedside ultrasound accuracy and applications of lung ultrasound to diagnose ventilator-associated pneumonia: a systematic review thoracic ultrasound for pleural effusion in the intensive care unit : a narrative review from diagnosis to treatment diagnostic bedside ultrasonography for acute respiratory failure and severe hypoxemia in the medical intensive care unit: basics and comprehensive approaches lung ultrasound for diagnosing pneumothorax in the critically ill neonate the definition and classification of pneumonia is procalcitonin useful in pediatric critical care patients procalcitonin versus c-reactive protein: review of kinetics and performance for diagnosis of neonatal sepsis procalcitonin: the marker of pediatric bacterial infection procalcitonin is a better biomarker than c-reactive protein in newborns undergoing cardiac surgery: the prokineca study procalcitonin-guided antibiotic discontinuation and mortality in critically ill adults: a systematic review and meta-analysis biomarkers kinetics in the assessment of ventilator-associated pneumonia response to antibioticsresults from the biovap study procalcitonin to stop antibiotics after cardiovascular surgery in a pediatric intensive care unit-the prosacab study procalcitonin-guided antibiotic discontinuation might shorten the duration of antibiotic treatment without increasing pneumonia recurrence the use of procalcitonin for prediction of pulmonary bacterial coinfection in children with respiratory failure associated with viral bronchiolitis bringing quality improvement into the intensive care unit prospective evaluation of point-of-care ultrasonography for the diagnosis of pneumonia in children and young adults international evidence-based recommendations for point-of-care lung ultrasound lung ultrasound compared to chest x-ray for diagnosis of pediatric pneumonia: a meta-analysis lung ultrasound for the diagnosis of community-acquired pneumonia in children accuracy of lung ultrasonography in the diagnosis of pneumonia in adults: systematic review and meta-analysis lung ultrasound combined with procalcitonin for a diagnosis of ventilator-associated pneumonia point-of-care lung ultrasound in young children with respiratory tract infections and wheeze point-of-care lung ultrasound in infants with bronchiolitis in the pediatric emergency department: a prospective study on the use of a friedman-type statistic in balanced and unbalanced block designs lung ultrasound for the diagnosis of pneumonia in children: a meta-analysis lung ultrasound to assess the etiology of persistent pulmonary hypertension of the newborn (lupphyn study): a pilot study lung ultrasound as a predictor of mechanical ventilation in neonates older than 32 weeks a multicenter lung ultrasound study on transient tachypnea of the neonate procalcitonin: a useful biomarker to discriminate infection after cardiopulmonary bypass in children procalcitonin-guidance reduces antibiotic exposure in children with nosocomial infection (prorani) publisher's note springer nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations received: 3 april 2020 accepted: 29 july 2020 key: cord-006708-nionk55w authors: aktaş, fatma; aktaş, turan title: the pulmonary findings of crimean–congo hemorrhagic fever patients with chest x-ray assessments date: 2019-03-25 journal: radiol med doi: 10.1007/s11547-019-01024-w sha: doc_id: 6708 cord_uid: nionk55w background: crimean–congo hemorrhagic fever (cchf), characterized by fever and/or hemorrhage, is a zoonotic viral disease with high mortality. the agent causing cchf is a nairovirus. the virus is typically transmitted to humans through tick bites. cchf is a life-threatening disease observed endemically over a wide geographical regions in the world, and there is limited information about pulmonary findings in cchf patients. purpose: we aimed to investigate the pulmonary findings belonging to a large cchf patient cohort and to determine if there is any relationship between laboratory findings and disease severity. materials and methods: a total of 165 patients who were diagnosed with cchf and examined through chest x-ray (cxr) due to respiratory symptoms at their first examination and/or during their hospitalization were included in this study. in addition to demographical and laboratory findings of the patients, chest x-rays were also examined. results: of the 165 patients examined, 96 were male (58.2%) and 69 were female (41.8%). the mean age was 51.64 ± 17.95 years (4–81 years). single and/or multiple pathological findings were detected in 93 patients (56.4%) as a result of chest x-ray during their first examination. on chest x-ray, consolidation in 74 patients (44.8%), pleural effusion in 64 patients (39.8%), ground glass opacity in 49 patients (29.7%), and atelectasis in 30 patients (18.2%) were detected. conclusion: according to the results of our study, it can be suggested that radiological examination in lungs should be performed primarily with cxr and pulmonary involvement (pleural effusion and consolidation) affects survival in cchf negatively. crimean-congo hemorrhagic fever (cchf), characterized by fever and/or hemorrhage, is a zoonotic viral disease with high mortality. the agent causing cchf is a nairovirus, which is a rna type from the bunyaviruses family. the virus is typically transmitted to humans through tick bites. rare transmissions to medical staff as a result of contact with blood or blood products have also been reported. the general non-specific symptoms of cchf are fever, malaise, anorexia, nausea/vomiting, and myalgia, which are observed after the contact. however, ecchymosis, petechia, vaginal bleeding, and epistaxis can also be observed. in addition, intracranial bleeding, gastrointestinal bleeding, and alveolar hemorrhage that are life-threatening and often responsible for mortality can be observed as well [1] [2] [3] [4] . the geographical location, exposure to animals with ticks, and contact with suspicious infected blood and blood products in the patients' history are helpful in diagnosis. cchf can cause sudden and severe bleedings and death in the end [1] . there have been many studies performed on cchf in the literature. however, the number of studies related to pulmonary imaging in cchf is very rare and limited despite the increasing number of cases in the world and in our country [1] [2] [3] [4] . on the other hand, we have a distinct advantage of studying the cchf patients due to the increase in the frequency of the disease in our region. there are four distinct phases in the clinical follow-up of cchf. these are incubation phase, pre-hemorrhagic phase, hemorrhagic phase, and convalescent phase [1, 14] . since patients have cold-like symptoms in pre-hemorrhagic phase, there may be some difficulties in differential diagnosis. therefore, examination with chest x-ray (cxr) (which is specific for pulmonary pathologies) must be considered. we may propose simple pulmonary imaging such as cxr to be used within the following 5 days after the first examination as an effective tool to evaluate clinical course and severity of the disease. the aim of this study was to investigate the chest x-ray findings of cchf patients with respiratory findings and to determine if there is any relationship between the laboratory findings and severity of the disease. a total of 464 patients with confirmed cchf were admitted to emergency department and department of infectious diseases and clinical microbiology between january 2011 and december 2016. the cases were evaluated retrospectively. a total of 165 patients, who had posteroanterior cxr due to respiratory symptoms at the first examination and/or during hospitalization, were included in this study. patients using medicines that may cause pulmonary toxicity and having acute or chronic lung diseases, malignancies, heart diseases, and blood diseases, and a history of thoracic radiotherapy or surgery in addition to cchf were excluded from the study. additionally, patients who were clinically diagnosed with pulmonary infection (such as cough sputum) and who were diagnosed with other microorganisms in their cultures were not included in the study as well. ethical approval was obtained from ethics committee of the university. in addition to demographical findings of the patients such as age, gender, occupation, place of residence, tick exposure, clinical findings, and medical history, laboratory findings and radiological images of lungs were evaluated. posteroanterior cxr was performed on all patients during the first examination. cxr was evaluated by two experienced radiologists. the radiologists evaluated each patient's cxr together in order to avoid interobserver variability and decided to share the findings jointly. the cchf diagnosis was made based on the detection of virus genome by using enzyme-linked immunosorbent assay (elisa) or polymerase chain reaction (pcr). data are expressed as mean ± standard deviation. independent sample t test was used to compare the continuous data between the groups. chi-square tests were used to compare the categorical data between/among the groups. categorical variables were presented as a count and percentage. kaplan-meier method was used for survival probabilities. a p value of less than 0.05 was considered significant. analyses were performed using spss 19 (ibm spss statistics 19, spss inc., an ibm co., somers, ny). of total patients, 96 were male (58.2%) and 69 were female (41.8%). the mean age was 51.64 ± 17.95 years (4-81 years). the majority of the patients [95 patients (57.6%)] lived in the city. there were no comorbidities in 124 patients (75.2%) in medical history examinations; on the other hand, there was chronic disease in 40 patients (24.8%). out of 40 patients with comorbidity, 23 were found to have hypertension, three of them had diabetes mellitus, 13 were diagnosed with both hypertension and diabetes mellitus, and one of them had hypothyroids. the demographical and clinical characteristics of the patients can be seen in table 1 . considering complaints and symptoms of the patients at the first examination, the most frequent symptoms were fever (98.8%), malaise (84.2%), and myalgia (35.8%). the frequency of general symptoms of the patients observed during the first examination is presented in table 2 . based on the laboratory findings, the mean values of liver-specific enzymes such as alkaline phosphatase (alp), alanine aminotransferase (alt), aspartate aminotransferase (ast), gamma-glutamyltransferase (ggt), creatine kinase (ck), and lactate dehydrogenase (ldh) were seen to be high. single and/or multiple abnormal imaging findings were detected in 93 patients (56.4%) as a result of cxr examination during the first examination. the cxr examination of 72 patients (43.6%) was normal. in addition, 36 patients (21.8%) found to have normal cxr findings at the first examination became symptomatic during clinical follow-up, and then, they were examined with cxr again within the first 5 days (fig. 1a, b) . those 36 patients whose cxr findings were obtained on the fifth day as a result of radiological evaluation were also included in this study. as a result of cxr findings obtained based on the first examination and clinical follow-up within the first 5 days, consolidation in 74 patients (44.8%), pleural effusion in 64 patients (39.8%), ground glass opacity in 49 patients (29.7%), and atelectasis in 30 patients (18.2%) were detected (fig. 2) . the cxr findings are demonstrated in table 3 in detail. in the comparative assessment of cxr findings and laboratory findings, there was a statistically significant relationship between cxr findings and laboratory findings (table 4) . twenty-three (17%) of the 165 patients who were diagnosed with pathology based on chest radiography had thorax 1 3 ct. the results of the thorax ct revealed that 20 of these patients (86%) had pleural effusion, 12 of them (52%) had consolidation, nine (39%) had atelectasis, and 14 (60%) had ground glass densities. moreover, two patients (8%) were identified with extensive alveolar density promoting alveolar hemorrhage (fig. 3) . four patients who were diagnosed with consolidation based on cxr were not observed to have any consolidation on ct. in a similar vein, four patients, claimed to have ground glass densities on cxr, and one patient, claimed to have pleural effusion, were not found to have these findings according to the results of the thorax ct. on the other hand, four patients who were not diagnosed with ground glass density in cxr were seen to have it after thorax ct was performed. as to two patients who were not found to have atelectasis in cxr, it was seen based on the thorax ct results that they had atelectasis. the relationship between the survival and cxr findings was discussed. there were statistically significant results in both groups (p < 0.001). moreover, the mortality rate was calculated as 11.5% in this study and there were statistically significant results in comparative assessment of each cxr findings with the survival separately. the comparison of survival and cxr findings of patients is shown in table 5 . cchf is a life-threatening zoonotic viral disease that can be seen endemically in many regions in the world. cchf is more common in males since contact with ticks is more frequent in males [5] [6] [7] [8] [9] . the most frequent clinical symptoms are fatigue (92.3%), fever (89.4%), myalgia (69.7%), headache (68.1%), and nausea (64.7%); 23% of patients had bleeding (petechiae, epistaxis, gingival bleeding, vaginal bleeding, and bleeding in internal organs) [10] . sumer [7] reported in their study that the majority of patients diagnosed with cchf were males (51.8%). bilgin et al. [11] also reported the ratio of males as 51.6% in a similar type of study. the male patient ratio (58.2%) was found to be higher in our study. this situation might be related to the possibility that males work in occupations with high risk of being in contact with ticks. as in many other viral diseases, cchf may cause symptoms of pulmonary involvement and may lead to death. although lung is one of the organs cchf affects frequently, the literature information regarding pulmonary involvement is very limited [12] [13] [14] . the study conducted by sannikova et al. in russia has the largest number of patients (283 patients) among those studies. the relationship between high inflammatory cytokine levels with ards and clinical severity of cchf was revealed in their study. hemoptysis was found to be the most frequent symptom in their study [14] . on the other hand, cough and dyspnea were indicated as the most common symptoms in the study conducted by doğan et al. in turkey [12] . the presence of symptoms such as dyspnea, chest pain, and hemoptysis was reported to be poor prognosis criteria in their study. there was no relationship between respiratory symptoms and the survival in the study with 128 patients by bilgin et al. [11] . there was a statistically significant relationship between cough symptom and survival in our study (p = 0.023). however, hemoptysis was detected in a total of nine out of 27 patients with hemorrhage (six of those patients died). there have been a very limited number of studies on pulmonary radiological findings in cchf in the literature. cxr is the first preferred imaging method in cchf since it is easily accessible and applicable. further investigation opportunities may be limited for patients due to the general condition disorder in cchf. the first and most frequent examination preferred in the studies by doğan et al. and bilgin et al. was cxr [11, 12] . all radiological imagings of lungs were performed according to cxr. in the study by sannikova et al. [14] , ards and alveolar hemorrhage diagnosis were generally established based on clinical findings and cxr. cxr was also used in all patients' examination of lungs in our study. there were abnormal radiological findings in 93 (56.4%) of the 165 patients in our study. there were abnormal cxr findings in 33 of 108 patients in the study by doğan et al. [12] and in 32 of 128 patients in the study by bilgili et al. [11] . the number of patients with cxr was higher in our study than the other studies in the literature. this may be caused by the inclusion of patients whose cxr findings emerged on the fifth day of clinical follow-up. there were 36 patients included in our study through follow-up. there was no significant relationship between the comparison of patients with pathological findings in cxr and survival in the studies by doğan et al. and bilgin et al. [11, 12] . on the other hand, there was a very statistically significant fig. 1 a 24 -year-old male patient complaint with fever and malaise. a chest x-ray at the initial hospital admission is normal, b massive pleural effusion is detected on chest x-ray at the fifth day relationship between the cxr findings and the survivals of patients in our study (p < 0.001). pleural effusion, one of the findings that can be seen in the clinical follow-up of cchf, was observed in 64 patients (37.8%) in our study. pleural effusion was observed bilaterally in 34 (20.6%) patients. there was no pleural effusion in any of the patients in the study of doğan et al., but pleural thickening was observed only in seven patients (6.5%). there was pleural effusion in only one patient in the study of bilgin et al. [11, 12] . interestingly, there was a very significant relationship between pleural effusion and survival in our study (p < 0.001). other cxr findings such as consolidation, ground glass opacity, and atelectasis were observed in 74, 49, and 30 patients, respectively. the number of patients with respiratory symptoms was much higher in our study than the number of patients in the other studies. the pathophysiological mechanism of pleural effusion has not been clearly demonstrated in cchf. however, it has been stated that it can share similarities with mechanism of pulmonary pathologies such as pleural effusion, pneumonitis, hemoptysis, and alveolar hemorrhage in other febrile viral hemorrhagic diseases [15, 16] . it has been suggested that pulmonary findings emerge due to the increase in capillary permeability in febrile viral hemorrhagic diseases. particularly, virus in the circulatory system causes an increase in pulmonary capillary permeability causing capillary endothelial damage [17] . in a study performed on dengue hemorrhagic fever, a total of 468 cxr taken from 363 patients were examined and parenchymal infiltration and pleural effusion were observed in more than half of the patients on the third day of follow-up. these findings were stated as the most frequent radiological findings. moreover, these radiological findings were found to be closely related to laboratory findings of leukocyte count, platelet count, aptt, alt, and albumin levels in the same study [18] . similar to other febrile viral hemorrhagic diseases (in particular to dengue hemorrhagic fever), clinical conditions must be kept in mind in cchf. replacement therapy was also performed with blood and blood products in some patients with sudden decrease in hemoglobin values and with alveolar hemorrhage findings in our study. in conclusion, pleural effusion and other findings related to pulmonary involvement can be seen in cchf due to endothelial damage caused by viral load in the circulation, increased capillary permeability, and hemorrhages due to low platelet. in the study by aktaş et al. [3] , thorax ct findings of 40 patients were evaluated, and then, ct and cxr findings of the patients were compared. the results of the study revealed that there was no statistically significant difference between the two examination techniques in terms of detecting parenchymal infiltration, alveolar infiltration, and pleural effusion. in our study, 23 of the 129 patients went through both ct and cxr examinations. when these patients were compared, it was found that there was consolidation in cxr of the four patients and pleural effusion in one patient while they were not observed in thorax ct. on the other hand, based on thorax ct results, ground glass density was detected in four patients and atelectasis in two patients, and these were not observed in cxr. however, there was a time period changing from 2 to 14 days between cxr and thorax ct examinations. therefore, there was a possibility of response to the treatment and spontaneous regression or progression in the findings. the mortality rate of cchf generally ranges from 2 to 80%. the mean mortality rate has been reported as 30-50% [19] [20] [21] [22] . the mortality rate was found to be 30-62% in the study by bakır et al. [15] . in our study, it was 11.5%. this rate was lower compared to the existing literature. this result can be associated with the fact that cchf is endemic in this region; therefore, medical staff and people are informed about the disease, and thus, they are alert. in addition, the experience in the clinical follow-up might also contribute to the decrease in mortality rate besides improvements in medical care conditions in cchf and increase in awareness of the seriousness of the disease. it is known that mortality and some laboratory parameters prolonged a ptt, increased inr, high alt, ast and ldh, low platelet, and increased leukocyte are related to each other [23, 24] . moreover, it is claimed that age, bleeding, melena, hepatomegaly, organ failure, and somnolence increase the mortality [23, [25] [26] [27] . apart from these findings, based on the results of our study, there was a statistically significant relationship between the cxr findings of cchf and mortality. for this reason, all of the cchf patients who were found to have pathologic cxr result should be considered at high risk. our choice to conduct a retrospective study is the first limitation; on the other hand, medical information of the patients and their radiological images has been recorded in a proper way. thus, there was an easy access to their data. the other limitation was the inability to confirm the cxr findings of the patients as not all of these patients had thorax ct. in conclusion, cchf is a serious public health problem that can be fatal. we investigated the specific pulmonary findings in cchf through cxr and the effects of these findings on the course of disease and survival in this study. according to the results of our study, we suggest that radiological imaging of pulmonary system should be performed primarily with cxr. pulmonary involvement (pleural effusion, consolidation, and ground glass opacity) affects the survival in cchf negatively, and further examination should be performed with thorax ct in case of necessity. radiological findings emerging on the fifth day of our study were demonstrated to be related closely to both survival and many laboratory findings. although not as detailed as chest ct, as the access to cxr is more common and it is easy to be applied (such as portable cxr in non-stabilized patients), it can be the first option to investigate pulmonary involvement in cchf patients. conflict of interest the authors declare that they have no conflict of interest. ethical approval all procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 helsinki declaration and its later amendments or comparable ethical standards. informed consent informed consent was obtained from all individual participants included in the study. crimean-congo haemorrhagic fever is hemorrhage the reason in crimean-congo hemorrhagic fever patients with neurological signs and symptomsa thorax ct findings in patients with crimean-congo hemorrhagic fever (cchf) the evaluation of abdominal findings in crimean-congo hemorrhagic fever crimean-congo hemorrhagic fever in turkey: current status and future challenges discharge criteria for crimean-congo haemorrhagic fever in endemic areas the evaluation of the patients who were admitted to the emergency department of kaş state hospital because of tick biting a preliminary report on crimean-congo haemorrhagic fever in turkey tick paralysis the effectiveness of routine laboratory findings in determining disease severity in patients with crimean-congo hemorrhagic fever: severity prediction criteria an investigation of pulmonary findings of crimean-congo haemorrhagic fever patients evaluation of respiratory findings in crimean-congo haemorrhagic fever crimean-congo haemorrhagic fever and diffuse alveolar hemorrhage respiratory lesions in crimean-congo haemorrhagic fever turkish cchf study group crimean-congo haemorrhagic fever outbreak in middle anatolia: a multicentre study of clinical features and outcome measures plasma activity of thrombin activatable fibrinolysis inhibitor in crimean-congo hemorrhagic fever viral hemorrhagic fever-a vascular disease? chest radiographic first examination in patients with dengue haemorrhagic fever crimean-congo hemorrhagic fever in iran and neighboring countries crimean-congo haemorrhagic fever outbreak investigation in the western region of afghanistan in 2008 characteristics of crimean-congo hemorrhagic fever virus (xinjiang strain) in china validation of a severity grading score (sgs) system for predicting the course of disease and mortality in patients with crimean-congo hemorrhagic fever (cchf) pathogenesis of crimean-congo hemorrhagic fever clinical and laboratory features of crimean-congo hemorrhagic fever: predictors of fatality evaluation of clinical and laboratory predictors of fatality in patients with crimean-congo haemorrhagic fever in a tertiary care hospital in turkey crimean-congo hemorrhagic fever in eastern turkey: clinical features, risk factors and efficacy of ribavirin therapy key: cord-303483-wendrxee authors: rubin, geoffrey d.; ryerson, christopher j.; haramati, linda b.; sverzellati, nicola; kanne, jeffrey p.; raoof, suhail; schluger, neil w.; volpi, annalisa; yim, jae-joon; martin, ian b. k.; anderson, deverick j.; kong, christina; altes, talissa; bush, andrew; desai, sujal r.; goldin, jonathan; goo, jin mo; humbert, marc; inoue, yoshikazu; kauczor, hans-ulrich; luo, fengming; mazzone, peter j.; prokop, mathias; remy-jardin, martine; richeldi, luca; schaefer-prokop, cornelia m.; tomiyama, noriyuki; wells, athol u.; leung, ann n. title: the role of chest imaging in patient management during the covid-19 pandemic: a multinational consensus statement from the fleischner society date: 2020-04-07 journal: radiology doi: 10.1148/radiol.2020201365 sha: doc_id: 303483 cord_uid: wendrxee with more than 900,000 confirmed cases worldwide and nearly 50,000 deaths during the first three months of 2020, the covid-19 pandemic has emerged as an unprecedented healthcare crisis. the spread of covid-19 has been heterogeneous, resulting in some regions having sporadic transmission and relatively few hospitalized patients with covid-19 and others having community transmission that has led to overwhelming numbers of severe cases. for these regions, healthcare delivery has been disrupted and compromised by critical resource constraints in diagnostic testing, hospital beds, ventilators, and healthcare workers who have fallen ill to the virus exacerbated by shortages of personal protective equipment. while mild cases mimic common upper respiratory viral infections, respiratory dysfunction becomes the principal source of morbidity and mortality as the disease advances. thoracic imaging with chest radiography (cxr) and computed tomography (ct) are key tools for pulmonary disease diagnosis and management, but their role in the management of covid-19 has not been considered within the multivariable context of the severity of respiratory disease, pre-test probability, risk factors for disease progression, and critical resource constraints. to address this deficit, a multidisciplinary panel comprised principally of radiologists and pulmonologists from 10 countries with experience managing covid-19 patients across a spectrum of healthcare environments evaluated the utility of imaging within three scenarios representing varying risk factors, community conditions, and resource constraints. fourteen key questions, corresponding to 11 decision points within the three scenarios and three additional clinical situations, were rated by the panel based upon the anticipated value of the information that thoracic imaging would be expected to provide. the results were aggregated, resulting in five main and three additional recommendations intended to guide medical practitioners in the use of cxr and ct in the management of covid-19. with more than 900,000 confirmed cases worldwide and nearly 50,000 deaths during the first three months of 2020, the covid-19 pandemic has emerged as an unprecedented healthcare crisis. the spread of covid-19 has been heterogeneous, resulting in some regions having sporadic transmission and relatively few hospitalized patients with covid-19 and others having community transmission that has led to overwhelming numbers of severe cases. for these regions, healthcare delivery has been disrupted and compromised by critical resource constraints in diagnostic testing, hospital beds, ventilators, and healthcare workers who have fallen ill to the virus exacerbated by shortages of personal protective equipment. while mild cases mimic common upper respiratory viral infections, respiratory dysfunction becomes the principal source of morbidity and mortality as the disease advances. thoracic imaging with chest radiography (cxr) and computed tomography (ct) are key tools for pulmonary disease diagnosis and management, but their role in the management of covid-19 has not been considered within the multivariable context of the severity of respiratory disease, pre-test probability, risk factors for disease progression, and critical resource constraints. to address this deficit, a multidisciplinary panel comprised principally of radiologists and pulmonologists from 10 countries with experience managing covid-19 patients across a spectrum of healthcare environments evaluated the utility of imaging within three scenarios representing varying risk factors, community conditions, and resource constraints. fourteen key questions, corresponding to 11 decision points within the three scenarios and three additional clinical situations, were rated by the panel based upon the anticipated value of the information that thoracic imaging would be expected to provide. the results were aggregated, resulting in five main and three additional recommendations intended to guide medical practitioners in the use of cxr and ct in the management of covid-19. on march 11, 2020 the world health organization (who) officially characterized the rapid global spread of coronavirus disease 2019 (covid-19) as a pandemic and called for urgent international action in four key areas: to prepare and be ready; detect, protect, and treat; reduce transmission; and innovate and learn (1) . at the time of writing (april 1, 2020), there are over 900,000 confirmed covid-19 cases and nearly 50,000 deaths in 205 countries around the world, with the majority of cases concentrated in 4 countries: united states, italy, spain, and china (2, 3) . with sustained community transmission now established in multiple countries on multiple continents, the who public health goal has changed from containment to mitigation of the pandemic's impact. consequently, strategies are now focused on efforts to reduce the incidence, morbidity, and mortality of covid-19 by breaking the chain of human transmission through social distancing and imposed quarantine. early detection and containment of infection caused by the novel coronavirus sars-cov2 has been hindered by the need to develop, mass produce, and widely disseminate the required molecular diagnostic test, a real-time reverse transcriptase-polymerase chain reaction (rt-pcr) assay. early reports of test performance in the wuhan outbreak showed variable sensitivities ranging from 37% to 71% (4, 5) . while laboratory-based performance evaluations of rt-pcr test show high analytical sensitivity and near-perfect specificity with no misidentification of other coronaviruses or common respiratory pathogens, test sensitivity in clinical practice may be adversely affected by a number of variables including: adequacy of specimen, specimen type, specimen handling, and stage of infection when the specimen is acquired (cdc guidelines for in-vitro diagnostics) (6, 7) . false negative rt-pcr tests have been reported in patients with ct findings of covid-19 who were eventually tested positive with serial sampling (8) . limited testing capacity due to insufficient specimen collection kits, lab test supplies, and testing equipment precluded early widespread testing and is believed to have contributed to rapid and unchecked transmission of infection within communities by undetected individuals with milder, limited, or no symptoms (9, 10) . for example, ct screening of 82 asymptomatic individuals with confirmed covid-19 from the cruise ship "diamond princess" showed findings of pneumonia in 54% (11) . provision of diagnostic imaging services to large numbers of patients suspected or confirmed to have covid-19 during an outbreak can be challenging, as each study is lengthened and complicated by the need for strict adherence to infection control protocols designed to minimize risk of transmission and protect healthcare personnel (12) . droplet transmission followed by contaminated surfaces are believed to be the main modes of spread for sars-cov2 in radiology suites; all patients undergoing imaging should be masked and imaged using dedicated equipment that is cleaned and disinfected after each patient encounter (13) . although personal protection equipment (ppe) recommendations vary between countries, the current centers of disease control (cdc) guidelines recommend radiology staff wear a mask, goggles or face shield, gloves, and an isolation gown. in countries with more stringent ppe protocols, a surgical cap and shoe covers may be added, while a surgical mask and goggles or face shield are suggested in some countries with less stringent ppe protocols (14) . additional precautions are required for specific situations that are more likely to generate aerosols, including patients receiving non-invasive ventilation, during intubation or extubation, throughout bronchoscopy, or when receiving nebulized therapies. portable imaging, including imaging patients through glass walls, has been used in some hospitals to further reduce the chance of spreading infection. written from multidisciplinary and multinational perspectives, this fleischner statement is intended to provide context for the use of imaging to direct patient management during the covid-19 pandemic in different practice settings, different phases of epidemic outbreak, and environments of varying critical resource availability. this document is structured around three clinical scenarios and three additional situations in which chest imaging is often considered in the evaluation of patients with potential covid-19 infection. the committee elected to present this document as a consensus statement rather than a guideline given the limited evidence base and the urgent need for direction on this topic for the medical community. imaging is also often considered (fig 4) . the entire panel was convened during a single session using a live audio and video interface (zoom video communications, san jose, ca). the three scenarios and three additional situations were presented, discussed, and refined. the panel independently and anonymously rated the appropriateness of imaging with chest radiography (cxr) or ct at each of these decision points on a five-point scale. at least 70% agreement on the direction of a recommendation was considered consensus. the scenarios are intended to support the management of adults only. children, who are typically spared from severe infections (15) , merit separate consideration, particularly with regard to use of radiationassociated procedures, and are beyond the scope of the current document. the final document was supported by a comprehensive literature search for relevant articles. using the search terms "((coronavirus or covid or sars-cov or *ncov*) and (ct or computed tomography or radio* or imag*))", a total of 137 english articles published between dec 1, 2019 and march 23, 2020 were identified. each article was assessed for relevance to the primary objective and a summary of key findings from relevant articles was created. the value of an imaging test relates to the generation of results that are clinically actionable either for establishing a diagnosis or for guiding management, triage, or therapy. that value is diminished by costs that include the risk of radiation exposure to the patient, risk of covid-19 transmission to uninfected healthcare workers and other patients, consumption of ppe, and need for cleaning and downtime of radiology rooms in resource-constrained environments. the appropriate use of imaging in each of the scenarios was considered on this basis. this statement focuses exclusively on the use of chest radiography (cxr) and computed tomography of the thorax (ct). while ultrasound has been suggested as a potential triage and diagnostic tool for covid-19 given the predilection for the disease in subpleural regions, there is limited experience at this time (16) , as well as infection control issues. cxr is insensitive in mild or early covid-19 infection (17) . however, with respect to the relative value of cxr or ct for detecting the presence of viral pneumonia, the experience is vastly different dependent upon community norms and public health directives. when patients are encouraged to present early in the course of their disease, as was the case in wuhan, china, cxr has little value. the greater sensitivity of ct for early pneumonic changes is more relevant in the setting of a public health approach that required isolation of all infected patients within an environment where the reliability of covid-19 testing was limited and turnaround times were long (4). alternatively, in new york city where patients were instructed to stay at home until they experienced advanced symptoms, cxr was often abnormal at the time of presentation. equipment portability with imaging performed within an infected patient's isolation room is another factor that may favor cxr in selected populations, effectively eliminating the risk of covid-19 transmission along the transport route to a ct scanner and within the room housing a ct scanner, particularly in environments lacking ppe. in hospitalized patients cxr can be useful for assessing disease progression and alternative diagnoses such as lobar pneumonia, suggestive of bacterial superinfection, pneumothorax and pleural effusion. ct is more sensitive for early parenchymal lung disease, disease progression, and alternative diagnoses including acute heart failure from covid-19 myocardial injury (18) and when acquired with intravenous contrast material, pulmonary thromboembolism. leveraging these superior capabilities depends upon the availability of ct capacity, particularly considering the potential reduction in ct scanner availability due to the additional time required to clean and disinfect equipment following imaging of patients with suspected covid-19. some centers rely on the improved depiction of covid-19 findings with ct relative to cxr (19) and their association with clinical worsening to determine patient disposition to home, hospital admission, or intensive care. in recognition of variance amongst local practice patterns and resource availability, it is important to state at the outset that the scenarios specify the use of imaging but do not articulate the relative merit of cxr versus ct. ultimately, the choice of imaging modality is left to the judgement of clinical teams at the point-of-care accounting for the differing attributes of cxr and ct, local resources, and expertise. the scenarios apply only to patients presenting with features consistent with covid-19 infection. the severity of respiratory disease and pre-test probability of covid-19 infection are specified for each scenario, with additional key considerations including the presence of risk factors for disease progression, evidence of disease progression, and the presence of significant critical resource constraints ( table 1) . the scenarios distinguish mild respiratory disease from moderate-to-severe respiratory disease based on the absence vs. presence of significant pulmonary dysfunction or damage. pre-test probability is defined by the background prevalence of infection and can be estimated by observed transmission patterns: low by sporadic transmission; moderate by clustered transmission; and high by community transmission (20) . individual pre-test probability is further modified if there is known exposure through contact with a confirmed case of covid-19 (21) . for health care providers, the cdc categorizes medical-related exposures into low, medium, and high-risk groups (22) . within a diagnostic radiology department, brief (a few minutes or less) unprotected interaction with a patient with covid-19 as well as prolonged close contact with a masked, infected patient by a medical provider wearing ppe are categorized as low-risk exposures (21, 22) . risk factors for poor outcomes in patients with covid-19 infection are considered separately from pre-test probability, with common risk factors including age > 65 years, cardiovascular disease, diabetes, chronic respiratory disease, hypertension, and immune-compromised (23) . identifying a patient as being at high risk for covid-19 progression is not necessarily a feature of any single risk factor, but is rather a clinical judgement based on the combination of underlying comorbidities and general health status that suggests a higher level of clinical concern. where appropriate, management variations based upon risk factors for disease progression are called out explicitly, as in scenario 1. all clinical scenarios begin by characterizing covid-19 status based upon the availability of laboratory test results. the first scenario (fig 1) addresses a patient presenting for evaluation at an outpatient clinic or via telehealth with mild respiratory features consistent with covid-19 infection, any pre-test probability of covid-19 infection, and no significant critical resource constraints. when covid-19 test results are unavailable, patients with moderate-to-high pre-test probability should be initially managed as if covid-19 testing is positive, while patients with low pre-test probability should be initially managed as if covid-19 testing is negative. imaging is advised for patients with risk factors for covid-19 progression and either positive covid-19 testing or moderate-tohigh pre-test probability in the absence of covid-19 testing (fig 1, q1) . imaging provides a baseline for future comparison, may establish manifestations of important comorbidities in patients with risk factors for disease progression ( table 1) , and may influence the intensity of monitoring for clinical worsening. imaging is not advised for patients with mild features who are covid-19 positive without accompanying risk factors for disease progression, or for patients with mild features who are covid-19 negative (fig 1, q2 & q3) . the panel felt that the yield of imaging in these settings would be very low and that it was safe for most patients to self-monitor for clinical worsening. regardless of covid-19 test results and risk factors, imaging is advised for patients with mild clinical features who subsequently develop clinical worsening (fig 1, q4 & q5) . in the absence of clinical worsening, management involves support and isolation of patients with positive covid-19 testing or patients with moderate to high pre-test probability without covid-19 test results available. although not specifically addressed by this scenario, in the presence of significant resources constraints, there is no role for imaging of patients with mild features of covid-19. the second scenario (fig 2) addresses a patient presenting with moderate-to-severe features consistent with covid-19 infection, any pre-test probability of covid-19 infection, and no significant critical resource constraints. separate ratings were obtained for covid-19 positive patients and either covid-19 negative patients or patients for whom covid-19 testing is unavailable (fig 2, q6 & q7) . imaging is advised regardless of the results or availability of covid-19 testing given the impact of imaging in both circumstances. for covid-19 positive patients, imaging establishes baseline pulmonary status and identifies underlying cardiopulmonary abnormalities that may facilitate risk stratification for clinical worsening. in the presence of clinical worsening, imaging is again advised to assess for covid-19 progression or secondary cardiopulmonary abnormalities such as pulmonary embolism, superimposed bacterial pneumonia, or heart failure that can potentially be secondary to covid-19 myocardial injury (fig 2, q8) . infection and covid-19 test availability. falsely negative covid-19 testing is more prevalent in high pre-test probability circumstances and repeat covid-19 testing is therefore advised if available. depending upon the imaging findings, other clinical investigations may be pursued. the third scenario (fig 3) the third scenario first considers the potential availability of poc covid-19 testing. imaging is advised when poc covid-19 testing is available and positive (fig 3, q9) for the same reasons as described for scenario 2. based upon imaging findings and clinical features, patients are subsequently supported and monitored with a level of intensity consistent with clinical features. imaging is again indicated if patients subsequently clinically worsen (fig 3, q11) . imaging is advised to support more rapid triage of patients in a resource-constrained setting when poc covid-19 testing is not available or negative (fig 3, q10) . imaging may reveal features of covid-19, which within this scenario may be taken as a presumptive diagnosis of covid-19 for medical triage and associated decisions regarding disposition, infection control, and clinical management. in this high pre-test probability environment, and as described for scenario 2, the possibility of falsely negative covid-19 testing creates a circumstance where a covid-19 diagnosis may be presumed when imaging findings are strongly suggestive of covid-19 despite negative covid-19 testing. this guidance represents a variance from other published recommendations which advise against the use of imaging for the initial diagnosis of covid-19 (28) and was supported by direct experience amongst panelists providing care within the conditions described for this scenario. the relationship between disease severity and triage may need to be fluid depending upon resources and case load. when imaging reveals an alternative diagnosis to covid-19, management is based upon established guidelines or standard clinical practice. multiple studies have shown no difference in important outcomes (mortality, length of stay, and ventilator days) for intensive care unit patients imaged on-demand as compared to a daily routine protocol (29) (30) (31) (32) . avoidance of non-value-added imaging is particularly important in the covid-19 patient population to minimize exposure risk of radiology technologists and to conserve ppe. with the recent emergence of sars-cov2 as a human pathogen, there are no long-term followup studies of survivors. postmortem evaluation of a single patient who succumbed to severe covid-19 showed pathologic findings consistent with diffuse alveolar damage, similar to findings previously described with severe acute respiratory syndrome (sars) and middle east respiratory syndrome (mers) (33) . patients with functional impairment following recovery from covid-19 should undergo imaging to differentiate between expected morphologic abnormalities as sequelae of infection, mechanical ventilation, or both versus a different and potentially treatable process. while ct findings of covid-19 infection are nonspecific, their presence in an asymptomatic patient with no or mild respiratory symptoms is concerning in a setting of known community transmission, particularly if there is no better alternative diagnosis. asymptomatic carriers of covid-19 have been estimated to comprise 17.9% -33.3% of all infected cases (34, 35) . asymptomatic infection with suggestive ct findings in the lung has been documented in screened cruise ship passengers (11) . it is believed that the presence of undetected infected and mildly symptomatic or asymptomatic individuals may be contributing to the rapid geographic spread of sars-cov2 (9) . rt-pcr testing in this scenario is important to potentially identify an occult infection and limit further transmission both within the community and in the environment where the patient is receiving medical care. in highly prevalent areas, an additional uncertainty is whether ct should be used as a screening tool either as a stand-alone or as an adjunct to rt-pcr to exclude occult infection prior to surgery or intensive immunosuppressive therapies. the panel's ratings are provided in figure 4 , and a summary of all recommendations is provided in table 2 . for purposes of image interpretation and reporting, readers are referred to a recently published systematic review of imaging findings of covid-19 (36) and a multi-society consensus paper on reporting chest ct findings related to covid-19 (37) . as an aid to improving radiologist and pulmonologist familiarity with the imaging findings of covid-19, we provide the following link (https://www.fleischner-covid19.org) to the fleischner society website where an educational repository of proven covid-19 cases can be found. this statement is intended to offer guidance to physicians on the use of thoracic imaging across a breadth of healthcare environments. it represents the collective opinions and perspectives of thoracic radiology, pulmonology, intensive care, emergency medicine, laboratory medicine, and infection control experts practicing in 10 countries, representative of the highest burden of covid-19 worldwide. it also represents opinion at a moment in time within a highly-dynamic environment where the status of regional epidemics and the availability of critical resources to combat those epidemics vary daily. the evidence base supporting the use of imaging across the scenarios presented is scant and the advice presented herein may undergo refinement through rigorous scientific investigation, exposing nuances of image interpretation that may lead to prognostic information and guide management decisions. at the time of this writing, no therapy has been confirmed to alter the course of covid-19, there is no known cure, and there is no vaccine for prevention. as effective treatments are developed, thoracic imaging may find new roles by establishing treatment response or characterizing patients as likely responders to novel therapies. ï�· daily chest radiographs are not indicated in stable intubated patients with covid-19 ï�· ct is indicated in patients with functional impairment and/or hypoxemia after recovery from covid-19 ï�· covid-19 testing is indicated in patients incidentally found to have findings suggestive of covid-19 on a ct scan world health organization: director-general's opening remarks at the media briefing on covid-19 -11 world health organization: coronavirus disease (covid-19) situation dashboard coronavirus covid-19 global cases by the center for systems science and engineering correlation of chest ct and rt-pcr testing in coronavirus disease 2019 (covid-19) in china: a report of 1014 cases stability issues of rt-pcr testing of sars-cov-2 for hospitalized patients clinically diagnosed with covid-19 detection of sars-cov-2 in different types of clinical specimens sars-cov-2 viral load in upper respiratory specimens of infected patients sensitivity of chest ct for covid-19: comparison to rt-pcr substantial undocumented infection facilitates the rapid dissemination of novel coronavirus (sars-cov2) transmission of 2019-ncov infection from an asymptomatic contact in germany chest ct findings in cases from the cruise ship "diamond princess radiology department preparedness for covid-19: radiology scientific expert panel covid-19) outbreak: what the department of radiology should know covid-19) interim infection prevention and control recommendations epidemiological characteristics of 2143 pediatric patients with 2019 coronavirus disease in china is there a role for lung ultrasound during the covid-19 pandemic? frequency and distribution of chest radiographic findings in covid-19 positive patients cardiovascular considerations for patients, health care workers, and health systems during the coronavirus disease 2019 (covid-19) pandemic radiology department preparedness for covid-19: facing an unexpected outbreak of the disease world health organization: critical preparedness, readiness and response actions for covid-19 covid-19) public health recommendations for community-related exposure guidance for risk assessment and public health management of healthcare personnel with characteristics of and important lessons from the coronavirus disease 2019 (covid-19) outbreak in china: summary of a report of 72314 cases from the chinese center for disease control and prevention first of 3 diagnostic tests for sars-cov-2 coronavirus available from biomã©rieux xpertâ® xpress sars-cov-2 has received fda emergency use authorization id now covid-19 molecular. in minutes statements/recommendations-for-chest-radiography-and-ct-for-suspected-covid19-infection abandoning daily routine chest radiography in the intensive care unit: metaanalysis comparison of routine and on-demand prescription of chest radiographs in mechanically ventilated adults: a multicentre, cluster-randomised, two-period crossover study azurea network for the radioday study g. chest radiographs in 104 french icus: current prescription strategies and clinical value (the radioday study) acr appropriateness criteria(r) intensive care unit patients pathological findings of covid-19 associated with acute respiratory distress syndrome. the lancet respiratory estimating the asymptomatic proportion of coronavirus disease 2019 (covid-19) cases on board the diamond princess cruise ship estimation of the asymptomatic ratio of novel coronavirus infections (covid-19) coronavirus disease 2019 (covid-19): a systematic review of imaging findings in 919 patients radiological society of north america expert consensus statement on reporting chest ct findings related to covid-19. endorsed by the society of thoracic radiology, the american college of radiology, and rsna moderate-to-severe features refer to evidence of significant pulmonary dysfunction or damage. high pre-test probability is based upon high background prevalence of disease associated with community transmission. rapid covid-19 test is a point-of-care test with a less than one-hour turnaround time. numbers in blue circles indicate key questions referenced in the text and presented in figure 4. contextual detail and considerations for imaging with cxr (chest radiography) versus ct key: cord-327257-doygrgrc authors: zhu, jocelyn; shen, beiyi; abbasi, almas; hoshmand-kochi, mahsa; li, haifang; duong, tim q. title: deep transfer learning artificial intelligence accurately stages covid-19 lung disease severity on portable chest radiographs date: 2020-07-28 journal: plos one doi: 10.1371/journal.pone.0236621 sha: doc_id: 327257 cord_uid: doygrgrc this study employed deep-learning convolutional neural networks to stage lung disease severity of coronavirus disease 2019 (covid-19) infection on portable chest x-ray (cxr) with radiologist score of disease severity as ground truth. this study consisted of 131 portable cxr from 84 covid-19 patients (51m 55.1±14.9yo; 29f 60.1±14.3yo; 4 missing information). three expert chest radiologists scored the left and right lung separately based on the degree of opacity (0–3) and geographic extent (0–4). deep-learning convolutional neural network (cnn) was used to predict lung disease severity scores. data were split into 80% training and 20% testing datasets. correlation analysis between ai-predicted versus radiologist scores were analyzed. comparison was made with traditional and transfer learning. the average opacity score was 2.52 (range: 0–6) with a standard deviation of 0.25 (9.9%) across three readers. the average geographic extent score was 3.42 (range: 0–8) with a standard deviation of 0.57 (16.7%) across three readers. the inter-rater agreement yielded a fleiss’ kappa of 0.45 for opacity score and 0.71 for extent score. ai-predicted scores strongly correlated with radiologist scores, with the top model yielding a correlation coefficient (r(2)) of 0.90 (range: 0.73–0.90 for traditional learning and 0.83–0.90 for transfer learning) and a mean absolute error of 8.5% (ranges: 17.2–21.0% and 8.5%-15.5, respectively). transfer learning generally performed better. in conclusion, deep-learning cnn accurately stages disease severity on portable chest x-ray of covid-19 lung infection. this approach may prove useful to stage lung disease severity, prognosticate, and predict treatment response and survival, thereby informing risk management and resource allocation. coronavirus disease 2019 (covid19) is an infectious disease that can cause severe respiratory illness [1, 2] . first reported in wuhan, china in december 2019 [3] , covid-19 was declared a pandemic on mar 11, 2020 (https://www.who.int). more than 2 million people have been infected and more than 120,000 have died of covid-19 (https://coronavirus.jhu.edu, a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 apr 14, 2020) . these numbers are predicted to continue to rise in the foreseeable future with a high likelihood of a second wave and recurrence. covid-19 has already overwhelmed many hospitals in many countries. portable chest x-rays (cxr) has become an indispensable imaging tool to imaging patients with contagious diseases because it is informative and the equipment can be readily disinfected. cxr provides both the extent and severity of lung infection, and is widely used to quantify disease severity in many lung diseases [4] . the hallmarks of covid-19 lung infection on cxr include bilateral, peripheral hazy opacity and airspace consolidation [5] . given the anticipated shortage of intensive care unit (icu) beds and mechanical ventilators in many hospitals, cxr has the potential to play a critical role in decision-making to determine which patients to put on a mechanical ventilator, monitor disease progression and treatment effects during mechanical ventilation, and determine when it is safe to extubate. similarly, computed tomography (ct), which offer better sensitivity than cxr, has been used in imaging suspected or confirmed covid-19 patients, albeit mostly in china early on in the pandemic [6] [7] [8] [9] . however, ct is less convenient, not practical in icu setting, and prone to cross contamination and, thus, it is not widely used in covid-19 infection or similar contagious disease in the united states. deep-learning artificial intelligence (ai) has become increasingly popular for analyzing diagnostic images [10, 11] . ai has the potential to facilitate disease diagnosis, diseases severity staging, and longitudinal monitoring of disease progression. one common machine-learning algorithm is the convolutional neural network (cnn) [12, 13] , which takes an input image, learns important features in the image such as size or intensity, and saves these parameters as weights and bias to differentiate types of images [14, 15] . cnn architecture is ideally suited for analyzing images. the majority of machine learning algorithms to date are trained to solve specific tasks, working in isolation. models have to be rebuilt from scratch if the feature-space distribution changes. transfer learning overcomes the isolated learning paradigm by utilizing knowledge acquired for one task to solve related ones. transfer learning in ai is particularly important for small sample size data because the pre-trained weights enables more efficient training and better performance [16, 17] . there are many ai algorithms that have already been developed for cxr applications (see review [18] ) and these ai algorithms can be repurposed to study covid-19 lung infection. a few studies have already reported ai applications to classify covid-19 versus non-covid-19 lung infection using cxr [19] [20] [21] [22] and ct [23] [24] [25] [26] . many of these referenced publications here are non-peer reviewed pre-prints. these studies aimed to classify covid-19 versus non-covid-19 images, not to stage lung disease severity. classification of covid-19 versus non-covid-19 images for the purpose of diagnosis made by ai unfortunately has poor specificity. the reverse-transcription polymerase chain reaction test of a nasopharyngeal or oropharyngeal swab is still needed for definitive diagnosis. by contrast, deep-learning ai is well suited to stage disease severity. to our knowledge, there have been no studies to date that use deeplearning ai to stage disease severity on cxr of covid-19 lung infection. the goal of this study was to employ deep-learning cnn to stage disease severity of covid-19 infection on cxr (as opposed to classify covid-19 versus non-covid-19 lung infection). the ground truths were the disease severity scores determined by expert chest radiologists. comparison was made with traditional and transfer learning. this approach has the potential to provide frontline physicians an efficient and accurate means to triage patients, assess risk, allocate resources, as well as to monitor disease progression and treatment response. this approach is timely and urgent because of the anticipated widespread shortage of icu beds and mechanical ventilators. the diagram of patient selection and experimental design is shown in fig 1. this is a retrospective study using publicly available de-identified data. images were downloaded from https:// github.com/ieee8023/covid-chestxray-dataset [27] on mar 30, 2020. the original download contained 250 images of covid-19 and sars. only anterior-posterior and posterior-anterior cxr from covid-19 were included in this study, resulting in final sample size of 131 cxrs from 84 covid-19 positive patients (51 males and 29 females; 4 with missing information). the mean and standard deviation of age was 55.1±14.9 years old for men and 60.1±14.3 years old for women. radiological scoring is widely used to stage disease severity in many lung diseases [4] . to establish the disease severity score, a group of 6 chest radiologists with at least 20 years of experience worked together to reach consensus by evaluating two dozen images of portable cxrs of covid-19 patients from stony brook university hospital (these testing cxrs were not used in this study). our scoring system was adapted from those by warren et al. [28] and wong et al [5] . two board-certified chest radiologists with 20+ years of experience and one radiology resident reviewed image quality and scored the cxr for disease severity using the following criteria based on degree of opacity and geographical extent. the degree of opacity score of 0-3 was assigned to each of the right and left lung as: 0 = no opacity; 1 = ground glass opacity; 2 = consolidation; 3 = white-out. the right and left lung were scored separately and were added together. the geographical extent score of 0-4 was assigned to each of the right and left lung depending on the extent of involvement with ground glass opacity or consolidation: 0 = no involvement; 1 = <25%; 2 = 25-50%; 3 = 50-75%; 4 = >75% involvement. the right and left lung were scored separately and were added together. sum of the two types of scores (0-14) and the product of the two types scores (0-48) were computed. the average of all radiologist scores were used for analysis. a convolutional neural network (fig 2a) [12, 13] with an additional regression layer was utilized to predict disease severity scores of cxr on a graded scale. the images were normalized, converted to rgb images, resized to 64x64 pixel images, and separated into 80% training and 20% testing datasets. this model architecture included convolutional, activation, batch normalization, max pooling, dense layers, and a final dense regression layer to fit the model to predict continuous values. specifically, for convolutional neural networks, deep learning models are well-suited to learn complex functions, and optimize performance of regularization methods. batch normalization layers served to stabilize the learning process by standardizing inputs, and standard rectified linear activation layers for cnns were used to optimize performance. in general, the number of layers and nodes depends heavily on the input data and prediction goals, therefore heuristics were used to generate a standard cnn model for image analysis, the model was then fine-tuned based on performance on the testing dataset. this resulted in three 3x3 kernel sized convolutional layers. l2 regularization and dropout layers, using a parametric probability of 0.3 as determined by analyzing performance results, were used to prevent overfitting. an optimal batch size of 8 was determined by five-fold cross validation, and the model was trained for 110 epochs. the loss function was measured by mean squared error as the model predicted continuous values, and the learning rate was set to 0.001, as it proved to perform more positively than the standard 0.01 rate in early training, and the adam optimizer proved the most successful in minimizing loss. given the limited dataset, a transfer learning method was also explored to optimize prediction. a vgg16 model was loaded with sample weights trained off the imagenet dataset [29] . to be compatible with vgg16, the data was normalized, converted to rgb images, and resized into 224x224 pixel images. the dataset was also split into 80% training and 20% testing. only the top layers were trained to prevent overfitting, and all convolutional layers were frozen. standard vgg16 architecture was utilized, with the exception of the fully connected layers ( fig 2b) . once out of the convoluted layers, the model output was flattened, then passed through three dense layers, including a regression dense layer. for this transfer learning method, cross validation revealed an optimal batch size of 8 and only 10 epochs were necessary to train the deep learning artificial intelligence staging of portable chest radiographs in covid-19 data, as opposed to the 110 needed for the traditional training method. adam optimizer and 0.001 learning rate were similarly employed in this model. the means and standard deviations of the radiologist scores, and the fleiss' kappa inter-rater agreement were calculated. correlation analysis between ai-predicted versus radiologist severity scores were analyzed with slopes, intercepts, correlation coefficients (r 2 ), p-values, and mean absolute errors (mae) reported. results using traditional and transfer learning were compared. all results are reported as means ± standard deviations. a p<0.05 was taken to be statistically significant unless otherwise specified. note that the prediction of graded values precluded receiver operating curve (roc) analysis. the final sample size consisted of 131 cxr from 84 covid-19 positive patients (51m and 29f; 4 missing information). the men were on average 55.1±14.9 years old and women were 60.1±14.3 years old. cxr examples demonstrating a range of severity scores which include extent and opacity score are illustrated in fig 3. cxr of covid-19 positive patients showed hazy opacities and/or airspace consolidation. it has a predominance of bilateral, peripheral deep learning artificial intelligence staging of portable chest radiographs in covid-19 and lower lung zone distribution. the results of three raters are summarized in table 1 . the mean opacity score was 2.52 (0-6), with a standard deviation of 0.25 (9.9%) across three readers. the mean geographic extent score was 3.42 (a range from 0 to 8), with a standard deviation of 0.57 (16.7%) across three readers. the low standard deviations across three raters suggest good agreement. fleiss' kappa for inter-rater agreement was 0.45 (z = 24.4) for opacity score and 0.71 (z = 29.5) for geographic extent score, indicating good to excellent agreement. the mean of sum of scores was 5.84 (0-14) and mean of product of scores was 11.86 (0-48). fig 4 shows the histograms of radiologist scores. the product score was slightly skewed toward lower values because numbers multiplied by a number close to zero tended to be of lower values. the data were separated into 80% training and 20% testing datasets. the loss function decreased and prediction accuracy improved with increasing epochs for both training and validation datasets. for the traditional training method, the loss function typically converged at training epochs of 30-50 epochs, and five-fold validation confirmed optimal hyperparameters of training on 110 epochs. for the transfer learning method employing a vgg16 model, 10 epochs were sufficient to train the model with high predictive ability; more epochs led to overfitting and increased computational time. fig 5 shows the ai-predicted scores versus radiologist severity scores of the "test" dataset, obtained for different types of scores using traditional and transfer learning. overall, there were strong correlations between the ai-predicted and radiologist scores. the slopes, intercepts, r 2 , p values, and mean absolute errors (mae) with traditional and transfer learning are summarized in table 2 . the slopes were generally (but not consistently across different scores) closer to unity, the intercepts were generally closer to zero, and r 2 are generally closer to unity for the transfer learning compared to traditional learning. the top model yielded a r 2 = 0.90 (range: 0.73-0.90 for traditional learning and 0.83-0.90 for transfer learning). the p-values for the transfer learning were consistently smaller than to those of the traditional learning. the maes for the transfer learning are consistently smaller than to those of the traditional learning. the top model yielded a mae of 8.5% (range: 17.2% to 21.3% for traditional learning, and 8.5% to 15.5% for transfer learning). transfer learning generally performed better than those traditional learning. this study tested the hypothesis that deep-learning convolutional neural networks accurately stage disease severity on portable chest x-rays using radiologists' severity scores as ground truths associated with covid-19 lung infection. disease severity scores by three radiologists yielded good inter-rater agreement. ai-predicted scores were highly correlated with radiologist scores. transfer learning improved efficiency and shorten computational time without compromising performance. although there are already over 5500 publications on covid-19 on pubmed (keyword "covid-19" apr 21, 2020), only a few studies have reported ai applications to classify covid-19 versus non-covid-19 lung infection on cxr [19] [20] [21] [22] and ct [23] [24] [25] [26] . all of these studies (including many non-peer-reviewed pre-prints cited here) aimed to classify covid-19 versus non-covid-19 images, not to stage disease severity. a major challenge for table 2 . summary of slope, intercepts, r 2 , and p values of the correlation analysis, and the mean absolute error (mae) of ai-predicted and radiologist scores with traditional and transfer learning. the % was obtained by dividing mae by the corresponding maximum score. table 2 ai in the application to disease diagnosis is that training datasets were limited to a few lung diseases, resulting in low generalizability. amongst these studies, most compare cxr of covid-19 infection with those of bacterial pneumonia and/or normal cxr. more importantly, ai diagnosis of disease in general, and covid-19 infection in particular, based on cxr has poor specificity because many pathologies have similar radiographic appearance as other infections and diseases on cxr. definitive diagnosis of covid-19 infection still requires a test by reverse transcription polymerase chain reaction of a nasopharyngeal or oropharyngeal swab. our study differs from these previous studies in that we used ai of cxr to stage disease severity on a graded scale in positive covid-19 patients. deep-learning ai, specifically a convolutional neural network, is well suited to extract information from cxr and stage disease severity by training using chest radiologist determination of disease severity scores. we evaluated individual and combination scores of severities as it is unknown which type of scores more accurately reflect lung infection disease severity. our findings showed that most scores yielded similar correlation coefficients. one possible explanation is that both the opacity score and geographic extent yielded similar information regarding disease severity. further studies are needed to determine which is superior. it is worth noting that performance by receiveroperating curve (roc) analysis (such as area under the curve (auc), accuracy, sensitivity and specificity) cannot be used for continuous variables. instead, correlation analysis and mean square error analysis were performed. there were strong correlations between the ai-predicted and radiologist scores, with the top model yielding a r 2 = 0.90 and a mean absolute error of 8.5%. this is remarkable performance. the ideal intercept should be zero and ideal slope should be unity. the intercepts were generally close to zero. however, all the slopes were consistently below the line of unity. a possible explanation is that the score distributions were skewed by low score values. as the models attempted to minimize loss, it tended to underestimate higher scores, resulting in slope less than unity. other explanations are possible. further studies are needed. another novelty of our study is that we compared traditional and transfer learning. transfer learning approach is expected to yield good performance with small datasets. most of the performance indices (such as r 2 , p values, and maes) were better with transfer learning compared to traditional learning. the training time with transferring learning was shorter without compromising performance. although radiology reports are informative, they are usually qualitative. a quantitative score of disease severity of portable cxr afforded by ai in an accurate and efficient manner should prove useful in clinical settings of covid-19 circumstance. with cxr being an indispensable imaging tool in the management of covid-19 lung infection, radiologists need an efficient and accurate means to stage disease severity and monitor disease progression and treatment response. furthermore, with the anticipated widespread shortage of icu beds and ventilators, frontline physicians need an efficient and accurate means to triage patients, assess risk and allocate resources. this ai approach to stage disease severity is ideally suited to tackle these challenges associated with the covid-19 pandemics or the like. this proof-of-concept pilot study has several limitations. first, this retrospective study was conducted on an open multi-institutional dataset with a small sample size. this topic is timely and only small dataset are currently available at the time of this writing. these findings need to be replicated on larger sample size with multi-institutional data. it may also be of important to investigate multiple longitudinal time points to evaluate disease progression. second, we did not study the correlation of radiographic score severity with clinical disease severity or clinical outcome which were unavailable. although radiographic score has been widely used as a surrogate marker of disease severity in a variety of lung diseases (see review [4] ), it is unknown at this time if radiographic scoring reflects functional or clinical outcome in the case of covid-19. third, a variety of radiographic scoring systems has been used in other studies [5, 28] , each has its advantages and disadvantages. a more sophisticated score system could be explored. future studies should also consider incorporating non-imaging data (such as demographics, co-morbidities, vitals, blood biomarkers). in conclusion, deep-learning convolutional neural networks accurately stage lung disease severity on portable chest x-rays associated with covid-19 lung infection. this approach has the potential to be used to prognosticate, stage disease severity, monitor disease progression and treatment response, which in turn could inform risk management and resource allocation associated with the covid-19 pandemic. conceptualization: jocelyn zhu, tim q. duong. the continuing 2019-ncov epidemic threat of novel coronaviruses to global health-the latest 2019 novel coronavirus outbreak in wuhan, china outbreak of pneumonia of unknown etiology in wuhan, china: the mystery and the miracle early transmission dynamics in wuhan, china, of novel coronavirus-infected pneumonia a chest radiograph scoring system in patients with severe acute respiratory infection: a validation study frequency and distribution of chest radiographic findings in covid-19 positive patients coronavirus disease 2019 (covid-19): a systematic review of imaging findings in 919 patients the clinical and chest ct features associated with severe and critical covid-19 pneumonia temporal changes of ct findings in 90 patients with covid-19 pneumonia: a longitudinal study ct imaging features of 2019 novel coronavirus (2019-ncov) deep learning using deep convolutional neural networks to identify and classify tumor-associated stroma in diagnostic breast biopsies imagenet classification with deep convolutional neural networks improving neural networks by preventing co-adaptation of feature detectors very deep convolutional networks for large-scale image recognition deep machine learning-a new frontier in artificial intelligence research multi-task transfer learning deep convolutional neural network: application to computer-aided diagnosis of breast cancer on mammograms transfer learning with deep convolutional neural network for liver steatosis assessment in ultrasound images a systematic review of the diagnostic accuracy of artificial intelligence-based computer programs to analyze chest x-rays for pulmonary tuberculosis goldgof gm. finding covid-19 from chest x-rays using deep learning on a small datase a modified deep convolutional neural network for detecting covid-19 and pneumonia from chest x-ray images based on the concatenation of xception and resnet50v2. informatics in medicine unlocked covid-19: automatic detection from x-ray images utilizing transfer learning with convolutional neural networks automatic detection of coronavirus disease classification of covid-19 patients from chest ct images using multiobjective differential evolution-based convolutional neural networks artificial intelligence distinguishes covid-19 from community acquired pneumonia on chest ct covid-net: a tailored deep convolutional neural network design for detection of covid-19 cases from covid-19 image data collection severity scoring of lung oedema on the chest radiograph is associated with clinical outcomes in ards very deep convolutional networks for large-scale image recognition key: cord-127759-wpqdtdjs authors: qi, xiao; brown, lloyd; foran, david j.; hacihaliloglu, ilker title: chest x-ray image phase features for improved diagnosis of covid-19 using convolutional neural network date: 2020-11-06 journal: nan doi: nan sha: doc_id: 127759 cord_uid: wpqdtdjs recently, the outbreak of the novel coronavirus disease 2019 (covid-19) pandemic has seriously endangered human health and life. due to limited availability of test kits, the need for auxiliary diagnostic approach has increased. recent research has shown radiography of covid-19 patient, such as ct and x-ray, contains salient information about the covid-19 virus and could be used as an alternative diagnosis method. chest x-ray (cxr) due to its faster imaging time, wide availability, low cost and portability gains much attention and becomes very promising. computational methods with high accuracy and robustness are required for rapid triaging of patients and aiding radiologist in the interpretation of the collected data. in this study, we design a novel multi-feature convolutional neural network (cnn) architecture for multi-class improved classification of covid-19 from cxr images. cxr images are enhanced using a local phase-based image enhancement method. the enhanced images, together with the original cxr data, are used as an input to our proposed cnn architecture. using ablation studies, we show the effectiveness of the enhanced images in improving the diagnostic accuracy. we provide quantitative evaluation on two datasets and qualitative results for visual inspection. quantitative evaluation is performed on data consisting of 8,851 normal (healthy), 6,045 pneumonia, and 3,323 covid-19 cxr scans. in dataset-1, our model achieves 95.57% average accuracy for a three classes classification, 99% precision, recall, and f1-scores for covid-19 cases. for dataset-2, we have obtained 94.44% average accuracy, and 95% precision, recall, and f1-scores for detection of covid-19. conclusions: our proposed multi-feature guided cnn achieves improved results compared to single-feature cnn proving the importance of the local phase-based cxr image enhancement. coronavirus disease 2019 (covid19) is an infectious disease caused by severe acute respiratory syndrome coronavirus 2 (sars-cov-2), a newly discovered coronavirus [1, 2] . in march 2020, the world health organization (who) declared the covid-19 outbreak a pandemic. up to now, more than 9.23 million cases have been reported across 188 countries and territories, resulting in more than 476,000 deaths [3] . early and accurate screening of infected population and isolation from public is an effective way to prevent and halt spreading of virus. currently, the gold standard method used for diagnosing covid-19 is real-time reverse transcription polymerase chain reaction (rt-pcr) [4] . the disadvantages of rt-pcr include its complexity and problems associated with its sensitivity, reproducibility, and specificity [5] . moreover, the limited availability of test kits makes it challenging to provide the sufficient diagnosis for every suspected patients in the hyper-endemic regions or countries. therefore, a faster, reliable and automatic screening technique is urgently required. in clinical practice, easily accessible imaging, such as chest x-ray (cxr), provides important assistance to clinicians in decision making. compared to computed tomography (ct) the main advantages of cxr are: enabling fast screening of patients, being portable, and easy to setup (can be setup in isolation rooms). however, the sensitivity and specificity (radiographic assessment accuracy) of cxr for diagnosing covid-19 is low compared to ct. this is especially problematic for identifying early stage covid-19 patients with mild symptoms. this causes larger intra-and inter-observer variability in reading the collected data by radiologists since qualitative indicators can be subtle. therefore, there is increased demand for computer aided diagnostic method to aid the radiologist during decision making for improved management of covid-19 disease. in view of these advantages and motivated by the need for accurate and automatic interpretation of cxr images, a number of studies based on deep convolutional neural networks (cnns) have shown quite promising results. ozturk et al. [6] proposed a cnn architecture, termed darkcovidnet, and achieved 87.02% three class classification accuracy. the method was evaluated on 127 covid-19, 500 healthy and 500 pneumonia cxr scans. covid-19 data was obtained from 125 patients. wang et al. [7] built a public dataset named covidx, which is comprised of a total of 13975 cxr images from 13870 patient case and developed covid-net, a deep learning model. their dataset had 358 covid-19 images obtained from 266 patients. their model achieved 93.3% overall accuracy in classifying normal, pneumonia, and covid-19 scans. in [8] a resnet-50 architecture was utilized to achieve a 96.23% overall accuracy in classifying four classes, where pneumonia was split into bacterial pneumonia and viral pneumonia. however, there were only eight covid-19 cxr images used for testing. in [9] , 76.37% overall accuracy was reported on a dataset including 1583 normal, 4290 pneumonia and 76 covid-19 scans. covid-19 data was collected from 45 patients. in order to improve the performance of the proposed method, data augmentation was performed on the covid-19 dataset bringing the total covid-19 datasize to 1,536. with data augmentation they have improved the overall accuracy 97.2%. in [10] , contrast limited adaptive histogram equalization (clahe) was used to enhance the cxr data. the authors proposed a depth-wise separable convolutional neural network (dscnn) architecture. evaluation was performed on 668 normal, 619 pneumonia, and 536 covid-19 cxr scans. average reported multi-class accuracy was 96.43%. number of patients for the covid-19 dataset was not available. in [11] , a stacked cnn architecture achieved an average accuracy of 92.74%. the evaluation dataset had 270 covid-19 scans from 170 patients, 1139 normal scans from 1015 patients, and 1355 pneumonia scans from 583 patients. in [12] , the reported multi-class average classification accuracy was s 94.2%. the evaluation dataset included 5000 normal, 4600 pneumonia, and 738 covid-19 cxr scans. the data was collected from various sources and patient information was not specified. in [13] transfer learning was investigated for training the cnn architecture. the evaluation dataset included 224 covid-19, 504 normal, and 700 pneumonia images. 93.48% average accuracy was reported for three-class classification. the average accuracy increased to 94.72% if viral pneumonia was included in the evaluation. in [14] , performance of three different, previously proposed, cnn architectures was evaluated for multi-class classification. with 2,265 covid-19 images, the study used the largest covid-19 dataset reported so far. average area under the curve (auc), for classification of covid-19 from regular pneumonia, was 0.73 [14] . although numerous studies have shown the capability of cnns in effective identification of covid-19 from cxr images, none of these studies investigated local phase cxr image features as multi-feature input to a cnn architecture for improved diagnosis of covid-19 disease. furthermore, except [14, 7] , most of the previous work was evaluated on a limited number of covid-19 cxr scans. in this work we show how local phase cxr features based image enhancement improves the accuracy of cnn architectures for covid-19 diagnosis. specifically, we extract three different cxr local phase image features which are combined as a multi-feature image. we design a new cnn architecture for processing multi-feature cxr data. we evaluate our proposed methods on large scale cxr images obtained from healthy subjects as well as subjects who are diagnosed with community acquired pneumonia and covid-19. quantitative results show the usefulness of local phase image features for improved diagnosis of covid-19 disease from cxr scans. our proposed method is designed for processing cxr images and consists of two main stages as illustrated in figure 1 : 1-we enhance the cxr images (cxr(x, y)) using local phase-based image processing method in order to obtain a multi-feature cxr image (m f (x, y)), and 2-we classify cxr(x, y) by designing a deep learning approach where multi feature cxr images (m f (x, y)), together with original cxr data (cxr(x, y)), is used for improving the classification performance. next, we describe how these two major processes are achieved. in order to enhance the collected cxr images, denoted as cxr(x, y), we use local phase-based image analysis [15] . three different cxr(x, y) image phase features are extracted: 1-local weighted mean phase angle (lwp a(x, y)), 2-lwp a(x, y) weighted local phase energy (lp e(x, y)), and 3-enhanced local energy attenuation image (elea(x, y)). lp e(x, y) and lwp a(x, y) image features are extracted using monogenic signal theory where the monogenic signal image (cxr m (x,y)) is obtained by combining the bandpass filtered cxr(x, y) image, denoted as cxr b (x, y), with the riesz filtered components as: here h 1 and h 2 represent the vector valued odd filter (riesz filter) [16] . α-scale space derivative quadrature filters (assd) are used for band-pass filtering due to their superior edge detection [17] . the lwp a(x, y) image is calculated using: ). we do not employ noise compensation during the calculation of the lwp a(x, y) image in order to preserve the important structural details of cxr(x, y). the lp e(x, y) image is obtained by averaging the phase sum of the response vectors over many scales using: in the above equation sc represents the number of scales. lp e(x, y) image extracts the underlying tissue characteristics by accumulating the local energy of the image along several filter responses. the lp e(x, y) image is used in order to extract the third local phase image elea(x, y). this is achieved by using lp e(x, y) image feature as an input to an l1 norm based contextual regularization method. the image model, denoted as cxr image transmission map (cxr a (x, y)), enhances the visibility of lung tissue features inside a local region and assures that the mean intensity of the local region is less than the echogenicity of the lung tissue. the scattering and attenuation effects in the tissue are combined as: here ρ is a constant value representative of echogenicity in the tissue. in order to calculate elea(x, y), cxr a (x, y) is estimated first by minimizing the following objective function [15] : in the above equation • represents element-wise multiplication, χ is an index set, and * is convolution operator. d j is calculated using a bank of high order differential filters [18] . the filter bank enhances the cxr tissue features inside a local region while attenuating the image noise. w j is a weighting matrix calculated using: equation the first part measures the dependence of cxr a (x, y) on lp e(x, y) and the second part models the contextual constraints of cxr a (x, y) [15] . these two terms are balanced using a regularization parameter λ [15] . after and is a small constant used to avoid division by zero [15] . combination of these three types of local phase images as three-channel input creates a new multi-feature image, denoted as m f (x, y). qualitative results corresponding to the enhanced local phase images are displayed in figure 2 . investigating figure 2 we can observe that the enhanced local phase images extract new lung features that are not visible in the original cxr(x, y) images. since local phase image processing is intensity invariant, the enhancement results will not be affected from the intensity variations due to patient characteristics or x-ray machine acquisition settings. the multi-feature image m f (x, y) and the original cxr(x, y) image are used as an input to our proposed deep learning architecture which is explained in the next section. our proposed multi-feature cnn architecture consists of two same convolutional network streams for processing cxr(x, y) images and the corresponding m f (x, y) respectively. strategies for the optimal fusion of features from multi-modal images is an active area of research. generally, data is fused earlier when the image features are correlated, and later when they are less correlated [19] . depending on the dataset, different types of fusion strategies outperform the other [20] . in [21] , our group has also investigated early, mid, and late-level fusion operations in the context of bone segmentation from ultrasound data. late-fusion operation has outperformed the other fusion operations. in [22] , authors have also used late-fusion network, for segmenting brain tumors from mri data, has outperformed other fusion operations. during this work we design mid-fusion and late-fusion architectures (fig.3) . as part of this work we have also investigate several fusion operations: sum fusion, max fusion, averaging fusion, concatenation fusion, convolution fusion. based on the performance of the fusion operations and fusion architectures, on a preliminary experiment, we use concatenation fusion operation for both of our architectures. we use the following network architectures as the encoder network: pretrained alexnet [23] , resnet50 [24] , sononet64 [25] , xnet(xception) [26] , inceptionv4(inception-resnet-v2) [27] and efficient-netb4 [28] . pretrained alexnet [23] and resnet50 [24] have been incorporated into various medical image analysis tasks [29] . sononet64 achieved excellent performance in implementation of both classification and localization tasks [25] . xnet(xception) [26] , inceptionv4 (inception-resnet-v2) [27] and ef-ficientnetb4 [28] were chosen due to their outstanding performance on recent medical data classification tasks as well as classification of covid-19 from chest ct data [30, 31] . we use the following datasets to evaluate the performance of proposed fusion network models: bimcv [32] , covidx [7] , and covid-cxnet [12] . covid-19 cxr scans from bimcv [32] and covidx [7] datasets were combined to generate the 'evaluation dataset' (table 1) . for normal and pneumonia datasets we have randomly selected a subset of 2567 images (from 2567 subjects) from the evaluation dataset (table 1 ). in total 2567 images from each class (normal, pneumonia, covid-19) were used during 5-fold cross validation. table 2 shows the data split for covid-19 data only. similar split was also performed for normal and pneumonia datasets. in order to provide additional testing for our proposed networks, we have designed a new test dataset which we call 'test dataset-2' ( table 3 ). the images from normal and pneumonia cases which were not included in the 'evaluation dataset' were part of the 'test dataset-2'. furthermore, we have included all the covid-19 scans from covid-cxnet [12] . in order to show the improvements achieved using our proposed multifeature cnn architecture we also trained the same cnn architectures using only m f (x, y) or cxr(x, y) images. we refer to these architectures as monofeature cnns. quantitative performance was evaluated by calculating average accuracy, precision, recall, and f1-scores for each class [9, 7] . the experiments were implemented in python using pytorch framework. all models were trained using stochastic gradient descent (sgd) optimizer, crossentropy loss function, learning rate 0.001 for the first epoch and a learning rate fig. 4 : grad-cam images [33] obtained by late fusion resnet50 architecture. decay of 0.1 every 15 epochs with a mini-batches of size 16. for local phase image enhancement, we have used sc = 2 and the rest of the assd filter parameters were kept same as reported in [15] . for calculating elea(x, y) images we used λ = 2, = 0.0001, η = 0.85, and ρ, the constant related to tissue echogenicity, was chosen as the mean intensity value of lp e(x, y). these values were determined empirically and kept constant during qualitative and quantitative analysis. qualitative analysis: gradient-weighted class activation mapping (grad-cam) [33] visualization of normal, pneumonia, and covid-19 are presented as qualitative results in figure 4 . investigating figure 4 we can see the discriminative regions of interest localized in the normal, pneumonia, and covid-19 data. quantitative analysis of evaluation dataset: table 4 shows average accuracy of the 5-fold cross validation on the 'evaluation dataset' for mono-feature cnn architectures as well as the proposed multi-feature cnn architectures. a box and whisker plot is presented in figure 5 . in most of the investigated network designs m f (x, y)-based mono-feature cnn architectures outperform cxr(x, y)-based mono-feature cnn architectures. the best average accuracy is obtained when using our proposed multi-feature resnet50 [24] architecture. all multi-feature cnns with mid-and late-fusion operation compared with mono-feature cnns, with original cxr(x, y) images as input, achieved statistically significant difference in terms of classification accuracy (p<0.05 using a paired t-test at %5 significance level). except sononet64 [25] , xnet(xception) [26] , and inceptionv4(inception-resnet-v2) [27] , all multi-feature cnns with mid-fusion operation compared with mono-feature cnns with m f (x, y) images as input show statistically significant difference in terms of classification accuracy (p<0.05 using a paired t-test at %5 significance level). we did not find any statistical significant difference in the average accuracy results between the middle-level and late-fusion networks (p>0.05 using a paired t-test at %5 significance level). figure 6 presents confusion matrix results together with average precision, recall, and f1-scores for all multi-feature late-fusion cnn architectures. one important aspect observed from the presented results we can see that almost all the investigated multi-feature networks achieved very high precision, recall, and f1-scores for covid-19 data indicating very few cases were misclassified as covid-19 from other infected types. quantitative analysis of test dataset-2: multi-feature resnet50 provides the highest overall accuracy shown in table 5 , which is consistent with the quantitative result achieved with the 'evaluation dataset'. figure 7 shows a box and whisker plot for each network. all multi-feature cnns with late-fusion operation compared with mono-feature cnns, with original cxr(x, y) imfig. 6 : confusion matrix, and average precision, recall and f1-scores obtained from 5-fold cross validation on 'evaluation data' using all multi-feature network models. ages as input, achieved statistically significant difference in terms of classification accuracy (p<0.05 using a paired t-test at %5 significance level). except xnet(xception) [26] , all the multi-feature cnns with mid fusion operation compared with mono-feature cnns with original cxr(x, y) images as input achived statistically significant difference in terms of classification accuracy (p<0.05 using a paired t-test at %5 significance level). except xnet(xception) [26] , all multi-feature cnns with mid-fusion operation compared with mono-feature cnns with m f (x, y) images as input show statistically significant difference in terms of classification accuracy (p<0.05 using a paired t-test at %5 significance level). similar to 'evaluation dataset' results, there was no statistically significant difference in the average accuracy results between the middle-level and late-fusion networks (p>0.05 using a paired t-test at %5 significance level) except resnet50 [24] , and xnet(xception) [26] architectures. confusion matrix results, together with average precision recall and f1-score values, for all multi-feature late-fusion cnn architectures evaluated are presented in fig-ure8 . similar to the results presented for 'evaluation dataset', high precision, recall, and f1-score values are obtained for the covid-19 data. development of a new computer aided diagnostic methods for robust and accurate diagnosis of covid-19 disease from cxr scans is important for improved management of this pandemic. in order to provide a solution to this need, in this work, we present a multi-feature deep learning model for classification of cxr images into three classes including covid-19, pneumonia,and normal healthy subjects. our work was motivated by the need for enhanced representation of cxr images for achieving improved diagnostic accuracy. to this end we proposed a local phase-based cxr image enhancement method. we have shown that by using the enhanced cxr data, denoted as m f (x, y), in conjunction with the original cxr data, diagnostic accuracy of cnn architectures can be improved. our proposed multi-feature cnn architectures were trained on a large dataset in terms of the number of covid-19 cxr scans and have achieved improved classification accuracy across all classes. one of the very encouraging result is the proposed models show high precision, recall, and f1-scores on the covid-19 class for both testing datasets. in addition, except for alexnet [23] , all multi-feature cnns with late fusion operation has less number of parameters compared with corresponding multi-feature cnns with middle fusion operation ( figure 9 ). since the image classifier of alexnet [23] is consist of three fully connected layers (fc), which store majority of parameters, alexnet [23] with late fusion operation almost double the number of parameters compared with middle fusion operation. the rest of networks have only one or no fc layer in the image classifiers. finally, compared to previously reported results, our work achieves the highest three class classification accuracy on a significantly larger covid-19 dataset (table 6 ). this will ensure few false positive cases for the covid-19 detected from cxr images and will help alleviate burden on the healthcare system by reducing the amount of ct scans performed. while the obtained results are very promising, more evaluation studies are required specifically for diagnosing early stage covid-19 from cxr images. our future work will involve the collection of cxr scans fig. 9 : model size vs. overall accuracy from early stage or asymptotic covid-19 patients. we will also investigate the design of a cxr-based patient triaging system. haghanifar et al. [12] unet+densenet training data: testing data: a review of coronavirus disease-2019 (covid-19) coronavirus disease 2019 an interactive web-based dashboard to track covid-19 in real time detection of sars-cov-2 in different types of clinical specimens development of reverse transcription (rt)-pcr and real-time rt-pcr assays for rapid detection and quantification of viable yeasts and molds contaminating yogurts and pasteurized food products automated detection of covid-19 cases using deep neural networks with x-ray images covid-net: a tailored deep convolutional neural network design for detection of covid-19 cases from chest x-ray images covid-resnet: a deep learning framework for screening of covid19 from radiographs covidiagnosis-net: deep bayes-squeezenet based diagnostic of the coronavirus disease 2019 (covid-19) from x-ray images covidlite: a depth-wise separable deep neural network with white balance and clahe for detection of covid-19 stacked convolutional neural network for diagnosis of covid-19 disease from x-ray images covid-cxnet: detecting covid-19 in frontal chest x-ray images using deep learning covid-19: automatic detection from x-ray images utilizing transfer learning with convolutional neural networks umls-chestnet: a deep convolutional neural network for radiological findings, differential diagnoses and localizations of covid-19 in chest x-rays localization of bone surfaces from ultrasound data using local phase information and signal transmission maps the monogenic signal α scale spaces filters for phase based edge detection in ultrasound images efficient image dehazing with boundary constraint and contextual regularization multimodal deep learning. in: icml a review: deep learning for medical image segmentation using multi-modality fusion automatic segmentation of bone surfaces from ultrasound using a filter-layer-guided cnn multi modal convolutional neural networks for brain tumor segmentation imagenet classification with deep convolutional neural networks deep residual learning for image recognition sononet: real-time detection and localisation of fetal standard scan planes in freehand ultrasound xception: deep learning with depthwise separable convolutions inception-v4, inception-resnet and the impact of residual connections on learning efficientnet: rethinking model scaling for convolutional neural networks a survey on deep learning in medical image analysis identifying melanoma images using efficientnet ensemble: winning solution to the siim-isic melanoma classification challenge automatic detection of coronavirus disease (covid-19) in x-ray and ct images: a machine learningbased approach bimcv covid-19+: a large annotated dataset of rx and ct images from covid-19 patients grad-cam: visual explanations from deep networks via gradient-based localization acknowledgements the authors are thankful to all the research groups, and national agencies worldwide who provided the open source x-ray images. funding: nothing to declare. conflict of interest the authors declare that they have no conflict of interest. key: cord-336843-c0sr3six authors: gerritsen, m. g.; willemink, m. j.; pompe, e.; van der bruggen, t.; van rhenen, a.; lammers, j. w. j.; wessels, f.; sprengers, r. w.; de jong, p. a.; minnema, m. c. title: improving early diagnosis of pulmonary infections in patients with febrile neutropenia using low-dose chest computed tomography date: 2017-02-24 journal: plos one doi: 10.1371/journal.pone.0172256 sha: doc_id: 336843 cord_uid: c0sr3six we performed a prospective study in patients with chemotherapy induced febrile neutropenia to investigate the diagnostic value of low-dose computed tomography compared to standard chest radiography. the aim was to compare both modalities for detection of pulmonary infections and to explore performance of low-dose computed tomography for early detection of invasive fungal disease. the low-dose computed tomography remained blinded during the study. a consensus diagnosis of the fever episode made by an expert panel was used as reference standard. we included 67 consecutive patients on the first day of febrile neutropenia. according to the consensus diagnosis 11 patients (16.4%) had pulmonary infections. sensitivity, specificity, positive predictive value and negative predictive value were 36%, 93%, 50% and 88% for radiography, and 73%, 91%, 62% and 94% for low-dose computed tomography, respectively. an uncorrected mcnemar showed no statistical difference (p = 0.197). mean radiation dose for low-dose computed tomography was 0.24 msv. four out of 5 included patients diagnosed with invasive fungal disease had radiographic abnormalities suspect for invasive fungal disease on the low-dose computed tomography scan made on day 1 of fever, compared to none of the chest radiographs. we conclude that chest radiography has little value in the initial assessment of febrile neutropenia on day 1 for detection of pulmonary abnormalities. low-dose computed tomography improves detection of pulmonary infiltrates and seems capable of detecting invasive fungal disease at a very early stage with a low radiation dose. neutropenic fever is one of the most important complications in cancer patients. [1] it is critical to rapidly localize infections and identify the organism involved, especially for fungal infections, which need a specific treatment approach. despite a standard diagnostic workup including chest radiograph (cxr) and microbiological screening, no focus is identified in up to 44% of patients. [2] this can partly be explained by the low sensitivity of radiographs for diagnosing pulmonary infections in neutropenic patients. [3] acquiring cxrs in respiratory asymptomatic patients with febrile neutropenia is therefore controversial. the esmo guidelines recommend performance of a cxr in every neutropenic patient with fever, whereas the idsa guidelines suggest its use only in patients with respiratory signs or symptoms. [1, 4] in febrile neutropenia a substantial part (~10%) of infections is caused by invasive fungal disease (ifd). [5] ifd usually presents as a pulmonary infection, which is rarely visible on cxr and is therefore often missed in the early phase of neutropenic fever. notably, a delayed start of appropriate treatment has a negative impact on clinical outcome. [6] in order to improve the early detection of pulmonary infiltrates in febrile neutropenia, other imaging tools have been investigated as an alternative to cxr. high-resolution computed tomography (hrct) improved pulmonary focus detection: pulmonary abnormalities were found in 60% of the neutropenic patients with persistent fever (>48 hours) and a normal cxr. [7] furthermore, direct initiation of effective antifungal treatment after the early detection of a halo sign on hrct images, had a positive effect on treatment response and survival in case of ifd, emphasizing the importance of early diagnosis. [8] however, due to costs and high radiation doses of approximately 7 msv average, hrct scanning is usually not incorporated in the initial febrile neutropenia workup. [9] another imaging modality that could be used for the evaluation of febrile neutropenia is low-dose ct scanning (ldct). ldct is performed with low mean radiation doses below 1.5 msv and without the use of contrast. two studies comparing ldct to cxr in patients with persistent febrile neutropenia demonstrated an increased detection of pulmonary abnormalities. [10, 11] therefore, we hypothesized that ldct already acquired on day 1 of febrile neutropenia would improve detection of pulmonary infections. the primary aim of our study was to compare the diagnostic performance of ldct and 2-view cxr for early detection of pulmonary infiltrates. the secondary aim was to explore its performance for early ifd detection. the study was approved by the local institutional review board (approval number nl41415.041.12) of the university medical center utrecht. written informed consent was obtained from all patients. this prospective study was conducted at the haematology ward of the university medical center utrecht, a tertiary care oncology center, between march 2013 and december 2014. patients with febrile neutropenia treated with intensive chemotherapy for haematological malignancies or receiving an autologous or myelo-ablative allogeneic stem cell transplant (sct) were eligible for inclusion. febrile neutropenia was defined as a single temperature measurement of !38.3˚c or a temperature of !38˚c for more than 1 hour, accompanied by an absolute neutrophil count of <0.5×10 9 /l or a neutrophil count <1.0×10 9 /l with a predicted decline to <0.5×10 9 /l within 3 days. patients were excluded in case of a known focus of infection unrelated to the lower respiratory tract at inclusion, active possible or probable fungal infection at inclusion, or concomitant participation in clinical research in which the subject was exposed to additional radiation. patients could participate only once in case of multiple febrile episodes. after conformation of neutropenic fever the diagnostic workup was started according to the esmo guidelines [1] and patients received broad spectrum antibiotics (imipenem) within 2 hours. all patients received selective digestive tract decontamination at the start of chemotherapy, comprising of ciprofloxacin 2dd 500 mg and fluconazole 1dd 150 mg. an additional ldct scan was made within 24 hours after the start of fever for research purposes only. all ldct scans remained blinded during the study. the study patients were monitored until they were 24 hours without fever. after this follow-up period and when all culture results were available a predefined consensus diagnosis of the cause of the fever episode was made by an expert panel consisting of 2 haematologists, a microbiologist and a radiologist. diagnostic workup for febrile neutropenia. the diagnostic workup included a clinical examination, a 2-view cxr and microbiological screening. standard microbiological screening included evaluation of urine and blood cultures. [1, 4] a respiratory microbiological evaluation was performed in case of a suspected pulmonary infection (clinical symptoms of dyspnea or cough, sputum production, or an oxygen saturation level <93%). for the respiratory evaluation a throat swab was taken and tested for viral pathogens with pcr pack 1 (influenza virus, respiratory syncytial virus (rsv), coronavirus, rhinovirus) and pcr pack 2 (adenovirus, human metapneumovirus (hmpv), parainfluenzavirus type 1-4, bocavirus and mycoplasma pneumoniae). in case of a possible atypical pneumonia the swab was additionally tested for atypical pathogens with pcr pack 3 (legionella pneumoniae, chlamydophila pneumoniae, chlamydophila psitacci, coxiella burnetti). if sputum was available this was tested for bacteria (gram stain, culture) and fungal pathogens (blancophore stain, culture). other microbiological testing was only performed if judged appropriate by the attending physician. furthermore, serum samples were taken twice weekly for galactomannan (gm) testing at the end of inclusion. if fever persisted for 4 days, a routine hrct scan was acquired to detect possible ifd. in case of abnormalities suspect for (fungal) pulmonary infection a broncho-alveolar lavage (bal) was performed. bal fluid was routinely tested for bacteria (gram, stain, culture) including haemophilus influenza, and fungal pathogens (blancophore stain, gm, microscopy and culture). on indication bal fluid was also tested for nocardia (culture), mycobacteria (culture, microscopy, pcr), pneumocystis jiroveci pneumonia (pcr, microscopy) and viral pathogens: pcr pack 1,2 and 3. in case of ifd anti-fungal treatment was started and a hrct was repeated after six weeks. consensus diagnosis. the consensus diagnosis of a fever episode was defined according to predefined categories as either pulmonary infection or non-pulmonary causes of fever such as line infection, mucositis, other infections (i.e. sinusitis), unknown focus of fever or any combination of the above. the expert panel diagnosis was based on information obtained from the clinical charts, microbiology results and imaging results: the 2-view cxr and the hrct. hrct was only available in case of persistent (4 days) fever, and ldct results were not used for the consensus diagnosis. clinical criteria for a pulmonary infection included either one of the following symptoms: coughing, sputum production, dyspnea or an oxygen saturation level <93%. microbiological criteria were a positive culture or pcr for bacterial, viral and fungal pathogens known to cause pulmonary infections, or in case of invasive aspergillosis a positive gm in bal (!0.8) or serum (!0.5). in case of abnormal imaging results without additional clinical or microbiological criteria these were considered false positive. fungal infections were classified as either possible, probable or proven in accordance with the revised european organization for research and treatment of cancer/mycosis study group (eortc/msg) criteria. [12] ifd was considered to be ruled out in all patients recovering from neutropenic fever within 3 days without ever receiving mould-active antifungal therapy, given the very low likelihood of spontaneous recovery during neutropenia, without appropriate treatment. central venous catheter infections were defined by criteria adapted from the dutch surveillance network of nosocomial infections (prezies). [13] mucositis was diagnosed based on clinical and radiological (typhlitis) criteria. clinical criteria included pain when swallowing, nausea, vomiting, abdominal pain/cramping and diarrhea, physical examination and the absence of a positive microbiological test. any other focus of infection was mainly determined by the judgment of the attending physician combined with the results of the diagnostic workup. fever of unknown origin was defined as fever without any focus or etiology identified by clinical, radiological or microbiological examination. the ldct scan was made during inspiration at the lowest achievable radiation dose. ldct images were acquired at a tube potential of 80 kvp and a tube current-time product of 10 mas or 20 mas depending on patient's weight using a 256-slice ct system (ict, philips healthcare, best, the netherlands), or at a tube current-time product of 15 mas or 30 mas with a 64-slice ct system (brilliance 64, philips healthcare, best, the netherlands). at the end of the study, anonymized cxr and ldct images were first independently evaluated by a board certified radiologist and a radiology resident who were not involved in the consensus diagnosis. the observers used a standard imaging scoring table (table 1) . second, cxr and ldct images were re-analysed if one or two observers gave a positive score. re-evaluation was performed during consensus reading of the study exams, for which additional historical radiological images were available to prevent false-positives as a result of pre-existing abnormalities. the primary objective was to investigate whether ldct is a better diagnostic tool than cxr for detection of pulmonary infections on the first day of neutropenic fever. we hypothesized that ldct could increase the detection rate of pulmonary infiltrates by 80%; from 15% with cxr (unpublished data from our institute) to 27%, considering an expected incidence of pulmonary focus of infection in febrile neutropenia of 30%. [14] we expected a proportion of discordant pairs of 15% with a two-sided test (alpha 0.05 and power 0.80). a sample size of 68 patients was esteemed sufficient according to a power calculation based on the mcnemar test. sensitivity, specificity, positive predictive value (ppv) and negative predictive value (npv) of both tests were calculated using contingency tables, the consensus diagnosis served as reference standard. an uncorrected mcnemar test was applied to determine the differences between the two modalities. [15] a p-value of 0.05 was considered significant. parametric data are presented as means ± standard deviation (sd) and non-parametric data as medians (range). cohen's kappa (κ) was used to assess the inter observer agreement and defined as excellent with κ > 0.80, good with κ between 0.61 and 0.80, moderate with κ between 0.41 and 0.60, and poor with κ 0.40. ibm spss statistics software version 22.0 (ibm, somers, ny, usa) was used for statistical analyses. the secondary objective was to explore the performance of ldct for detection of pulmonary lesions indicating ifd on day 1 of neutropenic fever. a total of 72 consecutive patients were recruited between march 2013 and december 2014. five patients were excluded: one because of a known infection unrelated to the respiratory tract at inclusion, one as a result of active fungal infection, one because the ldct was acquired more than 24 hours after the cxr, one was not able to undergo ldct scanning and for one patient cxr was not available (fig 1) . baseline characteristics of the 67 patients are listed in table 2 . according to the consensus diagnosis 11 patients (16.4%) had a pulmonary infection, of which 5 had ifd (2 possible, 3 probable), 1 had rhinovirus and 5 pneumonia of unknown aetiology. five of the patients with pneumonia underwent bronchoscopy due to persistent fever and abnormalities on hrct. in 3 of those patients a probable pathogen was identified after evaluation of bal-fluid, which was probable ifd in all 3. furthermore in 35 (55.2%) patients, overall nonspecific mucositis was the most likely cause of fever. in four of these patients abdominal imaging (ct) was performed showing no signs of typhlitis. in 15 patients (22.4%) a focus of infection was not found. the other consensus diagnosis results are listed in table 3 . the mean interval between cxr and ldct was 3.1 ± 6.3 hours. the mean radiation dose for ldct exams was 0.24 ± 0.15 msv, which is comparable to 2-3 chest radiographs (average dose posteroanterior and lateral chest radiograph: 0.1 msv). [9] none of the patients experienced adverse events as a result of ldct scanning or chest radiography. eight cxrs (11.9%) were indicative of pneumonia, 4 cxrs were false positive and 7 false negative as compared to the consensus diagnosis (table 4 ). this resulted in a sensitivity of 36%, a specificity of 93%, a ppv of 50% and a npv of 88% for cxr. none of the cxrs were suspect for ifd. thirteen ldct scans (19.4%) were suggestive of pneumonia. compared to the consensus diagnosis, 5 ldct scans were false positive and 3 false negative. sensitivity, specificity, ppv and npv were 73%, 91%, 62% and 94% respectively. (table 4) consensus diagnosis of the 5 false positive ldct scans were phlebitis and in 4 cases fever of unknown origin. no hrct scans were made because in all 5 patients the fever subsided within 48 hours after receiving broad spectrum antibiotics, therefore cxr was the only available imaging tool for the consensus meeting. only one of the patients with a false positive ldct had respiratory symptoms (cough), however a microbiological respiratory evaluation was negative. an uncontrolled mcnemar's test showed no statistically significant difference in the proportion of scans positive for pulmonary infection between cxr and ldct (p = 0.197), which was our primary endpoint. inter-observer agreement was moderate for both cxr and ldct (κ = 0.52 for cxr and κ = 0.52 for ldct, respectively). based on the consensus diagnosis 5 patients in this cohort were classified as either probable (n = 3) or possible (n = 2) ifd. the diagnosis of probable ifd was based on a positive bal gm test in all cases and 2 out of 3 also had a positive serum gm test. only 1 patient had a positive bal culture (a. versicolor). four of the patients with ifd (3 probable, 1 possible) had a ldct scan suspect for ifd on the first day of fever. in all these patients the abnormalities were also seen on the hrct performed as part of the diagnostic workup for persistent fever (mean 3.5 days later than ldct). (fig 2) none of the patients diagnosed with pulmonary ifd had abnormalities on their cxr on the first day of fever, and only one had respiratory signs or symptoms. the diagnosis of possible ifd in the patient with a negative ldct scan was based on abnormalities on hrct made on day 4 of fever. serum gm values were negative and bronchoscopy was not performed because of severe thrombocytopenia. finally one ldct scan with abnormalities suspect for ifd was considered false positive: the fever episode was classified as "fever of unknown origin". we conducted a prospective study to evaluate whether pulmonary focus detection would improve using a ldct scan instead of cxr on the first day of febrile neutropenia. we established an improved detection rate from 11.9% of radiographs to 19.4% of the ldct scans but our primary endpoint was not met. as a result of the lower than expected incidence of pulmonary infections in this study, the study was underpowered. [14] nevertheless, we demonstrated a clinically significant increased sensitivity of ldct (73% versus cxr 36%) in detecting a pulmonary focus on day 1 of neutropenic fever. the limited value of cxr as a standard diagnostic procedure is in line with previous results. [3] in a retrospective study 1083 adult sct patients were evaluated, but in none of the 242 cxrs performed in asymptomatic patients with febrile neutropenia pulmonary abnormalities indicative of infection were detected. in contrast, in 76 patients with respiratory symptoms 24 cxrs showed evidence of pneumonia. [3] in our study 4 patients had a true positive cxr; 1 of these patients had respiratory symptoms on the first day of fever. two previous studies have compared the use of cxr with ldct in febrile neutropenic patients. in the present study a much lower mean radiation dose of 0.24 msv was used compared to 0.6 msv and 1.5 msv in the other studies. [10, 11] still, reconstructed images were of diagnostic quality in all patients. the ldct scans in the previous studies were acquired using older generation ct systems (4 slice ct scanner by patsios et al. [10] and 16 slice ct scanner by kim et al. [11] .) we used two newer generation ct systems (64 slices and 256 slices) with newer x-ray tubes, detectors and software, resulting in the potential to further reduce the radiation dose without compromising on image quality. both previous studies demonstrated that ldct increased the detection rate but they differ in several aspects from our study. patsios et al. [10] compared cxr with ldct in neutropenic aml patients with already clinically suspected pneumonia, and found abnormal imaging results in 31 of 40 cxrs, whereas 38 out of 40 patients had abnormalities on ldct. instead of only focusing on pulmonary abnormalities suggestive of pulmonary infection, this study also included radiological signs of cardiac failure or fluid overload, which might explain the high number of patients with abnormalities in both cxr and ldct. [10] in a prospective study kim et al. [11] evaluated the use of ldct in a selection of patients with persistent neutropenic fever (>48 hours) regardless of the presence of respiratory symptoms. of the 207 included patients 150 were diagnosed with pneumonia (72%). differences between sensitivity of cxr and ldct for correctly diagnosing infectious pneumonia were a little less pronounced than in our study: cxr 39% and ldct 63%. to our knowledge this is the first study to evaluate ldct in all patients with febrile neutropenia on day 1 of fever. we demonstrated an increase in sensitivity and an improved npv when performing ldct in the detection of a pulmonary infection in febrile neutropenia. this is important because detection of a focus of infection in neutropenic fever is difficult and patients often undergo several diagnostic tests and uncertain treatments which can have negative side effects. therefore it can be expected that every improvement in the diagnostic workup will eventually lead to improved patient care. a major advantage of our study is the completely independent assessment of both index and reference test (ldct and cxr by independent radiologists) and the short time interval between cxr and ldct (3.1 ± 6.3 hours). however, the inter-observer agreement was moderate and should be improved when moving forward with the ldct technique. this points out that evaluation of ldct images in this neutropenic population at high-risk of developing pulmonary infections requires expert-thoracic radiologists, which may not yet be available in every hospital. a limitation of the study is the use of the reference standard (consensus diagnosis) which may have contributed to the lower test performance of ldct. in 38 of the 67 patients a cxr was the only imaging modality available for the consensus meeting, since hrct was only performed in case of persistent fever. considering the low sensitivity of cxr in this population, pulmonary abnormalities could have been missed, this may have led to an underestimation of the amount of patients with pneumonia, and ldcts might have incorrectly be judged as false positives. furthermore clinical symptoms of cough, sputum production and dyspnea are not always evident in a (bedridden) neutropenic patient. [16] an extensive microbiological evaluation for respiratory causes of fever, which was only performed upon clinical indication, might therefore not have been performed in all patients that had abnormalities suspect for pulmonary infection on ldct. several studies report on the limited diagnostic yield of a respiratory microbiological evaluation in neutropenic patients. in up to 70% of the lower airway infections a pathogen can not be identified. [17, 18] sputum cultures (which are often not available) only reveal a possible pathogen in less than 50% of cases. [18] bal seems to be the diagnostic procedure with the highest yield. despite the low detection rates of a probable pathogen in approximately 50% of cases, it is the most sensitive procedure for detecting ifd. [19] however, since routinely performance of bal in the initial assessment of fever is not performed due to its invasive nature, a consensus diagnosis as was used in our study is the best available option and complies with established guidelines. [1, 4] we were able to identify a probable pathogen in 36% of the patients with pneumonia (2 cases of possible ifd excluded), which is consistent with reports in literature.[17, 18] however we did not establish any case of bacterial pneumonia. this may be explained by the low amount of patients with sputum available for culture (36%, all culture-negative), or a possible treatment response to broad spectrum antibiotics, resulting in defervescence within 3 days, in which case hrct and bal were not performed. we included a heterogenic population with patients with prolonged neutropenia as well as patients with shorter neutropenic episodes. this could have had an effect on the incidence of pulmonary infections, because the risk of developing pneumonia (especially ifd) increases in case of prolonged neutropenia. the use of ldct scans for detection of pulmonary lesions indicating ifd at day 1 of neutropenic fever seems promising. out of 5 fever episodes classified as either possible or probable ifd, 4 patients already had abnormalities suspect for ifd on their ldct whereas these where not seen on cxr. importantly, only one of these patients had respiratory signs or symptoms and therefore we think that omitting chest imaging in patients without respiratory symptoms can lead to a delayed diagnosis of ifd. incorporation of ldct in the diagnostic workup of all patients with neutropenic fever will increase costs when compared to cxr. however, initiation of early and targeted treatment of ifd may reduce overall costs, for example by reducing length of hospital stay, intensive care unit admittance rates and the amount of diagnostic procedures required. therefore the performance of ldct in all patients with febrile neutropenia might still be cost-effective. this issue should be evaluated in future research. performance of cxr in the initial assessment of febrile neutropenia is of limited value for detection of pulmonary abnormalities. the introduction of ldct improved the detection of pulmonary infiltrates and there was a clear signal that ldct scanning is capable of detecting invasive fungal infections at a very early stage. therefore, the use of ldct in the initial assessment of febrile neutropenia is promising, and should be further evaluated in a larger study powered on ifd detection. furthermore it would be interesting to see whether ldct could replace hrct as imaging tool in patients with suspected ifd in order to decrease radiation exposure. management of febrile neutropenia: esmo clinical practice guidelines a prospective survey of febrile events in hematological malignancies the utility of routine chest radiography in the initial evaluation of adult patients with febrile neutropenia patients undergoing hsct clinical practice guideline for the use of antimicrobial agents in neutropenic patients with cancer: 2010 update by the infectious diseases society of america. clinical infectious diseases: an official publication of the infectious diseases society of america comparison of the occurrence of mold infection among patients receiving chemotherapy for acute leukemia versus patients undergoing stem cell transplantation delay of active antimicrobial therapy and mortality among patients with bacteremia: impact of severe neutropenia pneumonia in febrile neutropenic patients and in bone marrow and blood stem-cell transplant recipients: use of highresolution computed tomography imaging findings in acute invasive pulmonary aspergillosis: clinical significance of the halo sign. clinical infectious diseases effective doses in radiology and diagnostic nuclear medicine: a catalog chest low-dose computed tomography in neutropenic acute myeloid leukaemia patients. respiratory medicine ultra-low-dose chest ct in patients with neutropenic fever and hematologic malignancy: image quality and its diagnostic performance. cancer research and treatment: official journal of korean cancer association revised definitions of invasive fungal disease from the european organization for research and treatment of cancer/invasive fungal infections cooperative group and the national institute of allergy and infectious diseases mycoses study group (eortc/msg) consensus group definities ziekenhuisinfecties: lijnsepis -2014 comparison of infectious complications during induction/consolidation chemotherapy versus allogeneic hematopoietic stem cell transplantation. bone marrow transplantation interval estimation for the difference between independent proportions: comparison of eleven methods clinical presentation of infection in granulocytopenic patients the authors would like to thank all of the patients who participated in this study and all the consulting physicians on the ward. key: cord-018027-goxdiyv3 authors: heussel, claus peter title: diagnostic radiology in hematological patients with febrile neutropenia date: 2014-11-27 journal: infections in hematology doi: 10.1007/978-3-662-44000-1_7 sha: doc_id: 18027 cord_uid: goxdiyv3 radiologists have a special role in the management of neutropenic patients. the appropriate investigational technique, specific differential diagnoses, and particular risks of these patients need to be understood by referring physicians as well as by radiologists. thus, communication and cooperation, also including other clinical disciplines such as pulmonology, are required. early detection of an infectious focus is the major goal in febrile neutropenic patients. as pneumonia is the most common focus, chest imaging is a special radiological task. the sensitivity of chest x-ray, especially in supine position, is low. therefore, the very sensitive thin-section multislice ct became a gold standard in neutropenic hosts and might be cost effective in comparison to antibiotic treatment. ct-based localization can be used to guide invasive procedures in order to obtain samples for microbiological workup. furthermore, the radiological characterization of infiltrates gives a first and rapid hint to discriminate between infectious (viral, typical bacterial, atypical bacterial, fungal) and noninfectious etiologies. radiological follow-up has to take into account aspects according to disease, immune recovery, and treatment modalities. due to a high incidence of fungal-related lung infiltrates, interpretation of follow-up findings must include further parameters besides lesion size. apart from the lungs, also other organ systems such as the brain, liver, and paranasal sinuses need attention and are to be imaged with the appropriate technique. the necessity for early detection of the focus of infection is based upon high fatality of infections in immunocompromised hosts, increasing within hours of delayed appropriate treatment [ 48 ] . (this paper does not refer to immunocompromised patients; maybe greene 2007 [ 52 ] or cornely 2010 [ 58 ] could be used), a potential negative impact of delayed diagnosis (i.e., more advanced infection) on future antineoplastic treatment, and high costs of prolonged hospitalization. this has to be compared with the costs of a non-enhanced ct scan, which is around € 230 in german hospitals. after physical examination and interpretation of laboratory fi ndings, the search for an infectious focus starts with the identification of the most suspected organ system(s). the appropriate imaging technique has to be selected demanding for high sensitivity and clinically meaningful negative predictive value [ 4 ] . exact proportions of organ involvement are diffi cult to determine and may differ from clinical and pathological fi ndings, the latter often obtained from autopsies (i.e., negative selection). clinically, lungs are affected in 30 % of febrile neutropenic patients and allogeneic hematopoietic stem cell transplant (asct) recipients, paranasal sinuses in 3 % of neutropenic patients, and 30 % in the asct setting (concomitant to pneumonia), while the gastrointestinal tract, liver, spleen, central nervous system, and kidneys are less frequently involved [ 4 ] . chest x-ray (cxr) is still frequently used when pneumonia is suspected or should be ruled out [ 14 , 15 ] . cxr has several advantages: it is quick, widely available (even on the ward), inexpensive, and associated with a low radiation dose. cxr is occasionally done on the ward to keep neutropenic patients in protective isolation, even if performed in supine position. but cxr has the crucial disadvantage of superimposition and therefore a very limited sensitivity for the detection of pneumonia (figs. 7.2 and 7.6 ) [ 14 , 16 ] . especially if performed in supine position, lung infl ation is worse and lateral projection is lacking, which limits image quality besides other technical issues. in patients with fever of unknown origin (fuo) after sct, digital cxr in supine position achieves a sensitivity for early detection of pneumonia of only 46 % [ 17 ] . while cxr provides relevant clinical information concerning central venous catheters (cvc), pleural effusion, and pulmonary congestion [ 17 ] , it fails to enable early detection or exclusion of pneumonia, which is a major task in immunocompromised hosts. cxr in supine position alone is therefore not recommended for the early detection of pneumonia in these patients [ 5 ] . also, if an infi ltrate is apparent at cxr, the options for its characterization are very limited. if pneumonia is considered in these hosts, thin-section ct should be preferred at any time [ 18 ] . the radiation dose is of limited concern in patients who eventually underwent local or total body irradiation and received cytotoxic agents, etc., considering that actual ct techniques apply 1-10 msv per lung scan (in this diagnostic scenario: msv = mgy) [ 19 , 20 ] . the risk of developing a radiation-induced neoplasm, even after several diagnostic exposures, is low when compared to the high mortality associated to infection and disease as well as the risk of a secondary malignancy due to antineoplastic treatment. terms like incremental ct , highresolution ct (hrct), spiral ct , thinsection ct , multislice ct (msct), and lowdose ct are widely used and might confuse non-radiologists. to keep it simple, hrct is an incremental scanning technique with several respiratory breath-holds resulting in inaccuracy of repositioning of the anatomical lung position. the use of 1 mm sections and gaps in between (e.g., 10 mm) results in representative, detailed images of selected lung areas; however, the noncontiguous scanning has its limitations in nodule detection, quantifi cation, and monitoring. volumetric techniques as used in spiral ct and msct, acquired without gaps, are frequently reconstructed with larger thickness (e.g., 5 mm) resulting in spatial volume effects. this results in limitations to detect infl ammatory lung disease, especially ground-glass opacifi cation [ 22 ] . since no additional information is expected from supplemental spiral ct to hrct, as shown in aids patients [ 23 ] , hrct may be used as a diagnostic standard. in contrast, thin-section msct provides volumetric scanning as well as detailed images [ 24 , 25 ] . this technique also allows for an adequate monitoring of lung disease since the same anatomical position can be reidentifi ed in baseline and follow-up studies [ 24 -26 ] . while a rapid technical development in ct imaging is ongoing, the different techniques applied today are addressed as "ct" in this chapter. in general, contrast enhancement is not required for detecting and characterizing pneumonia [ 6 , 18 ] . only in special situations such as suspicion of pulmonary embolism or hemoptysis caused by vessel erosion is ct angiography benefi cial ( fig. 7 .1 ) [ 27 ] . in the asct setting, bronchiolitis obliterans is to be considered [ 24 , 28 ] where air-trapping is a relevant fi nding. here, an additional expiratory ct scan is helpful [ 24 , 28 ] . the advantage of hrct in comparison to cxr for the early detection of pneumonia was demonstrated in febrile neutropenic patients not responding to empirical antibiotic therapy [ 21 ] . in approximately 60 % of the patients with a normal cxr, hrct showed pulmonary infi ltrates ( fig. 7.2 ). in only 10 % of patients with a normal chest x-ray and a normal hrct, pneumonia occurred during follow-up [ 21 ] . exclusion of pneumonia is another clinically relevant information. thus, ct yields very useful results with good sensitivity (87 %) and negative predictive value (88 %). the early use of hrct achieves a gain of approximately 5 days during which pneumonia may be excluded [ 17 ] . in clinical practice, this may be very helpful for the management of immunocompromised hosts at high risk of life-threatening pulmonary infection [ 5 ] ( fig. 7. 2 ). mri has been evaluated for the investigation of pulmonary disease since it has a known benefi t in lesion characterization [ 29 , 30 ] . comparing ct to mri on an intraindividual basis, mri reveals comparable clinical results (sensitivity 95 %, specifi city 88 %, positive predictive value 95 %, negative predictive value 88 %) [ 53 ] . besides the lack of radiation, there is no clear advantage of mri in the early detection of pneumonia ( fig. 7.3 ). in advanced stages, ct and mri are comparable in the visualization of infi ltrates [ 30 ] . ct is widely available, easier, and faster to perform as well as less susceptible to breathing artifacts. mri is superior to ct in the detection of abscesses due to a clearer detection of central necrosis in t2-weighted images and rim enhancement after contrast application in t1-weighted images [ 29 ] . however, this fact has limited clinical impact and duration of mri, and required compliance is substantially higher compared to ct. mr has problems to detect small lesions and those which are adjacent to the left ventricle due to the cardiac motion [ 53 ] . in contrast to systemic infections, identifi cation of the underlying organism in pneumonia is more diffi cult and complex. attempts to reinforce pathogen identifi cation did not improve the clinical outcome signifi cantly [ 9 ] . therefore, a calculated (preemptive) decision on antimicrobial therapy in febrile immunosuppressed patients based on imaging studies is widely used. the use of ct is recommended for the early detection of pneumonia [ 9 ] . it may serve for indication and localization of invasive diagnostic procedures such as neutropenic febrile patient receiving broad-spectrum antibiotic therapy. cxr was normal at day 3 of fever ( a , b ). hrct performed the same day demonstrates bilateral infi ltrates, which were hidden behind the heart in posterior-anterior and the spine in lateral projection ( c ) bronchoscopy and bronchoalveolar lavage or ct-guided biopsy. on the other hand, the exclusion of pneumonia can be obtained with a higher reliability as compared to conventional cxr. the sequential cascade as shown in fig. 7 .4 can be modifi ed if the local institutional ct capacity allows for the skipping of cxr. an increasing size of pulmonary infi ltrates during hematopoietic recovery has been well described by [ 31 ] . caillot et al. evaluated hrct in neutropenic patients with proven pulmonary aspergillosis at weekly intervals [ 31 ] and documented the time points of different radiological patterns and evaluated the size of infi ltrates and documented the time points of different radiological patterns and evaluated the size of infi ltrates. they frequently found a "halo sign" (fig. 7 .5 ) on the fi rst ct scans and reported a low sensitivity of this pattern (68 %), which was no longer visible on follow-up scans. in contrast, the more specifi c "air-crescent sign" ( frequently during follow-up (up to 63 %). the size of infi ltrates increased by fourfold under successful antifungal treatment due to hematopoietic reconstitution. in this study, pneumonia was fi rst detected on day 19 of neutropenia. enlargement of infi ltrates is probably caused by the invasion of newly generated neutrophil granulocytes at the beginning of bone marrow recovery. in critically ill patients, leukocyte invasion has been described as a risk factor for the development of acute respiratory distress syndrome (ards) [ 14 ] ( fig. 7 .6 and 7.7 ). radiologists' dream is to be capable to identify the underlying microorganism in pneumonia of immunocompromised hosts with a suffi cient specifi city. in some cases, imaging can provide very fast useful hints, but no verifi cation. the quality of these clues [ 13 ] . since the initial cxr is of limited use, this diagnostic step is more and more omitted and ct is performed primarily depends on the cooperation between clinicians and radiologists and on the radiologists' experience with these complications. this requires an informational exchange concerning relevant individual patient data like severe neutropenia or allogeneic stem cell transplantation. for example, information on reactivation of cytomegalovirus (cmv) in a patient with graft-versus-host disease is very helpful for the correct interpretation of pulmonary hrct fi ndings. also the chemotherapy or total body irradiation applied for conditioning before asct may be relevant for differential diagnostic considerations in patients who might present with clinically similar signs and symptoms [ 6 , 7 , 32 , 33 ] . some of the most useful clues are listed in table 7 .1 . bacteria are causing approximately 90 % of infections during the early phase of neutropenia [ 8 ] . the radiological appearance of bacterial pneumonia includes consolidation, especially bronchopneumonia and positive pneumobronchogram ( fig. 7. 2 ) [ 33 , 34 ] . in contrast to immunocompetent patients, ground-glass opacifi cation is found more often and remains nonspecifi c. severe neutropenia lasting for more than 10-14 days is associated with an increasing risk of invasive fungal infection [ 3 ] , with aspergillus species being the primary pathogen, while candida species very rarely cause primary pneumonia ( fig. 7.8 ) [ 5 ] . typical radiological fi ndings of fungal and non-fungal pneumonia as well as of infi ltrates from noninfectious diseases have been reviewed in detail [ 35 ] . the typical appearance of pulmonary infi ltrates from fungal origin are as follows: early phase of fungal pneumonia: ill-defi ned nodules (figs. 7.3 , 7.5 and 7.8 ) [ 33 ] in combination with the halo sign (figs. 7.5 and 7.8 ) [ 33 ] , which is nonspecifi c late phase: air-crescent sign [ 36 ] cavitation ( fig. 7.8 ) a c d b fig. 7.6 the small ill-defi ned nodule in the right upper lobe ( c ) of this 34-year-old neutropenic aml patient was even retrospectively not visible on conventional chest x-ray done at the same day ( a , b ). amphotericin b treatment was started due to suspicion of fungal pneumonia; however, the nodule size increased during hematopoietic reconstitution 2 weeks later ( d ). in preparation of bone marrow transplantation, the lesion was surgically resected and aspergillus pneumonia was verifi ed for use in the context of clinical and epidemiological research in neutropenic patients, standards for the interpretation of radiological fi ndings in invasive fungal infections have been elaborated [ 10 , 51 ] ; newly emerged "typical" ct patterns (dense, well-circumscribed lesions with or without a halo sign, air-crescent sign) are classifi ed as a clinical criterion for fungal pneumonia figs. 7.5 and 7.8 . the halo sign, fi rst described in 1984 [ 49 , 50 ] , is nonspecifi c [ 33 ] and not a necessary part of the updated defi nitions [ 51 ] . a nonspecifi c infi ltrate, rated as a minor criterion in the fi rst version [ 10 ] , was abandoned in the update [ 51 ] . in a later workup of a pharmaceutical trial investigating response rates to antifungal treatment, the evidence of the halo sign was associated with an improved response rate (52 % vs. 29 %; p < 0.001), as well as a higher 3-month survival rate (71 % vs. 53 %; p < 0.01) [ 52 ] . this large ( n = 235) antemortem trial suffers, however, from systemic limitations like investigation of halo which was part of inclusion criteria and technical insuffi ciencies like usage of thick-section ct instead of appropriate thin-section ct and evaluation of hardcopies instead of monitor reading [ 52 ] . histopathological workup of lung biopsies verifi ed fungal pneumonia in 56 % of cases in another study [ 37 ] . relevant differential diagnoses for the halo sign, such as cryptogenic organizing pneumonia (cop, formerly known as bronchiolitis obliterans organizing pneumonia, boop), pulmonary hemorrhage, pulmonary manifestation of the underlying malignancy, lung cancer, and nonfungal infections (cmv, tuberculosis, abscesses (fig. 7.7 ) , etc.) or candida (fig. 7.8 ) , have to be considered [ 37 ] . thus, diagnostic clarifi cation will frequently be necessary, particularly when antifungal treatment is not successful. non-fungal lung infi ltrate: ill-defi ned nodules with cavitation on ct scans done due to repeated febrile episodes. they appeared like fungal pneumonia and were not visible on baseline ct. after removal of a central venous port system, both the pulmonary lesions as well as the febrile episodes, disappeared air-crescent sign and cavitation occur with hematopoietic reconstitution during the late phase of fungal infection (fig. 7.8 ) [ 36 ] . both radiological signs are known to be associated with a favorable prognosis. however, the specifi cities of these fi ndings are limited, and relevant differential diagnoses have to be considered (fig. 7.7 ) [ 37 ] . there are other useful patterns for the identifi cation of fungal pneumonia, e.g., ggo ground-glass opacifi cation, tbi total body irradiation distribution along the bronchovascular bundle resulting in the feeding vessel sign with an angiotrophic location. the ongoing development of antifungal therapy may have an important impact on the radiological appearance of fungal pneumonia. thus, in the near future radiologists will not only be confronted with the question for "breakthrough fungal pneumonia," but also for fungal pneumonia caused by non-aspergillus pathogens. pneumocystis jiroveci pneumonia (pcp) [ 38 ] is a typical fi nding in hematological patients affected by severe cellular (t-cell) immunosuppression and those with graft-versus-host disease after asct, if they are not protected by effective chemoprophylaxis [ 8 ] . despite standard trimethoprim-sulfamethoxazole prophylaxis, 8 % of the patients develop pcp, while among patients without prophylaxis, the incidence may reach 29 % [ 8 ] . up to 15 % of these patients will have a fatal outcome [ 8 ] . ct provides a valuable characterization for pcp [ 6 , 7 , 18 , 33 ] and is a reliable method for discriminating it from other infectious processes [ 33 , 39 ] of ground-glass opacities and intralobular septae sparing out the subpleural space (i.e., perihilar distribution) is very typical for pcp (fig. 7.9 ) tuberculosis (tb) has to be considered as a rare but relevant differential diagnosis. in immunocompromised hosts, tb appears different compared to immunocompetent hosts (e.g., gangliopulmonary (primary) forms) [ 41 ] . more widespread lymphogenic and hematogenous dissemination can occur, and therefore, the clinical course might be fulminant [ 41 , 42 ] . on the other hand, tb might mimic or come along with other infections like pulmonary aspergillosis or systemic candidiasis [ 42 ] . in immunocompromised hosts a segmental bronchial spread (resulting in a "treein-bud" sign) of small, sometimes, cavitated ill-defi ned nodules can be obtained as well as a miliar distribution [ 41 , 42 ] . gangliopulmonary (primary) forms, however, present with nonhomogenous consolidation and necrotizing mediastinal or hilar lymphadenopathy [ 41 ] . interstitial pneumonia caused by viral infection may occur primarily in asct recipients but also in neutropenic and t-cell-immunosuppressed patients. mortality rate may be up to 50 %. most frequently, cytomegalovirus (cmv) is suspected; however, other herpesviruses, infl uenza, parainfl uenza, adenovirus, or respiratory syncytial (rsv) viruses have to be considered as well. there are no specifi c radiological patterns available to differentiate various forms of viral pneumonia. however, confi rming the suspicion of a viral pneumonia may be a clinically useful information, since effective drugs are available for some of these viruses. the typical appearance of viral pneumonia in the early stage is ground-glass opacifi cation [ 33 ] and a mosaic pattern with affected and non-affected secondary lobules lying adjacent to one another (fig. 7.10 ). certain noninfectious diseases have to be considered in hematological patients: graft-versus-host disease (gvhd), radiation or drug toxicity, pulmonary congestion, bleeding, or progressive underlying malignancy. fever, dyspnea, or clinical chemistry fi ndings (c-reactive protein, elevation of liver function tests) might be a b fig. 7. 10 bilateral ground-glass opacifi cation and mosaic pattern in both patients. however, pneumonia in patient a is caused by cytomegalovirus ( cmv ), and patient b by respiratory syncytial virus ( rsv ). note the mosaic pattern resulting from affected and non-affected secondary lobules lying adjacent to one another caused by some of these processes and obscure the differentiation from infection. ct may help to detect and discriminate these diseases [ 6 , 7 , 18 , 32 ] . pulmonary manifestation of chronic gvhd occurs in approximately 10 % of patients after allogeneic hematopoietic stem cell transplantation ( fig. 7.11 ) [ 43 ] . bronchiolitis obliterans is the pulmonary manifestation of this rejection [ 28 ] . the radiological appearance is similar to viral pneumonia, and clinical appearance and time point for both diseases are often similar (fig. 7.11 ). ground-glass opacifi cation and mosaic pattern, as well as signs of bronchiolitis obliterans such as air-trapping [ 24 , 28 ] and bronchus wall thickening, occur during the early stage of pulmonary gvhd (fig. 7.11 ), whereas intralobular septae and tree-in-bud sign follow in later stages [ 7 , 43 , 44 ] . an incidence of 5-25 % pulmonary radiogenic toxicity even after total body irradiation (tbi) applied for conditioning prior to stem cell transplantation is reported fig. 7 .11 a 28-year-old male after allogeneic retransplantation. hrct was performed due to fever, cough, and dyspnea. peripheral intralobular septae ( arrow ) and ground-glass opacifi cation were seen on day 91 posttransplant. tree-in-bud pattern ( arrowhead ) points at bronchiolitis obliterans. acute gvhd was diagnosed from transbronchial biopsy. under appropriate immunosuppression, clinical symptoms and radiological signs disappeared. note the similarity to fig. 7.13 fig. 7 .12 three weeks after radiation therapy for malignant spine destruction, this patient suffered from fever and dyspnea. perihilar infi ltrates appeared suddenly. intralobular septae, consolidation, and ground-glass opacifi cation were visible on hrct. especially the paramediastinal distribution of the infi ltrates led to the differential diagnosis of radiation pneumonitis. after failure of antibiotic treatment, steroids were applied, resulting in improvement of symptoms and reduction of infi ltrates [ 45 ] . it emerges approximately 3 weeks after exposure but can also occur several months later [ 44 , 45 ] . on ct scans, radiation-induced toxicity is characterized by ground-glass opacities with transition to consolidations (fig. 7 .12 ) [ 44 , 45 ] . the key fi nding is the limitation of these patterns to the exposed parenchyma. for tbi, lungs are shielded, while paramediastinal and apical lung parenchyma is affected from radiation. of note, demarcation of initially exposed from nonexposed lung parenchyma is blurred frequently due to deformation of lung parenchyma and to bridle. chemotherapy protocols may lead to pulmonary toxicity. some of the frequently used agents are bleomycin, high-dose methotrexate (mtx) or cytarabine (ara-c), or carmustine (bcnu) (fig. 7.13 ) [ 46 ] . radiologists should be informed of previous exposure of patients to these agents when evaluating ct scans for pulmonary abnormalities. the term "drug-induced pneumonitis" includes mainly nonspecifi c interstitial pneumonia (nsip) and cryptogenic organizing pneumonia (cop, formerly known as bronchiolitis obliterans organizing pneumonia, boop) [ 46 ] . the ct appearance of nsip consists of ground-glass opacities with transition to consolidations, intralobular septae, traction bronchiectasis, air-trapping, and in a later phase the nonspecifi c "crazy-paving" pattern [ 44 , 46 ] . this is quite similar to radiation toxicity but without being limited to the radiation fi eld. dyspnea and infi ltration are frequent in patients suffering from pulmonary congestion. extensive hydration for renal protection during chemotherapy or to overcome renal impairment may cause pulmonary congestion also in younger patients. it is one of the most frequent disorders in patients undergoing intensive care. at cxr, pulmonary congestion might be combined with infi ltration. ct shows thickening of lymphatic vessels, corresponding to classical "kerley lines" on conventional chest radiographs (fig. 7 .14 ). leukemic pulmonary infi ltration is a less common clinical fi nding. especially the perilymphatic pulmonary interstitium is involved [ 47 ] . this can be visualized on ct scans as thickening of bronchovascular bundles and interlobular septae. besides fig. 7.13 a 40 -year-old male who received chemotherapy including bleomycin for testicular cancer and had fever, cough, and dyspnea. ct revealed peripheral intralobular septae and ground-glass opacifi cation ( a ). bleomycin toxicity was suspected and histologically confi rmed. symptoms disappeared and fi ndings decreased after application of steroids. note the similarity to fig. 7.11 this, non-lobular and non-segmental ground-glass opacifi cations can be seen [ 32 ] . this pattern arrangement might mimic pulmonary congestion (fig. 7 .14 ). in pancytopenia, pulmonary bleeding may occur spontaneously, secondary to invasive infections, after interventions (e.g., bronchoscopy and bal), or during marrow recovery particularly in patients with fungal pneumonia [ 27 ] . pulmonary bleeding might cause a focal or diffuse pattern, and the phenomenon of sedimentation within the secondary lobules can sometimes be depicted for few days (fig. 7.15 ). suspicious clinical symptoms or unexplained laboratory fi ndings may suggest an involvement of the liver and spleen [ 13 ] , particularly secondary to fungemia [ 54 ] . in addition to candidiasis, also mycobacteriosis, bacterial granulomatous hepatitis, viral hepatitis, and noninfectious organ involvement such as drug-related hepatotoxicity, gvhd, veno-occlusive disease (vod), or relapse of the underlying disease have to be considered [ 12 ] . due to its microbial fl ora and the chemotherapy-induced mucosal injury, the gastrointestinal tract is particularly exposed to infection. however, without any history of surgical intervention, gastrointestinal involvement is rare. the main affections of the gastrointestinal tract, such as cmv colitis, pseudomembranous enterocolitis, enterocolitis in the context of rejection (gvhd), appendicitis, and diverticulitis, can be seen in ct as bowel wall thickening even without intravenous contrast after adequate oral, rectal contrast application [ 55 ] . cerebral infection is a rare complication of myelosuppressive chemotherapy. it is more likely in the asct setting than after conventional chemotherapy [ 2 ] . besides infectious diseases (e.g., herpesvirus group, toxoplasmosis, aspergillosis, mucormycosis, listeriosis), diagnoses such as bleeding, ischemia, drug toxicity (cyclosporine, ribavirin, voriconazole, etc.), and electrolyte disorders have to be taken into consideration. ct is helpful in emergency situations, while mri is the method of choice in brain imaging in terms of sensitivity and specifi city for detection and characterization of brain abnormalities [ 12 ] . the bilateral ground-glass opacifi cation has an anterior-to-posterior gradient (1) over the whole lung and (2) within certain secondary lobules. this gravity-dependent sedimentation phenomenon can also occur temporarily and may be localized, e.g., after bronchoscopy and bal since the sinuses are part of the respiratory tract, there is a coincidence of pneumonia [ 56 ] . since the risk for sinusitis is up to 30 % in allogeneic stem cell transplant recipients, paranasal sinuses are often screened by ct prior to transplantation [ 57 ] . bone erosion and orbital or brain invasion are classifi ed by the 2008 eortc guideline as clinical criteria for probable invasive fungal disease [ 51 ] . evolving risk factors for infectious complications of cancer prevention, diagnosis and therapy of infections in patients with malignant diseases empiric antibiotic and antifungal therapy for cancer patients with prolonged fever and granulocytopenia pulmonary infi ltrations in febrile neutropenic patients. risk factors and outcome under empirical antimicrobial therapy in a randomized multicenter trial pneumonia in febrile neutropenic patients: radiologic diagnosis high-resolution ct of diffuse interstitial lung disease: key fi ndings in common disorders hrct fi ndings of chest complications in patients with leukemia epidemiologie und interventionelle therapiestrategien infektiöser komplikationen nach allogener stammzelltransplantation diagnostik und therapie von lungeninfi ltraten bei febrilen neutropenischen patienten -leitlinien der arbeitsgemeinschaft infektionen in der hämatologie und onkologie der deutschen gesellschaft für hämatologie und onkologie defi ning opportunistic invasive fungal infections in immunocompromised patients with cancer and hematopoietic stem cell transplants: an international consensus multiple pulmonary nodules in aids: usefulness of ct in distinguishing among potential causes magnetic resonance imaging (mri) of liver and brain in hematologic-oncologic patients with fever of unknown origin looking for the cause in neutropenic fever. imaging diagnostics deterioration of previous acute lung injury during neutropenia recovery conventional chest radiography in the initial assessment of adult cancer patients with fever and neutropenia high-resolution ultrafast chest ct in the clinical management of febrile bone marrow transplant patients with normal or nonspecifi c chest roentgenograms importance of digital thoracic radiography in the diagnosis of pulmonary infi ltrates in patients with bone marrow transplantation during aplasia thoracic disease in the immunocompromised patient patient doses from chest radiography in victoria multi-slice computed tomography as a screening tool for colon cancer, lung cancer and coronary artery disease pneumonia in febrile neutropenic patients, bone-marrow and blood stem-cell recipients: use of high-resolution ct computed tomography assessment of ground-glass opacity: semiology and signifi cance pulmonary manifestations in hiv patients: the role of chest fi lms, ct and hrct new frontiers in ct imaging of airway disease new technical developments in multislice ct -part 1: approaching isotropic resolution with sub-millimeter 16-slice scanning thorakale computertomographie von lungeninfi ltraten aneurysms complicating infl ammatory diseases in immunocompromised hosts: value of contrast-enhanced ct noninfectious lung disease in immunocompromised patients mr imaging of pneumonia in immunocompromised patients: comparison with helical ct mri of the lung parenchyma increasing volume and changing characteristics of invasive pulmonary aspergillosis on sequential thoracic computed tomography scans in patients with neutropenia ct fi ndings of leukemic pulmonary infi ltration with pathologic correlation pneumonia: high-resolution ct fi ndings in 114 patients bacterial pneumonia in immunocompromised patients fungal pneumonia crescent sign in invasive pulmonary aspergillosis: frequency and related ct and clinical factors importance of open lung biopsy in the diagnosis of invasive pulmonary aspergillosis in patients with hematologic malignancies a new name (pneumocystis jiroveci) for pneumocystis from humans accuracy of high-resolution ct in distinguishing between pneumocystis carinii pneumonia and non-pneumocystis carinii pneumonia in aids patients viral and pneumocystis carinii infections of the lung in the immunocompromised host imaging of pulmonary tuberculosis ct of tuberculosis and nontuberculous mycobacterial infections late cd8+ lymphocytic alveolitis after allogeneic bone marrow transplantation and chronic graft-versus-host disease organizing pneumonia: the many morphological faces clinical radiation pneumonitis and radiographic changes after thoracic radiation therapy for lung carcinoma high-resolution ct of drug-induced lung disease pulmonary leukemic infi ltrates: high-resolution ct fi ndings in 10 patients delaying the empiric treatment of candida bloodstream infection until positive blood culture results are obtained: a potential risk factor for hospital mortality new observations on the signifi cance of the air crescent in invasive pulmonary aspergillosis invasive pulmonary aspergillosis in acute leukemia: characteristic fi ndings on ct, the ct halo sign, and the role of ct in early diagnosis european organization for research and treatment of cancer/ invasive fungal infections cooperative group; national institute of allergy and infectious diseases mycoses study group (eortc/msg) consensus group. revised defi nitions of invasive fungal disease from the european organization for research and treatment of cancer/ invasive fungal infections cooperative group and the national institute of allergy and infectious diseases mycoses study group imaging fi ndings in acute invasive pulmonary aspergillosis: clinical signifi cance of the halo sign pulmonary abnormalities in immunocompromised patients: comparative detection with parallel acquisition mr imaging and thin-section helical ct hepatosplenic yeast infection in patients with acute leukemia: a diagnostic problem acute graft-versushost disease in children: abdominal ct fi ndings paranasal sinusitis following allogeneic bone marrow transplant klinische relevanz der nnh-ct vor knochenmarktransplantation treatment outcome of invasive mould disease after sequential exposure to azoles and liposomal amphotericin b key: cord-312251-t6omrr07 authors: vancheri, sergio giuseppe; savietto, giovanni; ballati, francesco; maggi, alessia; canino, costanza; bortolotto, chandra; valentini, adele; dore, roberto; stella, giulia maria; corsico, angelo guido; iotti, giorgio antonio; mojoli, francesco; perlini, stefano; bruno, raffaele; preda, lorenzo title: radiographic findings in 240 patients with covid-19 pneumonia: time-dependence after the onset of symptoms date: 2020-05-30 journal: eur radiol doi: 10.1007/s00330-020-06967-7 sha: doc_id: 312251 cord_uid: t6omrr07 objective: to analyze the most frequent radiographic features of covid-19 pneumonia and assess the effectiveness of chest x-ray (cxr) in detecting pulmonary alterations. materials and methods: cxr of 240 symptomatic patients (70% male, mean age 65 ± 16 years), with sars-cov-2 infection confirmed by rt-pcr, was retrospectively evaluated. patients were clustered in four groups based on the number of days between symptom onset and cxr: group a (0–2 days), 49 patients; group b (3–5), 75 patients; group c (6–9), 85 patients; and group d (> 9), 31 patients. alteration’s type (reticular/ground-glass opacity (ggo)/consolidation) and distribution (bilateral/unilateral, upper/middle/lower fields, peripheral/central) were noted. statistical significance was tested using chi-square test. results: among 240 patients who underwent cxr, 180 (75%) showed alterations (group a, 63.3%; group b, 72%; group c, 81.2%; group d, 83.9%). ggo was observed in 124/180 patients (68.8%), reticular alteration in 113/180 (62.7%), and consolidation in 71/180 (39.4%). consolidation was significantly less frequent (p < 0.01). distribution among groups was as follows: reticular alteration (group a, 70.9%; group b, 72.2%; group c, 57.9%; group d, 46.1%), ggo (group a, 67.7%; group b, 62.9%; group c, 71%; group d, 76.9%), and consolidation (group a, 35.5%; group b, 31.4%; group c, 47.8%; group d, 38.5%). alterations were bilateral in 73.3%. upper, middle, and lower fields were involved in 36.7%, 79.4%, and 87.8%, respectively. lesions were peripheral in 49.4%, central in 11.1%, or both in 39.4%. upper fields and central zones were significantly less involved (p < 0.01). conclusions: the most frequent lesions in covid-19 patients were ggo (intermediate/late phase) and reticular alteration (early phase) while consolidation gradually increased over time. the most frequent distribution was bilateral, peripheral, and with middle/lower predominance. overall rate of negative cxr was 25%, which progressively decreased over time. key points: • the predominant lung changes were ggo and reticular alteration, while consolidation was less frequent. • the typical distribution pattern was bilateral, peripheral, or both peripheral and central and involved predominantly the lower and middle fields. • chest radiography showed lung abnormalities in 75% of patients with confirmed sars-cov-2 infection, range varied from 63.3 to 83.9%, respectively, at 0–2 days and > 9 days from the onset of symptoms. since december 2019, when the first cluster of cases of coronavirus disease 2019 (covid19) was reported in wuhan, hubei province, china, the widespread transmission of coronavirus type 2 (sars-cov-2) has reached pandemic proportions [1] . the manifestations of sars-cov-2 infection in humans range from mild respiratory symptoms to severe acute respiratory syndrome [2, 3] . according to the who, several molecular assays based on reverse transcription polymerase chain reaction (rt-pcr) for the individuation of sars-cov-2 genes are recommended for the confirmation of covid-19 [4] . the role of imaging in the diagnosis of covid-19 is still under debate. several early radiological studies emphasized the role of radiological imaging in the early detection and management of covid-19. these studies analyzed and described the chest computed tomography (ct) findings at the presentation and at different times throughout the disease course [5, 6] . during the subsequent outbreak in the western world, chest x-ray (cxr), together with arterial blood gas analysis and clinical presentation, in patients positive to rt-pcr, was recommended as a useful and easily available tool to support the initial diagnosis and for the subsequent management of covid-19 patients [7, 8] . nonetheless, data specifically addressing cxr findings in covid-19 are still limited [9] [10] [11] . also, the role of lung ultrasound is probably valuable but still unclear [12] . in order to create reasonable imaging workflows, it is important to evaluate the diagnostic potential of cxr and that radiologists become familiar with the presentation pattern of covid-19 in cxr. our study aimed to evaluate the percentage of abnormal chest radiographs at different time intervals from the onset of symptoms and to identify the type and distribution of radiographic alterations and their frequency at different times throughout the disease course of covid-19 pneumonia. the inclusion criteria of this retrospective study were as follows: (1) patients with confirmed sars-cov-2 infection, in which the sars-cov-2 infection was confirmed by real-time rt-pcr on nasopharyngeal swab or bronchoalveolar lavage (bal) specimens, according to international guidelines [4] ; (2) knowledge of the precise date of onset of symptoms considered linked to the onset of viral pneumonia such as fever (> 37.5°c), cough, and dyspnea [2] and asymptomatic patients were excluded from the study; and (3) patients who underwent cxr examination. all patients underwent cxr and rt-pcr on the day of admission in the emergency department. two investigators collected the clinical data regarding the time of onset of symptoms from the digital archive of the emergency department of all the patients suspected for covid-19 admitted at fondazione irccs policlinico san matteo between february 21 and march 8, 2020. the same two investigators subsequently noted the results and the date of rt-pcr of each patient. irb approval was obtained for this study. informed consent for processing personal data for research purposes was obtained from each patient. all the patients underwent a baseline digital anteroposterior bedside chest radiography at full inspiration using a portable radiography unit (fdr go, fujifilm corporation). x-ray examinations were reviewed independently by two experienced thoracic radiologists (r.d. and l.p., with more than 30 and 15 years of experience, respectively). results were compared, and when disagreement was found, final decisions were determined by consensus. the findings considered in the evaluation of cxr were the presence or absence and the type of pulmonary alterations, and their distribution. cxr alterations (fig. 1) were defined according to the fleischner society's nomenclature, available in the glossary of terms for thoracic imaging [13] : & reticular alteration, as a collection of innumerable small linear opacities that, by summation, produce an appearance of a net; & consolidation, as a homogeneous increase in pulmonary parenchymal attenuation that obscures the margins of the vessels and airway walls; & ground-glass opacity (ggo), as an area of hazy, increased lung opacity, usually extensive, within which margins of pulmonary vessels may be indistinct. the distribution of alterations was classified as unilateral or bilateral. within each hemithorax, the craniocaudal distribution of the lesions was evaluated on the basis of the involvement of the upper, middle, and lower fields. the middle field was defined as the lung area delimited by (included between) two horizontal lines at the level of the superior and inferior hilar horns, respectively; the superior field was defined as the lung area included between the horizontal line at the level of the upper hilar horn and the apical pleura; the lower field was defined as the lung area included between the horizontal line at the level of the inferior hilar horn and the diaphragm. the horizontal distribution of the lesions was evaluated on the basis of the involvement of the peripheral zone only, the central zone only, or both. central zones were defined as the central area within 2 cm from the lobar bronchial structures as far as visible; peripheral zones were defined as the remaining lung area between the central zones and the pleura [14] . the patients were clustered into four groups based on the number of days between the onset of symptoms and the chest radiography: group a (patients with chest radiographs acquired 0-2 days from the onset of symptoms), group b (patients with chest radiographs acquired 3-5 days from the onset of symptoms), group c (patients with chest radiographs acquired 6-9 days from the onset of symptoms), and group d (patients with chest radiographs acquired over 9 days from the onset of symptoms). statistical analyses were performed using medcalc for windows, version 15.0 (medcalc software). continuous fig. 1 the three main alterations on chest radiography (upper line) and the corresponding findings on chest ct (lower line). left: diffuse reticular alteration (arrows). the corresponding ct shows diffuse increased lung attenuation and interlobular septal thickening (arrows). middle: peripheral ground-glass opacities (arrows). right: extensive consolidations (arrows). the corresponding ct shows predominant consolidative alterations (arrows) variables were expressed as mean ± sd values. the frequency of the radiographic findings was expressed as the number of occurrences and percentage in every single cluster. frequencies in the different groups were compared using the chi-square test; p values < 0.05 were considered significant. among 305 patients initially considered, 60 patients were excluded because the onset of symptoms was uncertain and 5 patients were excluded because they were asymptomatic. a total of 240 patients was considered; among them, 169 (70%) were men and 71 (30%) were women. the average age was 65 ± 16 years (range 16-97 years). numerosity and demographics of the subgroups are summarized in table 1 . among the total of 240 patients, 180 (75%) had at least one alteration in one lung field. in 60/240 patients (25%), cxr was normal without any lesion. cxr showed at least one alteration in at least one lung field in 31/49 patients (63%) in group a, in 54/74 patients (72%) in group b, in 69/85 patients (81%) in group c, and in 26/31 patients (83%) in group d. the negative rate of chest radiographs was 18/49 patients (36.7%) in group a, 21/75 (28%) in group b, 16/ 85 (18.8%) in group c, and 5/31 (16.1%) in group d. alterations were bilateral in 132/180 patients (73.3%) and unilateral in 48/180 (26.6%) (24 in the left lung and 24 in the right lung). ggo, alone or in combination with other alterations, was present in 124/180 patients (68.8%); reticular alteration, alone or in combination with other findings, was present in 113/180 patients (62.7%); consolidation, alone or in combination with other findings, was present in 71/180 patients (39.4%). ggo and reticular alteration were significantly more frequent than consolidation (p < 0.01 in both cases) table 2 . a significantly higher frequency of involvement of the lower fields compared to the middle fields and of the lower and middle fields compared to the upper fields was observed (p < 0.01 in all cases) table 3 . the exclusive involvement of the central zones was significantly less frequent than the exclusive involvement of the peripheral zones and than the involvement of both peripheral and central zones (p < 0.01 in both cases) table 3 . pleural effusion was observed in 12/180 patients (6.6%), unilateral in all cases (7 on the right side and 5 on the left side). (fig. 4 ). in early clinical experience, chest ct has been considered the most reliable imaging modality to support the diagnosis in suspected cases of covid-19 [15] . chest ct showed high sensitivity in detecting ggo, which is considered a typical finding in covid-19 pneumonia and, in some cases, may be the only alteration present in the early phases of the disease [3, 16] . cxr was not recommended as a first-line imaging examination due to its low sensitivity in detecting alterations [5, 15] . conversely, recent statements of several influential radiological societies [7, 8, [17] [18] [19] [20] [21] [22] recommend that ct should not be used as a first-line tool to support the diagnosis of covid-19, and encourage the use of cxr in combination with rt-pcr test. our study is one of the first to address this issue and systematically describe the cxr findings at the admission in covid-19 pneumonia in a vast population [9, 10] . salehi et al [11] analyzed several studies on initial and follow-up imaging in covid-19, performed a meta-analysis on 919 patients who underwent chest ct, and reported general considerations on cxr. wong et al [9] reviewed 255 cxr in 64 patients, describing the time course of the radiographic findings of covid-19 pneumonia and reporting an overall rate of negative baseline cxr examinations of 31%. similarly, in our experience, the overall rate of normal cxr findings in patients with positive rt-pcr was 25%, versus an overall rate of normal chest ct of 0-22% [23] [24] [25] reported in the literature. 60-year-old woman, acquired within 6-9 days since the onset of symptoms, showing bilateral and symmetrical mixed patterns with groundglass opacification, patchy consolidations (arrows); reticular alteration in the right lower and middle fields. d group d, usual distribution. chest radiography of a 61-year-old woman, acquired after 10 days since the onset of symptoms, showing bilateral and symmetrical extensive consolidations (arrows) fig. 3 the trend of the radiographic alterations in the different groups bernheim et al [23] reported the percentage of normal chest ct in patients clustered on the base of the timing of symptom onset and observed a rapid decrease from 56% at 0-2 days to 9% at 3-5 days and 4% at 6-12 days; initial rt-pcr was negative in 12% of the patients included in the study. clustering our patients on the base of the timing of symptom onset, the rate of normal cxr progressively decreased from 36.7% (0-2 days after the onset of symptoms) to 28% (3-5 days), 18.8% (6-9 days), and 16.1% (> 9 days). although the two cohorts are not directly comparable due to different selection criteria and different diagnostic methods, we observed a decreasing trend in the rate of negative cxr through the different time intervals. in our radiographic series, the most frequent alterations were ggo and reticular pattern, alone or in combination with other alterations, resembling the radiographic appearance described in other coronavirus-related pneumonias [26] [27] [28] . furthermore, the distribution of the lesions in the middle fields and that of the relative sparing of the superior fields are similar to the pattern of distribution described in h1n1 influenza pneumonia [29] . in the first 5 days from the onset of the symptoms, reticular alteration was slightly more frequent than ggo, while after this period, ggo came to be predominant. consolidation was constantly less frequent than the other two alterations, especially in the early phase of the disease. our data suggest that early alterations are predominantly reticular, intermediate alterations are predominantly ggo with a period of overlap between these two, whereas consolidations increase in the late phase. as known, the interobserver reliability of cxr is fair to good for the detection of pulmonary infiltrates, while this is poor for determining the pattern of alteration [30] . in line with this observation, in our study when disagreement was present between the two observers, it concerned in most cases the type of radiographic alteration, rather than the detection of alterations or their location. considering the distribution of the lesions on both the lungs and on the axial and craniocaudal plane, our results are in line with the data reported in cxr studies and in ct studies in the literature [3, 9, 16, 25] . in accordance with the observations of wong et al [9] , in the majority of our patients, the alterations were bilateral, and fig. 4 the horizontal distribution (orange) and the craniocaudal distribution (green) of the alterations in the different groups in patients with unilateral lesions, no predominance was observed between left and right. the exclusive peripheral involvement and the combination of peripheral and central distributions were significantly more frequent than exclusive central distribution, without significant modifications of their proportions in the different time intervals. several studies reported a predominant, although not significant, localization of the lesions in the lower lobes [15, 23] ; our results confirm this observation and measured a significantly higher frequency of involvement of the lower fields compared to the middle fields, and of the lower and middle fields compared to the upper fields. our patients had a higher mean age than those reported in the literature, reflecting demographic differences between china and the western countries. the purpose of this study was to analyze and describe the type, frequency, and distribution of cxr findings in covid-19 pneumonia. therefore, the main limitation of the study is the lack of data about the specificity of the cxr findings in covid-19 towards its main differential diagnoses (other viral pneumonia, interstitial lung disease, cardiogenic pulmonary edema, acute lung injury), and the lack of correlation between cxr and ct findings. the description of the radiographic changes during the course of time was beyond the purpose of our study. our study focused on the cxr at the admission, and the subsequent examinations of each patient were not considered; hence, the radiographic differences between the groups do not reflect the course of the radiographic changes over time, but rather the different radiographic presentations at different time intervals since the onset of symptoms. another limitation of the present study is due to the lower quality of bedside radiographs compared to the pa radiographs. on the other hand, the advantage of portable bedside cxr consists of fewer requirement for decontamination of equipment, rooms, and hallways and, eventually, a lower risk of infection for other patients and healthcare workers. in conclusion, the most frequent cxr lesions were reticular alteration and ggo, the first prevailing in the early phase, the following prevailing in the late phase. consolidation was less frequent but, in agreement with other study observations, showed an increasing trend over time [16, 23] . our observations are consistent with previous cxr and ct studies about the bilateral, peripheral, middle, and lower field predominant patterns of distribution of the lesions [9, 16, 23, 25] . in a pandemic scenario, with a high number of inpatients and a growing number of suspected cases, cxr should be considered as a feasible and easy-to-use method to assess lung involvement. our results confirm the most recent recommendations, to employ cxr as a first-line imaging modality in the diagnostic workflow of patients with suspected covid-19 pneumonia. funding information the authors state that this work has not received any funding. guarantor the scientific guarantor of this publication is prof. lorenzo preda. conflict of interest cb is a consultant for bracco imaging italia and doc. congress. the other authors have nothing to disclose. statistics and biometry one of the authors has significant statistical expertise. informed consent written informed consent was obtained from all subjects (patients) in this study. ethical approval institutional review board approval was obtained. • retrospective • observational • performed at one institution world health organization (2020) who director-general's opening remarks at the media briefing on covid-19. world health organization clinical features of patients infected with 2019 novel coronavirus in wuhan radiological findings from 81 patients with covid-19 pneumonia in wuhan, china: a descriptive study coronavirus disease (covid-19) technical guidance: laboratory testing for 2019-ncov in humans. world health organization imaging profile of the covid-19 infection: radiologic findings and literature review emerging 2019 novel coronavirus (2019-ncov) pneumonia the role of chest imaging in patient management during the covid-19 pandemic: a multinational consensus statement from the fleischner society società italiana di radiologia medica e interventistica (2020) documento intersocietario sirm di marzo 2020. società italiana di radiologia medica e interventistica frequency and distribution of chest radiographic findings in covid-19 positive patients chest radiographic and ct findings of the 2019 novel coronavirus disease (covid-19): analysis of nine patients treated in korea coronavirus disease 2019 (covid-19): a systematic review of imaging findings in 919 patients is there a role for lung ultrasound during the covid-19 pandemic? fleischner society: glossary of terms for thoracic imaging excessive toxicity when treating central tumors in a phase ii study of stereotactic body radiation therapy for medically inoperable early-stage lung cancer coronavirus disease 2019 (covid-19): a perspective from china time course of lung changes on chest ct during recovery from 2019 novel coronavirus (covid-19) pneumonia. radiology ct) for suspected covid-19 infection. american college of radiology rcr (2020) rcr position on the role of ct in patients suspected with covid-19 infection. royal college of radiologists society of thoracic radiology (2020) str/aser covid-19 position statement. society of thoracic radiology covid-19 essential role of clinical radiology services position statement. royal australian and new zealand college of radiologists canadian society of thoracic radiology and the canadian association of radiologists (2020) canadian society of thoracic radiology and the canadian association of radiologists' statement on covid-19. canadian society of thoracic radiology and the canadian association of radiologists chest ct findings in coronavirus disease-19 (covid-19): relationship to duration of infection chest ct findings in 2019 novel coronavirus (2019-ncov) infections from wuhan, china: key points for the radiologist ct imaging features of 2019 novel coronavirus (2019-ncov) severe acute respiratory syndrome: radiographic and ct findings imaging of pulmonary viral pneumonia radiographic and ct features of viral pneumonia h1n1 influenza: initial chest radiographic findings in helping predict patient outcome interobserver reliability of the chest radiograph in community-acquired pneumonia publisher's note springer nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations key: cord-275974-uqd30v7b authors: shorfuzzaman, mohammad; hossain, m. shamim title: metacovid: a siamese neural network framework with contrastive loss for n-shot diagnosis of covid-19 patients date: 2020-10-17 journal: pattern recognit doi: 10.1016/j.patcog.2020.107700 sha: doc_id: 275974 cord_uid: uqd30v7b various ai functionalities such as pattern recognition and prediction can effectively be used to diagnose (recognize) and predict coronavirus disease 2019 (covid-19) infections and propose timely response (remedial action) to minimize the spread and impact of the virus. motivated by this, an ai system based on deep meta learning has been proposed in this research to accelerate analysis of chest x-ray (cxr) images in automatic detection of covid-19 cases. we present a synergistic approach to integrate contrastive learning with a fine-tuned pre-trained convnet encoder to capture unbiased feature representations and leverage a siamese network for final classification of covid-19 cases. we validate the effectiveness of our proposed model using two publicly available datasets comprising images from normal, covid-19 and other pneumonia infected categories. our model achieves 95.6% accuracy and auc of 0.97 in diagnosing covid-19 from cxr images even with a limited number of training samples. the coronavirus disease 2019 initially identified in december 2019 in the city of wuhan in china has rapidly spread throughout the world within a very short period of time resulting in an ongoing pandemic. since the outbreak it has affected over two hundred countries and territories across the globe with more than 14.5 million cases reported. fig. 1 depicts the global trend of covid-19 as of july 20, 2020 including total number of confirmed, active, death, and recovered cases [1] . the outbreak was declared as public health emergency of international concern (pheic) by the world health organization (who) [2] on january 30, 2020. the virus is very contagious and primarily transmitted between people through close contact. to stop its rapid transmission, it is crucial to gain a good understanding of the genetic characteristics of the virus. the genome size of the virus varies from approximately 26 to 32 kilobases being one of the largest among single stranded rna (ribonucleic acid) viruses. the average diameter of the particle of the virus is around 120 nm [3] . various common symptoms are found in the infected patients such as cough, fever, shortness of breath, fatigue, loss of smell, and pneumonia. the complications include pneumonia, acute respiratory distress syndrome, and other infections. precise and on time diagnosis are being hampered due to undiscovered treatment, scarcity of resources, and harsh conditions of laboratory environment. this has increased the challenges to curb the spread of the virus. accurate and speedy identification of suspected patients at the early phase may possibly play a critical role in timely quarantine and progressive cure. thus, swift identification of potential infection by coronavirus is incredibly crucial for timely control of epidemic and public health welfare. identification of coronavirus infection is primarily done by nucleic acid test also called a pcr (polymerase chain reaction) test which examines for the existence of antibodies for an infection. however, results from recent studies show that this type of pathogenic laboratory testing though being a diagnostic gold standard suffers from limitations since it is timeconsuming and produces high false negative cases [5] . furthermore, deploying covid-19 tests at a large scale is very expensive and is not affordable by many developing and underdeveloped countries. hence, development of artificial intelligence (ai) based diagnosis and testing methods will be very beneficial. variety of such ai based medical applications include cancer detection [15] , diabetic retinopathy diagnosis [16] , multi-modal skin lesion classification [17] , polyp detection [23] and so on. in favor of this, researchers are taking global initiatives to use ai as a potentially powerful tool to come up with cost-effective and fast diagnostic procedures to control the ongoing epidemic [32] . the key research goals include covid-19 transmission, its early diagnosis, development of effective treatment, and understanding its socio-economic impact [4] . computer aided diagnosis (cad) systems capable of processing chest x-ray (cxr) images and computed tomography (ct) scans along with state-of-the-art deep learning techniques could be very beneficial for the health professionals in diagnosing covid-19 cases. some studies in the literature have already demonstrated the effectiveness of using various deep learning techniques to identify positive covid-19 cases from chest x-ray (cxr) images [27] and computed tomography (ct) scans [24] and to monitor the disease progress over time. since deep learning algorithms generally require huge amount of training data to produce effective prediction results, the existing methods trained on limited training samples (due to the lack of large covid-19 public dataset availability) are likely to suffer from model generalizability to new data. to alleviate this problem of data scarcity, researchers adopted various techniques such as data augmentation and generative adversarial network (gan) [6, 7] . nevertheless, these techniques are highly dependent on the appropriate selection of parameters. various hand-tuned data augmentation technique suffers from over-fitting problem [6] whereas techniques related to generating images through gan face challenges in emulating real patient data which leads to unanticipated bias during model testing [7] . furthermore, some studies have adopted transfer learning technique by using various pre-trained cnn models. since these cnn models are pretrained on a large non-medical dataset (i.e., imagenet), substantial amount of fine-tuning which generally requires longer training period is necessary to produce promising diagnostic results. lately, n-shot (specially, one-shot and few-shot) learning has gained immense popularity in research community for analyzing medical images with a limited sample size. in general terms, for example, one-shot learning refers to the task of classifying an image to a particular class given a single (or few) training samples of each class. specifically, one or more samples from each image class are used to prepare (train) the model which in turn can classify unseen images in future. one of the meta learning models that recently gained success in implementing few-shot learning (especially one-shot) in various domains is siamese network. in a siamese network architecture, identical deep convolutional neural networks (cnns) are trained to extract feature vectors discriminating between samples of each image class which are then contrasted to verify the similarity of the input images. this paper presents a trainable n-shot deep meta learning framework to classify covid-19 cases with limited training cxr images. we use a fine-tuned cnn model called vgg16 [8] as backbone encoder network to generate feature embeddings from the input images and leverage pairwise contrastive loss function to adjust the network weights. more specifically, we have used cxr images from two public datasets to pre-train the embedding cnn network to generate feature representations that are used by the siamese network which learns a metric space for n-shot classification of unseen images without any retraining. in summary, following are the contributions of our work: (a) a meta learning framework called metacovid based on siamese neural network is presented for diagnosis of covid-19 patients from chest x-ray images, (b) the proposed work focuses on the benefit of using contrastive loss and n-shot learning in framework design, (c) a fine-tuned pre-trained vgg encoder is used to capture unbiased feature representations to improve feature embeddings from the input images, (d) the covid-19 diagnosis problem is formulated as a k-way, n-shot classification problem where k and n represent the number of class labels and data samples used for model training, (e) performance evaluation is presented to demonstrate the efficacy of the proposed framework with a limited dataset. the rest of the paper is organized as follows. related work is presented in section 2. section 3 and 4 present method, dataset, and experiments with performance results. finally, section 5 concludes the paper with future work. relevant to the proposed research, our literature study will largely contain existing research effort in the area of covid-19 diagnosis using ai techniques. deep learning which is a specialized form of machine learning in the domain of ai has shown great potential in medical image analysis during the last decade [9] . substantial research has been conducted using deep learning in various medical fields such as disease prediction, diagnosis of pulmonary nodules, and classification of benign and malignant tumors and so on. according a recent study from the researchers at un global pulse [10] , it is shown that ai applications can be as accurate as humans in detecting covid-19 and offer faster and cheaper solutions in diagnosing the virus than standard test kits thus saving radiologists' valuable time. as part of this, researchers are primarily concentrating on techniques based on statistical learning for the detection of potential coronavirus infection from cxr images and computed tomography (ct scans). some research initiatives in progress are provided below. a relatively earlier (in the beginning of the outbreak) effort done by a group of researchers in renmin university of wuhan, china [11] proposed an ai model for the diagnosis of covid-19 cases using ct scans. the model uses unet++ [12] architecture for coronavirus detection using ct scan features and makes use of more than 40,000 images from 106 patients for model training. experimental results demonstrate that the radiologists' efforts in terms of time can substantially be decreased by using this model. xu et al. [13] presented a deep learning model to screen coronavirus disease from viral pneumonia (of type influenza-a) and normal cases with pulmonary ct scans. they have first since it is relatively easier to find cxr images than ct scans especially in rural areas, they can be a viable alternative to ct images. wang and wong [18] proposed an ai system called covid-net to diagnose covid-19 from chest x-ray images containing samples from healthy, covid-19 and other pneumonia infected patients. the limitation of this study is that the authors trained and tested their model using an imbalanced dataset which contains very few (less than 100) covid-19 images as opposed to about 16,000 images from healthy and other non-covid pneumonia patients. additionally, authors in [19] presented a similar study that uses cxr images to detect coronavirus infection through transfer learning mechanism. chakraborty [20] also developed a cxr based model using deep neural network that can achieve significant performance improvement even when the size of the dataset is limited. nevertheless, the model lacks generalizability and needs fine-tuning to produce more stable results. another laudable effort made by researchers from delft imaging project [21] which developed an ai model for diagnosing covid-19 from cxr images. their model is called cad4covid which is built upon an existing ai model previously developed for diagnosing tuberculosis. it triages covid-19 suspected patients. hossain et al. [38] presented a healthcare framework based on 5g network that makes use of cxr and ct images for covid-19. abbas et al. [22] have also leveraged deep learning techniques for the diagnosis of covid-19 from cxr images using cnn and other pre-trained models such as resnet [25] . besides, karim et al. [26] have made contribution in interpreting their predictions by extracting critical features related to covid-19 to gain deeper understanding. although a recent report [28] has shown the success of a number of chinese hospitals in deploying ai-assisted radiology technologies in combating covid-19, radiologists have shown their concern [29] that the shortage of available data to train the ai diagnostic models is a major challenge. this is substantiated by the fact that large body of the ai models in the literature have used datasets containing limited covid-19 samples. to address the challenge of limited training samples, he et al. [30] proposed a framework based on deep neural network that is able to attain significant improvement in accuracy in covid-19 detection with a limited dataset consisting of ct scans. they have developed a synergistic approach to combine transfer learning with self-supervised contrastive learning to extract unbiased features to avoid overfitting problem. the experimental results demonstrate the superiority of their approach in comparison with several state-of-the-art models. in a subsequent effort, chen et al. [31] develop a meta learning model with prototypical network to predict coronavirus infections from chest ct images that requires a small dataset for training. they have used momentum based contrastive learning to extract feature vectors form the input images that are used by the prototypical network to make predictions on unseen ct images for potential covid-19 encounters. validation results with two publicly available ct scans datasets suggest the effectiveness of their model compared to several other relevant methods. existing techniques in the literature that use cxr images for covid-19 diagnosis mostly use custom cnn architecture or pre-trained transfer learning models which require large training data to produce rich feature encoding. in contrast, we have proposed an end-to-end trainable nshot deep meta learning framework based on siamese neural network to classify covid-19 cases with limited training cxr images. our proposed model is computationally efficient that can achieve better or the same level of performance as the pre-trained and other custom cnn models that require longer training time. also, techniques in the literature mostly use de-facto categorical or binary cross entropy loss function. in contrast, we have used contrastive loss function which results in faster model convergence with fewer experiments and hyperparameter updates. furthermore, most of the existing models use image augmentation to improve model generalizability even with prolonged training time. alternatively, the proposed meta learning framework shows faster model convergence and greater generalizability by using contrastive loss function and appropriate hyper-parameter optimization such as learning schedule and regularization through dropout technique used in various layers in the model. the shortage of covid-19 cxr images brings substantial challenges in developing tools for effectively diagnosing covid-19 cases using deep learning based techniques. to tackle this challenge, we leverage deep meta learning and devise the diagnosis problem of covid-19 from cxr images as a n-shot classification problem which can be considered as an instantiation of meta learning in the paradigm of supervised learning. meta learning has recently emerged as a trending research area in the field of artificial intelligence (ai) and is believed to be a steppingstone for attaining artificial general intelligence (agi). it is often referred to as "learning to learn" and is capable of learning new skills and generalizing to new tasks quickly by means of limited training samples. the aim of n-short learning is to make classification of unseen data given only a limited the problem of n-shot image classification can be defined as k-way n-shot episodic task where k denotes the number of target class labels in the dataset and n denotes the number of available images (samples) for each of the classes. given, we have a dataset, d, we sample n data points (images) from each of the k classes present in our data set and we call it as support set. similarly, we sample q different images from each of the classes and call it as query set. the goal is to classify the images of the query set based on k classes and kn total images available in the support set. figure 3 demonstrates an n-shot classification scenario in visual form. in our case, we have three different types cxr images in our dataset, namely, normal, covid-19 positive, and non-covid pneumonia cases. hence, we view the diagnosis problem both as 3-class (images from all three categories) and 2-class (normal and covid-19 positive images) classification problem. in this context, our covid-19 diagnosis problem can be identified as three-way n-shot and two-way n-shot learning problem. this section presents the generic high-level architecture of our deep siamese network (as shown in fig. 4) for n-shot learning to diagnose covid-19 cases which is widely known as metric learning-based approach to meta learning. we will describe all major components of the architecture including base cnn encoder for feature embedding, contrastive learning, and training strategies. in the metric-based meta learning setting, the goal is to learn the appropriate metric space. the fundamental concept is mostly related to nearest neighbors techniques and kernel density estimation where the predicted probability is calculated across a set of given output labels, y, as a weighted sum of labels of the examples from the support set. a kernel function, f ï´ , is used to create the weight which essentially estimates the similarity between two different input data points. the predicted probability over the samples of a support set, s, can be formulated as: the performance of a metric-based meta learning model largely depends on learning an appropriate kernel. a useful metric would represent the relationship among the inputs such as similarity in the latent space to facilitate highly accurate predictions. in our case, we want to learn the similarity between two images. for this purpose, a convolutional neural network is used to extract the features from two images and finds the similarity by computing the distance between features of these two images. this approach is widely used in metric-based learning algorithms such as siamese networks that we have used in our meta learning framework besides other metric-based networks such as matching, relation and prototypical networks. as shown in the preceding diagram, a siamese network contains two identical parallel networks both sharing the same weights and architecture where each of the networks accepts a different input image and the output from them are combined to make the final prediction. more specifically, the goal is to have two identical base neural networks that take an actual image and another candidate image as input and can learn a function to produce the similarity output between these two images. the concern is that how we can essentially train such a neural network encoder that can learn this similarity function. ideally, a convolutional neural network could be used without any constraints. apparently, it would be desirable to use a custom cnn model which is smaller and computationally efficient and can achieve the same level of performance in feature encoding as the pre-trained models. however, such a cnn model requires large training data to produce rich feature encoding. since we have a limited dataset, we leverage the power of cnn models pre-trained on large imagenet [39] data which in recent times have shown promising results in solving computer vision problems such as medical imaging. hence, we have used a fine-tuned pre-trained vgg-16 [8] as base encoder to obtain feature embeddings from the input images to ultimately compute similarity among them. let's consider that we have two input images, x 1 and x 2 . after passing the image first image, x 1 , through the top encoder, we receive a feature embedding of x 1 denoted as z 1 (x 1 ) = vgg16 (x 1 ) where z 1 (x 1 ) is the output generated from the average pooling layer. similarly, the second image, x 2 is fed to the identical bottom encoder sharing the same weights, w, to get a different feature embedding of x 2 denoted as z 2 (x 2 ). then, in the latent space of feature embeddings, we feed these two embeddings to an energy function, e, which will give us the similarity between the two inputs. we use l 1 component wise distance as our energy function which can be expressed as follows: ( 1 , 2 ) = ( 1 , 2 ) = â�� 1 ( 1 , 2 ) â�� 2 ( 1 , 2 )â�� (2) the value of e will be smaller if the input images (x 1 and x 2 ) are similar and vice versa. in reality, if it is less than a supplied threshold value the images are similar and, if not, they are different. finally, this distance value can be incorporated in loss function through the use of an appropriate loss function, the base encoder network can learn parameters to obtain a better encoding of the input image. since siamese networks make binary classification by classifying if the input images are similar or not, using binary cross-entropy loss function would be a natural choice. however, we have also considered using contrastive loss function due to the nature of classification strategy adopted by siamese networks based on the similarity of pairs of input images. hence, we have used both the loss functions in this study for performance evaluation. originally proposed by hadsell et al. [33] , the contrastive loss function requires pairs of input samples as opposed to individual samples. the idea is that the base encoder in the preceding equation, the value of y is the true label, which will be 0 when the two input images are similar and 1 if they are dissimilar, and is the distance measure between feature embeddings of the input images. now, y equals to 0 implies that the amount of loss contributed by similar pairs would be simplified to the first term only and is minimized. on the contrary if y=1then the loss will be simplified to the second term and is maximized to m, a hyperparameter called margin. thus, when input pairs are dissimilar, and if their distance is greater than the margin, they do not incur a loss (as shown in fig. 5 ) binary cross entropy loss also called log loss is used to estimate the performance of a classifier with an output probability ranging from 0 to 1. the loss value will increase if the predicted probability deviates from the true label. this can be formulated as follows where y and p represent the class label and prediction probability, respectively: this loss function is used to train the network so that it can differentiate between similar and dissimilar images if we provide one training example from positive and negative categories and aggregate both the losses as below: formally, we have in our hand a k-way, n-shot classification problem for classifying cxr images. we have our dataset, d, with a training (d train ) and a test (d test ) split. now, the training set will contain n samples form each of k classes thus totaling k.n examples in d train and d test consists of several samples for evaluation. we train our model in an episodic fashion which dictates that in each episode, we sample a few data points from our meta-training dataset, d train , prepare our support set and query set, and train on the support set and test on the query set. so, over series of episodes, our model will learn how to learn from a smaller dataset. in this way, our model will learn to solve an unseen task by gaining experience through a series of training tasks. an example n-shot classification training scenario is given in fig. 6 where we create episodes consisting of tasks each of which is defined by a support set and a query set containing sample images from meta-training dataset. moreover, each of these tasks is similar to the test n-shot classification task containing images for support and query sets. finally, the detailed training strategy of our model is shown in algorithm 1. to demonstrate the effectiveness of our n-shot meta learning approach to covid-19 diagnosis and inspect the effects of using different k-way, n-shot variants and loss functions, we extensively evaluate our proposed siamese network model with two publicly available cxr datasets. doctors frequently use cxr and ct scans for the diagnosis of various common diseases such as pneumonia, cancers, lung inflammation, and internal organ injuries. given the fact that cxr imaging machines are available in nearly all hospitals we decide to use cxr images instead of using ct scans or other types of image data. in the subsequent sections, we will present the datasets used with preprocessing, experimental settings, and results with discussion. in this study, we use covid19 update parameter ï´ b with new weight, w end for end for a final balanced dataset containing a total of 678 cxr. we divide the dataset for generating support set and query set based on various values of n in n-shot classification task. to pretrain the base vgg-16 encoder to be used in generating feature embeddings on the input images to the siamese network, we use the same dataset consisting of cxr images as mentioned above and split the dataset into training and validation sets with a ratio of 0.7:0.3. table 1 shows the statistics of the dataset split. due to the fact that the images in the dataset were collected from different locations with various clinical settings, the intensity and quality of images vary considerably. nevertheless, we avoid extensive pre-processing of our cxr images in the dataset to gain improved generalization ability of our proposed siamese network model. this in turn makes our model further robust to artifacts and noises present in the images while extracting feature embeddings from the input images. thus, we only used few standard pre-processing tasks including image resizing, normalization, and histogram-equalization to optimize the model training method. the size of the cxr images in the dataset varies from 365 ã� 465 to 1125 ã� 859 pixels. hence, we re-scale all images to a size of 100 ã� 100 pixels to get a consistent image dimension for the entire dataset. additionally, we perform intensity normalization also called scaling which is an important pre-processing task to expedite model convergence by eliminating feature biases and attaining a uniform distribution for the dataset. we convert the image pixel values from [0, 255] to [0, 1] to obtain a standard normal distribution by using min-max normalization technique. finally, we apply histogram equalization on the input images in all three rgb channels to improve image contrast. this is usually done by effectively stretching out the most often used intensity values which allows the areas with poorer local contrast to achieve a better contrast. the pre-trained base encoder network and the proposed siamese network models are implemented using tensorflow. to evaluate the efficacy of our proposed meta-learning model in diagnosing covid-19 cases we took the following approach. we start by evaluating our model for various 3-way, n-shot learning settings with both contrastive and cross-entropy losses. a comparison of performance results is also done with different pre-trained cnn models such as inception [36] , xception [37] , inception resnet v2 [36] , vgg-16 [8] . finally, we present the results obtained from 2-class (normal, covid-19) variant of our classification problem with different n-shot learning settings. since our method is based on multi-shot learning, we are interested in investigating how does model performance change with the number of shots. as such, we carry a number of experiments to observe the relationship between the number of shots and the performance. we also produce performance results (as shown in table 3 ) with similar learning settings but with binary cross-entropy loss which is usually deemed to be a natural choice for classification problem. generally, the performance results obtained with contrastive loss function seem to be better than the results obtained with cross-entropy loss function. this is due to the fact that our siamese network model works based on similarity of pairs of images and contrastive loss function is reported in the literature to be more effective than crossentropy loss. moreover, as shown in fig. 7 finally, to evaluate the effectiveness of our model in identifying covid-19 cases from normal cxr images only (2-class problem) we perform experiments with similar settings in 3-class problem. table 5 shows the performance results with contrastive loss and various nshot settings in classifying healthy and covid-19 patients. as expected, the model shows better performance in all metrics for diagnosing covid-19 patients in 2-class scenario. this study is one step towards better understanding of the dynamics of covid-19 pandemic and proposing a state-of-the-art ai based solution for efficient and fast diagnosis system for covid-19 infections which is the need of the time. the proposed research aims to achieve this through the integration of a meta learning network model with contrastive loss and pre-trained cnn encoder. specifically, we use a fine-tuned pre-trained vgg-16 network encoder to capture unbiased feature representations that are robust to overfitting and leverage a siamese network for final classification of covid-19 cases. we show that our proposed model with contrastive loss and various n-shot learning settings offer a highly accurate yet practical solution for automatically diagnosing covid-19 cases to accelerate line of treatment for patients. our best model with 10-shot learning setting achieves an accuracy of 95.6% in diagnosing covid-19 cases with impressive values of sensitivity (96.0%) and specificity (98.0%) which are deemed to be very critical performance estimates for applications in medical settings. furthermore, our proposed model exhibits comparable or in some cases better performance than the studied fine-tuned pre-trained cnn models. this is promising due to the fact that our meta learning model is trained only with a limited sample (10) of training examples from each category of cxr images. simultaneously, it is essential to pinpoint some of the shortcomings of this work which can possibly be tackled in future research. the major drawback is the inadequate interpretability of our model since effective diagnosis requires that results obtained from such interpretability study should be clinically verified by an expert radiologist. as an immediate future work, we plan to extend our work by producing qualitative results with the aid of a model interpretation tool to gain deeper understanding of what our model is learning from the input data during training and validation. we also plan to better tackle covid-19 diagnosis problem as a multi-modal data fusion problem where various types of clinical data such as patient vitals, location, and population density will be used in addition to image data. covid-19 dashboard, coronaboard statement on the second meeting of the international health regulations (2005) emergency committee regarding the outbreak of novel coronavirus (2019-ncov)". world health organization covid-19 infection: origin, transmission, and characteristics of human coronaviruses b5g and explainable deep learning assisted healthcare vertical at the edge: covid-i9 perspective a deep learning algorithm using ct images to screen for corona virus disease (covid-19) data augmentation using learned transformations for one-shot medical image segmentation brain tumor segmentation using convolutional neural networks in mri images very deep convolutional networks for large-scale image recognition deep learning and medical diagnosis mapping the landscape of artificial intelligence applications against covid-19 deep learning-based model for detecting 2019 novel coronavirus pneumonia on high-resolution computed tomography: a prospective study. medrxiv unet++: a nested u-net architecture for medical image segmentation deep learning system to screen coronavirus disease 2019 pneumonia, engineering going deeper with convolutions detection and classification of cancer in whole slide breast histopathology images using deep convolutional networks exudate segmentation using fully convolutional neural networks and inception modules multi-label classification of multi-modality skin lesion via hyper-connected convolutional neural network covid-net: a tailored deep convolutional neural network design for detection of covid-19 cases from chest radiography images covid-19: automatic detection from x-ray images utilizing transfer learning with convolutional neural networks an attempt-detection of covid-19 presence from chest x-ray scans using cnn & class activation maps delft imaging, cad4covid: triage for covid-19 using artificial intelligence on chest x-rays classification of covid-19 in chest x-ray images using detrac deep convolutional neural network polyp detection during colonoscopy using a regression-based convolutional neural network with a tracker ai augmentation of radiologist performance in distinguishing covid-19 from pneumonia of other etiology on chest ct deep residual learning for image recognition deep covid explainer: explainable covid-19 predictions based on chest x-ray images coronet: a deep neural network for detection and diagnosis of covid-19 from chest x-ray images artificial intelligence assisted radiology technologies aid covid-19 fight in china debate flares over using ai to detect covid-19 in lung scans sample-efficient deep learning for covid-19 diagnosis based on ct scans momentum contrastive learning for few-shot covid-19 diagnosis from chest ct images covid-19 networking demand: an auction-based mechanism for automated selection of edge computing services dimensionality reduction by learning an invariant mapping covid-19 image data collection inception-v4, inception-resnet and the impact of residual connections on learning xception: deep learning with depthwise separable convolutions explainable ai and mass surveillance systembased healthcare framework to combat covid-i9 like pandemics imagenet: a large-scale hierarchical image database the authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. key: cord-292341-uo54ghf3 authors: cocconcelli, elisabetta; biondini, davide; giraudo, chiara; lococo, sara; bernardinello, nicol; fichera, giulia; barbiero, giulio; castelli, gioele; cavinato, silvia; ferrari, anna; saetta, marina; cattelan, annamaria; spagnolo, paolo; balestro, elisabetta title: clinical features and chest imaging as predictors of intensity of care in patients with covid-19 date: 2020-09-16 journal: j clin med doi: 10.3390/jcm9092990 sha: doc_id: 292341 cord_uid: uo54ghf3 coronavirus disease 2019 (covid-19) has rapidly become a global pandemic with lung disease representing the main cause of morbidity and mortality. conventional chest-x ray (cxr) and ultrasound (us) are valuable instruments to assess the extent of lung involvement. we investigated the relationship between cxr scores on admission and the level of medical care required in patients with covid-19. further, we assessed the cxr-us correlation to explore the role of ultrasound in monitoring the course of covid-19 pneumonia. clinical features and cxr scores were obtained at admission and correlated with the level of intensity of care required [high(himc) versus low-intensity medical care (limc)]. in a subgroup of patients, us findings were correlated with clinical and radiographic parameters. on hospital admission, cxr global score was higher in himcs compared to limc. smoking history, po(2) on admission, cardiovascular and oncologic diseases were independent predictors of himc. the us score was positively correlated with fio(2) while the correlation with cxr global score only trended towards significance. our study identifies clinical and radiographic features that strongly correlate with higher levels of medical care. the role of lung ultrasound in this setting remains undetermined and needs to be explored in larger prospective studies. since december 2019, when the first cases of coronavirus disease 2019 (covid19) were reported, the diffusion of the severe acute respiratory syndrome coronavirus type 2 (sars-cov-2) has rapidly spread from the hubei province in china to involve up to 213 states and territories to date, reaching pandemic proportions [1] . despite epidemiological reports showing that approximately half of the infected people are asymptomatic [1] , the spectrum of respiratory manifestations may range from mild symptoms, such as dry cough, fever, and fatigue, to acute respiratory distress syndrome (ards), requiring admission to intensive care unit (icu) and mechanical ventilation (mv). in this scenario, thoracic radiology plays a key role in early detection of lung involvement from covid-19. chest computed tomography (ct) is the technique with the highest sensitivity, but the risk of contamination and the need for a dedicated hospital organization makes ct hardly available in an emergency setting. portable chest x-ray (cxr) and ultrasonography (us) are quicker, safer and less expensive alternatives [2] . cxr is recommended as first level assessment by several scientific societies (american college of radiology, society of thoracic radiology) in the context of the sars-cov-2 pandemic [3] . predominant cxr features in patients with covid-19 include lower lobe, peripheral, bilateral ground glass opacities (ggo) or consolidations [2] , similar to other forms of viral pneumonias, such as the h1n1 strain [4] . yet, cxr could be normal in as many as 31% cases, peaking its sensibility in patients with advanced disease [5] [6] [7] . in the last three decades, lung us (lus) has become increasingly important in clinical practice, particularly in the assessment of patients with pneumonia, with sensitivity and specificity of 92% and 93%, respectively, especially when performed by experienced operators [8] . in the covid-19 pandemic, lus has been used in multiple centers as first radiological approach in patients with suspected pneumonia. the main ultrasound findings include multiple b-lines (separated or coalescent), peripheral consolidations and thickened pleural lines [9] , which however are nonspecific and found in a number of infectious and non-infectious diseases [10] . the use of lus and cxr in combination has the potential to facilitate the identifications of ards [11] . with this background, we investigated the relationship between cxr severity score on admission and the level of medical care required in patients with covid-19. further, we assessed the radiographic-ultrasound correlation with the aim to explore the value of ultrasound in monitoring the course of covid-19 pneumonia. in this longitudinal retrospective study, we identified a cohort of clinically well-characterized patients with sars-cov-2 infection referred to the university hospital of padova (division of infectious and tropical diseases, respiratory disease unit and intensive care unit) between march and may 2020. one hundred and two patients were included in the study (table 1 ) since the diagnosis of sars-cov-2 infection was made based on nasopharyngeal swab positivity. clinical and demographics data, and cxrs were obtained on admission. a subset of 25 patients (25/102, 24.5%), who were hospitalized in a low-intensity care setting, underwent a bedside lus and a cxr in the late phase of covid-19 pneumonia. the aim of performing lus and cxr in parallel was to explore the relation between these two procedures. the need for invasive/non-invasive ventilation or high-flow nasal cannula (hfnc), which required admission to icu or to the respiratory icu, was considered as high-intensity medical care (himc), while the need for low flow oxygen supplementation through nasal cannula or face mask, which required the setting of a general ward, was considered as low-intensity medical care (limc). the level of care could change over time based on patient's clinical conditions. for all patients, clinical data (demographics and comorbidities), gas exchange values (fio 2 , po 2 and po 2 /fio 2 ) were collected on admission (table 1) . we have categorized the five most frequent type of comorbidities: cardiovascular diseases (cvd), respiratory diseases, metabolic diseases, autoimmune diseases and oncologic diseases. among the metabolic comorbidities, we have considered diabetes mellitus, obesity and dyslipidemia (13%). oncologic history mentioned the different organs affected (i.e., lung, prostate, pancreas. breast, colon). this was a retrospective study on anonymized patient's data collected from electronic medical records. the study protocol complies to the ethical guidelines of the 1975 declaration of helsinki and, in agreement with national regulation on retrospective observational studies, it was notified and approved by the local ethics committee (n • 46430/03.08.2020) and the need for patient's informed consent was waived. we retrieved data on patients hospitalized for covid-19 between march and may 2020 at the university hospital of padova, one of the most affected areas in north-east of italy. we screened records of all patients admitted to our hospital with a diagnosis of sars-cov2 infection. for each patient, a single image plane cxr was available on hospital admission. two radiologists (c.g., g.b.) with more than ten years of experience in the thoracic field, who were blind to clinical data, scored the images independently using a semi-quantitative scale. this represented a modification of previously reported scoring systems that allowed to evaluate the extension of ground glass opacities (ggo) and consolidation (co) [6, 12, 13] . for each lung lobe, the two radiologists assessed the extent of ggo and co using the following scale: 0 (normal), 1 (up to 30% of the lobe involved), 2 (30% to 60% of the lobe involved), and 3 (more than 60% of the lobe involved). the sum of the scores for each lung lobe and a final value of ggo and co score for each patient was then calculated ( table 2 ). the cxr "global" score was calculated as the sum of the ggo and co scores of each patient, with a maximum score of 36. finally, each patient was classified as "normal", "ggo prevalent", "co prevalent", or "mixed" based on the prevalent cxr pattern [14] . table 2 . baseline radiological scores of the overall population hospitalized for sars-cov-2 related infection, and of the two subgroups categorized in low (limc) and high (himc) intensity medical care. a subset of 25 patients underwent bed-side lus. the examination was conducted with a portable mylab tm 25 gold ultrasound unit (esaote, genova, italy) and a dedicated ca631 convex transducer (range of frequency 1-8 mhz). we used low frequency and a single-focal modality at the pleural line. the depth was arranged on 6-7 cm and the harmonic-imaging system was deactivated. the lus score was calculated across 12 chest zones (six on each hemithorax) using a scale from 0 (normal pattern, a-lines or non-significant b-lines), 1 (significant b-lines ≥3 per rib space), and 2 (coalescent b-lines with or without small consolidations) to 3 (consolidation), as previously reported [15] . a final "us global score" was calculated for each patient with a maximum score of 36. categorical variables were described as absolute (n) and relative values (%), whereas continuous variables were described as median and range. to compare demographic data and baseline clinical characteristics between limc and himc groups, chi square test and fisher's exact test for categorical variables and mann-whitney u test for continuous variables were used, as appropriate. the correlation between cxr global score and po 2 , fio 2 , p/f ratio on admission was assessed for the entire study population and in the limc and himc groups using the nonparametric spearman's rank method. univariate logistic regression analysis, followed by a multivariate logistic regression, was performed to detect the strongest predictors of level of care. the covariates included in the final model were those that were significant in the univariate regression analyses. the correlation between lus global score and the corresponding cxr global score and fio 2 was calculated using the nonparametric spearman's rank method. all data were analyzed using spss software version 25.0 (us: ibm corp., new york, ny, usa). p-values < 0.05 were considered statistically significant. the graphs were obtained using the statistical package graphpad prism 7.0 (graphpad software, inc., la jolla, san diego, ca, usa). demographic and clinical characteristics at baseline (i.e., on hospital admission) are summarized in table 1 . most patients were male (73%) with a median age on admission of 68 years. seventy-one patients required limc during hospitalization and thirty-one himc. patients requiring himc (himcs) were mainly male (87 vs. 67%; p = 0.05) and older [74 vs. 63 (22-94) years; p = 0.03], with a higher body mass index (bmi) [31 (21-43) vs. 24 (16-31) kg/m 2 ; p = 0.02]. moreover, they had a heavier smoking history (10 (0-60) vs. 0 (0-60) pack/year (py); p = 0.01) and were mainly former smokers (61%). the most common presenting symptoms were fever (92%), cough (61%) and shortness of breath (34%), and with 5% of patients complaining of impaired sensory. the frequency of these symptoms did not differ between himcs and limcs. interestingly, although the time interval between onset of respiratory symptoms and admission to the emergency unit was similar, himcs showed a greater impairment of respiratory gas exchange with a lower po 2 on room air on admission (60 (21-90) vs. 90 (54-119) mmhg; p < 0.0001), greater fio 2 requirement at the time of admission (39 (21-100) vs. 21 (21-51) %; p < 0.001) and worse p/f (158 (33-429) vs. 429 (106-567); p < 0.0001) compared to limcs. in the overall population, cvds were the most frequent comorbidities (55%) that we observed. among the metabolic comorbidities, diabetes mellitus was the most prevalent (26%), followed by obesity (8%) and dyslipidemia (13%). hypothyroidism was the most frequent condition among the autoimmune diseases (3%). oncologic diseases (13%) were equally distributed among organs affected (i.e., lung, prostate, pancreas. breast, colon). himcs reported more comorbidities, in particular cardiovascular diseases (cvds) (80 vs. 49% of cases; p = 0.002), metabolic diseases (61 vs. 37%; p = 0.002) and oncologic diseases (22 vs. 8%; p = 0.05). furthermore, this patient group showed a higher frequency of bacterial co-infections (42 vs. 15%; p = 0.002) during hospitalization. finally, the hospitalization time was significantly longer for himcs compared to limcs [26 (7-119) vs. 8 (2-50) days; p < 0.0001], with 4 patients dying among himcs and only one among limcs (p = 0.01). on admission, himcs showed a more severe radiological impairment compared to limcs, with higher x-ray global score [8 (0-35) vs. 3 (0-22); p < 0.0001], ggo score (5 (0-15) vs. 1 (0-18); p < 0.001) and co score (0 (0-35) vs. 0 (0-10); p = 0.02), respectively. when considering the prevalent cxr pattern, only one patient among himcs had a normal cxr on admission compared to limcs (14; p = 0.003), with similar proportion of patients with "ggo prevalent" and "co prevalent" patterns in the himc and limc groups. in the overall study population, a positive correlation was observed between cxr global score and fio 2 on admission (r = 0.6, p < 0.0001). when stratified by level of care, the correlation between cxr global score and fio 2 on admission was confirmed in limcs (r = 0.51, p < 0.0001) but not in himc (figure 1a ). in the overall study population, we observed a negative correlation between cxr global score and po 2 on admission (r = −0.6, p < 0.0001). when stratified by level of care, the correlation between cxr global score and po 2 on admission was confirmed in limcs (r = −0.37; p = 0.02) but not in himcs (figure 1b) . finally, in the overall study population, we observed a negative correlation between cxr global score and p/f on admission (r = −0.6, p < 0.0001). when stratified by level of care, the correlation between cxr global score and p/f at admission was confirmed in both limcs (r = −0.40; p = 0.0003) and himcs (r =−0.37; p = 0.04) (figure 1c) . univariate logistic regression analysis of factors associated with level of care revealed that sex, age, smoking history, fio 2 , po 2 in room air at admission, bacterial co-infections developed during hospitalization, cvds, metabolic and oncologic diseases and chest x-ray global score had significant positive association with a higher level of care in the entire study population (table 3) . multivariate analysis performed using variables with statistical significance in univariate analysis revealed that smoking history (odds ratio 6.55; 95% ci: 1.15-52.09; p = 0.04), po 2 values are expressed as odds ratio (95% confidence interval). logistic regression analysis in relation to level of care was used to determine the relationship of clinical and radiological characteristics with higher level of care needed during hospitalization. a subset of 25 patients underwent a bed-side lus after a median time of 11 days from admission. in parallel, cxrs were performed in the same patients at the same time point. the median lus global score was 7 , whereas the median cxr global score was 9 (3-13). the lus global score positively correlated with the fio 2 requirement at the time of the us examination (r = 0.36; p = 0.03) ( figure 2) . conversely, the correlation between lus global score and cxr global score only trended towards statistical significance (r = 0.36, p = 0.07) ( figure 3) . finally, the lus global score positively correlated with the cxr co score (r = 0.38; p = 0.05) (figure 4) but not with the ggo score. j. clin. med. 2020, 9, x for peer review 7 of 12 univariate logistic regression analysis of factors associated with level of care revealed that sex, age, smoking history, fio2, po2 in room air at admission, bacterial co-infections developed during hospitalization, cvds, metabolic and oncologic diseases and chest x-ray global score had significant positive association with a higher level of care in the entire study population (table 3) . multivariate analysis performed using variables with statistical significance in univariate analysis revealed that smoking history (odds ratio 6.55; 95% ci: 1.15-52.09; p = 0.04), po2 (36.7, 3.64-681.4; p = 0.005), cvds (10.89, 1.44-112; p = 0.02), and oncologic diseases (17.13, 1.76-242.6; p = 0.02) were independent predictors of higher level of care in patients with sars-cov-2 infection. values are expressed as odds ratio (95% confidence interval). logistic regression analysis in relation to level of care was used to determine the relationship of clinical and radiological characteristics with higher level of care needed during hospitalization. a subset of 25 patients underwent a bed-side lus after a median time of 11 days from admission. in parallel, cxrs were performed in the same patients at the same time point. the median lus global score was 7 , whereas the median cxr global score was 9 (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) . the lus global score positively correlated with the fio2 requirement at the time of the us examination (r = 0.36; p = 0.03) ( figure 2) . conversely, the correlation between lus global score and cxr global score only trended towards statistical significance (r = 0.36, p = 0.07) ( figure 3) . finally, the lus global score positively correlated with the cxr co score (r = 0.38; p = 0.05) (figure 4) but not with the ggo score. this is a retrospective analysis of clinical features and radiographic severity scores in patients with covid-19 and how these parameters on hospital admission correlate with different levels of medical care (i.e., himc vs. limc). a subgroup of patients also underwent lus, which was correlated with chest radiographs. our study revealed that patients with covid-19 who required a himc are mainly men, former smokers with a higher pack/year of smoking history, older and with a higher bmi compared to patients requiring limc. furthermore, the majority of them reported at least one comorbidity (i.e., cardiovascular, metabolic, or oncologic) and required on emergency room oxygen supplementation due to low alveolar oxygen partial pressure (pao2). moreover, using a multivariate analysis, we found that a heavier smoking history, po2 level on room air, and presence of cardiovascular or oncological disease on admission were independent predictors of the need of himc. our findings mirror those from previous studies indicating that older male patients with comorbidities are at higher risk of pulmonary infection and fatal consequences from covid-19 [16, 17] . in our study, we show that the number of pack-years was significantly higher in former smokers who required intensive care compared to those requiring limc. moreover, the proportion of former smokers was markedly increased among severe patients, whereas nonsmokers with covid-19 experienced a milder illness, which required low-flow oxygen supplementation. this is in line with other reports that explored the association between smoking and progression of covid-19 pneumonia [18] . notably, in our study, multivariate analysis revealed that smoking this is a retrospective analysis of clinical features and radiographic severity scores in patients with covid-19 and how these parameters on hospital admission correlate with different levels of medical care (i.e., himc vs. limc). a subgroup of patients also underwent lus, which was correlated with chest radiographs. our study revealed that patients with covid-19 who required a himc are mainly men, former smokers with a higher pack/year of smoking history, older and with a higher bmi compared to patients requiring limc. furthermore, the majority of them reported at least one comorbidity (i.e., cardiovascular, metabolic, or oncologic) and required on emergency room oxygen supplementation due to low alveolar oxygen partial pressure (pao2). moreover, using a multivariate analysis, we found that a heavier smoking history, po2 level on room air, and presence of cardiovascular or oncological disease on admission were independent predictors of the need of himc. our findings mirror those from previous studies indicating that older male patients with comorbidities are at higher risk of pulmonary infection and fatal consequences from covid-19 [16, 17] . in our study, we show that the number of pack-years was significantly higher in former smokers who required intensive care compared to those requiring limc. moreover, the proportion of former smokers was markedly increased among severe patients, whereas nonsmokers with covid-19 experienced a milder illness, which required low-flow oxygen supplementation. this is in line with other reports that explored the association between smoking and progression of covid-19 pneumonia [18] . notably, in our study, multivariate analysis revealed that smoking this is a retrospective analysis of clinical features and radiographic severity scores in patients with covid-19 and how these parameters on hospital admission correlate with different levels of medical care (i.e., himc vs. limc). a subgroup of patients also underwent lus, which was correlated with chest radiographs. our study revealed that patients with covid-19 who required a himc are mainly men, former smokers with a higher pack/year of smoking history, older and with a higher bmi compared to patients requiring limc. furthermore, the majority of them reported at least one comorbidity (i.e., cardiovascular, metabolic, or oncologic) and required on emergency room oxygen supplementation due to low alveolar oxygen partial pressure (pao 2 ). moreover, using a multivariate analysis, we found that a heavier smoking history, po 2 level on room air, and presence of cardiovascular or oncological disease on admission were independent predictors of the need of himc. our findings mirror those from previous studies indicating that older male patients with comorbidities are at higher risk of pulmonary infection and fatal consequences from covid-19 [16, 17] . in our study, we show that the number of pack-years was significantly higher in former smokers who required intensive care compared to those requiring limc. moreover, the proportion of former smokers was markedly increased among severe patients, whereas nonsmokers with covid-19 experienced a milder illness, which required low-flow oxygen supplementation. this is in line with other reports that explored the association between smoking and progression of covid-19 pneumonia [18] . notably, in our study, multivariate analysis revealed that smoking history was an independent risk factor for himc. we speculate that cigarette smoke upregulates the expression of angiotensin-converting enzyme 2 receptors, which in turn facilitate sars-cov-2 entry in the respiratory epithelium; this implies that smoking habit may represent a risk factor for developing severe illness even among former smokers. in other words, having quit smoking does not seem to prevent the risk of severe covid-19 pneumonia [19] . chronic respiratory disease, including, among others, chronic obstructive pulmonary disease (copd), carry a worse prognosis when associated with chronic conditions, such as cardiovascular diseases [20] [21] [22] . interestingly, in our cohort, concomitant cvds and neoplasms were independent risk factors for hospitalization in himc, with up to 80% of patients who required himc reporting an history of cvd (mainly arterial hypertension). a recent meta-analysis of 1576 patients concluded that hypertension, chronic respiratory disease and cvd are risk factors for severe covid-19 disease [23] . considering our study population, we observed that cvds are the most frequent comorbidities (55% of cases), 27% of patients suffered from diabetes mellitus, 13% showed blood tests positive for dyslipidemia, and 8% of our patients were obese. we, therefore, are in line with an italian nationwide observational study of covid-19 inpatients which reported a linear direct relationship between the number of comorbidities and the risk of death [24] . all these findings emphasize the importance of past medical history and comorbidities in the disease course of covid-19 patients, as they may predispose to worse outcome and higher intensity of care. ppo 2 level < 90 mmhg on admission to emergency room was an additional independent predictor of himc requirement. this is interesting, as the duration of symptoms (i.e., median of 4 days) did not differ between patients requiring himc and patients requiring limc. thirty-one subjects required subsequent admission to icu due to worsening of pneumonia and gas exchange. on admission, these patients displayed extensive radiological impairment in terms of both ggo score and consolidation. in the overall population gas exchange parameters correlate significantly with radiological scores but, interestingly enough, this correlation was mainly due to patients who remained in the limc group. indeed, in this group, radiological score correlated negatively with ppo 2 levels and positively with fio 2 reflecting exact correspondence between respiratory failure and radiologic impairment. conversely, among patients who subsequently required himc, cxr at baseline showed a variety of radiologic impairment, ranging from normal to highly abnormal however without a concurrent relation with gas abnormality. this result might arise attention to that patients who display discrepancies between gas exchange parameters and cxr. pevious reports on cxr findings in covid-19 patients focused on the distribution and type of lung abnormalities. wong and coauthors demonstrated that cxr at baseline has a sensitivity of 69% for a diagnosis of covid-19 pneumonia, corroborating the utility of cxr in the initial evaluation of subjects with suspected covid-19 pneumonia, thus obviating the need for ct [6] . toussie and colleagues have recently reported that initial cxr severity score is also an independent predictor of outcome in covid-19 patients [3] . we could not replicate this finding, but our study population was older than that studied by toussie et al. the prognostic role of cxr in covid-19 pneumonia therefore needs to be clarified in larger studies. lung ultrasound has been suggested as a potential diagnostic tool for covid-19 pneumonia given the predominant involvement of the lung periphery [7] ; lung ultrasound is a relatively simple technique that can be easily applied at patient bedside [25] . in our study, we investigated its role in the late phase of covid-19 pneumonia and its relation with cxr in a subgroup of patients hospitalized in a low-intensity care setting. we found a significant correlation between lus features and fio 2 level, suggesting these two parameters can be integrated into the evaluation of patients with covid-19 pneumonia. lus global score positively correlated with cxr consolidation score while the correlation with cxr global score only trended towards statistical significance. although only exploratory, these findings may anticipate further studies mainly focused on the utility of lus as a monitoring tool, possibly limiting the use of serial cxr, at least in the advanced phase of covid-19 pneumonia. in this regard, lus has been suggested as a potential substitute for cxr in the follow-up of various lung diseases in icu [26] , reducing the number of cxrs performed and relative medical costs without affecting patient outcome. of interest in a recent study by møller-sørensen and colleagues, the usefulness of bed-side lus in icu patients treated with extracorporeal membrane oxygenation (ecmo) was assessed during the covid-19 pandemic. authors used a three-zone score for each lung (anterior, posterior and lateral) with a maximum of 102 points for patient. lus score demonstrated a strong correlation with compliance during mechanical ventilation. moreover, a lower lus score advanced weaning capacity from ecmo [27] . soldati and colleagues have also suggested that lus can be useful in covid-19 pneumonia by identifying disease extension and specific patterns, as well as their evolution toward the consolidation phase [28] , thus providing further support to the role of lus in the follow-up of patients with covid-19 pneumonia. at present, however, the majority of studies performed during the covid-19 pandemic focused on ultrasonographic signs and disease patterns at presentation rather than overtime [29] [30] [31] [32] [33] . accordingly, the role of lus in monitoring the evolution of covid-19 pneumonia needs to be confirmed in larger studies. the results of our study should be interpreted in the light of important limitations. first, this is not a longitudinal study and we retrospectively collected all clinical and radiological data; therefore, the accuracy of the clinical information depends on medical records, which may introduce inaccuracies. however, every effort was made to limit this risk, even asking to the patients to fill all the missing data when possible. second, the study population was relatively small, particularly the subset of patients for whom lus data were available, although this was an exploratory analysis, and its findings should be viewed as such. clearly, these data need to be validated in larger, independent, prospectively collected populations of patients. in summary, our study identified clinical features that strongly predict the level of medical setting required by patients with covid-19 pneumonia (himc or limc). these findings allow the identification of patients at risk for severe disease and worse outcome already on hospital admission. the correlation of lus with clinical parameters and radiological score provides the basis for future studies on the utility of lus in the follow-up of patients with covid-19 pneumonia. conflicts of interest: p.s. has received personal fees and non-financial support from roche, boehringer-ingelheim, zambon, and ppm services. e.b. has received personal fees from roche and boehringer-ingelheim. m.s. has received research grants for the department (not personal) to her institution from takeda ltd., chiesi farmaceutici and laboratori guidotti spa. these funds were not used to support this project. coronavirus suppression of a sars-cov-2 outbreak in the italian municipality of vo' review of chest radiograph findings of covid-19 pneumonia and suggested reporting language clinical and chest radiography features determine patient outcomes in young and middle age adults with covid-19 presenting cxr phenotype of h1n1. flu compared with contemporaneous non-h1n1, community acquired pneumonia, during pandemic and post-pandemic outbreaks' covid-19) infection: findings and correlation with clinical outcome frequency and distribution of chest radiographic findings in covid-19 positive patients authors the role of chest imaging in patient management during the covid-19 pandemic lung ultrasound in diagnosing pneumonia in the emergency department: a systematic review and meta-analysis why, when, and how to use lung ultrasound during the covid-19 pandemic: enthusiasm and caution clarifying the role of lung ultrasonography in covid-19 respiratory disease chest radiography versus lung ultrasound for identification of acute respiratory distress syndrome: a retrospective observational study covid-19 outbreak in italy: experimental chest x-ray scoring system for quantifying and monitoring disease progression chest radiological findings of influenza a h1n1 pneumonia validation of a composed covid-19 chest x-ray score: the care project point-of-care lung ultrasound in patients with covid-19 -a narrative review epidemiological and clinical characteristics of 99 cases of 2019-novel coronavirus (2019-ncov) pneumonia in wuhan clinical course and outcomes of critically ill patients with sars-cov-2 pneumonia in wuhan, china: a single-centered, retrospective, observational study covid-19 and smoking: a systematic review of the evidence cigarette smoke exposure and inflammatory signaling increase the expression of the sars-cov-2 receptor ace2 in the respiratory tract chronic obstructive pulmonary disease severity and cardiovascular outcomes association between chronic obstructive pulmonary disease and systemic inflammation: a systematic review and a meta-analysis prevalence of comorbidities and its effects in coronavirus disease 2019 patients: a systematic review and meta-analysis comorbidities, cardiovascular therapies and covid-19 mortality: a nationwide, italian observational study (italico). front. cardiovasc the use of lung ultrasound images for the differential diagnosis of pulmonary and cardiac interstitial pathology could the use of bedside lung ultrasound reduce the number of chest x-rays in the intensive care unit? cardiovasc covid-19 assessment with bedside lung ultrasound in a population of intensive care patients treated with mechanical ventilation and ecmo is there a role for lung ultrasound during the covid -19 pandemic? sonographic signs and patterns of covid-19 pneumonia a preliminary study on the ultrasonic manifestations of peripulmonary lesions of non-critical novel coronavirus pneumonia (covid-19) diagnostic accuracy of lung ultrasonography combined with procalcitonin for the diagnosis of pneumonia: a pilot study can lung us help critical care clinicians in the early diagnosis of novel coronavirus (covid-19) pneumonia? findings of lung ultrasonography of novel corona virus pneumonia during the 2019-2020 epidemic this article is an open access article distributed under the terms and conditions of the creative commons attribution (cc by) license key: cord-269014-ck27fm58 authors: vo, luan nguyen quang; codlin, andrew james; huynh, huy ba; mai, thuy doan to; forse, rachel jeanette; truong, vinh van; dang, ha minh thi; nguyen, bang duc; nguyen, lan huu; nguyen, tuan dinh; nguyen, hoa binh; nguyen, nhung viet; caws, maxine; lonnroth, knut; creswell, jacob title: enhanced private sector engagement for tuberculosis diagnosis and reporting through an intermediary agency in ho chi minh city, viet nam date: 2020-09-14 journal: trop med infect dis doi: 10.3390/tropicalmed5030143 sha: doc_id: 269014 cord_uid: ck27fm58 under-detection and -reporting in the private sector constitute a major barrier in viet nam’s fight to end tuberculosis (tb). effective private-sector engagement requires innovative approaches. we established an intermediary agency that incentivized private providers in two districts of ho chi minh city to refer persons with presumptive tb and share data of unreported tb treatment from july 2017 to march 2019. we subsidized chest x-ray screening and xpert mtb/rif testing, and supported test logistics, recording, and reporting. among 393 participating private providers, 32.1% (126/393) referred at least one symptomatic person, and 3.6% (14/393) reported tb patients treated in their practice. in total, the study identified 1203 people with tb through private provider engagement. of these, 7.6% (91/1203) were referred for treatment in government facilities. the referrals led to a post-intervention increase of +8.5% in all forms tb notifications in the intervention districts. the remaining 92.4% (1112/1203) of identified people with tb elected private-sector treatment and were not notified to the ntp. had this private tb treatment been included in official notifications, the increase in all forms tb notifications would have been +68.3%. our evaluation showed that an intermediary agency model can potentially engage private providers in viet nam to notify many people with tb who are not being captured by the current system. this could have a substantial impact on transparency into disease burden and contribute significantly to the progress towards ending tb. tuberculosis (tb) is a curable disease, yet an estimated 10 million people develop active tb and 1.5 million people succumb to tb each year [1] . it remains the deadliest disease caused by a single promising results in multiple sites throughout india [28, 29] and has been recognized as one avenue of sustainably scaling private sector engagement for tb worldwide [30] . in 2017, friends for international tb relief piloted a private-sector engagement initiative called proper care private sector (pcps), modeled after the successful ppia pilots from india [27] . this pilot investigated the feasibility of building a portfolio of private providers and measured the outputs of incentivizing and supporting referral and reporting of private tb treatment. this pilot was conducted in two districts of ho chi minh city (hcmc), viet nam-district 10 and go vap-between july 2017 and march 2019. the intervention area had a combined population of 1.2 million people and notified 1070 people with all forms of tb in the 12 months preceding the study. in each district, there is a district tb unit (dtu) responsible for managing diagnosis, treatment and notification of tb according to ntp guidelines and for coordinating patient management with primary health facilities. there were no official private sector tb-reporting entities in the evaluation area before this study's implementation. we obtained lists of licensed private healthcare providers from each intervention district's regulatory authority. these providers included pharmacies, single-doctor practices and multi-doctor clinics. in collaboration with licensing, health, and tb authorities, through consensus we conducted a mapping exercise to identify priority providers with a high likelihood of encountering people who had pulmonary tb, while categorically excluding certain specialists, such as dermatologists, obstetricians, and gynecologists. through repeated in-person and telephonic engagement, we recruited eligible providers. interested providers were invited to capacity building events organized in collaboration with the pham ngoc thach provincial lung hospital (pnt). the scope of these training events included new diagnostic tests for tb and specifically xpert mtb/rif (xpert) and the newly recommended mtb/rif ultra assay [31] , standardized tb treatment regimens, and follow-up schedules according to ntp guidelines. we complemented these formal training events with one-on-one provider detailing activities [32] to elaborate on the study's procedures, the provider's role and responsibilities, and the benefits of participation. providers were eligible to participate through two principal strategies: diagnostic referral and private tb treatment reporting ( figure 1 ). in this strategy, participating providers verbally screened their customers for tb symptoms and distributed referral vouchers to anyone reporting at least one tb symptom, i.e., (productive) cough or hemoptysis, weight/appetite loss, fatigue, fever, night sweats, chest pain, dyspnea. symptomatic persons could use the voucher to access a chest x-ray (cxr) subsidy of vnd 50,000 (usd 2.20 at an exchange rate of vnd 22,700 = usd 1) at one of the study's 12 participating radiology sites. as the cost per cxr charged by these radiology sites ranged from vnd 80,000 (usd 3.52) to vnd 120,000 (usd 5.29), the radiography site collected the balance payment from the health-seeking person. in figure 1 . schematic of the two private sector engagement strategies; the grey boxes show in which parts of the tuberculosis (tb) care cascade private providers were engaged by the study. in this strategy, participating providers verbally screened their customers for tb symptoms and distributed referral vouchers to anyone reporting at least one tb symptom, i.e., (productive) cough or hemoptysis, weight/appetite loss, fatigue, fever, night sweats, chest pain, dyspnea. symptomatic persons could use the voucher to access a chest x-ray (cxr) subsidy of vnd 50,000 (usd 2.20 at an exchange rate of vnd 22,700 = usd 1) at one of the study's 12 participating radiology sites. as the cost per cxr charged by these radiology sites ranged from vnd 80,000 (usd 3.52) to vnd 120,000 (usd 5.29), the radiography site collected the balance payment from the health-seeking person. in comparison, the price for one cxr at the district tb unit was vnd 49,000 (usd 2.16) at the start of the study and was subsequently raised to vnd 69,000 (usd 3.04). patients who elected to take their tb treatment with a private provider were charged a consultation fee of between vnd 80,000 and vnd 150,000 (usd 6.61) in addition to drugs and other services. according to field staff estimates, the approximate average cost per visit per person at private facilities was vnd 200,000 (usd 8.81). persons assessed with parenchymal abnormalities on cxr by the x-ray technician and verified by the attending radiologist at the radiography site provided a sputum sample for free follow-on testing with the xpert assay. at selected sites, health-seeking persons also underwent smear microscopy, in which case these results were requested from the participating provider as well. sputum was collected at the radiography site or by the referring private provider. study staff collected sputum specimens for transport to a designated government xpert laboratory in go vap district. people with xpert-positive results were encouraged to take treatment at their closest dtu, or at pnt if their xpert result showed rifampicin resistance. when an individual was diagnosed and treated for tb via this strategy, the private provider making the initial referral received a vnd 500,000 (usd 22.07) payment or approximately 2.5x the estimated average cost per visit per person. if the person chose to take tb treatment with a private provider, the treatment was recorded through the study's second strategy. the second strategy focused on documenting private tb treatment practices. once a month, study staff collected tb treatment information from participating private providers. this information included individuals diagnosed through the diagnostic referral strategy above that elected treatment outside of the ntp. providers were paid vnd 500,000 (usd 22.07) for each complete patient report, which included the patient's name, age, sex, address, cxr results, sputum test results (xpert, smear, culture, other), type of tb (pulmonary, extra-pulmonary), treatment regimen, and initiation dates. treatment outcomes were not systematically assessed in this pilot study due to resource limitations and data provided by providers were sparse as providers did not conduct post-treatment follow-up with patients. despite the attempts to characterize these treatment reports in detail, they were not recognized by the ntp for official notification for several reasons. the primary reason was that these providers were not registered as official ppm model 4 participants in accordance to 02/2013/tt-byt and therefore had not undergone required capacity building and site assessment by the ntp. we tabulated descriptive statistics for private provider engagement and participation, the number and proportion of referred people progressing through the study's tb care cascade by intervention district and the private tb treatment reported to our study. we calculated the ratio of bacteriologic confirmation over the number of successful cxr referrals. official tb notifications were collected from the two intervention districts for three years prior to the study and during the study period to analyze trends of official tb notifications before and during the pilot. additional notifications and percent change from baseline were calculated using a pre-/post-intervention comparison of official notification data in the intervention districts. due to barriers outlined above, the collected private tb treatment cases were not included in the official ntp notification statistics, so that a second additionality model was constructed to assess the impact of including these privately treated individuals in official tb statistics for the intervention districts. statistical analyses were performed on stata version 13 (statacorp, college station, tx, usa). the institutional review boards of pham ngoc thach hospital (155/nckh-pnt) and the hanoi school of public health (324/2019/ytcc-hd3) granted scientific and ethical approval for this study. the ho chi minh city provincial people's committee approved the implementation of the intervention (4699/qd-ubnd). participating private providers granted permission to use data for the analyses based on the terms and conditions of their practice. all personally identifying information was removed prior to analysis. the study enumerated 1107 licensed private providers in the two intervention districts (table 1) . of these, 67.0% (742/1107) were targeted for recruitment based on the initial mapping exercise and 53.0% (393/742) of those targeted agreed to participate. among participants, at least one staff member of 48.6% of centers (191/393) attended a capacity building event. by the end of the study, we recorded at least one referral for cxr from 32.1% (126/393). of the 126 private providers with at least one successful cxr referral (table 2) , 58.7% were multi-doctor clinics and 25.4% were single-doctor practices. these two provider types accounted for 70.0% and 18.9% of referrals, respectively. the remaining referrals were from pharmacies, hospitals or could not be traced to the source. the bacteriologic positivity rate among successful cxr referrals was highest among single pulmonologist practices at 58.6%, followed by multi-doctor clinics at 21.9% and single-doctor practices with no specialty focus at 11.2%. eighty-two point two percent of the people diagnosed with tb via the diagnostic referral strategy were referred by just ten private providers constituting 7.9% (10/126) of those making at least one successful cxr referral and 2.5% (10/393) of those signing participation agreements. the study received tb diagnosis and treatment data from 3.6% (14/393) of participating private providers. these consisted of 71.4% (10/14) single-doctor practices and 28.6% (4/14) multi-doctor clinics. the top five providers supplying tb diagnosis and treatment data reported 81.7% (907/1112) of patients on private tb treatment. the study identified 1203 people with tb of whom 7.6% (91/1203) were referred and linked to care with the ntp (figure 2 ), while 92.4% (1112/1203) consisted of private tb treatment reports and remained un-notified (table 3) . all 91 tb patients linked to care with the ntp were bacteriologically confirmed. among persons treated in the private sector, the proportion with bacteriologic confirmation was 30.5% (339/1112). together, the total proportion of tb patients with bacteriologic confirmation was 35.7% (430/1203). overall, 1.2% (15/1203) were people with multi-drug resistant tb (mdr-tb). patients diagnosed with rifampicin resistance were largely referred by private providers to ntp facilities. particularly, diagnostic referrals generated 93.3% (14/15) of persons detected with rifampicin resistance ( figure 2 ). meanwhile, private tb treatment reports included one mdr-tb case (table 3 ). in addition to persons treated for active tb, four persons were treated for latent tb infection by private providers. (figure 2 ). meanwhile, private tb treatment reports included one mdr-tb case ( table 3 ). in addition to persons treated for active tb, four persons were treated for latent tb infection by private providers. the results of the study's diagnostic referral strategy are in figure 2 . the 12 radiology centers recorded 4984 cxr results, of which 817 were abnormal (16.4% of those with cxr results). sputum specimens were collected from 65.4% (534/817) of these individuals and tested on the xpert assay with a positivity of 25.8% (138/534) including 14 individuals with rifampicin-resistant tb (14/138 = 10.1%). an additional 528 smear microscopy tests were conducted for individuals who did not get a cxr or presented no radiographic abnormalities suggestive of tb but still reported tb symptoms, resulting in the detection of 31 (31/528 = 5.9%) people with smear-positive tb. of the total 169 people diagnosed with bacteriologically-confirmed tb, 95.9% (162/169) were linked to care, corresponding to a ratio of 3.2% among successfully referred persons with a cxr screen. among patients linked to care, 56.2% (91/162) were initiated on treatment at a ntp facility, while 43.8% (71/162) elected to take treatment with the initially referring private provider. these patients are included in the private tb treatment reports. the characteristics of the privately-treated, un-notified 1112 individuals are in table 3 the results of the study's diagnostic referral strategy are in figure 2 . the 12 radiology centers recorded 4984 cxr results, of which 817 were abnormal (16.4% of those with cxr results). sputum specimens were collected from 65.4% (534/817) of these individuals and tested on the xpert assay with a positivity of 25.8% (138/534) including 14 individuals with rifampicin-resistant tb (14/138 = 10.1%). an additional 528 smear microscopy tests were conducted for individuals who did not get a cxr or presented no radiographic abnormalities suggestive of tb but still reported tb symptoms, resulting in the detection of 31 (31/528 = 5.9%) people with smear-positive tb. of the total 169 people diagnosed with bacteriologically-confirmed tb, 95.9% (162/169) were linked to care, corresponding to a ratio of 3.2% among successfully referred persons with a cxr screen. among patients linked to care, 56.2% (91/162) were initiated on treatment at a ntp facility, while 43.8% (71/162) elected to take treatment with the initially referring private provider. these patients are included in the private tb treatment reports. the characteristics of the privately-treated, un-notified 1112 individuals are in table 3 . of these, 30.5% (339/1112) had either a positive smear microscopy, xpert, and/or culture result. just 29.0% (322/1112) of those taking private tb treatment lived inside the study's intervention area, with another 41.1% (455/1112) living in one of hcmc's other 22 districts. about 28.7% (319/1112) of privately treated persons were registered residents of other provinces, while the remaining 1.3% (14/1112) of people had no documented address. overall, 68.3% (759/1112) of people privately treated for tb were prescribed a standard first-line regimen as per ntp guidelines, while the records for another 27.6% (307/1112) of people showed the correct drugs but were modified from the standard regimen or missing information on duration. three percent (33/1112) of treatments included streptomycin, and 1.1% (12/1112) included levofloxacin. table 4 and figure 3 summarize changes in the ntp's tb case notifications in the study's intervention area and present the modeled impact of including private tb treatment on official notification statistics. bacteriologically-confirmed and all forms tb notifications increased by +17.0% (+177 tb cases) and +8.5% (+158 tb cases), respectively, over six quarters of implementation. if private tb treatment had been eligible for inclusion in the official notification statistics, bacteriologically-confirmed and all forms of tb notifications would have increased by +49.7% (+516 tb cases) and +68.3% (+1270 tb cases), respectively. our pilot study showed that the ppia model was effective in engaging a large number of private providers in the vietnamese urban setting to contribute to tb care and prevention efforts. we found a substantial number of persons treated for tb in the private sector of hcmc, the vast majority of whom were not known to the ntp. this indicates that creating enabling mechanisms, as well as our pilot study showed that the ppia model was effective in engaging a large number of private providers in the vietnamese urban setting to contribute to tb care and prevention efforts. we found a substantial number of persons treated for tb in the private sector of hcmc, the vast majority of whom were not known to the ntp. this indicates that creating enabling mechanisms, as well as further scale-up and evaluation of private tb treatment reporting approaches, should be a critical component of the tb response in viet nam's urban areas. numerous studies have shown that effective engagement of private providers to screen for tb and refer presumptive cases for diagnostic testing can be an efficient way to close the detection gap [33] [34] [35] . this was corroborated by the results of our study and particularly by the increase in all forms tb notifications compared to the baseline period. moreover, this share of private provider contribution to notifications (+8.5%) was over five times viet nam's 2017 national average private sector contribution rate (1727/105,733 = 1.6%) [36] . lastly, and perhaps most telling, un-notified private tb treatment reports corresponded to about 70% of the officially notified patient load in these two districts managed by the ntp. even though these districts are not representative of the average district in viet nam, they present a compelling argument to expand novel private provider engagement models in the country's urban areas. meanwhile, the efficiency of this approach was evidenced by the high ratio of positively detected cases among those successfully referred. this high ratio suggests a pre-screening step performed by these healthcare professionals or self-selection by patients. the high ratio consequently implies the risk of false-negative assessments and missed opportunities to engage persons with tb. therefore, more advocacy for providers and the general population to raise top-of-mind awareness about tb is warranted. as observed on our study and documented by ppm projects in other settings, a referral strategy in isolation remains limited in both novelty and impact [37] . a more comprehensive engagement strategy is required to identify tb patients accessing treatment via the private sector. including the reported private tb treatments into the ntp's routine surveillance would have represented a substantial increase in case notifications in the two study districts. however, since these providers did not complete the ntp's registration process as an accredited ppm partner, the private tb treatment records were not recognized as official notifications. the registration process is arduous and accompanied by external inspections and laborious reporting requirements, which can inhibit ppm participation for tb in viet nam [16] . this suggests the need for bold policies that promote private provider participation. this need is well-understood and has shown substantial impact in other settings once addressed [38, 39] . notification gains represent only the initial milestone. while all people with tb detected and notified through the referral strategy were bacteriologically confirmed, we observed low levels of bacteriologic confirmation among private-sector tb treatments, as only one-third was substantiated by a positive sputum test. we further observed that clinical diagnoses and follow-up testing for bacteriologically-confirmed patients oftentimes did not follow national treatment guidelines. as this study focused on case detection, treatment outcomes were optional to report and sparse when collected. private providers did not employ a systematic follow-up process but also did not permit the study to directly engage their customers for household contact investigations due to fears of reputational damages from breaching patient confidentiality. this has also been observed in other settings [35] and represents a crucial opportunity to improve quality of private-sector tb care. this is particularly the case in light of the low attendance rate on the capacity building sessions offered by the study, as they were not mandatory for study participation. consequently, while the goal of policy reform should be to remove unnecessary bureaucratic barriers to promote private provider participation, this reform should be designed with the long-term goal of improving quality of care among all stakeholders in mind. meanwhile, access to xpert testing constituted a unique selling proposition of the ppia to these providers, which they could pass on to their clientele. this study was the first to enable commercial access to xpert testing for non-ppm providers in viet nam, so that the consistent message across size and geography of providers was that the ability to offer naat to their clients was a critical catalyst for participation. while this dynamic may be a temporary effect until market access is established through registration and formalization of a commercial distribution channel, intermediary agencies in other settings should leverage these dynamics to build the private provider network. increased acceptance of xpert testing has also been observed to result in a reduction of clinical diagnosis [40] , so that increasing private-sector xpert uptake could substantially reduce the rate of over-diagnosis and contribute to improved individual and public health outcomes. efforts to optimize naat access have proven effective in several settings through the initiative for promoting affordable, quality tb tests [24, 38, 41] . an important lesson across both strategies was the need to sufficiently power monetary and non-monetary incentives. evidence suggests that referral and notification incentives can represent a welcome income generation opportunity [42, 43] . however, determining the appropriate threshold at which the individual cost-benefit analysis turns favorable is critical. the level of usd 22.07 proved sufficient to elicit private tb treatment reports among some, but it is safe to say that the 14 reporting providers in our study did not constitute the entire spectrum of private tb treatment. for example, risk-averse providers and those with a small caseload may have found the incentive to be insufficient to offset the risk exposure and expected value of penalties of un-notified tb treatment. these incentives may have also created inefficiencies whereby pulmonologists referred persons with tb through our study that would also have been referred in our absence as this level of incentive was high compared to traditionally paid amounts in viet nam [19, 44, 45] . nevertheless, the costs of incentives paid by our study to detect a person with tb were a fraction of estimated total costs of detecting a new case through other systematic screening strategies [46] and warrant further optimization and evaluation. a key success factor of the study was the broad coverage and participation of a diverse set of private providers. this was evidenced by the fact that we received referrals from all types of providers listed above and detected tb cases from most provider types. this effectiveness in generating leads and detecting tb patients also suggests that we were able to target the right providers. one reason for this was likely the detailed a priori landscaping and targeting, which allows implementers to have a better sense of the options people have for care seeking and coverage of their interventions [22, 47] . our study faced several limitations. with respect to private-sector tb treatment, our study was observational in nature, so that we did not attempt to change clinical practices. similarly, we did not systematically incentivize and collect treatment outcomes in this study, but we intend to do so in future engagements. as such, provider willingness to alter behavior to meet international standards of tb care and the extent to which previously mentioned aspiration of improving diagnostic and treatment quality are feasible remain critical research questions to be answered on future studies. another limitation was that we were only able to verify private tb treatment through reviews and abstractions of data, which were only available in patient records, as private providers did not permit direct engagement of their customers. the study's implementation area was limited, so that it is necessary to test the model at a greater scale to strengthen the generalizability of these results. lastly, it also remains unclear, if this model or an adaptation thereof were appropriate in non-urban areas. nevertheless, this pilot study has elucidated the potential gains inherent in effective private sector engagement to national and provincial stakeholders in viet nam. as has been noted elsewhere, future work should focus on strengthening data systems, including the use of direct electronic data capture to track referrals and loss to follow up between referral and cxr [38] . this work should also employ mechanisms to verify that private tb treatment reports are genuine individuals who have not already been reported elsewhere in the tb notification system. finally, policy changes are required to facilitate the scale-up of this approach. private providers in hcmc are treating many people with tb who are not reported to the national program, and it is critical to improve engagement approaches that arrive at a system, which allows private providers to notify through the ntp. to achieve public health targets, this system will also need to ensure the highest level of care adherent to national standards. scaling effective private-sector engagement efforts, such as this enhanced intermediary model, could have a strong impact on the progress towards ending tb, and we recommend the ntp to scale up the model and through it to build capacity for improvements in quality of tb diagnosis and care. potential impact of the covid-19 response on tuberculosis in high-burden countries: a modelling analysis the global burden of latent tuberculosis infection: a re-estimation using mathematical modelling approval of the national strategy for tb prevention and control until 2020 with vision to the second national tuberculosis prevalence survey in vietnam changes in the health care system of vietnam in response to the emerging market economy vietnam's health care system: a macroeconomic perspective perspective how could private healthcare better contribute to healthcare coverage in vietnam ? risks and benefits of private health care: exploring physicians' views on private health care in ho chi minh city national survey of tuberculosis prevalence in viet nam private pharmacies and tuberculosis control: a survey of case detection skills and reported anti-tuberculosis drug dispensing in private pharmacies in ho chi minh city diagnosis and treatment of tuberculosis in the private sector viet nam national assembly. law on the management and prevention of infectious diseases world health organization. first meeting of the public-private mix subgroup for dots expansion. public-private mix for dots expansion; world health organization circular about the decision on coordination between health care providers in the management of tb [vietnamese]; 02/2013/tt-byt; ministry of health progress and plans viet nam. ppm progress and plans viet nam; world health organization: kuala lumpur private tuberculosis care provision associated with poor treatment outcome: comparative study of a semi-private lung clinic and the ntp in two urban districts in ho chi minh city public-private mix for improved tb control in ho chi minh city, vietnam: an assessment of its impact on case detection progress and plans for ppm in the western pacific region delay and discontinuity-a survey of tb patients' search of a diagnosis in a diversified health care system map, know dynamics and act; a better way to engage private health sector in tb management scale and ambition in the engagement of private providers for tuberculosis care and prevention enhancing the role of private practitioners in tuberculosis prevention and care activities in india altruistic capital improving tuberculosis control through public-private collaboration in india: literature review private sector engagement for tb control in mumbai-private provider interface agency (ppia) universal access to tb care: ppia patna. in south asian public-private initiative experience sharing; world health partners tapping private health sector for public health program? findings of a novel intervention to tackle tb in mumbai achieving systemic and scalable private sector engagement in tuberculosis care and prevention in asia meeting report of a technical expert consultation: non-inferiority analysis of xpert mtb / rif ultra compared to xpert mtb / rif; world health organization the impact of a physician detailing and sampling program for generic atorvastatin: an interrupted time series analysis hard gains through soft contracts: productive engagement of private providers in tuberculosis control public-private mix for dots implementation: what makes it work? an evaluation of systematic tuberculosis screening at private facilities in karachi missing tuberculosis patients in the private sector: business as usual will not deliver results. public health action public-private mix for tuberculosis care and prevention. what progress? what prospects? introducing new tuberculosis diagnostics: the impact of xpert mtb/rif testing on case notifications in nepal initiative for promoting affordable and quality tuberculosis tests (ipaqt): a market-shaping intervention in india does cash incentive effect tb case notification by public private mix-general practitioners model in pakistan? engaging the private sector to increase tuberculosis case detection: an impact evaluation study revised national tuberculosis control programme. national strategic plan for tuberculosis elimination what would it cost to scale-up private sector engagement efforts for tuberculosis care? evidence from three pilot programs in india how much is tuberculosis screening worth? estimating the value of active case finding for tuberculosis in south africa, china, and india guide to develop a national action plan on public-private mix for tuberculosis prevention and care the authors express their sincere gratitude to the viet nam national tuberculosis control programme, the pham ngoc thach hospital and the staff working at the district tb units in the study's intervention areas (district 10 and go vap) for their participation. the authors also wish to thank giang t. le, thanh n. vu and the ho chi minh city public health association and all participating private providers. the authors declare no conflict of interest. the funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. key: cord-346942-88l03lf0 authors: kerpel, ariel; apter, sara; nissan, noam; houri-levi, esther; klug, maximiliano; amit, sharon; konen, eli; marom, edith m. title: diagnostic and prognostic value of chest radiographs for covid-19 at presentation date: 2020-08-17 journal: west j emerg med doi: 10.5811/westjem.2020.7.48842 sha: doc_id: 346942 cord_uid: 88l03lf0 introduction: pulmonary opacities in covid-19 increase throughout the illness and peak after ten days. the radiological literature mainly focuses on ct findings. the purpose of this study was to assess the diagnostic and prognostic value of chest radiographs (cxr) for coronavirus disease 2019 (covid-19) at presentation. methods: we retrospectively identified consecutive reverse transcription polymerase reaction-confirmed covid-19 patients (n = 104, 75% men) and patients (n = 75, 51% men) with repeated negative severe acute respiratory syndrome coronavirus 2 (sars-cov-2) tests. two radiologists blindly and independently reviewed the cxrs, documented findings, assigned radiographic assessment of lung edema (rale) scores, and predicted the patients’ covid-19 status. we calculated interobserver reliability. the score use for diagnosis and prognosis of covid-19 was evaluated with the area under the receiver operating characteristic curve. results: the overall rale score failed to identify covid-19 patients at presentation. however, the score was inversely correlated with a covid-19 diagnosis within ≤2 days, and a positive correlation was found six days after symptom onset.interobserver agreement with regard to separating normal from abnormal cxrs was moderate (k = 0.408) with low specificity (25% and 27%). definite pleural effusion had almost perfect agreement (k = 0.833) and substantially reduced the odds of a covid-19 diagnosis. disease distribution and experts’ opinion on covid-19 status had only fair interobserver agreement. the rale score interobserver reliability was moderate to good (intraclass correlation coefficient = 0.745). a high rale score predicted a poor outcome (intensive care unit hospitalization, intubation, or death) in covid-19 patients; a score of ≥5 substantially increased the odds of having a poor outcome. conclusion: chest radiography was found not to be a valid diagnostic tool for covid-19, as normal or near-normal cxrs are more likely early in the disease course. pleural effusions at presentation suggest a diagnosis other than covid-19. more extensive lung opacities at presentation are associated with poor outcome in covid-19 patients. thus, patients with more than minimal opacities should be monitored closely for clinical deterioration. this clinical application of cxr is its greatest strength in covid-19 as it impacts patient care. coronavirus disease 2019 (covid-19) is spreading globally. 1 the world health organization (who) declared covid-19 a pandemic on march 11, 2020. 2 the most common presenting clinical symptoms are fever, cough, dyspnea, myalgia, and fatigue. [3] [4] [5] older age and medical comorbidities are linked to more severe disease. 4, [6] [7] [8] men are over-represented among covid-19 patients. 3, 4, 6, 7 although the radiological literature mainly focuses on computed tomography (ct) findings, 9, 10 many patients are imaged solely with chest radiography 10, 11 primarily as an adjunct to reverse transcription polymerase chain reaction (rt-pcr) but in some scenarios as a triage tool, 12, 13 especially in resourceconstrained environments where the supply of laboratory pcr kits cannot meet the demand. although there are nonspecific respiratory symptoms commonly observed in covid-19 patients at presentation, some patients with covid-19 do not present with these classic clinical manifestations, which further complicates triage and diagnosis. 4 the chest radiograph (cxr) was reported as having a sensitivity of 69% for covid-19 in one study of 64 patients. 9 in that study, the common findings were bilateral peripheral opacities with a predilection to the lower lung zones. opacities increased throughout the illness, with a peak in severity at 10-12 days after symptom onset; this was shown by documenting lung opacities using a simplified radiographic assessment of lung edema (rale) score. 9, 14 when the fleischner society consensus statement was created, which specified that chest radiography has little value early in the course of the disease, there were limited data available on the accuracy of chest radiography for the diagnosis of covid-19. 13 data on the strengths and weaknesses of chest radiography for the diagnosis of covid-19 are important, as cxrs are the most commonly used triage imaging tool in any patient presenting with respiratory symptoms. 12 this is especially important because experts suggest that the second wave of coronavirus is likely to be even more devastating. 15 our aim was to assess the diagnostic accuracy and reliability of cxrs in patients suspected of having covid-19 at presentation to the emergency department (ed) and to assess the prognostic value of the rale score in patients with covid-19. this retrospective study was approved by our institutional review board, and informed consent was waived. we identified our study population by extracting severe acute respiratory syndrome coronavirus 2 (sars-cov-2) rt-pcr test results (positive or negative) of nasopharyngeal swabs from all consecutive patients older than 18 years analyzed at our hospital's laboratory from the ed from march 6-31, 2020, who had a cxr at presentation (within 24 hours of the first rt-pcr). we extracted data by a database search (query) using the mdclone platform (mdclone ltd, be'er sheva, israel), a big data system for healthcare. we were granted access to the raw data in order to validate the quality and reliability of the information in the database source underlying the study. of the rt-pcr test kits used, 90% (161/179) were allplex 2019-ncov assay kits (seegene inc. seoul, korea), and 10% (18/179) were kits produced in our hospital laboratory. the patients were then divided into two groups: those who had covid-19 and those who did not. the former group comprised patients who had a positive rt-pcr test. the latter, control group comprised patients who had a negative rt-pcr result on at least two separate occasions, more than 24 hours apart (without a previous positive test result). this methodology is similar to that of previously published studies, 16 as we tried to avoid the imperfect gold standard bias. we excluded patients who underwent sars-cov-2 testing due to an abnormal cxr and not due to clinical suspicion (n = 1 positive, n = 3 negative) based on the patients' electronic health records (ehr) (figure 1 ) to avoid partial verification bias (referral bias). 17 the patients' ehrs were reviewed to obtain demographics and clinical data. the primary outcomes were intensive care unit (icu) hospitalization, intubation, and mortality. covid-19 severity was classified as severe or non-severe based on respiratory distress (≥30 breaths per minute) or oxygen saturation ≤93% on room air. 18 although lung opacities are included in some published severity criteria, we did not use cxr findings to define severity to avoid incorporation bias. 17 the data cutoff date was april 21, 2020. we extracted the overall number of ed visits at our hospital during the study period using the mdclone platform database search. overall covid-19 new cases in israel for the study period (26 days), and for an equal time span before and after the study period, were extracted from israel's ministry of health website. 19 cxrs were acquired as computed radiographs (n = 127) or digital radiographs (n = 52) from multiple vendors. the projections were posterior-anterior (pa) (n = 108), and anterior-posterior (ap) (n=71). two radiologists (emm, a thoracic radiologist with 28 years of experience, and sa, an oncology imaging radiologist with 40 years of experience) independently reviewed all cxrs using a communication system search (pacs), carestream, pacs vue v12.1.5 (carestream health, inc, rochester, ny), while blinded to the rt-pcr results and clinical data. the cxrs of covid-19 patients and the control patients were in random order. both readers recorded pulmonary opacity characteristics, including their distribution (peripheral, perihilar or diffuse), zonal predominance (upper, lower, or equal), and laterality (bilateral or unilateral). pleural effusion presence was recorded. disagreements between reader 1 (r1) and reader 2 (r2) regarding the categorization of a pleural effusion as definite or questionable were resolved by an independent and blinded third reader (ek, a cardiothoracic radiologist with 21 years of experience). r1 and r2 calculated the rale scores 14 ( figure 2 ). the rale score, which is used to quantitate lung opacities, 14 is calculated by dividing each radiograph into quadrants and multiplying the extent (0 = no involvement, 1 = <25%, 2 = 25-50%, 3 = 50-75%, 4 = >75%) by the density (1 = hazy, 2 = moderate, 3 = dense) for each quadrant and then summing them (maximum score = 48). 14 for the purposes of our study, the following density definitions were used: hazy, ranging from barely noticeable opacities to mild or veiling opacities, through which the lung vessels can be clearly seen; moderate, in which opacities are identified, but the blood vessels are still visible; and dense, in which consolidation is apparent, and the blood vessels are not visible. for rale scoring, we excluded cxrs with one of the following overshadowing radiopaque abnormalities: pleural effusion; pleural plaques; and pulmonary nodules or masses, whether due to lung cancer or metastatic disease. finally, the readers gave their expert opinion regarding patient covid-19 status based on imaging alone. all previous imaging tests were available to the readers for comparison, and any changes were recorded. to evaluate the sensitivity and specificity of categorical variables to discriminate between patients with and without covid-19, assuming sensitivity and specificity of 80% and a 95% confidence interval (ci) of 0.2, 140 patients were needed. to evaluate the use of rale for determining covid-19 diagnosis using the area under the receiver operating characteristic curve (auc), assuming an area of 0.8 with a 95% ci width of 0.2 and an equal number of participants with and without covid-19, 78 participants were needed. we assumed that the mean rale score for patients without poor prognosis was 2, with a mean score of 10 for patients with poor outcomes. we assumed that the standard deviation of the rale score was 8 (range 0-48, divided by six). using a significance level of 5% and power of 80%, and assuming a proportion of patients having poor outcomes to be 20%, a total of 53 patients were needed. we evaluated continuous variables for normal distributions using histograms. variables that were close to being normally distributed are reported as the means and standard deviations (sd), while skewed variables are reported as the medians and interquartile ranges. categorical variables are reported as frequencies and percentages. we used independent samples t-tests and mann-whitney tests to compare normally distributed variables and skewed variables between groups, respectively. chi-square tests and fisher's exact tests were applied to compare categorical variables between patients with positive and negative tests. the kappa statistic was used to evaluate the agreement between readers 20 and was interpreted according to landis and koch. 21 when a kappa of 0.4 was reached, accuracy was evaluated. diagnostic accuracy parameters were calculated by crosstabulation and included the following: sensitivity, specificity, and positive (lr+) and negative (lr-) likelihood ratios. we used the intraclass correlation coefficient (icc) to evaluate the agreement of the two readers with regard to the rale score. 22 the auc 23 was used to evaluate the ability of the rale score to discriminate between covid-19 and control patients and between poor and favorable outcomes in covid-19 patients. the discriminatory ability was also evaluated in patients who presented at early (0-2 days), intermediate (3-5 days) , and late (≥6 days) time points from symptom onset. for prognostic ability, we used a rale score cutoff threshold of 5. all statistical tests were two-sided, and p<0.05 was considered statistically significant. for statistical analyses, we used spss software (ibm spss statistics for windows, ibm corp., armonk, ny). during the study period, 105 patients had positive rt-pcr results and had a cxr, and 78 patients had repeated negative results and had a cxr. after excluding patients who had the rt-pcr ordered due to an abnormal cxr (n = 1 covid-19 patient, n = 3 control patients), our study group included 104 covid-19 patients (men 78/104, 75%, mean age 57.0, sd 15.7 years) and 75 control patients (men 38/75, 51%, mean age 65.6, sd 21.4 years) ( figure 1 ). table 1 shows patient characteristics and outcomes with a comparison of covid-19 to control and non-severe to severe covid-19 patients. the overall number of ed visits at our hospital during the study period was 8025 (all causes). the number of new cases of covid-19 in israel during the study period (26 days) was 5699. the number for the period immediately preceding was 17. the number for the period immediately ensuing was 9723. these numbers show that our study took place at the beginning of the first wave of covid-19 in israel. the mortality rate in the control group was significantly higher: 27% (20/75) vs 7% (7/104) in covid-19 patients (p<0.001). heart disease and active cancer were more common in the control group. heart disease was present in 44% (n = 33/75) of the control patients compared to 16% (n = 17/104) of covid-19 patients (p<0.001). active cancer, defined as malignancy in the prior 12 months, was present in 24% (n =18/75) of the control patients compared to 4% (n = 4/104) there was no significant difference (p>0.05) in the prevalence of other comorbidities between the control and covid-19 patients: diabetes mellitus (34%, 29%); hypertension (32%, 38%); obesity (12%, 13%); dyslipidemia (35%, 29%); smoking (19%, 11%); respiratory disease (13%, 7%); and chronic renal failure (7%, 7%). most covid-19 patients underwent a pa cxr (78/104, 75%) in a dedicated radiography room of the corona section emergency department (ed), while most control group patients underwent an ap cxr (45/75, 60%) (p<0.001). among the covid-19 patients, most patients with non-severe disease had a pa cxr (64/75, 85%), while most patients with severe disease had an ap cxr (15/29, 52%) (p<0.001). the majority of both the covid-19 and control groups underwent computerized radiography (cr) (77/104, 74% and 50/75, 67%, respectively) (p=0.284). similar proportions were observed between patients with non-severe and severe disease. the identification of any opacity on cxrs had a moderate interobserver agreement (kappa = 0.408). when assuming that any parenchymal lung opacity could represent covid-19 pneumonia, the diagnostic accuracy for the diagnosis of covid-19 for both readers was sensitivity (r1-87%; r2-69%) and specificity (r1-25%; r2-27%), and both lr+ and lr-showed the poor diagnostic performance of cxrs for covid-19, as most crossed or included 1 ( when interobserver reliability did not reach a kappa of 0.4, diagnostic accuracy parameters were not calculated. . *laterality = bilateral or unilateral. **change = change from previous radiograph when comparison was available. lr+, positive likelihood ratio; lr-, negative likelihood ratio; ci, confidence interval. kerpel et al. pleural effusion had almost perfect interobserver agreement (kappa = 0.833). the accuracy parameters of the presence of a pleural effusion for the diagnosis of covid-19 were as follows: sensitivity (r1 and r2 -0.01%), specificity (r1-81%; r2-77%), and very low positive likelihood ratio (lr+) (r1-0.05; r-0.04); thus, the presence of definite pleural effusion at presentation makes the diagnosis of covid-19 very unlikely (see table 2 ). with regard to rale scoring, 103 cxrs were available in the covid-19 group after excluding one cxr due to pleural effusion, and 55 cxrs were available in the control group after excluding cxrs with the following overshadowing radiopaque abnormalities: pleural effusion (n = 17); lung cancer (n = 1); multiple metastases (n = 1); and calcified pleural plaques (n =1) (figure 1 ). the rale score interobserver reliability was moderate to good, with an icc of 0.745 (0.665 -0.806, p<0.001). see table 3 for the auc assessment summary. the auc for all patients (overall) showed no significant difference from sheer chance (r1-p = 0.010; r2-0.865). the evaluation of the discriminatory ability of the rale score in patients who presented early (0-2 days) showed an inverse correlation with covid-19 diagnosis. simply put, in patients presenting within 0-2 days of symptom onset who were clinically suspected of having covid-19, pulmonary opacities were more likely to be due to a diagnosis other than covid-19. for patients presenting within three to five days from symptom onset, only r1 achieved statistical significance, while for patients presenting more than six days from symptom onset, both readers reached significant discrimination ability. thus, for patients presenting later after symptom onset, especially from day six, the higher the rale score, the more likely a diagnosis of covid-19. an example is seen in figure 3 , showing the sensitivity of the rale score with a threshold of 5 for the diagnosis of covid-19 increasing as the patients arrive later in the disease course. see figure 4 for cxr examples of patients presenting at different timeframes from symptom onset. when the rale score was evaluated as a prognostic indicator within the covid-19 patient group, both readers had statistically significant discriminatory accuracy for severe disease and poor outcomes (table 3) . when a rale score of 5 was used as a threshold for severe disease and for poor outcome, sensitivity was moderate to good, and specificity was moderate. however, lrs were encouraging, as lr+ ranged from 2.21 to 2.59 and lr-ranged from 0.10 to 0.45 (supplemental table) . hence, a rale score <5 in covid-19 patients at presentation substantially reduces the odds of having severe covid-19 or poor outcome (intensive care unit hospitalization, intubation, or death), whereas a rale score ≥5 substantially increases those odds. in this study we assessed the diagnostic value of the initial cxr for diagnosing covid-19 in patients clinically suspected of having covid-19, as well as the prognostic value of this cxr in covid-19 patients. the study took place in a single hospital in israel at the beginning of the covid-19 pandemic first wave. our study showed that the reliability of radiographs is only moderate for any opacity and moderate to good for the rale score. overall, chest radiography was found not to be a valid diagnostic tool for covid-19. however, the diagnosis of covid-19 pneumonia by cxrs reached significant diagnostic accuracy when performed at least six days after symptom onset. for patients presenting early (0-2 days from symptom onset), a normal or near-normal cxr is more likely to be seen in a patient with covid-19, although opacities early in the disease course do not completely rule out this condition. the presence of a definite pleural effusion indicates that the diagnosis is unlikely to be covid-19. more extensive lung opacities are associated with poor outcome in covid-19 patients. previous covid-19 studies mainly concentrated on computed tomography (ct) findings and indicated that kerpel et al. diagnostic and prognostic value of chest radiographs for covid-19 at presentation is in contrast to previous studies that did not have a control group 9,11 and were only able to assess sensitivity. moreover, lrs showed the cxr is ineffective in the ed setting as it failed to meaningfully change the estimation of disease probability from pretest to posttest. this, at the very least, raises doubts about the utility of the cxr as a triage tool. it is perhaps not surprising that the quantification of pulmonary opacities, as performed in our study with the rale score, was not useful for assessing the entire cohort when trying to distinguish between patients with and without covid-19, but when interpreted in the context of time from symptom onset, the accuracy improved. opacities are usually bilateral, with a peripheral distribution and lower zones predominance. 24 we found only fair agreement with regard to the opacity predominance, distribution, and laterality, which probably relates to the lower sensitivity of cxrs compared with ct for pulmonary opacities. a previously published study reported 69% sensitivity for diagnosis on the baseline cxr, 9 similar to our findings. on the other hand, we found that this high sensitivity had a trade-off with low specificity, which represents the reader's avoidance of falsenegative results, offsetting with more false-positive results. this observation can only be made with a control group. this diagnostic and prognostic value of chest radiographs for covid-19 at presentation kerpel et al. highly experienced radiologists' expert opinions for guessing covid-19 status were not reliable and did not reach a high enough interobserver agreement to discuss the accuracy parameters. however, poor interobserver agreement regarding specific disease status on cxrs was documented in previous studies. 25, 26 despite the limited role of imaging in the diagnosis of covid-19 as expressed by leading societies worldwide, 12, 13, 27 the cxr is still the recommended imaging tool for any patient presenting at the ed with an acute respiratory illness. 28 future covid-19 patients will continue to have cxrs at presentation before their disease status is known to the referring clinicians. to complicate matters, even in the ideal setting, when rt-pcr is available and results are delivered within minutes to hours, the sensitivity of the rt-pcr for sars-cov2 is poor, 29 leaving emergency clinicians with a dilemma as to how to manage patients with non-specific presenting symptoms suggestive of covid-19 with a negative initial rt-pcr test. this dilemma emphasizes the need to maximize available knowledge in the ed setting. time from symptom onset is available data in this setting, and applying it to cxr interpretation may improve diagnostic accuracy. despite not being recommended for diagnosis of covid-19, the cxr is a tool used for the risk stratification of patients with covid-19 and is often used as an aid to decision-making with regard to discharge vs hospitalization and the amount of close monitoring needed for specific patients. 13, 18, 20 our study validates this approach and shows that the amount of pulmonary opacities, as quantified by the rale score, correlates with poor outcome. the knowledge gained from this study allows for a better understanding of the diagnostic and prognostic value of cxrs in covid-19 patients and can aid emergency physicians in clinical decision-making. the added information can also serve educators and future researchers in understanding the strengths and weaknesses of cxrs, as this "classic" imaging modality is also the most frequently performed. this study has several sources of bias. differential verification bias (double gold standard bias) 17 was present in our study, as we selected patients with only one rt-pcr test for the covid-19 group, whereas we selected only patients with two negative rt-pcr tests for the control group. lack of clinical follow-up to confirm the absence of covid-19 precluded incorporation of this patient population with only one negative test into our study. in our opinion, the bias reduced specificity, as the patients in the control group were sicker with almost four times the mortality rate and a higher prevalence of heart disease and active cancer. thus, the patients in the control group probably had more lung opacities than would be expected in the general population. similarly, spectrum bias potentially influenced our results because the control group was enriched with many "sickest of the sick," whose clinical condition influenced the decision to repeat the test and, hence, could underestimate the specificity. 17 even though this methodology is well accepted, 16 and the motivation was to ensure having only truly non-covid-19 patients in the control group, the trade-off eliminated many non-covid-19 patients who might have had less remarkable radiographs. all these biases do not impact the results regarding prognosis, as these did not relate to the control group. the study's results can be generalized to the ed setting. in a community setting, in which fewer non-covid-19 patients have competing conditions, lrs will move further away from 1, and the test will appear more useful. 31 chest radiography was found not to be a valid diagnostic tool for covid-19. however, sensitivity increased in patients presenting later in the disease course. when presenting early, a normal or near-normal cxr is more likely in covid-19. when a pleural effusion is present, the diagnosis is unlikely to be covid-19. furthermore, more extensive lung opacities at presentation are associated with poor outcome in covid-19 patients. thus, patients with more than minimal opacities should be monitored closely for clinical deterioration. this clinical application of chest radiography is its greatest strength in covid-19 as it impacts patient care. johns hopkins coronavirus resource center world health organization . who director-general's opening remarks at the media briefing on covid-19-11 clinical features of patients infected with 2019 novel coronavirus in wuhan clinical characteristics of coronavirus disease 2019 in china chest ct findings in coronavirus disease-19 (covid-19): relationship to duration of infection characteristics of and important lessons from the coronavirus disease 2019 (covid-19) outbreak in china risk factors associated with acute respiratory distress syndrome and death in patients with coronavirus disease 2019 pneumonia in wuhan, china preliminary estimates of the prevalence of selected underlying health conditions among patients with coronavirus disease 2019 -united states frequency and distribution of chest radiographic findings in patients positive for covid-19 the chest radiograph in the covid-19 pandemic: role, standardized reporting, and ct correlation chest x-ray findings in 636 ambulatory patients with covid-19 presenting to an urgent care center: a normal chest x-ray is no guarantee american college of radiology. acr recommendations for the use of chest radiography and computed tomography (ct) for available at: https://www. acr.org/advocacy-and-economics/acr-position-statements/ recommendations-for-chest-radiography-and-ct-for-suspected-covid19-infection the role of chest imaging in patient management during the covid-19 pandemic: a multinational consensus statement from the fleischner society severity scoring of lung oedema on the chest radiograph is associated with clinical outcomes in ards cdc director warns second wave of coronavirus is likely to be even more devastating accuracy and reproducibility of low-dose submillisievert chest ct for the diagnosis of covid-19 understanding the direction of bias in studies of diagnostic test accuracy national health commission & state administration of traditional chinese medicine. diagnosis and treatment protocol for novel coronavirus pneumonia (trial version 7) state of israel ministry of health coronavirus information dashboard reader agreement studies a one-way components of variance model for categorical data a guideline of selecting and reporting intraclass correlation coefficients for reliability research receiver operating characteristic curves and their use in radiology time course of lung changes on chest ct during recovery from interobserver reliability of the chest radiograph in community-acquired pneumonia interobserver reliability of the chest radiograph in pulmonary embolism radiological society of north america expert consensus statement on reporting chest ct findings related to covid-19. endorsed by the society of thoracic radiology, the american college of radiology, and rsna acr appropriateness criteria; acute respiratory illness in immunocompetent patients should rt-pcr be considered a gold standard in the diagnosis of covid-19 an update on covid-19 for the radiologist: a british society of thoracic imaging statement users' guides to the medical literature iii. how to use an article about a diagnostic test b. what are the results and will they help me in caring for my patients? key: cord-006683-7rsmbk3j authors: coppola, m.; porto, a.; de santo, d.; de fronzo, s.; grassi, r.; rotondo, a. title: influenza a virus: radiological and clinical findings of patients hospitalised for pandemic h1n1 influenza date: 2011-01-12 journal: radiol med doi: 10.1007/s11547-011-0622-0 sha: doc_id: 6683 cord_uid: 7rsmbk3j purpose: this paper describes the radiological and clinical findings identified in a group of patients with h1n1 influenza. materials and methods: between may and mid-november 2009, 3,649 patients with suspected h1n1 influenza presented to our hospital. our study population comprised 167 (91 male, 76 female patients, age range 11 months to 82 years; mean age 29 years) out of 1,896 patients with throat swab positive for h1n1 and clinical and laboratory findings indicative of viral influenza. all 167 patients were studied by chest x-ray (cxr), and 20 patients with positive cxr and worsening clinical condition also underwent computed tomography (ct). the following findings were evaluated on both modalities: interstitial reticulation (ir), nodules (n), ground-glass opacities (ggo), consolidations (cons), bacterial superinfection and pulmonary complications. results: ninety of 167 patients had positive cxr results. abnormalities identified on cxr, variously combined and distributed, were as follows: 53 ir, 5 n, 13 ggo, 50 cons; the predominant combination was represented by six ggo with cons. of the 20 cxr-positive cases also studied by ct, 17 showed pathological findings. the abnormalities identified on ct, variously combined and distributed, were as follows: 14 ir, 2 n, 5 ggo; the predominant combination was 10 ggo with cons. despite the differences between the two modalities, the principle radiological findings of bacterial superinfection were tree-in-bud pattern, consolidation with air bronchogram, and pleural and pericardial effusion. fifteen of the 20 patients studied by both cxr and chest ct showed respiratory complications with bilateral and diffuse cons on cxr and ct. six of 15 died: 4/6 of acute respiratory distress syndrome and 2/6 of multiple organ failure. conclusions: our study describes the radiological and clinical characteristics of a large population of patients affected by h1n1 influenza. cxr and chest ct identified the site and extent of the pulmonary lesions and documented signs of bacterial superinfection and pulmonary complications. broncogramma aereo, versamento pleurico e pericardico. dei 20 pazienti studiati con rx e tc, 15 hanno presentato complicanze respiratorie con un quadro rx e tc di ogg e cm diffusi e bilaterali. sei/15 sono deceduti: 4/6 per acute respiratory distress syndrome (ards), 2/6 per multiple organ failure (mof) . conclusioni. il nostro studio ha delineato le caratteristiche radiologiche e cliniche di un'ampia popolazione di pazienti con infl uenza h1n1. la rx e la tc del torace hanno identifi cato sede ed estensione delle lesioni polmonari, documentando i segni di sovrainfezione batterica e le complicanze polmonari. [1, 2] . il virus infl uenzale h1n1 è un sottotipo di infl uenzavirus a appartenente alla famiglia delle orthomyxoviridae, di cui sono note numerose varianti che, negli animali, causano forme infl uenzali pandemiche, come l'infl uenza aviaria e la febbre suina. analisi genetiche suggeriscono che la comparsa della nuova linea h1n1 dell'infl uenza suina nell'uomo sia diretta conseguenza di un riassortimento dei genomi virali dell'infl uenza suina, aviaria ed umana del nord america e dell'eurasia [3] [4] [5] [6] . nel mese di giugno 2009, in seguito alla documentazione pubblicata da parte della organizzazione mondiale della sanità (oms) del primo contagio interumano in due stati e alla rapida diffusione del virus, è stato dichiarato lo stato di emergenza mondiale di scala 6 (pandemia infl uenzale) [6, 7] . febbre elevata, faringodinia, tosse, astenia, artromialgie, nausea, vomito e/o diarrea hanno costituito i principali sintomi d'esordio clinico dell'infl uenza h1n1; nei casi più complessi è insorta insuffi cienza respiratoria acuta [1, 6, 8] . le principali alterazioni del quadro ematochimico sono state leucopenia, trombocitopenia, aumento degli indici di funzionalità epatica [1, 6, 8] . il peculiare tropismo virale per l'apparato respiratorio ha determinato il ricorso alle tecniche di imaging per meglio caratterizzare sede, distribuzione e tipologia delle lesioni elementari polmonari [9] . l'esame radiografi co del torace (rx) e la tomografi a computerizzata (tc) del torace hanno rappresentato le indagini bronchogram, and pleural and pericardial effusion. fifteen of the 20 patients studied by both cxr and chest ct showed respiratory complications with bilateral and diffuse cons on cxr and ct. six of 15 died: 4/6 of acute respiratory distress syndrome and 2/6 of multiple organ failure. conclusions. our study describes the radiological and clinical characteristics of a large population of patients affected by h1n1 infl uenza. cxr and chest ct identifi ed the site and extent of the pulmonary lesions and documented signs of bacterial superinfection and pulmonary complications. keywords infl uenza a virus · h1n1 subtype · computed tomography · x-ray · pneumonia · viral swine-origin infl uenza -or swine infl uenza; swine fl u -is an acute respiratory infection caused by the novel infl uenza a, subtype h1n1, which was transmitted to humans from pigs in mexico and the united states in march and april 2009 and rapidly spread worldwide [1, 2] . h1n1 virus is a subtype of the infl uenza a virus that belongs to the orthomyxoviridae family, of which numerous variants exist that cause fl u pandemics among animals, such as bird fl u and swine fl u. genetic analyses suggest that the emergence of the novel h1n1 strain of swine fl u in humans is the direct consequence of a reassortment of the viral genomes of swine fl u, bird fl u and human fl u in north america and eurasia [3] [4] [5] [6] . in june 2009, following the report of the fi rst human-to-human transmission in two countries in one world health organization (who) region and the rapid spread of the virus, the who declared a level-6 pandemic alert [6, 7] . high fever, sore throat, cough, fatigue, joint pain, nausea, vomiting and/or diarrhoea are the principal symptoms of h1n1 infl uenza; more complex cases present with respiratory failure [1, 6, 8] . the main abnormalities in blood chemistry are leucopenia, thrombocytopenia and elevated liver function tests [1, 6, 8] . the specifi c tropism of the virus for the respiratory tract has led to the use of imaging to better characterise the location, distribution and type of the primary pulmonary lesions [9] . chest x-ray (cxr) and chest computed tomography (ct) have been the investigations of choice for identifying and quantifying, respectively, pulmonary damage in patients affected by infl uenza a [10] [11] [12] . the aim of this paper is to describe the radiological and clinical fi ndings in patients hospitalised for a diagnosis of suspected h1n1 swine-origin infl uenza. di riferimento ai fi ni rispettivamente dell'individuazione e quantifi cazione del danno polmonare nei pazienti affetti da infl uenza virale a [10] [11] [12] . scopo del nostro studio è defi nire il quadro radiologico e clinico dei pazienti ospedalizzati per diagnosi di infl uenza virale h1n1 di origine suina. cxr was performed on all 167 patients, 20 of whom (11.9%) were also studied by ct to assess equivocal fi ndings on cxr or discrepancies between the cxr and clinical fi ndings. cxr was performed using the standard technique (posteroanterior projection: 10 mas, 70 kv; lateral projection: 60 mas, 80 kv; fi lm-focus distance 180 cm) with the patient in standing position in 150 cases (89.9%); only 17 patients (10.1%) were imaged in a sitting or lying position and with anteroposterior projection due to an inability to maintain a standing position and/or inspiration. ct study was performed with a 64-detector-row ct scanner immediately after the cxr in 11/20 cases and 24 h after cxr in the remaining 9/20 cases. on clinical request, 18/20 patients -12/20 large and 6/20 poorly cooperative patients -underwent unenhanced chest and abdominal ct with the spiral technique: breath-hold volumetric scans, slice thickness 1 mm, kv 120, ma 180, pitch 1. only 2/18 patients with suspected pulmonary embolism were subsequently imaged after intravenous administration of nonionic iodinated contrast material with the sure start technique at a fl ow rate of 3 ml/s and dose of 100 ml. the resulting ct images were viewed with a lung window (width 1,500 hu, level -700 hu) and mediastinal window (width 350 hu, level 40 hu). the remaining 2/20 patients were studied with high-resolution ct of the chest for clinical suspicion of concurrent interstitial pulmonary disease using a standard protocol: collimation 1 mm, interval 10 mm, acquisition time 1-2 s, high-spatial-frequency reconstruction algorithm, matrix 512 × 512, 120 kv, 200 ma, fov encompassing both lungs, window level -600 hu (range -500/-900 hu), window width 1,500 hu (range 1,100/2,000 hu), expiratory scans and patient in prone position when needed. two radiologists independently examined and interpreted the cxr and ct images stored in the picture archiving and communications system (pacs). the primary lesions assessed at cxr and chest ct were [13] and recent publications [10, 11] describing the main radiological pulmonary manifestations of a/h1n1 infl uenza: interstitial reticulation (ri; linear opacities of the central and peripheral interstitium appearing as radio-opaque lines on cxr and hyperdensities on ct), nodules (n; well-or illdefi ned, rounded opacities/hyperdensities, with maximum diameter of 3 cm.), ground-glass opacities (ggo; heterogeneous increase in parenchymal opacity with preservation of bronchial and vascular margins), consolidation (cons; homogenously increased parenchymal attenuation that obscures the margins of the bronchial and vessels walls). images were also assessed for signs of bacterial suprainfection: consolidation with air bronchogram (area of radiolucency at cxr and low attenuation at ct, refl ecting the air-fi lled bronchi on a background of opaque or high-attenuation airless lung), cavitations (gas-fi lled spaces, seen as lucencies or low-attenuation areas within a parenchymal consolidation), tree-in-bud pattern (branching centrilobular structures), pleural and/or pericardial effusion (fl uid in the pleural/pericardial cavity), lymphadenopathy (short-axis diameter >1 cm for mediastinal nodes and >3 mm for hilar nodes). in patients with suspected bacterial suprainfection, blood and bronchoaspirate cultures were ordered to search for pathogenic microorganisms. lastly, the images were assessed for possible pulmonary complications. extent of pulmonary damage was defi ned univocally at cxr and chest ct: unilateral or bilateral; symmetrical or asymmetrical; focal, multifocal or diffuse; with predominant distribution in the upper, middle or lower lobes. on ct we also determined the predominant distribution of lesions as being central (perihilar) or peripheral (subpleural). multiplanar reconstructions (mpr) and maximum intensity projections (mip) were obtained through postprocessing in all patients studied with volumetric ct scans. table 2) : ir was seen in 14 (82.3%), n in 2 (11.7%), ggo in 5 (29.4%). the predominant combination was ggo with cons, seen in 10 (58.8%). none of the patients showed isolated pulmonary consolidation, and none of those studied with contrast-enhanced multidetector-row ct (mdct) showed direct and/or indirect signs of pulmonary embolism. on cxr, fi ndings suggestive of bacterial suprainfection were seen in various combinations in 14/90 patients (15.5%) ( table 1) : pleural effusion in 9/14 (64.2%), consolidation with air bronchogram in 4 (28.5%), lymphadenopathy in 2 (14.2%), cavitation in 1 (7.1%), hydropneumothorax in 1 (table 2) : tree-in-bud pattern in 9/9 (100%), consolidation with air bronchogram in 6 (66.6%), pleural effusion in 4 (44.4%) and pericardial effusion in 3 (33.3%), lymphadenopathy >1 cm in 2 (22.2%), cavitation in 1 (11.1%), and hydropneumothorax in 1 (11.1%). ct confi rmed the cxr fi ndings of bacterial suprainfection in 9 cases (9/14, 64.2%) and identifi ed as false positive fi ve cases of suspected pleural effusion detected on cxr. in all fi ve cases, the radiograms had been obtained in the supine position and with anteroposterior projection only in poorly cooperative patients. in addition, in 2/9 cases in which ct confi rmed cxr signs of bacterial suprainfection, it also revealed consolidations with air bronchogram, which had gone undetected on cxr. these were located in the basal and retrocardiac regions in radiograms acquired with anteroposterior projection only. finally, in 3/9 patients in whom ct confi rmed the cxr signs of bacterial suprainfection, it also identifi ed the presence of pericardial effusion, which had been missed at cxr in all cases owing to very small size. blood and bronchoaspirate culture identifi ed staphylococcus aureus in 3/9 patients and a mixed bacterial fl ora in 6/9 cases. in 15/167 patients (8.98%), worsening clinical and radiological features required orotracheal intubation and mechanical ventilation after admission to the intensive care unit. all of these patients were affected by an underlying condition: arterial hypertension complicated by placenta previa and postpartum uterine atony (1/15; 6.6%), hypertensive cardiopathy (3/15; 20%), emphysema (3/15; 20%), bullous dystrophy (1/15; 6.6%), rib-cage malformations (2/15; 13.3%), obesity (4/15; 26.6%), diabetes (2/15; 13.3%), and drug abuse (1/15; 6.6%). in all cases, cxr and ct fi ndings were characterised by bilateral and diffuse ggo with cons (figs. 3a, b, 4a-c) . death occurred in 6/15 (40%) patients due to acute respiratory distress syndrome (ards) in four cases and multiple organ failure in two. human infl uenza pandemics are caused by infl uenza viruses from nonhuman reservoirs: among the infl uenza pandemics [14] and the other two, in 1957 and 1968, were caused by new strains resulting from the combination of avian and human viruses through a reassortment process [15, 16] . viral infl uenza a is a pandemic caused by a novel infl uenza virus a/h1n1, which spread worldwide from mexico in march 2009. the infection is due to pigto-human transmission of a viral pathogen produced by the triple genetic reassortment of human, swine and avian viral strains in north america and eurasia; human-to-human transmission occurs through respiratory droplets or contact with infected surfaces [9, 14, 16] . according to the who, from the beginning of the pandemic to 15 november 2009, >78,000 cases of infl uenza a h1n1 were notifi ed in europe and 190,765 in the americas, with a death toll of at least 350 and 4,806, respectively [17] . from 19 october, when infl unet monitoring began in italy [18] , to 8 november, there were an estimated 1,521,000 cases of infl uenza a/h1n1 in italy [19] . in un recente lavoro [8] è stato confermato che il virus h1n1 si trasmette caratteristicamente tra bambini e giovani adulti, colpendo nel 45% dei casi giovani di età <18 anni e solo nel 5% dei casi individui di età >65 anni. sebbene nella nostra personale esperienza la gravità dell'impatto epidemiologico dell'infl uenza h1n1 sia stato ampiamente ridimensionato rispetto alle aspettative globali, anche la nostra popolazione di studio ha presentato un prevalente interessamento delle fasce di età più giovanili and 53 deaths according to the ecdc (european centre for disease prevention and control (ecdc) [19] . the italian regions that recorded the highest incidence of the virus were: marche (2.9%), followed by emilia romagna (1.8%), latium (1.7%), abruzzo (1.6%) and campania (1.6%). as of 15 november 2009, when the infection reached its peak, fig. 3a,b a 63 year-old man with obesity and copd treated with oxygen therapy. clinical fi ndings: high fever and acute respiratory failure. a chest x-ray, posteroanterior projection: cons, preferentially distributed in the lower and middle lung fi elds. b axial chest mdct: bilateral pulmonary cons associated with diffuse ggo are preferentially distributed peripherally and posteriorly and in the apices of the lower lobes. [8, 19] : il 68,7% era <18 anni, il 3,5% degli individui era di età >65 anni. è stato ipotizzato che gli individui anziani siano dotati di anticorpi neutralizzanti cross-reattivi verso il virus h1n1 [20] . la più alta incidenza nelle fasce di età inferiori ai 18 anni potrebbe essere legata, invece, in particolar modo nei bambini, a meccanismi di immunodefi cienza e/o di immaturità immunologica [21, 22] . le principali manifestazioni cliniche dell'infl uenza virale h1n1 descritte in letteratura [8] sono rappresentate da: febbre (95%), tosse (88%), cefalea (34%), faringodinia (31%), vomito (29%), diarrea (25%). in accordo con tali dati, i segni clinici di infl uenza a identifi cati nel nostro gruppo di studio sono stati febbre (96,4%), tosse (85,6%), angina (41,3%), vomito (34,1%) e diarrea (41,3%). in accordo con i dati della letteratura [8] abbiamo osservato un elevazione degli indici di funzionalità epatica (44,9%), dei leucociti (17,9%) e trombocitopenia (2,9%) . le principali patologie concomitanti sono state bronchite asmatica the number of infl uenza a victims had risen to 53, 23 of whom were in campania, seven in emilia romagna, and fi ve in lombardy. all but three were affected by severe underlying conditions. the most affected age groups were children and teenagers from birth to 14 years of age (incidence 3.6%), and, to a lesser extent, individuals aged 15-64 years (0.7%) and >65 years (0.1 %) [19] . a recent paper [8] confi rmed that the h1n1 virus is typically transmitted among children and young adults, affecting individuals <18 years in 45% of cases and those >65 years in 5% of cases only. although in our personal experience the epidemiological impact of h1n1 infl uenza has been substantially milder than expected, our study population showed a prevalent involvement of the younger age groups [8, 19] : 68.7% <18 years of age, 3.5% >65 years. it has been suggested that older individuals have cross-reactive neutralising antibodies to the h1n1 virus [20] . the higher incidence among individuals <18 years of age may, instead, be related -espe(45,5%), con valori percentuali più alti rispetto ai dati della letteratura [8] , cardiopatie (38,9%) e diabete (14,9%) . allo stato attuale sono stati realizzati pochi studi di imaging del torace nei pazienti affetti da infl uenza virale a/ h1n1 [10] [11] [12] . la presentazione della polmonite virale h1n1, sia alla radiografi a del torace che alla tc sembra rispecchiare le caratteristiche generali delle polmoniti virali [23] . alcuni autori [11] hanno descritto le principali alterazioni radiografi che e tc in 7 pazienti affetti da infl uenza a/h1n1: opacità ground-glass bilaterali, più di frequente associate ad aree di consolidamento a distribuzione multifocale, talora anche focali. all'indagine tc le opacità ground-glass e le aree di consolidamento presentavano una predominante distribuzione peribroncovascolare e subpleurica. agarwal et al. [10] in accordo con i dati della letteratura [10, 11] , le principali lesioni elementari da noi individuate alla rx del torace (53%) (tabella 1) sono state cm (55,5%), ad estensione prevalentemente bilaterale, simmetrica, diffusa/multifocale (14%/14%) ed a distribuzione predominante basale (46%) (figg. 2a, 3a ) ma anche ogg (14,4%) (fig. 1a) . queste ultime presentavano, tuttavia, un'estensione prevalentemente monolaterale e focale (61,5%) e distribuzione più evidente in sede medio-basale (46,1%). meno rappresentativa (6,6%) è stata, nella nostra casistica, l'associazione di cm e ogg che mostravano estensione diffusa, simmetrica e multifocale (83,3%) e predominante distribuzione medio-basale (50%) (fig. 4a) . alla rx del torace abbiamo registrato, caratteristicamente, un'elevata percentuale di pazienti con ri (58,8%) (fig. 1a) 17/20 (85%) pazienti reperti patologici. in accordo con i dati della letteratura [10, 11] le lesioni elementari evidenziate in tc (tabella 2) sono state ogg associate a cm (58,8%), ambedue ad estensione bilaterale, cially as regards children -to mechanisms of immunodeficiency and/or immunological immaturity [21, 22] . the most important reported [8] clinical manifestations of h1n1 virus infl uenza are: fever (95%), cough (88%), headache (34%), sore throat (31%), vomiting (29%) and diarrhoea (25%). in agreement with these data, the clinical signs of infl uenza a identifi ed in our study population were fever (96.4%), cough (85.6%), angina (41.3%), vomiting (34.1%) and diarrhoea (41.3%). in addition, similar to previous reports [8] , we found elevated liver function tests (44.9%), leucocytes (17.9%) and thrombocytopenia (2.9%). the main underlying conditions were asthmatic bronchitis (45.5%), which was more frequent than reported in the literature [8] , heart disease (38.9%) and diabetes (14.9%). to date, few studies addressing chest imaging in patients affected by infl uenza a/h1n1 have been published [10] [11] [12] , and the presentation of h1n1 virus pneumonia on both cxr and chest ct seems to refl ect the general features of viral pneumonia [23] . one study [11] reported on the main cxr and chest ct fi ndings in seven patients affected by infl uenza a/h1n1: bilateral ggo, more frequently associated with focal or multifocal areas of consolidation. at ct, the ggo and the areas of consolidation had a predominant peribronchovascular and subpleural distribution. agarwal et al. [10] conducted a larger study involving 222 patients with infl uenza a/h1n1 seen between may and july 2009. of the 66 (30%) patients studied with cxr, 28 (42%) had consolidations (50%), more frequently distributed in the lower lobes. of the 15/66 (22.7%) patients who underwent ct, 9/15 (60%) had ggo combined with consolidation, with diffuse or lobar extension in 70% of cases. thromboembolic complications occurred in 8% of cases, and 8% of the patients died. in agreement with the literature [10, 11] , the main primary pulmonary lesions we identifi ed on cxr (53%) ( table 1) were consolidations (55.5%), with prevalent bilateral, symmetrical, diffuse/multifocal extension (14%/14%) and predominant basal distribution (46%) (figs. 2a, 3a) , and ggo (14.4%) (fig. 1a) . the latter had, however, prevalent unilateral and focal extension (61.5%) and predominant distribution in the middle-basal region (46.1%). in our series, we had fewer cases (6.6 %) of consolidation combined with ggo, which showed diffuse, symmetrical and multifocal extension (83.3%) and predominant middlebasal distribution (50%) (fig. 4a) . at cxr we found a typically high proportion of patients with interstitial reticulation (58.8%) (fig. 1a) , which showed bilateral, symmetrical and diffuse extension (90.6%) and predominant basal distribution (86.8%). it is likely that the alarmism regarding infl uenza infection prompted many patients to seek early medical attention, thus allowing detection of interstitial reticulation, an early fi nding in viral disease. only 5% of patients had con una maggiore tendenza alla distribuzione asimmetrica e multifocale (80%) (fig. 2b) , piuttosto che diffusa e simmetrica (20%) (fig. 4b,c) . la distribuzione delle ogg associate ai cm è stata predominante ai lobi inferiori (60%) ed in sede subpleurica, associata nel 60% dei casi ad omologhe lesioni in sede peribroncovascolare. inoltre non sono stati evidenziati cm isolati in assenza di ogg verosimilmente per la relativa precocità di osservazione delle lesioni elementari polmonari: nella fase iniziale di infezione le ogg, lesioni più precoci in cui sono ancora distinguibili bronchi e vasi, si manifestano insieme ai cm, rispetto a fasi più tardive della patologia, non necessariamente evolventi in ards, in cui esse aumentano la loro densitometria e confl uiscono interamente in consolidamenti. all'indagine tc sono stati identifi cati, inoltre, 82,3% casi di ri, a prevalente estensione bilaterale, diffusa e simmetrica (78,6%) con predominante distribuzione medio-basale (fig. 1b) [10] , negli esami eseguiti con mezzo di contrasto endovena per quesito clinico di embolia polmonare, non abbiamo evidenziato fenomeni di natura tromboembolica a carico delle arterie polmonari e dei suoi rami. la letteratura moderna [24] [25] [26] ha ampiamente descritto le principali alterazioni polmonari identifi cabili in caso di infezione batterica: consolidazioni con broncogramma aereo, tree-in-bud, cavitazioni,versamento pleurico e/o pericardico, linfoadenomegalie. in accordo con i dati della letteratura [24] [25] [26] , sebbene differentemente identifi cati nelle due metodiche di studio (tabelle 1 e 2), i segni di sovrainfezione batterica più frequenti nel nostro gruppo di studio sono stati tree-in-bud, cm con broncogramma aereo, versamento pleurico e pericardico. all'esame colturale e nel bronco aspirato, in 3/9 casi (33,3%) è stato identifi cato il batterio s. aureus, in 6/9 casi (66,7%) una fl ora batterica mista. dei figg. 3 e 4) . il nostro studio ha permesso di determinare le principali nodules, which showed unilateral and focal extension and basal distribution in 60% of cases. of 20/167 (11.9%) patients studied with ct, 17/20 (85%) showed pathological abnormalities. in agreement with the literature [10, 11] the primary lesions identifi ed on ct (table 2) were ggo combined with consolidation (58.8%), both with bilateral extension and a tendency to asymmetrical and multifocal (80%) (fig. 2b) , rather than diffuse and symmetrical distribution (fig. 4b,c) (20%). the distribution of ggo combined with consolidation was predominant in the lower lobes (60%) and subpleural regions and was associated in 60% of cases with similar peribronchovascular lesions. additionally, there were no cases of isolated consolidation without ggo, probably owing to the relatively early observation of the primary pulmonary lesions: in the initial phase of infection, ggos -earlier lesions in which bronchial and vessel margins are still discernible -manifest alongside consolidations compared with the later phases of disease (not necessarily evolving to ards) in which they increase in attenuation and coalesce into consolidations. on ct we also identifi ed 82.3% cases of interstitial reticulation, with prevalent bilateral, diffuse and symmetrical extension (78.6%) and predominant middle-basal distribution (fig. 1b) . these fi ndings appear to corroborate the cxr results and refl ect, similarly to cxr, the same early observation of the radiographic fi ndings. only two cases showed parenchymal nodules, which were focal and unilateral and distributed in the lower lobe; these nodules had already been identifi ed at cxr and were referable to an underlying infectious disease. at variance with previous reports [10] , in the contrast-enhanced examinations requested for suspected pulmonary embolism, we found no thromboembolic phenomena involving the pulmonary arteries or their branches. recent literature [24] [25] [26] has extensively described the principal pulmonary abnormalities seen in bacterial infections: consolidations with air bronchogram, tree-in-bud pattern, cavitation, pleural and/or pericardial effusion and lymphadenopathy. in agreement with these data [24] [25] [26] , although differently identifi ed by the two imaging modalities (tables 1 and 2) , the most common signs of bacterial suprainfection in our series were tree-in-bud pattern, consolidation with air bronchogram and pleural and pericardial effusion. blood and bronchoaspirate culture revealed s. aureus in 3/9 cases (33.3%) and mixed bacterial fl ora in 6/9 cases (66.7%). of the 15/167 patients (8.98%) who received mechanical ventilation due to worsening clinical and radiological features, all had an underlying condition, and in particular, copd (33.3%). six of these 15 patients died (40%): four (66.6%) due to ards. in agreement with the literature [27] , the radiographic and ct fi ndings in these 15 patients were characterised by diffuse and bilateral ggo and consolidation (figs. 3, 4) . caratteristiche radiologiche e cliniche di un'ampia popolazione di pazienti ospedalizzati in un centro di riferimento per le malattie infettive con diagnosi accertata, mediante tampone faringeo, di infl uenza virale h1n1. l'ampia casistica radiologica e l'integrazione, quando necessaria, con esame tc hanno consentito di determinare le principali alterazioni polmonari dell' infl uenza virale h1n1. l'esame radiografi co standard e la tomografi a computerizzata del torace hanno rappresentato le indagini di riferimento nell'individuazione della sede ed estensione delle lesioni elementari polmonari, documentando i segni di sovrainfezione batterica e le complicanze polmonari dell'infl uenza h1n1, ai fi ni di un corretto inquadramento diagnostico, prognostico e terapeutico. our study allowed us to determine the main clinical features of a large population of patients admitted to an infectious disease referral centre with a diagnosis of h1n1 virus infl uenza proved by pharyngeal swab. the large patient sample, and supplementation when needed with ct, allowed us to defi ne the main pulmonary abnormalities seen in h1n1 virus infl uenza. standard cxr and chest ct are the reference investigations in identifying the location and extension of primary pulmonary lesions and documenting the signs of bacterial suprainfection and pulmonary complications of h1n1 infl uenza, thus allowing correct diagnostic, prognostic and therapeutic management. pneumonia and respiratory failure from swine-origin infl uenza a (h1n1) in mexico clinical management of pandemic (h1n1) infection swine infl uenza a (h1n1) infection in two children -southern california update: infections with a swine-origin infl uenza a (h1n1) virus -united states and other countries update: swine infl uenza a (h1n1) infections -california and texas emergence of a novel swine-origin infl uenza a (h1n1) virus in humans global alert and response: pandemic (h1n1) 2009: update 64 hospitalized patients with 2009 h1n1 infl uenza in the united states an update on swine-origin infl uenza virus a/h1n1: a review chest radiographic and ct fi ndings in novel swine-origin infl uenza a (h1n1) virus (s-oiv) infection swine-origin infl uenza a (h1n1) viral infection: radiographic and ct fi ndings pulmonary complication of novel infl uenza a (h1n1) infection: imaging features in two patients fleischner society: glossary of terms for thoracic imaging characterization of the 1918 infl uenza virus polymerase genes genetic analysis of human h2n2 and early h3n2 infl uenza viruses, 1957-1972: evidence for genetic divergence and multiple reassortment events triple-reassortant swine infl uenza a (h1) in humans in the united states pandemia da infl uenza umana da virus a/h1n1v-aggiornamento 75 ministero del lavoro, della salute e delle politiche sociali, infl uenza a/h1n1, il punto della situazione al 15 novembre cross-reactive antibody responses to the 2009 pandemic h1n1 infl uenza virus immunocompetence of children with frequent respiratory infection pediatric hospitalizations associated with 2009 pandemic infl uenza a (h1n1) in argentina viral pneumonias in adults: radiologic and pathologic fi ndings radiology of bacterial pneumonia imaging of pneumonia: trends and algorithms imaging fi ndings in a fatal case of pandemic swine-origin infl uenza a (h1n1) key: cord-103840-2diao7zh authors: mungai, b. n.; joekes, e.; masini, e.; obasi, a.; manduku, v.; mugi, b.; ongango, j.; kirathe, d.; kiplimo, r.; sitienei, j.; oronje, r.; morton, b.; squire, s. b.; macpherson, p. title: it's not tb but what could it be? abnormalities on chest x-rays taken during the kenya national tuberculosis prevalence survey date: 2020-08-22 journal: nan doi: 10.1101/2020.08.19.20177907 sha: doc_id: 103840 cord_uid: 2diao7zh background: the prevalence of diseases other than tuberculosis (tb) detected during chest x-ray (cxr) screening is unknown in sub-saharan africa. this represents a missed opportunity for identification and treatment of potentially significant disease. our aim was to quantify and characterise non-tb abnormalities identified by tb-focused cxr screening during the 2016 kenya national tb prevalence survey. methods: we reviewed a random sample of 1140 adult ([≥]15 years) cxrs classified as "abnormal, suggestive of tb" or "abnormal other" during field interpretation from the tb prevalence survey. each image was read (blinded to field classification and study radiologist read) by two expert radiologists, with images classified into one of four major anatomical categories and primary radiological diagnosis. a third reader resolved discrepancies. prevalence and 95% confidence intervals of abnormalities diagnosis were estimated. findings: cardiomegaly was the most common non-tb abnormality at 259/1123 (23*1%, 95% ci 20*6%-25*6%), while cardiomegaly with features of cardiac failure occurred in 17/1123 (1*5 %, 95% ci 0.9%-2*4%). we also identified chronic pulmonary pathology including suspected chronic obstructive pulmonary disease in 3*2% (95% ci 2*3%4*4%) and non-specific patterns in 4*6% (95% ci 3*5%6*0%). prevalence of active-tb and severe post-tb lung changes was 3*6% (95% ci 2*6%4*8%) and 1*4% (95% ci 0*8%2*3%) respectively. interpretation: based on radiological diagnosis, we identified a wide variety of non-tb diagnoses during population-based tb screening. tb prevalence surveys and active case finding activities using mass cxr offer an opportunity to integrate disease screening efforts. funding national institute for health research (impala-grant reference 16/136/35). evidence before this study tuberculosis (tb) remains the leading adult infectious killer in the world. the world health organization (who) recommends the use of chest x-ray (cxr) as a mass screening tool in tb prevalence surveys and active case finding activities to identify patients eligible for bacteriological investigation. mathematical modelling suggests that an algorithm incorporating a screening cxr to direct subsequent xpert mtb/rif testing is the optimal pathway with lowest number needed to test at acceptable programme costs in active case finding mass screening activities. increased digital x-ray availability, coupled with the development of computer aided detection (cad) software for identification of tb, could enable widespread use of cxr in screening for tb in areas with limited access to radiologists. in addition, cxr has an advantage of detecting conditions besides tb. however, the prevalence of diseases other than tb identified by cxr during tb mass screening or tb prevalence survey activities is unknown. we systematically searched medline, cinhal, global health and google scholar databases from 1940-2019 to identify studies that described the prevalence of non-tb cxr findings during tb prevalence surveys or mass screening activities. the who stop tb department website and reference lists from relevant reviews and studies were used to supplement the search. the search strategy included mesh terms: "chest xray" or "chest-xray" or "chest radiograph" or "mass screening" or "mass radiography" and "tuberculosis screening" or "tuberculosis triaging" or "tb screening" or "tb triaging" and "non-tuberculous" or "non-tb pathology" or "other pathology". our search yielded a number of studies using cxr screening in prevalence surveys as well as active case finding activities. however, studies describing non-tb pathology during mass radiography were few and mostly in the 20 th century. a report in europe between 1946 and 1948 documented 35% non-tb pathology on mass miniature x-rays. our search did not identify any evidence pertinent to the sub-saharan african context. in this cross-sectional study, we analysed individual-level participant cxr data from the 2016 kenya national tb prevalence survey. our aim was to quantify and characterise non-tb abnormalities identified by tb-focused cxr screening during the survey. we hypothesised that non-tb abnormalities requiring further clinical review are highly prevalent and need to be considered when implementing cxr screening for tb. our study identified multiple non-tb diagnoses. the most prevalent was cardiomegaly at 23•1% (95% ci 20•6%-25•6%). we also identified chronic pulmonary pathology including suspected chronic obstructive pulmonary disease (copd) and non-specific interstitial patterns. mediastinal masses, excluding goitres, occurred in 0•8% (95% ci 0•4%-1•5%). tb related abnormalities, which may cause chronic respiratory symptoms, such as severe bronchiectasis and/or destroyed lung were present in 1•4%(95% ci 0•8%-2•3%). median cad4tb scores were low for the non-tb abnormalities. our study demonstrated a high prevalence of cxr-identified non-tb abnormalities, including cardiomegaly, chronic pulmonary diseases, post-tb lung disease and non-specific lung diseases. implementation of cxr tb screening in this context requires detailed health system planning to incorporate provision of care to people with non-tb abnormalities. this could include incorporation of additional tests such as blood pressure monitoring and spirometry as part of community tb screening interventions. tuberculosis (tb) remains the leading adult infectious killer in the world. 1 despite an estimated nine percent relative increase in case detection in 2018, there are still three million (30%) people with tb who are undiagnosed or not reported to national tb programmes. 1 in an effort to identify missing people with tb, countries have adopted more sensitive diagnostic tools, including xpert mtb/rif, scaled up intensified active case finding (acf), and adapted their screening and diagnostic algorithms to include chest x-ray (cxr) as a sensitive and efficient highthroughput initial screening test. 2 historically, miniature radiography for mass tb screening activities was widely utilized in high-income countries throughout the 20 th century. 3-6 in lower-and middle-income countries (lmic), however, cxr has been used primarily as a complementary tool to support clinical diagnosis of patients who are sputum smear negative. 7 following the findings from national tb prevalence surveys that have employed cxr for screening, there is renewed interest in the utility of cxr for tb screening, and to triage people seeking care with symptoms for further tb investigations. [8] [9] [10] in tb prevalence surveys conducted in lmics, cxr has shown high sensitivity for pulmonary tb (ptb) (94%, 95% ci 88-98) but poor specificity (73%, 95% ci 68-77), necessitating confirmation with a microbiological test. 11-13 mathematical modelling of various tb screening algorithms as well as diagnostic algorithms shows that cxr followed by xpert mtb/rif, though resource intensive, has the lowest number needed to screen to identify a case. 14 increased digital cxr availability, coupled with the development of computer aided detection (cad) software for identification of tb, has enabled widespread use of cxr in screening for tb in areas with limited access to radiologists or expert clinicians. 15, 16 use of cxr for mass tb screening will identify other conditions, especially those related to the rising burden of non-communicable diseases (ncds) in lmics, including cardiovascular disease, chronic respiratory disease and cancer. 17 a short narrative report from europe in the 1940s highlighted a significant number 1225/3423 (35%) of non-tuberculous findings in mass radiography screening. 5 however, there is no contemporaneous evidence about the prevalence of non-tb abnormalities identified during tb prevalence surveys and mass radiographic tb screening interventions. tb prevalence surveys focus on accurate estimation of tb prevalence encouraging intentional over-reading of cxrs for identification of participants eligible for bacteriological testing and not on identifying other abnormalities. 9 systematic screening/acf programs also focus on early detection of active tb. 4 computer-aided detection for tuberculosis (cad4tb) demonstrates high sensitivity for cxr tb diagnosis but is not calibrated for detection of non-tb abnormalities. 18 countries are currently adopting cxr screening in combination with cad software systems, both for mass community tb screening activities, as well as in healthcare settings. 16 however, the individual and health system implications of the presence of these non-tb cxr abnormalities are unknown. we therefore set out to characterise and quantify the nature of non-tb abnormalities on abnormal cxrs taken during the 2016 kenya national tb prevalence survey. 13 the secondary aim was to calculate cad4tb v6.0 software analysis scores of the images and compare to expert radiologist diagnosis. 19 we hypothesized that the use of cxr during tb screening would identify a substantial number of people with non-tb abnormalities who may require further clinical attention. a cross-sectional study using individual-level participant cxr data from adult community members who took part in the 2016 kenya national tb prevalence survey. 13 we conducted the study in two parts: an initial pilot study (n=484) to refine tools and estimate the full sample size required for precision, and the main study (n=1140). the main aim was to estimate prevalence and uncertainty for cxr-identified non-tb disease pathology within this population. the prevalence survey was a population-based cross-sectional study conducted in 2015-2016. the aim was to determine the prevalence of bacteriologically confirmed ptb among adults (≥15 years) and to assess their health seeking behaviour. the survey used the who recommended screening strategy comprising symptom questionnaire and cxr. 9 there were 63,050 enrolled participants, 62,484 (99%) underwent cxr screening. the survey identified 305 tb cases; weighted national prevalence of 558 [95% ci 455-662] per 100,000 adult population. 12, 13 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 22, 2020. . https://doi.org/10.1101/2020.08.19.20177907 doi: medrxiv preprint this study was conducted as part of the kenya prevalence survey ethics approval reference number ssc 2094 by kenya medical research institute. the cxr study used anonymized prevalence survey database and individual consent to participate in this secondary analysis was not required. digitally-acquired posterior-anterior cxrs were uploaded to a digital archive. independent, blinded reading of each film was conducted by two clinical officers in the field who had undergone cxr training. each image was classified as either: a) normal; b) "abnormal, suggestive of tb"; or c) "abnormal other". any participant with a cxr classified as "abnormal, suggestive of tb" by either one of the clinical officers, or with a cough of more than two weeks, was eligible for sputum collection. those confirmed to have tb were referred for treatment and those with other cxr abnormalities were to be linked to a health facility. the prevalence survey has been reported fully elsewhere. 13 we obtained an anonymised line list of all the participants from prevalence survey database; our sampling frame included all cxrs classified as "abnormal, suggestive of tb" or "abnormal other" by the survey field readers. images selected for inclusion in this study were uploaded to a web-based picture archiving and communication system. ten specialist radiologists (five kenyan, five global north) with median experience of 12•5 years in tb/chest radiology were recruited (appendix 1). one-to-one training of the radiologists was provided on the online reporting tool and diagnostic case definitions. 20 based on previous reporting tools, we developed radiological diagnostic case definitions, comprised of specific diagnoses within four major anatomical areas (lung parenchyma, heart and great vessels, pleura, and mediastinum). 21 for each major anatomical area, a list of most common diagnoses was developed, taking into consideration kenyan disease epidemiology (appendix 2). radiologists were able to select one primary diagnosis. differential diagnoses could be added where a single, confident primary diagnosis could not be made. each radiologist was randomly assigned cxrs for review. after completion of each reading, the image was released into a pool for second reading. the readers were blinded to each other's report, but not to clinical information (sex, age, hiv status and symptoms). finally, 10% of images were re-allocated to the original readers, for assessment of intra-observer variation. where pairs of radiologists had discrepant primary diagnoses, one of two additional radiologists undertook a consensus read, with knowledge of the first two radiologists' classification. based on the pilot study findings (appendix 3), 1140 cxr images (390 "abnormal, suggestive of tb" and 750 "abnormal other") would be required to estimate the prevalence of cardiomegaly (the most common diagnosis in the pilot study) within 3•5% percentage points of the true value with 95% confidence. for the main study, we included images classified as either "abnormal, suggestive of tb" and "abnormal other", excluding those sampled in the pilot study. these images were grouped into 99 strata as per the prevalence survey clusters, and sampled without replacement from each stratum. statistical analysis used r v3.6.2 (the r foundation for statistical computing, vienna). inter-and intra-reader agreement was calculated using the cohens kappa statistic. study participant characteristics were calculated as medians or percentages. the prevalence of primary diagnoses was calculated as the number of cxrs depicting the abnormality divided by the total number of images that were readable; 95% confidence intervals were estimated using the binomial exact method. a number of final diagnoses were combined or removed before analysis, based on their prevalence and likelihood of clinical relevance. the 1140 cxrs were analysed using cad4tb v6.0 (delft imaging systems, netherlands). 19 median scores and interquartile ranges (iqr) were calculated for each primary diagnostic group. out of 63,050 participants in the prevalence survey, 62,484 (99%) underwent cxr, with 50,935 (81.5%) reported as normal by survey field staff, 6,425 (10.3%) as "abnormal, suggestive of tb", and 5,124 (8.2%) as "abnormal other" (figure 1 ). . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 22, 2020. . https://doi.org/10.1101/2020.08.19.20177907 doi: medrxiv preprint out of 1140 images included in the main study, 1123 (98•5%) were read by two radiologists and 17 (1•5%) were classified as unreadable. the median patient age was 51•0 years (iqr 36•5-66•0) and 720 (64•2%) were female. two-hundred and fifty (22•3%) had reported cough, 40 (3•6%) chest pain and 42 (3•7%) self-reported hiv positive status. six (0•5%) reported current and 34 (3•0%) reported previous tb treatment. genexpert and/or culture results were positive for 27/456 (5•9%) of those that had sputum tested (table 1) . the overall agreement between pairs of readers was moderate with kappa = 0•41 (appendix 4). there was perfect intra-reader agreement at kappa=1. overall, six-hundred (53•4%) images were classified by study radiologists as having any abnormality. of the images classified as "abnormal, suggestive of tb" by field interpretation in the survey, 203 (64•9%) were classed by expert reviewers as abnormal, whereas among the "abnormal other" category 397 (49•0%) were abnormal by expert radiologist read (table 1) . overall prevalence of abnormalities in the major diagnostic categories were: heart and/or great vessels 26•3% (95% ci 23•7%-28•9%), lung parenchyma 26•1% (95% ci 23•5%-28•8%), pleura 7•6% (95% ci 6•1%-9•3%) and the mediastinum 3% (95% ci 2•1%-4•2%) (figure 2 ). among the 600 abnormal images, 21% (127/600) had multiple abnormalities, cardiomegaly accounted for 259/600, 43•2% (39•2%-47•2%) followed by mild/moderate post-tb lung changes at 85/600, 14•2% (11•5%-17•2%) (figure 3 ). among the clinically relevant non-tb abnormalities, cardiomegaly was the most prevalent at 23•1% (95% ci 20•6%-25•6%), while cardiomegaly combined with features of cardiac failure occurred in 1•5% (95% ci 0•9%-2•4%). non-specific patterns were noted in 4•6% (95% ci 3•5%-6•0%), while suspected chronic obstructive pulmonary disease (copd), including emphysema, was present in 3•2% (95% ci 2•3%-4•4%). mediastinal masses, excluding goitres, occurred in 0•8% (95% ci 0•4%-1•5%). for tb related abnormalities, prevalence of minor post-tb lung changes, such as old/latent tb involving fewer than two lobes of damage/scarring was 7•6% (95% ci 6•1%-9•3%), active-tb was 3•6% (95% ci 2•6%-4•8%) and severe post-tb lung changes, i.e. bronchiectasis and/or destroyed lung, 1•4% (95% ci 0•8%-2•3%). between a quarter and a third of females had cardiomegaly (29%). for males 11•7% had cardiomegaly and 11•2% mild/moderate post tb lung changes. history of cough was a common feature across all diagnosis types with 69/250 (27•6%) of coughers having cardiomegaly. out of the 40 participants with chest pain, 22•5 % (10•8-38•5%) had cardiomegaly and 20% (9•1-35•6%) had minor post-tb lung changes. bacteriological confirmation of tb was found in all categories of reported tb-related lung abnormalities and 4/27, (14•8%, 4•2%-33•7%) of images reported as non-specific patterns ( table 2) . out of the 34 participants with a history of previous tb treatment, features consistent with minor post-tb lung changes were reported in 11(32•4%, 95% ci 17•4%-50•5%), active ptb in 6 (17•6%, 95% ci 6•8%-34•5%) and severe post-tb lung changes in 4 (11•8%, 95% ci 3•3%-27•5%). the median score for all study cxrs was 50 (iqr 45-56.75) while for images classified as normal and abnormal by expert radiologists scores were 47 (iqr44-51) and 54 (iqr48-62) respectively. there was no difference between the sexes. the score for severe post-tb lung changes was highest at 76 (iqr 71-81). active-tb and minor post-tb change scores were similar at 66 (iqr 55-73) and 67 (iqr58-79) respectively. the xpert positive participants' images had scores of 72 (iqr63-80). abnormalities with lower scores included suspected copd at 57 (iqr 52-61), non-specific patterns at 56 (iqr 50-61), mediastinal mass, excluding goitres, at 52 (iqr 47-59), cardiomegaly alone at 50 (iqr 46-56), and cardiomegaly with features of cardiac failure at 57 (iqr55-60). . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 22, 2020. . https://doi.org/10.1101/2020.08.19.20177907 doi: medrxiv preprint the main finding from this analysis of x-ray images from the 2016 kenya tb prevalence survey was that the use of cxr for tb population-based studies identified a large number of patients with abnormalities, including noncommunicable diseases (ncds) such as cardiovascular abnormalities and chronic respiratory diseases that require clinical attention. clinically relevant cardiac and chronic pulmonary diseases accounted for 66% of the non-tb abnormalities in our setting.to our knowledge, this is the first study in sub-saharan africa to characterise and quantify non-tb cxr findings among participants who underwent mass screening as part of a population-based tb prevalence survey. tb prevalence surveys and acf activities using mass cxr therefore offer opportunities to integrate disease screening efforts. the findings of our study are also timely in the wake of disruption in the health system caused by coronavirus disease (covid-19). 22 as countries accelerate acf activities using cxr screening to make up for reductions in tb case-notification rates due to covid-19, they could plan to integrate screening for other diseases. at the outset of the study, we expected abnormalities to be primarily related to non-tb pulmonary disease. however, the most prevalent finding was cardiomegaly at 23•1% (95% ci 20•6%-25•6%). this is higher than in a smaller south african (n=450) study, which reported cardiomegaly in 12•7% (95% c.i. 9•6%-15•8%) of cxrs taken during a vaccine study, among hiv-positive participants. their participants had a much lower median age (34 vs 51 years) than in our study. 17 further, our study analysed images which had all originally been classified as abnormal. this, together with higher median age could explain our higher observed prevalence. calculation of the cardio-thoracic ratio (ctr) on cxr read by humans is a well-described affordable and reproducible screening method for cardiomegaly. 17, 23 however, current studies, including our own, use ctr cut off values developed in caucasian populations and there will be a need for robust validation of baseline ctr values for healthy populations in sub-saharan africa. 24 cad for detection of ctr is under development. 25 in sub-saharan africa, the commonest causes of cardiomegaly are conditions of significant public health importance associated with premature mortality, including: hypertensive heart disease; cardiomyopathies; cor pulmonale; chronic rheumatic heart diseases; and ischaemic heart diseases. 23, 26 cardiomegaly has been associated with both higher body mass index (bmi) and higher median systolic blood pressure (bp). 17 the high prevalence of cardiomegaly in our study supports exploration of the benefits of cvd screening during tb cxr screening as a potentially affordable public health intervention. 5, 23 this could include adding relevant questions about previous hypertension diagnosis and treatment, measurement of bp and calculation of bmi. we recognize that a single bp reading is not diagnostic, but it could serve as an indicator for further follow-up. 27 health messaging on prevention of ncds through recommendations on diet, such as reduction of commercial sugar and high salt diet, could be considered for integration in such programmes. 23 non-tb related respiratory pathology, including chronic respiratory diseases (crd) were another significant finding in our cohort. in a vancouver study in 1960s, three cases of significant previously unknown non-tb lung disease were identified for every new tb case; in our study, this figure was approximately 2:1. 6 it should be noted that cxr alone has limited specificity for many of these conditions, especially in this cohort where very limited clinical information was available. the diagnoses of "non-specific airspace opacification" and "interstitial pattern" cover a range of possible pathologies, varying from incidental acute or chronic infective changes, not typical of tb, to noninfective pathology. copd and emphysema cannot be diagnosed reliably on cxr alone, requiring spirometry and referral for further confirmation. however, crd morbidity and mortality is on the rise, with the prevalence of copd shown to range between 4%-25% in one systematic review in sub-saharan africa comparable to 3•6% (2•3%-4•4%) in our study. 28 as expected, our study confirmed that screening for tb will detect alternative lung abnormalities in a significant number of non-tb cases and spirometry will be required for a subset of these patients. forty-four percent of the participants in our study with reported post-tb lung changes had a history of tb treatment. among those, cxr revealed bronchiectasis and/or destroyed lung in 11% (95% ci 3•3%-27•5%), which is lower than reported in a prospective study in malawi, that used computed tomography and reported >40% bronchiectasis and 10% lobar destruction post-tb treatment. 21 bronchiectasis has a lower detection rate on cxr than ct, and will likely have been underestimated in our study. ptb is a risk factor for crd and in the malawi study ongoing clinical symptoms were associated with damage of three or more lobes, 21, 29 which is comparable to the "destroyed lung" category in our cxr study. unfortunately, patients with ptld and chronic symptoms are likely to be treated empirically for recurrent tb. 30 we therefore recommend that tb screening programs include protocols for ptld . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 22, 2020. . https://doi.org/10.1101/2020.08.19.20177907 doi: medrxiv preprint management at primary care level to ensure patients are not unnecessarily retreated for tb. this also underscores the importance of bacteriological confirmation in patients with high cad4tb scores. for those eligible, latent tb infection treatment should be offered. 31 our study also identified other less common findings for which interventions may be costly. for example, mediastinal masses that may represent lymphoma/ malignancy and would need referral for definitive diagnosis and management. cad4tb has been developed to rapidly identify people with cxr abnormalities indicative of tb. 18 our study had high median cad4tb scores for all active ptb images at 66 (iqr 55-73) as defined by radiologist interpretation, as well as by bacteriological confirmation. images with lower scores, including those with cardiomegaly (50, iqr 46-56) require review to ensure important non-tb pathology is detected. analysis and modelling of the non-tb abnormalities cad4tb scores is required. this will enable quantification of patients with non-tb pathology who would be flagged depending on various threshold cut-offs. this will translate to the numbers of patients with controllable ncds missed per 100 000 population hence justifying further refinement of cad algorithms to include non-tb diagnoses. our study used population-based national prevalence data and an explicit sampling approach to select images for review. each image was read by two expert radiologists. moderate inter-reader variability was mitigated by applying a third reader to resolve discrepancies. however, low specificity is an acknowledged issue with radiological classification. this was a retrospective study and we had limited clinical information available. hiv results were self-reported. important information such as smoking history and pre-existing medical conditions were not collected during the survey. we were therefore not able to adequately correlate clinical symptoms or hiv serostatus with our findings. though the prevalence survey protocol required those with other cxr abnormalities to be linked to a health facility within the cluster, we had no way of ascertaining if this was done, or obtaining data on final diagnosis and clinical outcome. our findings are strikingly similar to those of the 1940s study in europe; that mass radiography can be used to tackle "fundamental problems of disease in the chest, both of the respiratory system and also of the heart" and "aid in detection of early and treatable non-tb disease". 5 currently, tb screening activities using cxr and cad software are focused on finding abnormalities consistent with tb. 16 as countries embark on tb acf activities, they need to be aware that other respiratory and non-respiratory pathologies are likely to be as, or more prevalent, than active tb. mass screening with cxr therefore offers opportunity to screen for and address multiple important diseases. 5, 6 even though the algorithms or protocols for example in tb prevalence surveys do recommend that any other abnormalities should be referred as appropriate, there is no structured system for the detection and referral of such patients. 9, 15 we recommend a patient-centered approach incorporating ncds screening and health promotion during tb acf activities. clear referral pathways and follow-up plans for non-tb pathology could be incorporated during the planning of tb prevalence surveys and acf activities. 6 prospective data collection about non-tb conditions identified during screening and economic impact could assist with health system planning. our findings indicate a high prevalence of ncds in the population. at primary care health facilities, prevention efforts for ncds could be strengthened including health messaging, bp and bmi monitoring. bnm and sbs were responsible for the study conceptualization. bnm, sbs, ao, ej and pm designed the study. em, jo and js as key investigators of the kenya prevalence survey, contributed in the study protocol development and approval for use of the data for this study. bnm, ej, sbs, vm, bm, jo, em and ao developed the study protocol. ej, bnm, vm and bm developed study methodology, data collection tools and conducted the study. dk and rk developed the online reporting tool, conducted the sampling and managed the study data. pm conducted the data analysis and development of the figures. bnm, ej, b morton, sbs and pm played a major role in data interpretation and the writing of the manuscript. ao, b morton, ro critically reviewed the manuscript. sbs was director of the impala global health research unit at the time of this work and played a major role in securing the funding for the work. as joint senior authors, sbs and pm provided final approval of the version to be published. we declare no competing interests. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 22, 2020. . https://doi.org/10.1101/2020.08.19.20177907 doi: medrxiv preprint the kenya national tuberculosis, leprosy and lung disease program is the custodian of the 2016 kenya tuberculosis prevalence survey data. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. the copyright holder for this preprint this version posted august 22, 2020 . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted august 22, 2020. . https://doi.org/10.1101/2020.08.19.20177907 doi: medrxiv preprint figure 2 : prevalence of abnormalities by major diagnostic categories *active ptb was captured in the major diagnostic category (a) as this is the key diagnosis in any tb screening program. . cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted august 22, 2020. . https://doi.org/10.1101/2020.08.19.20177907 doi: medrxiv preprint cc-by-nc-nd 4.0 international license it is made available under a is the author/funder, who has granted medrxiv a license to display the preprint in perpetuity. (which was not certified by peer review) the copyright holder for this preprint this version posted august 22, 2020 geneva: world health organisation performance of algorithms for tuberculosis active case finding in underserved high-prevalence settings in cambodia: a cross-sectional study systematic screening for active tuberculosis principles and recommendations discussion on non-tuberculous conditions revealed by mass radiography non-tuberculous chest disease found in a mass x-ray survey in vancouver, b.c implementing the who stop tb strategy: a handbook for national tuberculosis control programmes how to identify tuberculosis cases in a prevalence survey [educational series: prevalence surveys. serialised guidelines. assessing tuberculosis prevalence through populationbased surveys tuberculosis prevalence surveys: a handbook screening strategies for tuberculosis prevalence surveys: the value of chest radiography and symptoms kenya tuberculosis prevalence survey 2016: challenges and opportunities of ending tb in kenya chest radiography in tuberculosis detection-summary of current who recommendations and guidance on programmatic approaches. switzerland: world health organization cardio-thoracic ratio is stable, reproducible and has potential as a screening tool for hiv-1 related cardiac disorders in resource poor settings computer aided detection of tuberculosis on chest radiographs: an evaluation of the cad4tb v6 system2019 computer-aided detection for tuberculosis (cad4tb) the netherlands: delft imaging fleischner society: glossary of terms for thoracic imaging patient outcomes associated with post-tuberculosis lung damage in malawi: a prospective cohort study the potential impact of the covid-19 response on tuberculosis in high-burden countries: a modelling analysis mortality from cardiovascular diseases in sub-saharan africa, 1990-2013: a systematic analysis of data from the global burden of disease study establishing the cardiothoracic ratio using chest radiographs in an indigenous ghanaian population: a simple tool for cardiomegaly screening automatic heart localization and radiographic index computation in chest x-rays. procspie cardiomegaly in ghana: an autopsy study international society of hypertension global hypertension practice guidelines chronic obstructive pulmonary disease in sub-saharan africa: a systematic review a systematic review of the prevalence and pattern of imaging defined post-tb lung disease who. latent tuberculosis infection updated and consolidated guidelines for programmatic management we are grateful to the 2016 kenya tuberculosis prevalence survey team and the division of national tuberculosis, leprosy and lung disease program whose data we used for this secondary study. we would like to thank jamilah meghji and jeremiah chakaya who gave input during the conceptualisation of the study. we also acknowledge the team of expert radiologists engaged in the interpretation of the study x-ray images: kenya team-beatrice mulama, eva maxine angoro, caroline kebuka; uk team through worldwide radiology, uk: john curtis, alberto alonso, nigel marchbank, charlie sayer, laura cormack and ting ting zhang. this research was funded by the national institute for health research (nihr) (impala, grant reference 16/136/35) using uk aid from the uk government to support global health research. the views expressed in this publication are those of the author(s) and not necessarily those of the nihr or the uk department of health and social care. page 15 of 20 appendix 3: we conducted a pilot study from december 2018 to january 2019, to trial and refine a list of selected chest x-ray diagnoses, to refine standard operating procedures for reporting and to estimate a sample size for the main study. we developed an on-line, study specific cxr reporting tool which comprised four major diagnostic categories: lung parenchyma, heart and great vessels, and the pleura and mediastinum. under each heading, a pick list of most common expected diagnoses was given, taking into consideration kenyan disease epidemiology and prevalence survey cohort. during reporting of the cxrs, readers were required to select one or more primary diagnoses as participants could have more than one diagnosis. , followed by the option of selecting up to two differential diagnoses. due to limited specificity of cxr in many disease presentations, allowing for alternative diagnostic options was designed to capture those cases where a single, confident primary diagnosis could be not made. this study was aimed at deriving the prevalence of a final diagnosis rather than exploring image characteristics for each disease, the proforma did not capture detailed descriptions of each film.five radiologists were selected for the pilot: three consultant radiologists (mv, bm, bmm) from kenya who had been part of the national tb prevalence survey team and two (jc, ej) from the united kingdom. prior to the pilot readings, a test set of 20 x-rays was read independently by all radiologists, followed by discussion in a consensus meeting to ensure uniformity in reporting and application of the tool. subsequently, each pilot image was read by a single radiologist only. we were unable to identify recent and relevant estimates from the literature to inform sample size estimates for prevalence of pulmonary abnormalities from similar settings in the pilot study. we therefore pragmatically set out to read 500 images (150 abnormal suggestive of tb; and 350 abnormal other) during the pilot, after which the detected prevalence of a set of predetermined pathologies was used to calculate a representative sample size for the main study. once sampled, the labels of "abnormal, suggestive of tb" and "abnormal other" were removed to reduce the risk of bias. a total of 484 (97%) images were reported, 16 images were not read within the set time frame. the reporting was as follows: 288 (60%) as abnormal, 178 (37%) as normal, 8 (2.3%) were not interpretable. the abnormalities included: heart and/or great vessel abnormalities 174 (37%), lung parenchyma abnormalities116 (24%), pleural abnormalities 39 (8.2%) and mediastinal abnormalities 8 (1.7%). in the images in the "abnormal other" category n=344, cardiomegaly was most prevalent at 122 (36%) and great vessel abnormalities at 79 (23%). in the "abnormal suggestive of tb" category n=122, old or latent tb was the most prevalent diagnosis at 17(13.9%) and active ptb at 16(13.1) %. cardiomegaly in this category was at 9(7%). appendix 5: cohen`s kappa scores for the expert radiologists inter-reader variability key: cord-157444-huvnyali authors: nabulsi, zaid; sellergren, andrew; jamshy, shahar; lau, charles; santos, eddie; kiraly, atilla p.; ye, wenxing; yang, jie; kazemzadeh, sahar; yu, jin; kalidindi, raju; etemadi, mozziyar; vicente, florencia garcia; melnick, david; corrado, greg s.; peng, lily; eswaran, krish; tse, daniel; beladia, neeral; liu, yun; chen, po-hsuan cameron; shetty, shravya title: deep learning for distinguishing normal versus abnormal chest radiographs and generalization to unseen diseases date: 2020-10-22 journal: nan doi: nan sha: doc_id: 157444 cord_uid: huvnyali chest radiography (cxr) is the most widely-used thoracic clinical imaging modality and is crucial for guiding the management of cardiothoracic conditions. the detection of specific cxr findings has been the main focus of several artificial intelligence (ai) systems. however, the wide range of possible cxr abnormalities makes it impractical to build specific systems to detect every possible condition. in this work, we developed and evaluated an ai system to classify cxrs as normal or abnormal. for development, we used a de-identified dataset of 248,445 patients from a multi-city hospital network in india. to assess generalizability, we evaluated our system using 6 international datasets from india, china, and the united states. of these datasets, 4 focused on diseases that the ai was not trained to detect: 2 datasets with tuberculosis and 2 datasets with coronavirus disease 2019. our results suggest that the ai system generalizes to new patient populations and abnormalities. in a simulated workflow where the ai system prioritized abnormal cases, the turnaround time for abnormal cases reduced by 7-28%. these results represent an important step towards evaluating whether ai can be safely used to flag cases in a general setting where previously unseen abnormalities exist. chest radiography (cxr) is a crucial thoracic imaging modality to detect, diagnose, and guide the management of numerous cardiothoracic conditions. approximately 837 million cxrs are obtained annually worldwide 1 , resulting in a high reviewing burden for radiologists and other healthcare professionals. 2, 3 in the united kingdom, for example, a shortage in the radiology workforce is limiting access to care, increasing wait times, and delaying diagnoses. 4 the need to reduce radiologist workload and improve turnaround time has sparked a surge of interest in developing artificial intelligence (ai)-based tools to interpret cxrs for a broad range of findings. [5] [6] [7] many algorithms have been shown to detect specific findings, such as pneumonia, pleural effusion, and fracture, with comparable or higher performance than radiologists. [5] [6] [7] [8] [9] [10] however, by virtue of being developed to detect specific findings, these algorithms are unlikely to properly report other abnormalities that they were not trained to detect. [11] [12] [13] for example, interstitial lung disease may not necessarily trigger a pneumonia detector. if these detectors are indeed highly specific, they can only be used to detect specific diseases, and are not suitable as comprehensive prioritization tools. moreover, because developing accurate ai algorithms generally requires large labeled datasets, developing algorithms for every potential abnormality that may be encountered in a broad clinical setting is impractical. therefore, a different problem framing is required for use as an effective prioritization tool: algorithms are needed to distinguish normal versus abnormal cxrs more generally. a reliable ai system for distinguishing normal cxrs from abnormal ones can contribute to prompt patient workup and management. there are several use cases for such a system. first, in scenarios with a high reviewing burden for radiologists, the ai algorithm could be used to identify cases that are unlikely to contain findings, empowering healthcare professionals to quickly exclude certain differential diagnoses and allowing the diagnostic workup to proceed in other directions without delay. cases that are likely to contain findings can be also grouped together for prioritized review, reducing the turnaround time. second, in settings when clinical demand outstrips availability of radiologists (for example, in the midst of a large disease outbreak), such a system might be used as a frontline point-of-care tool for non-radiologists. importantly, the ai needs to be evaluated on cxrs with "unseen" abnormalities (i.e. those that it had not encountered during development), to validate its robustness towards new diseases or new manifestations of diseases. in this work, we developed a deep learning system (dls) that classifies cxrs as normal or abnormal with data from 5 clusters of hospitals from 5 cities in india. we then evaluated the dls for its generalization to unseen data sources and unseen diseases using 6 independent datasets from india, china, and the united states. these datasets comprise of two broad clinical datasets, two tuberculosis (tb) datasets with microbiologically confirmed positive and negative cases, and two coronavirus disease 2019 (covid-19) datasets with reverse transcription polymerase chain reaction (rt-pcr)-confirmed positive and negative cases. dataset curation figure 1 shows the overall study design. our training set consisted of 250,066 cxrs of 213,889 patients from 5 clusters of hospitals from 5 cities in india (supplementary table 1 , supplementary figure 1 ). in the training set, all known tb cases were excluded and covid-19 cases were absent. to evaluate the trained dls, we used 6 datasets with a total of 11,576 cxrs from 11,298 patients ( table 1 , supplementary figure 1 ). this includes 2 broad clinical datasets (dataset 1 and chestx-ray14 , n=8,557 total cases) with 2,423 abnormal cases, 2 datasets (tb-1 and tb-2, n=595 total cases) with 294 tb-positive cases, and 2 datasets (cov-1 and cov-2, n=2,424 total cases) with 873 covid-19 positive cases. ds-1, cov-1, and cov-2 were obtained from a mixture of general outpatient and inpatient settings and thus represent a wide spectrum of cxrs seen across different populations. evaluation on these broad datasets mitigates the risk of selecting only the most obvious cases while excluding more difficult images. cxr-14, tb-1, tb-2 were enriched for rare conditions and were publicly available. evaluation on these datasets specifically validates the dls's performance on rarer conditions, and enables benchmarking with other studies using the same data. to define high-sensitivity and high-specificity operating points for the dls, we created four small operating point selection datasets for four scenarios: ds-1, cxr-14, tb, and covid-19; n=200 cases each (see figure 1b and "operating point selection datasets" section in methods). across these datasets, we collected 48,877 labels from 31 radiologists for either the reference standard or to serve as a comparison for the dls (see "labels" section in methods). the dls was first evaluated for its ability to classify cxrs as normal or abnormal on the test split of ds-1 and an independent test set cxr-14. we obtained the normal and abnormal labels from the majority vote of three radiologists (see "labels" section in methods). the percentage of abnormal images were 24% and 71% in ds-1 and cxr-14, respectively ( figure 2a ). to have a comprehensive understanding of the dls, we measured sensitivity, specificity, negative predictive value (npv), positive predictive value (ppv), percentage of predicted positives and the percentage of predicted negatives at a high-sensitivity operating point and a high-specificity operating point ("evaluation metrics" section in methods). with the high-sensitivity operating point (see "operating point selection" section in methods), the dls predicted 29.9% of ds-1 and 24.0% of cxr-14 as normal, with npvs of 0.98 and 0.85, respectively (table 2) . with the high-specificity operating point, the dls predicted 22.2% of ds-1 and 11.7% of cxr-14 as abnormal, with ppvs of 0.68 and 0.99, respectively ( table 2 ). the npvs and ppvs across different operating points are plotted in figure 3 . to put the performance of the dls in context, two independent board-certified radiologists reviewed both the test split of ds-1 and cxr-14. the radiologists had average npvs of approximately 0.87 and 0.70 and ppvs of 0.75 and 0.96 on ds-1 and cxr-14, respectively ( table 3 ). the radiologists' sensitivity and specificity are illustrated on the roc curves ( figure 2a ). radiographic findings vary in their difficulty and importance of detection. thus we next conducted subgroup analyses for each abnormality listed in supplementary the dls was next evaluated on two diseases that it had not been trained to detect (tb and covid-19) across four disease-specific datasets: tb-1, tb-2, cov-1, and cov-2. in these analyses, the dls was evaluated against the reference standard for each specific disease (tb or covid, respectively, see "labels" section in methods). for tb (where percentage of disease-positive images were 52% and 40% in tb-1 and tb-2; table 1 ), the aucs were 0.95 (95%cis: 0.93-0.97) in tb-1 and 0.97 (95%cis: 0.94-0.99) in tb-2 (table 2, figure 2b ). at the high-sensitivity operating point, the dls predicted 43.1% of tb-1 and 38.3% of tb-2 as negative, with npvs of 0.88 and 0.98, respectively (table 2a) . the npvs and ppvs across different operating points are also plotted in figure 3 . however, cxrs that were labeled (tb) negative could nonetheless contain other abnormalities (see "labels" section in methods). hence ppvs (table 2a-b) need to be interpreted with the context that low ppvs for identifying tb-positive radiographs as abnormal do not necessarily reflect the ppv for correctly identifying images with other findings in those datasets (see "distributional shift between datasets" below). every image in tb1 and tb2 was also annotated as normal or abnormal by one radiologist from a cohort of 8 consultant radiologists from india. the radiologist npvs were 0.74 and 0.88 and their ppvs were 0.93 and 0.93 on tb-1 and tb-2, respectively (table 3 and figure 2b ). for covid-19 (where percentage of disease-positive images were 32% and 48% in cov-1 and cov-2; table 1 ), the aucs were 0.68 (95%cis: 0.66-0.71) in cov-1 and 0.65 (95%cis: 0.60-0.69) in cov-2 (table 2, figure 2a ). with a high-sensitivity operating point, the dls predicts 5.9% of cov-1 and 9.8% of cov-2 as negatives with npvs of 0.85 and 0.56, respectively ( table 2 ). the npvs and ppvs for different operating points are plotted in figure 3 . similar to the tb case above, images that were negative for covid-19 often contained other abnormalities (see "distributional shift between datasets" section below) . every image in cov-1 and cov-2 was also reviewed by one radiologist from a cohort of four us board-certified radiologists. the radiologist npvs were 0.78 and 0.62 and their ppvs were 0.51 and 0.60 on cov-1 and cov-2, respectively (table 3 and figure 2c ). finally, to better understand the potential impact of the algorithm in the setting of imperfect rt-pcr sensitivity, we conducted a subanalysis of covid-19 cases that had a "false negative" rt-pcr test result on initial testing, defined as a negative rt-pcr test followed by a positive one within five days. in the 21 such cases, the dls achieved a 95.2% sensitivity, with the cxr taken at the time of the negative test. to better understand the data shifts between applications (general clinical setting in ds-1 vs. the enriched cxr-14; the broad clinical settings vs. tb; and the broad clinical settings vs. covid-19), we next examined the distributions of the dls predictive scores across all 6 test datasets and their corresponding operating point selection sets (figure 4 , see "operating point selection datasets" in methods). we observed similarly peaked dls prediction score distributions (near 1.0) for positive cases --whether for general abnormalities, specific conditions, tb, or covid-19 (see red histograms in figure 4a -c). however, although the distributions for "negative" cases were mostly similar, they did have a small degree of variability, even among datasets of the same scenario from different sites. for example, comparing tb-1 and tb-2 which have similar cxr findings (tb) but were from two independent sites, negative cases in tb-2 had higher scores than in tb-1. similarly, comparison between cov-1 and cov-2 also shows slight differences in the scores for negative cases. these observations confirm the existence of data shifts, suggesting that the scenario-specific operating points are essential, and that even having site-specific operating points may further improve the dls's performance. although scores for positive and the negative cases in ds-1, cxr-14, tb-1, and tb-2 were well-separated, there was significant overlap between the distributions of positive and negative cases for the covid-19 datasets. in fact, further review of the images revealed that 24.9% of negatives in cov-1 and 31.5% of negatives in cov-2 had other cxr findings, and were thus abnormal. a breakdown of the type of finding in these "negatives" is presented in supplementary figure 5 . examples of challenging cases of each condition and associated saliency maps highlighting the regions with the greatest influence on dls predictions are presented in figure 5 . to understand how the developed dls can assist practicing radiologists, we investigated two simulated dls-based workflows. in the first setup, to assist radiologists in prioritizing review of abnormal cases, the dls sorted cases by the predicted likelihood of being abnormal ( figure 1d ). we measured the differences in expected turnaround time for the abnormal cases with and without dls prioritization. for simplicity, in this simulation, we assume the same review time for each case, and that the review time per case does not vary based on review order. the dls-based prioritization reduced the mean turnaround time of abnormal cases by 8-29% for ds-1 and cxr-14, 21-28% for tb-1 and tb-2, and 8-13% for cov-1 and cov-2 ( figure 6 ). in the second setup, we investigated a simulated sequential reading setup where the dls identified cases that were unlikely to contain findings, and the radiologist reviewed only the remaining cases ( figure 1d ). though the deprioritized cases could be reviewed at a later time, we computed the effective immediate performance assuming the dls-negatives were not yet reviewed by radiologists and considered them to be interpreted as "normal" for evaluation purposes. there were minimal performance differences between radiologists and the sequential dls-radiologists setup, but the effective "urgent" caseload reduced by 25-30% for ds-1 and cxr-14, about 40% for the tb datasets, and about 5-10% for the covid-19 datasets (supplementary table 7 ). we have developed and evaluated a dls for interpreting cxrs as normal or abnormal, instead of detecting individual abnormalities. we further validated that it generalized with acceptable performance using six datasets: two broad clinical datasets (auc: 0.87 and 0.94), two datasets with one unseen disease (tb; auc: 0.95 and 0.97), and two datasets with a second unseen disease (covid-19; auc: 0.68 and 0.65). generalizability to different datasets and patient populations is critical for evaluation of ai systems in medicine. studies have shown that many factors might lead to challenges of generalization of ai systems to new populations, such as dataset shift and confounders. 14 furthermore, with cxrs, as with all medical imagery, the number of potential manifestations is unbounded, especially with the emergence of new diseases over time. understanding model performance on this set of unseen diseases is an imperative step in developing a robust and clinically useful model that can be trusted in real world situations. in this work, we evaluated the dls's performance on 6 independent test sets consisting of different patient populations, spanning three countries, and with two unseen diseases (tb and covid-19). the dls's high sensitivity operating point for ruling out normal cxrs performed on par with board-certified radiologists, with the dls npvs of 0.85-0.95 (general abnormalities), 0.88-0.98 (tb), and 0.56-0.85 (covid-19), comparable to radiologist npvs of 0.67-0.87 (general abnormalities), 0.74-0.88 (tb), and 0.62-0.78 . these results highlight the dls's generalizability across real-world dataset shifts, increasing the likelihood of such a system to also generalize to new datasets and new manifestations. the "lower" observed aucs of the dls on the covid-19 datasets were likely caused by our deliberate application of a general abnormality detector to a cohort enriched for patients with a clinical presentation consistent with covid-19 and thus tested for covid-19. however, as other acute diseases may share a similar clinical presentation, many cases negative for covid-19 exhibited abnormal cxr findings that likely triggered the dls ( figure 5, supplementary figure 5 ). in addition, a substantial number of covid-19 patients can present with a normal cxr 15 , which would also contribute to a lower observed auc. the variability in patient population and clinical environment across different datasets also meant that the same operating point was unlikely to be appropriate across all settings. for example, a general outpatient setting is substantially less likely to contain cxr findings compared to a cohort of patients with respiratory symptoms or fevers in the midst of the covid-19 pandemic. similarly, datasets that are deliberately enriched for specific conditions (cxr-14 and tb) are skewed and are not representative of a general disease screening population. thus, we used a small number of cases (n=200) from each setting to determine the operating points specific to that setting. consistent with this hypothesis, these operating points then generalized well to another dataset, such as from tb-1 to tb-2 and from cov-1 to cov-2. however, further performance improvement is likely possible with site-specific operating point selection sets. we anticipate that this simple operating point selection strategy using a small number of cases may be useful when evaluating an ai system in a new setting, institution, or patient population. in addition to general performance across the 6 datasets, subgroup analysis of the dls' performance on each specific abnormal cxr finding of ds-1 and cxr-14 (supplementary tables 4 and 5 ) revealed consistently high npvs, suggesting that the dls was not overtly biased towards any particular abnormal finding. in addition, the dls outperformed radiologists on atelectasis, pleural effusion, cardiomegaly / enlarged cardiac silhouette, and lung nodulessuggesting that the dls as a prioritization tool could be particularly valuable in emergency medicine where dyspnea, cardiogenic pulmonary edema, and incidental lung cancer detection are commonly encountered. furthermore, the dls also outperformed radiologists in settings where an abnormal chest radiographic finding was present but the abnormality was not one of the predefined chest radiographic findings (e.g. perihilar mass) or radiologists agreed on the presence of a finding but disagreed as to its characterization (indicating case ambiguity; see "other" in supplementary tables 4 and 5 ). this suggests that the dls may be robust in the setting of chest radiographic findings that are uncommon or difficult to reach consensus on. to further evaluate the potential utility of our system, we simulated a setup where the dls prioritizes cases that are likely to contain findings for radiologists' review. our evaluation suggests a potential reduction in turnaround time for abnormal cases by 7-28%, indicating the dls's potential to be a powerful first-line prioritization tool. whether deployed in a relatively healthy outpatient practice or in the midst of an unusually busy inpatient or outpatient setting, such a system could help prioritize abnormal cxrs for expedited radiologist interpretation. in radiology teams where cxr interpretation responsibilities are shared between general and subspecialist (i.e. cardiothoracic) radiologists, such a system could be used to distribute work. for non-radiologist healthcare professionals, a rapid determination regarding the presence or absence of an abnormality on cxr prevents releasing of a patient who needs care and enables alternative diagnostic workup to proceed without delay while the case is pending radiologist review. finally, a radiologist's productivity might increase by batching negative cxrs for streamlined formal review. finally, to facilitate the continued development of ai models for chest radiography, we are releasing our abnormal versus normal labels from 3 radiologists (2430 labels on 810 images) for the publicly-available cxr-14 test set. we believe this will be useful for future work because label quality is of paramount importance for any ai study in healthcare. in cxr-14, the binary abnormal labels were derived through an automated natural language processing (nlp) algorithm on the radiology report. 7 however, editorials have questioned the the quality of labels derived from clinical reports. 16 hence, in this study we obtained labels from multiple experts to establish the reference standard for evaluation, and a confusion matrix of our majority vote expert labels against the public nlp labels is shown in supplementary table 6 . prior studies have demonstrated an algorithm's potential to differentiate normal and abnormal cxrs. [17] [18] [19] [20] [21] hwang et al. evaluated a commercially available system with comparison to radiology residents. 19 annarumma et al. further demonstrated the system's utility in a simulated prioritization workflow using held-out data from the same institution as the training dataset. 18 our study complements prior works by performing extensive evaluations on model generalizability, including generalization to multiple datasets in different continents, different patient populations settings, and with the presence of unseen diseases. in addition, we also obtained radiologist reviews as benchmarks to understand the dls's performance. lastly, we presented two simulated workflows; one demonstrated reduced turnaround time for abnormal cases, and the other showed comparable performance while reducing effective caseload. our study has several limitations. first, there are a wide range of abnormalities and diseases that were not represented among the cxrs available for this study. although it's infeasible to exhaustively obtain and annotate datasets for every possible finding, further increasing the conditions and diseases considered in this study could help both in the dls development and evaluation. second, we only had labeled data regarding disease-positive and disease-negative for tb and covid-19. the absence of normal and abnormal labels for the tb and covid-19 datasets led to added complexity in understanding the performance metrics of ppvs and specificities for these scenarios. third, to provide a comparison with the dls, which only had cxrs as input, the radiologists reviewed the cases solely based on cxrs without referencing additional clinical or patient data. in a real clinical setting, this information is generally available, and likely influences a radiologist's decisions. lastly, the results were based on retrospective data. the utility of the dls-assisted workflows were based on simulation with many assumptions, such as identical radiologist diagnosis regardless of the review order and identical review time across normal and abnormal cases. hence, the true effects will need to be determined through future evaluation in a prospective setting. in conclusion, we have developed and evaluated a clinically relevant artificial intelligence model for chest radiographic interpretation and evaluated its generalizability across a diverse set of images in 6 distinct datasets. these results suggest the potential for the ai system to generalize to new patient populations and unseen diseases. using the ai system in a simulated workflow to prioritize abnormal cases, the turnaround time for abnormal cases reduced by 7-28%. lastly, we hope that the performance analyses reported here on the publicly available datasets can serve as a useful resource to facilitate the continued development of clinically useful ai models for cxr interpretation. in this study, we utilized 6 independent datasets for dls development and evaluation. the dls was evaluated in two ways: distinguishing normal vs. abnormal cases in a general setting with multiple radiologist-confirmed abnormalities (first 2 datasets), and in the setting of diseases that the dls was not exposed to during training (tb was excluded from the train set and covid-19 was not present; last 4 datasets). all data were stored in the digital imaging and communications in medicine (dicom) format and de-identified prior to transfer to study investigators. details regarding these datasets and patient characteristics are summarized in table 1, supplementary table 1 , and supplementary figure 1 . this study using de-identified retrospective data was reviewed by advarra irb (columbia, md), which determined that it was exempt from further review under 45 cfr 46. the first dataset (ds-1) was from five clusters of hospitals across five different cities in india (bangalore, bhubaneswar, chennai, hyderabad, and new delhi). 5 ds-1 consisted of images from consecutive inpatient and outpatient encounters between november 2010 and january 2018, and reflected the natural population incidence of the abnormalities in the populations. all tb cases were excluded and covid-19 cases were not present. in total, ds-1 originally contained 1,052,274 cxrs from 794,501 patients before exclusions (supplementary figure 1a) . this dataset was randomly split into training, tuning, and testing sets in a 0.775:0.1:0.125 ratio while ensuring that images from the same patient remained in the same split. the split is consistent with our previous study. 5 the dls was developed solely using the training and tuning splits of ds-1. because outpatient management is primarily done using posterior-anterior (pa) cxrs, while inpatient management is primarily done on anterior-posterior (ap) cxrs, we emphasized pa cxrs in the tune split to better represent an outpatient use case. both pa and ap images are used in the test datasets. to select operating points for each of the four scenarios (two general abnormalities, tb, covid-19), 200 images were randomly selected as the operating point selection sets. for general abnormalities, we selected two independent operating points using 200 randomly sampled images from the ds-1 tune set and 200 randomly sampled images from cxr-14's publicly-specified combined train and tune set 7, 22 . for tb, 200 randomly sampled images from tb-1 were used. for covid-19, 200 randomly sampled images from cov-1 were used. these images were only used to determine an operating point for that scenario, and once used for operating point selection, were excluded from the test set (supplementary figure 1 ). two datasets were used to evaluate the dls's performance in distinguishing normal and abnormal findings in a general abnormality detection setting. the first dataset contains 7,747 randomly selected pa cxrs from the original test split of the ds-1. 5 these sampled images were expertly labelled as normal or abnormal for the purposes of this study. the second dataset contains 2,000 randomly selected cxrs from the publicly-specified test set (25,596cxrs from 2,797 patients) of cxr-14 from the national institute of health. 7, 22 from these 2,000 cxrs (also used in prior work 5 ), we removed all the patients younger than 18 years of age and all the ap scans (to focus on an outpatient setting, see tune split procedure above), leaving us with 810 images. to evaluate the dls performance in unseen diseases, we curated 2 datasets for tb and 2 datasets for covid-19 (1 cxr per patient, supplementary figure 1c for development and evaluation of the dls, we obtained labels to indicate whether abnormalities were present in each cxr. each image was annotated as either "normal" or "abnormal", where an "abnormal" scan is defined as a scan containing at least one clinically-significant finding that may warrant further follow-up. for example, degenerative changes and old fractures were not labeled abnormal because no further management is required. for the train and tune split of ds-1, we obtained the abnormal and normal labels using nlp (regular expressions) on the radiology reports (supplementary table 2 ). for the normal images, radiology report templates were often used, meaning the same report indicating a normal scan was often used for numerous images. we extracted the most commonly used radiology reports, manually confirmed those that indicated normal reports, and obtained all images that used one of these normal template reports. examples of these radiology reports along with their frequencies are shown in supplementary table 2 . for the abnormal images, we obtained all images that did not contain keywords indicating the scan is normal in their respective radiology reports. for the test sets of ds-1 and cxr-14, a group of us board-certified radiologists reviewed the images to provide reference standard labels. for each image in ds-1, three readers were randomly assigned from a cohort of 18 us board-certified radiologists (range of experience 2-24 years in general radiology). for cxr-14, we obtained labels from three us board-certified radiologists (years of experience: 5, 12, and 24). in both cases, the majority vote of the three radiologists was taken to determine the final reference standard label. for both ds-1 and cxr-14, in addition to the normal versus abnormal label, we also obtained labels for a selected set of findings present in the abnormal images for subgroup analysis (supplementary table 3 ). note that the lists of findings for ds-1 and cxr-14 differ. for ds-1,we selected a slightly different list of findings to represent conditions that were more clinically reliable, mutually exclusive, and for which the cxr is reasonably sensitive and specific at characterizing (supplementary methods and supplementary table 3 ). similarly to the normal versus abnormal label, the majority vote was taken for each specific finding. for cxr-14, the differences between the majority voted labels and the publically available labels are shown in a confusion matrix in supplementary table 6. tb labels tb positive cases were microbiologically confirmed. the first tb dataset 23 for the covid-19 datasets cov-1 and cov-2, patients with rt-pcr tests and cxrs were included (supplementary figure 1) . the covid-19-positive labels were derived from positive rt-pcr tests. in accordance with current centers for disease control and prevention (cdc) guidelines 26 , covid-19-negative labels consisted of cxrs from patients with at least two consecutive negative rt-pcr tests and no positive test. as false negative rates for rt-pcr have been reported to be â�¥20% in symptomatic covid-19-positive patients, cxrs from patients with only one negative rt-pcr test were excluded. 27 we trained a convolutional neural network (cnn) with a single output to distinguish between abnormal and normal cxrs. the cnn uses efficientnet-b7 28 as its feature extractor, which was pre-trained on 300 million natural images 29 . since the cnn was pre-trained on three-channel rgb natural images, we tiled the single channel cxr image to three channels for technical compatibility. we trained the cnn using the cross-entropy loss and the momentum optimizer 30 with a constant learning rate of 0.0004 and a momentum value of 0.9. during training, all images were scaled to 600x600 pixels with bilinear interpolation and image pixel values were normalized on a per-image basis to be between 0 and 1. the original bit depth for each image was used (table 1) . for regularization, we applied dropout 31 , with a dropout "keep probability" of 0.5. furthermore, data augmentation techniques were applied to the input images, including horizontal flipping, padding, cropping, and changes in brightness, saturation, hue, and contrast. all hyperparameters were selected based on the empirical performance on the ds-1 tuning set. we developed the network using tensorflow and used 10 nvidia tesla v100 graphics processing units for training. given a cxr, the dls predicts a continuous score between 0 and 1 representing the likelihood of the cxr being abnormal. for making clinical decisions, operating points are needed to threshold the scores and produce binary normal or abnormal categorizations. in this study, we selected two operating points (see "operating point selection datasets" section above), a high sensitivity operating point (95% sensitivity) and a high specificity operating point (95% specificity) for each scenario: general abnormalities for a general clinical setting in ds-1, general abnormalities for an enriched dataset in cxr-14, tb, and covid-19. to compare the dls with radiologists in classifying cxrs as normal versus abnormal, additional radiologists reviewed all test images without referencing additional clinical or patient data. all images in the ds-1 and cxr-14 test set were independently interpreted by two board-certified radiologists (with 2 and 13 years of experience), who classified each cxr as normal or abnormal. these radiologists were independent from the cohort of radiologists who contributed to the reference standard labels. each image in tb-1 and tb-2 were reviewed by a random radiologist from a cohort of 8 consultant radiologists in india. each image was annotated as abnormal or normal. each image in cov-1 and cov-2 was reviewed by one of four board-certified radiologists (with 2, 5, 13, and 22 years of experience). similarly, each image was annotated as abnormal or normal. we simulated two setups in which the dls was leveraged to optimize radiologists' workflow ( figure 1d ). in the first setup, we randomly sampled 200 cxrs from each of our 6 datasets to simulate a "batch" workload for a radiologist in a busy clinical environment. for these cxrs, we compared the turnaround time for the abnormal cxrs when (1) they were sorted randomly (to simulate a clinical workflow without the dls) and (2) when the cxrs were sorted in descending order based on the dls-predicted scores, such that cases with higher scores appeared earlier. we repeated each simulation 1,000 times per dataset to obtain the empirical distribution of turnaround differences. in the second setup, we analyzed an extreme use case where the dls identified cxrs that were unlikely to contain findings using a high sensitivity threshold, and the radiologists only reviewed the remaining cases. all cases skipped by radiologists were labeled negative. we compared the sensitivity between this simulated "reduced workload" workflow and a normal workflow in which the radiologists reviewed all cases. to evaluate the dls across different operating points, we calculated the areas under receiver operating characteristic curves (area under roc, auc). to evaluate the performance of the dls in classifying cxrs as normal or abnormal, we measured negative predictive values (npv), positive predictive values (ppv), sensitivity, specificity, percentage of predicted negatives, and percentage of predicted positives at a high specificity and a high sensitivity operating point chosen for each scenario (see "operating point selection" in deep learning system development. for evaluating the dls for each individual type of finding, we considered a "each abnormality versus normal" setup where negatives consisted of all normal cxrs, and positives consisted of only the cxrs with that particular finding. as such, specificity values were the same across all findings in a given dataset. we measured the same set of metrics to evaluate the dls performance with unseen diseases (tb and covid-19). however, the ground truth here was defined by either the respective tb or covid-19 tests, and not whether each image contained any abnormal finding. thus "negative" tb and covid-19 cases could still contain other abnormalities. confidence intervals (ci) for all evaluation metrics were calculated using the non-parametric bootstrap method with n=1,000 permutations at the image level. to compare the performance of dls with the radiologists in a dls-assisted workflow, non-inferiority tests with paired binary data were performed using the wald test procedure with a 5% margin. 32 to correct for multiple hypothesis testing, we used bonferroni correction, yielding î±=0.003125 (one-sided test with î±=0.025 divided by 8 comparisons). 33 to provide a visual explanation of how the dls makes predictions, we utilized gradient-weighted class activation mapping (grad-cam) 34 to identify the image regions critical to the model's decision-making process ( figure 5 ). because overlaying activation maps on an image obscures the original image, a common grad-cam visualization shows two images: the original image, and the image with the overlaid activation maps. here, to balance brevity and clarity, we present the activation maps as outlines highlighting the regions of interest. the outlines were obtained by taking a horizontal cross-section of the activated maps' three-dimensional contour plot, where the x and y axes represent the spatial location, and the z-axis represents the magnitude of activation. many of the datasets used in this study are publicly available. cxr-14 is a public dataset provided by the nih. 7,22 tb-1 and tb-2 are publicly available. 23 other than these public datasets, ds-1, cov-1, and cov-2 are owned by their respective institutions and are not publicly available. the deep learning framework used here (tensorflow) is available at https://www.tensorflow.org/ and the neural network architecture is available at https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet. the python libraries used for computation and plotting of the performance metrics (scipy, numpy, lifelines, and matplotlib) are available from https://www.scipy.org/, http://www.numpy.org/, and https://matplotlib.org/, respectively. n/a indicates information was not available. * abnormal images in the disease-specific datasets include both those positive for tb or covid-19, and those with other findings; the numbers of images that contained other findings were not available. table 1 and supplementary table 1 table 3 ). positive cxrs in the two tb datasets are from patients with tuberculosis. positive cxrs in the two covid-19 datasets are from patients with reverse transcription polymerase chain reaction (rt-pcr)-verified covid-19. radiologists' performances in distinguishing the test cases as normal or abnormal are also highlighted in the figures. each image has the saliency presented as red outlines that indicate the areas the dls is focusing on for identifying abnormalities, and yellow outlines representing regions of interest indicated by radiologists. text descriptions for each cxr are below the respective image . note that the general abnormality false negative example is shown with abnormal saliency maps. however, the dls predictive score on the case was lower than the selected threshold; hence the image was classified as "normal". *note that the tb false positive image was saved in the system with inverted colors, and presented to the model that way. colors have been uninverted for visualization purposes. supplementary figure 1 . the stard diagrams with inclusion and exclusion criteria for the 6 datasets. *for covid-19, the first cxr during the patient's hospital encounter was selected. â� negative tests had to be administered at least 12 hours apart. tables supplementary table 1 united nations scientific committee on the effects of atomic radiation. sources and effects of ionizing radiation radiologist supply and workload: international comparison training for rural radiology and imaging in sub-saharan africa: addressing the mismatch between services and population clinical radiology uk workforce census 2019 report. the royal college of radiologists chest radiograph interpretation with deep learning models: assessment with radiologist-adjudicated reference standards and population-adjusted evaluation deep learning for chest radiograph diagnosis: a retrospective comparison of the chexnext algorithm to practicing radiologists chestx-ray8: hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases deep learning at chest radiography: automated classification of pulmonary tuberculosis by using convolutional neural networks development and validation of deep learning-based automatic detection algorithm for malignant pulmonary nodules on chexpert: a large chest radiograph dataset with uncertainty labels and expert comparison likelihood ratios for out-of-distribution detection concrete problems in ai safety machine learning for covid-19-asking the right questions key challenges for delivering clinical impact with artificial intelligence clinical characteristics of coronavirus disease 2019 in china assessing radiology research on artificial intelligence: a brief guide for authors, reviewers, and readers-from the machine learning 'red dot': open-source, cloud, deep convolutional neural networks in chest radiograph binary normality classification automated triaging of adult chest radiographs with deep artificial neural networks deep learning for chest radiograph diagnosis in the emergency department automated abnormality classification of chest radiographs using deep convolutional neural networks training and validating a deep convolutional neural network for computer-aided detection and classification of abnormalities on frontal chest radiographs nih chest x-ray dataset of 14 common thorax disease categories two public chest x-ray datasets for computer-aided screening of pulmonary diseases automatic tuberculosis screening using chest radiographs lung segmentation in chest radiographs using anatomical atlases with nonrigid registration criteria for return to work for healthcare personnel with sars-cov-2 infection (interim guidance variation in false-negative rate of reverse transcriptase polymerase chain reaction-based sars-cov-2 tests by time since exposure rethinking model scaling for convolutional neural networks revisiting unreasonable effectiveness of data in deep learning era the authors thank the members of the google health radiology and labeling software teams for software infrastructure support, logistical support, and assistance in data labeling. for tuberculosis data collection, thanks go to sameer antani, stefan jaeger, sema candemir, zhiyun xue, alex karargyris, george r. thomas, pu-xuan lu, yi-xiang wang, michael bonifant, ellan kim, sonia qasba, and jonathan musco. sincere appreciation also goes to the radiologists who enabled this work with their image interpretation and annotation efforts throughout the study, jonny wong for coordinating the imaging annotation work, and david f. steiner, kunal nagpal, and michael d. howell for providing feedback on the manuscript. list of specific findings for ds1 we modified the list of findings from cxr-14 to include conditions that were more likely to be clinically actionable, mutually exclusive, and for which cxr is reasonably sensitive and specific for characterizing (supplementary table 3 ). for example, findings in cxr-14 such as "emphysema" (for which cxr lacks both sensitivity and specificity) and "infiltration" (an ambiguous term that overlaps other cxr-14 findings such as "pneumonia" and "atelectasis") were replaced by more specific terms. on the other hand, clinically relevant and distinct findings commonly encountered on cxr were also introduced (e.g. "hilar enlargement", "acute fracture") or augmented (e.g. "abnormal mediastinal mass/widening" rather than "hiatal hernia"). our choice of findings for the ds-1 dataset also recognized inherent limitations of cxr for reliably distinguishing between some conditions; hence "focal/multifocal lung opacity" was adopted as a single finding, rather than distinct findings for "consolidation", "atelectasis", and "fibroconsolidative opacity". *note, "other" was not part of the public labels, and one that we added to indicate findings not covered by cxr-14's original 14 conditions, and for cxrs where the radiologists did not have a majority opinion regarding the specific finding. key: cord-355218-eici4eit authors: punn, narinder singh; agarwal, sonali title: automated diagnosis of covid-19 with limited posteroanterior chest x-ray images using fine-tuned deep neural networks date: 2020-10-17 journal: appl intell doi: 10.1007/s10489-020-01900-3 sha: doc_id: 355218 cord_uid: eici4eit the novel coronavirus 2019 (covid-19) is a respiratory syndrome that resembles pneumonia. the current diagnostic procedure of covid-19 follows reverse-transcriptase polymerase chain reaction (rt-pcr) based approach which however is less sensitive to identify the virus at the initial stage. hence, a more robust and alternate diagnosis technique is desirable. recently, with the release of publicly available datasets of corona positive patients comprising of computed tomography (ct) and chest x-ray (cxr) imaging; scientists, researchers and healthcare experts are contributing for faster and automated diagnosis of covid-19 by identifying pulmonary infections using deep learning approaches to achieve better cure and treatment. these datasets have limited samples concerned with the positive covid-19 cases, which raise the challenge for unbiased learning. following from this context, this article presents the random oversampling and weighted class loss function approach for unbiased fine-tuned learning (transfer learning) in various state-of-the-art deep learning approaches such as baseline resnet, inception-v3, inception resnet-v2, densenet169, and nasnetlarge to perform binary classification (as normal and covid-19 cases) and also multi-class classification (as covid-19, pneumonia, and normal case) of posteroanterior cxr images. accuracy, precision, recall, loss, and area under the curve (auc) are utilized to evaluate the performance of the models. considering the experimental results, the performance of each model is scenario dependent; however, nasnetlarge displayed better scores in contrast to other architectures, which is further compared with other recently proposed approaches. this article also added the visual explanation to illustrate the basis of model classification and perception of covid-19 in cxr images. coronaviruses are a large family of viruses that can cause severe illness to the human being. the first known severe epidemic is severe acute respiratory syndrome ( [58] . the coronavirus (covid19) outbreak was declared a public health emergency of international concern by who on january 30, 2020 [59] . on march 11, as the number of covid-19 cases has increased thirteen times apart from china with more than 118,000 cases in 114 countries and over 4,000 deaths, who declared this a pandemic [38] . globally, many researchers of medicine, clinical and artificial intelligence areas are trying hard to mobilize preventive action plans for covid-19 with identified research priorities. since this disease is highly contagious, the most desirable preventive measure is to identify the infected people to control the spread. unfortunately, there is no well-known treatment available to cure covid-19, therefore the identified infected person must be kept in isolation to break the transmission chain as this patient may become the source of community transfer. till now, the testing kit is the only available option for diagnosis of covid-19. unavailability of testing kits due to excessive demand all over the world is a severe problem in the mission against this pandemic. though several healthcare organizations are claiming for successful development of testing kits, there is a huge gap in demand and supply. the healthcare agencies have accelerated the rate of development of low-cost testing kits, but the inability to diagnose at early-stage and due to exponential growth of covid-19 cases, medical professionals are bound to rely on other diagnostic measures. clinical studies have shown that most covid-19 patients suffer from lung infection [13] . although chest ct is a more effective imaging technique for lung-related disease diagnosis; cxr is preferred because it is widely available, faster and cheaper than ct. since covid-19 infection attacks the epithelial cells that line our respiratory tract, hence x-rays images can be used to analyse the lungs to diagnose pneumonia, lung inflammation, abscesses, and/or enlarged lymph nodes [2] . due to its easy transmission, developing techniques to accurately and easily identify the presence of covid-19 and distinguish it from other forms of flu and pneumonia is crucial. biomedical image analysis (segmentation and classification) is an admired area of research to make the healthcare system more promising [11] . in this area, advancement in computing infrastructure makes it possible to deploy the deep learning techniques for complex medical image analysis tasks. recent works have shown that the chest x-rays of patients suffering from covid-19 depicts certain abnormalities in the radiography [39] . for medical image analysis; deep learning techniques, specifically, convolutional neural networks (cnn) are very effective and efficient in feature extraction and learning, hence becoming the most popular choice among researchers [25] . cnns have been successfully deployed in the analysis of video endoscopy [12] and ct images, and also used for the diagnosis of pediatric pneumonia via chest x-ray images [6, 22] . chouhan et al. [7] proposed a transfer learning based deep network approach pre-trained on imagenet [9] for pneumonia detection. wang et al. [57] proposed a customized vgg16 model for lung regions identification to classify different types of pneumonia. later, ronneburger et al. [42] demonstrated the effectiveness of image augmentation with cnn in the presence of a small set of images. in the area of biomedical image classification, rajpurkar et al. [40] proposed a dense cnn with 121-layers to detect several pathologies including pneumonia using chest x-rays. lakhani et al. [26] obtained an area under the curve (auc) of 0.95 in pneumonia detection using alexnet and googlenet along with image augmentation. recently, several deep-learning based covid-19 detection techniques have been proposed [20, 55, 56] . linda et al. [55] introduced a deep cnn, named covid-net for the detection of covid-19 cases from the chest x-ray images. shuai et al. [56] achieved accuracy, specificity and sensitivity of 89.5%, 88% and 87% respectively for covid-19 identification using ct images. there are many datasets available about chest x-rays for the detection of pneumonia [8, 18, 32, 50] ; but in present research work, covid-19 x-ray chest images, [8] and radiological society of north america (rsna) images [50] are utilized to generate all possible samples of chest infection and also to make the study comparable with other research works. the number of covid-19 infected samples present in this dataset is very limited that may lead to biased outcome, hence the objective of this research is to maximize the learning ability in presence of a small set of positive class samples. for early diagnosis of covid-19, this article presents the effectiveness of random oversampling and weighted class loss function approaches for unbiased fine-tuned learning (transfer learning) in various state-of-the-art deep learning techniques. rest of the manuscript is organized as follows: recent research articles are discussed in section 2, and section 3 briefs the dataset. section 4 contains the proposed methodology followed by the evaluation metrics in section 5. results are discussed in section 6 whereas the last section contains concluding remarks. due to the ample availability of x-ray machines, disease diagnosis using cxr images are widely used by healthcare experts. in case of any suspect of covid-19; instead of using test kits, an alternate way to detect pneumonia from the cxr images is required, so that further investigation can be narrowed down for covid-19 identification. many studies have been performed on similar ground with several cxr datasets for diagnosis of pneumonia and other complications [8, 18, 32, 50] . these studies also advocate the need of an automated system for quick diagnosis, because the manual methods of x-ray analysis are time consuming and unable to serve the purpose due to limited availability of x-ray machine operators or radiologists. amid the covid-19 outbreak, many companies at the global level around the world embraced a flurry of artificial intelligence (ai) based solutions to detect covid-19 on chest x-ray scans. it is evident that deep learning tools are effectively used to screen mild cases, triage new infections, and monitor disease advancements. this way of diagnosis can reduce the growing burden on radiologists, and also supplant standard nucleic acid tests as the primary diagnostic tool for coronavirus infection. it is also reported that a swab test needs isolation for testing procedure, whereas chest x-ray based detection can be easily manageable. kermany et al. [22] proposed cxr imagebased deep learning model to detect pneumonia and classify other diseases using different medical datasets with testing accuracy of 92.80%. in another similar research, stephen et al. [51] illustrated an efficient deep learning approach for pneumonia classification, by using four convolutional layers and two dense layers in addition to classical image augmentation and achieved 93.73% testing accuracy. later, saraiva et al. [44] experimented convolutional neural networks to classify images of childhood pneumonia by using a deep learning model with seven convolutional layers along with three dense layers while achieving 95.30% testing accuracy. liang and zheng [27] demonstrated a transfer learning method with a deep residual network for pediatric pneumonia diagnosis with 49 convolutional layers and two dense layers and achieved 96.70% testing accuracy. in similar research, wu et al. [60] focused on convolutional deep neural learning networks and random forest to propose a pneumonia prediction using cxr images and achieved 97% testing accuracy. afterwards, narin et al. [31] proposed a deep convolutional neural network based automatic prediction model of covid-19 with the help of pre-trained transfer models using cxr images. in this research, authors used resnet50, inceptionv3 and inception-resnetv2 pre-trained models to obtain a higher prediction accuracy for a subset of x-ray dataset. apostolopoulos et. al. [3] in their study, utilised state-of-the-art convolutional neural network architectures for classifying the cxr images. transfer learning was adopted to handle various abnormalities present in the dataset. two datasets from different repositories have been used to study images of three classes: covid-19, bacterial/viral pneumonia and normal condition. the article establishes the suitability of the deep learning model with the help of accuracy, sensitivity, and specificity parameters. in another research, generative adversarial networks (gan) are used by khalifa et al. [23] to detect pneumonia from cxr images. the authors addressed the overfitting problem and claimed its robustness by generating more images through gan. the dataset containing 5863 cxr images of two categories: normal and pneumonia, has been used with typical deep learning models such as alexnet, googlenet, squeeznet and resnet18 to detect pneumonia. this research highlights that the resnet18 outperformed among other deep transfer models in combination with gan. further, sethy et al. [46] proposed a deep learning based model to identify coronavirus infections using cxr images. deep features from cxr images have been extracted and support vect or machine (svm) classifier is used to measure accuracy, false positive rate, f1 score, matthew's correlation coefficient (mcc) and kappa. it is found that resnet50 in combination with svm is statistically superior when compared to other models. later, bukhari et al. [4] also used resnet-50 cnn architectures on 278 cxr images, partitioned under 3 groups as normal, pneumonia and covid-19. this approach gave promising results and indicated substantial differentiation of pulmonary changes caused by covid-19 from the other types of pneumonia. recently, in another research work, an improved resnet-50 cnn architecture named covidresnet has been proposed [10] , where conventional resnet-50 model is applied with different training techniques including progressive resizing, cyclical learning rate finding, and discriminative learning rates to gain fast and accurate training. the experiment is performed through progressively re-sizing of input images to 128 × 128 × 3, 224 × 224 × 3 and 229 × 229 × 3 pixels, and automatic learning rate selection for fine-tuning the network at each stage. this work claimed to be computationally efficient and highly accurate for multi-class classification. a new deep anomaly detection model is developed by zhang et. al. [62] for fast and more reliable screening. to evaluate the model performance, cxr image data of covid-19 cases and other pneumonia has been collected from two different sources. to eliminate the data imbalance problem in the collected samples, authors proposed a cxr based covid-19 screening model through anomaly detection task [35] . following this context, this article proposes to contribute for early diagnosis of covid-19 using the state-of-the-art deep learning architectures, assisted with transfer learning and class imbalance learning approaches. in this research three datasets are utilized for experiments: covid-19 image [8] , radiological society of north america (rsna) [50] and u.s. national library of medicine (usnlm) collected montgomery country -nlm(mc) [19] . covid-19 image dataset is a public database of pneumonia cases with cxr images related to covid-19, mers, sars, and ards collected by cohen et al. [8] from multiple resources available at public domains without infringing patient's confidentiality (fig. 1b) . it is claimed that this dataset can help to identify characteristics of covid-19 in contrast to other types of pneumonia; therefore it can play a major role in predicting survival rate. the dataset includes the statistics up to march 25, 2020 consisting of 5 types of pneumonia such as sarsr-cov-2 or covid-19, sarsr-cov-1 or sars, streptococcus spp., pneumocystis spp. and ards with following attributes: patient id, offset, sex, age, finding, survival, view, modality, date, location, filename, doi, url, license, clinical notes, and other notes. another dataset utilized in this study is published under rsna pneumonia detection challenge is a subset of 30,000 examinations taken from the nih cxr14 dataset [50] . out of 30,000 selected images, 15,000 examinations had positive cases of pneumonia and from the remaining 15000 cases, 7500 cases had no findings and other 7500 cases had symptoms other than pneumonia. all these images are annotated by a group of experts including radiologists in two stages. a sample image is shown in fig. 1c . this dataset has been published in two stages. in stage one, 25,684 training images were considered to test 1,000 images. later in stage two 1000 testing samples were added to the training set to form the dataset of 26,684 training images and a new set of 3,000 radiographs were introduced for the test. for robust testing and comprehensive coverage of the comparative analysis, nlm(mc) [19] dataset is also utilized that consists of 138 chest posterior-anterior xrays samples of tuberculosis and normal cases. a sample image is represented in fig. 1d . table 1 presents the class summary details of the fused dataset resulting from the above discussed datasets which is utilized for training, testing and validation of the proposed approach. the fused dataset is composed of 1214 posteroanterior chest x-ray samples with classes labeled as covid-19 (108), other pneumonia (515), tuberculosis (58) and normal (533). the generated fused dataset is publicly available [36] . the era of artificial intelligence has brought significant improvements in the living society [30] . the recent advancements in deep learning have extended its domain in various applications such as healthcare, pixel restoration, visual recognition, signal processing, and a lot more [28] . in healthcare domain, the deep learning based image processing approaches for classification and segmentation are applied for faster, efficient, and early diagnosis of the deadly diseases e.g. breast cancer, brain tumor, etc. by using different imaging modalities such as x-ray, ct, mri, [47] and fused modalities [37] along with its future possibilities. the success of these approaches is dependent on the large amount of data availability, which however is not in the case of automated covid-19 detection. the main contribution of the work is divided into the components as shown in fig. 2 . it has two concrete components: data preprocessing and classification. covid-19 image, rsna and nlm(mc) datasets are used to generate the final working set. the newly generated dataset contains cxr images of the following classes: coronavirus caused diseases, pneumonia, other diseases and normal cases. further, binary classification (covid-19 vs others) and multi-class classification (covid-19, other types of pneumonia, tuberculosis and normal) are achieved using random oversampling and weighted class loss function approaches for unbiased fine-tuned learning (transfer learning) in various state-of-the-art deep learning approaches such as baseline resnet, inception-v3, inception resnet-v2, densenet169, and nasnetlarge [17, 52, 54, 63] . the trained models are utilized for identification and classification of covid-19 in novel samples. later, visualization techniques are utilized to understand and elaborate the basis of the classification results. class balancing techniques are necessary when minority classes are more important. the dataset used in this research is highly imbalanced which may lead to biased learning of the model. number of coronavirus infected cxr images are very less compared to other classes, hence class balancing techniques must be insured to smoothen the learning process. this section discusses two approaches to handle the class imbalance problem: weight class approach and random oversampling [21] . in this approach, the intention is to balance the data by altering the weights that each training sample class carries when computing the loss. normally, each class carries equal weights, but sometimes certain classes with minority samples are required to hold more weights if they are more important because training examples within that class should have a significant effect on the loss function. in the used dataset the coronavirus infected image class samples must be given more weights as they are more significant. in this article, the weights for each class is generated based on the eq. (1). where c c is the class constant for a class c, n is the number of classes, and n c is the number of samples in a class c. the computed class weights are later fused with the objective function (loss function) of the deep learning model in order to heavily penalize the false predictions concerned with the minority samples, which in this case is coronavirus. in this approach, the objective is to increase the number of minority samples by utilizing the existing samples belonging to the minority class. the minority samples are increased until the samples associated with every class become equal. hence the procedure follows by identifying the difference between the number of samples in majority and minority class. to fill this void of difference, the samples are generated from the randomly selected sample belonging to the minority class by applying certain statistical operations. in this work, the samples of cxr image of covid-19 positive cases are less as compared to other classes, therefore, these minority class images are randomly oversampled by means of rotation, scaling, and displacement with the objective to achieve equal distribution of classes and accommodate unbiased learning among the deep learning models. based on the type of data samples availability of cxr images the covid-19 classification is divided into two following schemes: -binary classification -in this classification scheme, the coronavirus positive samples labelled as "1" (covid-19) are identified against the rest of the samples labelled as "0" (non covid-19 case) which involves other cases e.g. chlamydophila, sars, streptococcus, tuberculosis, etc., along with the normal cases. -multi-class classification-in this classification scheme, the aim is to distinguish and identify the covid-19 samples from the other pneumonia cases along with the presence of tuberculosis and normal case findings. the multi-class classification is performed with three and four classes. the three classes are provided with labels as "0" being a normal case, "1" being a covid-19 case, and "2" being other pneumonia and tuberculosis cases, whereas four classes are labeled as "0" being a normal case, "1" being a covid-19 case, and "2" being other pneumonia case and "3" as tuberculosis case. in both the classification strategies, the deep learning models are trained with the above discussed imbalanced learning approaches using the weighted categorical cross entropy (wce) loss function as given by eq. (2) and eq. 3 [24] : n. s. punn and s. agarwal in categorical cross entropy, the distribution of the predictions (the activations in the output layer, one for each class) is compared with the true distribution only, to ensure the clear representation of the true class as one-hot encoded vector; here, closer the model's outputs are to that vector, the lower the loss. in this article, due to the limited samples of posteroanterior chest x-ray images concerned with positive covid-19 [8] cases, the data samples are mixed with the other randomly selected cxr images selected from other datasets-, rsna [50] and nlm(mc) [19] . the rsna and nlm(mc) datasets consists of posteroanterior cxr images covering sample cases labelled as pneumonia and tuberculosis respectively along with normal samples. table 2 describes the distribution of training, testing, and validation sets using the fused dataset for binary and multiclass classification along with different class imbalance strategies i.e. class weighted loss function that penalizes the model for any false negative prediction and random oversampling [61] of minority classes which in this case is covid-19. the cxr images in the aggregated dataset also consists of unwanted artifacts such as bright texts, symbols, varying resolutions and pixel level noise, which necessitates its preprocessing. in order to suppress the highlighted textual and symbolic noise, the images are inpainted with the image mask generated using binary thresholding [34] as given by eq. (4), followed by resizing the images to a fixed size resolution of 331 × 331 × 3. m(x, y) = max th, i(x, y) ≥ min th. 0, otherwise. where i(x,y) is an input image, max th and min th are max and min thresholds to design the mask. despite filtering the unwanted information, there is still the possibility of uncertainty at the deep pixel level representation [15] . the denoising or removal of such uncertainty is carried through the adaptive total variation method [53] while preserving the original distribution of pixel values. let for a given grayscale image f, on a bounded set over r 2 , where ⊂ r 2 , denoising image u that closely matches to observed image x = (x 1 , x 2 ) -pixels, given as where ω(x) = 1 1+k mod g σ * u , g ρ -the gaussian kernel for smoothing with σ variance, k > 0 is contrast parameter and * is convolution operator. figure 3 illustrates the data preprocessing stages by considering an instance of covid-19 case consisting of textual and symbolic artifacts from the generated dataset. the resulting distributed pixels histograms at each stage of preprocessing shown in fig. 3 , illustrates that the preprocessing approach tends to preserve the original nature of distribution of the pixels while removing the irregular intensities. the preprocessed images are then divided into training, testing, and validation set for training and evaluation of the state-of-the-art deep learning classification models. this section incorporates the state-of-the-art deep learning models utilized in the present research work as shown in table 3 along with their respective contribution, parameters, and performance on the standard benchmark datasets. the inception deep convolutional architectures proposed by googlenet are considered as the state-ofthe-art deep learning architectures for image analysis and object identification with the basic model as inception-v1 [5] . later, this base model was refined by introducing the batch normalization and established as the inception-v2 [54] . in further iterations, additional factorisation was introduced and released as the inception-v3. it is one of the pre-trained models to perform two types of specific tasks: dimensionality reduction using cnn and classification using fully-connected and softmax layers. since it is originally trained on over a million images consisting of 1,000 classes of imagenet, its head layers can be retrained for the generated dataset using historical knowledge to reduce the extensive training and computational power. later, inception-resnet-v2 was proposed by szegedy et al. [52] . this hybrid model is a combination of residual connections and a recent version of inception architecture. it is intended to train very deep convolutional models by the additive merging of signals, both for image recognition and object detection. this network is more robust and learns rich feature representations. afterwards, densenet was proposed by huang et al. [17] . it works on the concept of reuse, in which each layer receives inputs from all previous layers and yields a condensed model to pass its own featuremaps to all subsequent layers. this makes the network thinner and compact with the fewer number of channels, while improving variation in the input of subsequent layers, and becomes easy to train and highly parameter efficient. google brain research team proposed nasnet model based on reinforcement learning search methods [63] . it creates search space by factoring the network into cells and further dividing it into a number of blocks. each block is supported by a set of popular operations in cnn models with various kernel size e.g: convolutions, max pooling, average pooling, dilated convolution, depth-wise separable convolutions etc. training deep learning models require (inception-v3, inceptionresnet-v2, etc.) exhaustive amount of resources and time. while these networks, attain relatively excellent performance on imagenet [9] , training them on a cpu is an exercise in futility. these cnns are often trained for a couple of weeks or more using arrays of gpus to get good results on the complex and complicated datasets. in most deep cnns the first few convolution layers learn low-level features (edges, curves, blobs) and with progress, through the network, it learns more mid/high-level features or patterns associated with the on-going task. in fine tuning, the aim is to keep or freeze these trained low-level features, and only train the high-level features needed for our new image classification problem. in this article, the trials of training the deep learning classification models initiates with the baseline residual network that is composed of five residual blocks [16] , defined by two convolutions whose rectified linear unit (relu) activation is concatenated with the input of the first convolution layer, followed by max-pooling and instance normalization, except the final output layer which uses softmax activation function. later, transfer learning is utilized on the state-of-the-art architectures discussed in table 3 to utilize the pre-trained models while fine tuning the head layers to serve the purpose of classifying the covid-19 samples, as illustrated in fig. 4 . this follows by enabling the four head layers of the network to adjust its trainable parameters along with the addition of fully connected layers with 128 neurons and 2 or 3 neurons in the output layer depending on binary or multi-class classification, accompanied with the softmax activation function. the deep learning models are trained using the training and validation set under consideration of each possible combination of the discussed approaches of class imbalance learning and classification strategy, thereby making four possible scenarios for training a model. later, the trained models are evaluated on the test set using the standard benchmark performance metrics such as accuracy, precision, recall (selectivity), area under curve (auc), specificity and f1 score (dice coefficient) as shown in fig. 5 . the proposed approach is trained and tested on the above discussed fused dataset using mini-batch gradient descent with learning rate optimizer as adam [43] via n-vidia titan gpu. the training, testing, and validation sets are highlighted in table 2 . the training phase is assisted with the batch size of 10, 4-fold cross validation, switch normalization [29] to adapt to instance or batch or layer normalization, and earlystopping technique that aims to stop the training process if the validation error stops decreasing. for each epoch the above discussed evaluation metrics are computed on the training and validation set to analyse the model's performance and improve the classification results. later, the test set is utilized to evaluate the results of the proposed approach. s with extensive trials, it is observed that there is no individual model that displayed best performance for all the scenarios in terms of accuracy, precision, recall, specificity, auc, and f1-score. however, the nasnetlarge mostly claimed the best scores, thus making it best fit for the classification of covid-19 samples. it is also observed that the results of binary classification (covid-19 vs non-covid-19) are better than the multi-class classification (covid-19 vs other classes). with this it is evident that by grouping other diseases together as non-covid-19 samples, models can efficiently learn features and patterns belonging to the covid-19. figure 6 shows the average performance curves of the nasnetlarge model evaluated and monitored on the test set for each iteration during the training phase for classification of covid-19 samples, generated using the tensorboard. since the models are initialized with the trained weights on the imagenet and head layers are fine-tuned for classification, the training of the models initiates with descent metrics values. the best scores of the models corresponding to each scenario are highlighted in table 4 bold numbers highlight the best scores of the models corresponding to each scenario approaches to detect covid-19 positive symptoms via xray images as shown in table 5 . it is observed that the proposed approach outperforms other approaches, whereas apostolopoulos et al. [3] achieved similar results with the help of vgg19 model [49] but with approximately double trainable parameters (144m) as compared to the proposed approach (83m). training the model and getting the results is not sufficient unless it is understood that what is triggering the concerned output. to handle this blackbox, visualization techniques assist in illustrating the basis of prediction of the model. there are many visualization techniques for example class activation maps (cam) [45] , saliency maps (sm) [48] , local interpretable model-agnostic explanations (lime) [41] , and a lot more. in this article, activation maps and lime techniques are utilized to present the model perception of identifying and classifying the covid-19 samples from cxr images. cam aims at understanding the feature space of an input image that influences the prediction, whereas lime is an innovative explanation technique to represent the model prediction with local fidelity, interpretability and model agnostic. for instance, fine-tuned nasnetlarge architecture is considered to generate lime explanations for some samples taken from the test set (see fig. 7 ), whereas class activation maps tend to present the patterns learned by the model for classification of samples as shown in fig. 8 . figure 7 presents the lime technique applied to four samples belonging to four distinct classes as covid-19, other types of pneumonia, tuberculosis and normal cases. the red and green areas in the lime generated explanation correspond to the regions that contributed against the predicted class and towards the predicted class respectively. this article proposes to leverage the state-of-the-art deep learning models to aid in early identification and diagnosis of covid-19 virus by using the limited posteroanterior chest x-ray images. each trained model was evaluated using benchmark performance metrics e.g. accuracy, precision, recall, area under curve, specificity, and f1 score under four different scenarios concerned with imbalanced learning and classification strategy. with extensive trials, it was observed that models achieve different scores in different scenarios, among which nasnetlarge displayed better performance specially in binary classification of covid-19 samples. the visual representation based on local interpretable model agnostic explanations is utilized to understand the basis of prediction of the model. as an extension to this work more deep learning models and preprocessing techniques can be explored to achieve better results. covid-caps: a capsule network-based framework for identification of covid-19 cases from x-ray images eosinophilic lung diseases covid-19: automatic detection from x-ray images utilizing transfer learning with convolutional neural networks the diagnostic evaluation of convolutional neural network (cnn) for the assessment of chest x-ray of patients infected with covid-19 adaptive total variation denoising based on difference curvature deep learning-based image conversion of ct reconstruction kernels improves radiomics reproducibility for pulmonary nodules or masses a novel transfer learning based approach for pneumonia detection in chest x-ray images covid-19 image data collection imagenet: a large-scale hierarchical image database covid-resnet: a deep learning framework for screening of covid19 from radiographs deep learning in medical image analysis: a third eye for doctors low-light image enhancement of high-speed endoscopic videos using a convolutional neural network the origin, transmission and clinical therapies on coronavirus disease 2019 (covid-19) outbreak-an update on the status finding covid-19 from chest x-rays using deep learning on a small dataset computer vision: a reference guide deep residual learning for image recognition densely connected convolutional networks chexpert: a large chest radiograph dataset with uncertainty labels and expert comparison two public chest x-ray datasets for computer-aided screening of pulmonary diseases using deep learning to detect pneumonia caused by ncov-19 from x-ray images survey on deep learning with class imbalance identifying medical diagnoses and treatable diseases by image-based deep learning detection of coronavirus (covid-19) associated pneumonia based on generative adversarial networks and a fine-tuned deep transfer learning model using chest x-ray dataset loss functions in classification tasks imagenet classification with deep convolutional neural networks deep learning at chest radiography: automated classification of pulmonary tuberculosis by using convolutional neural networks a transfer learning method with deep residual network for pediatric pneumonia diagnosis a survey of deep neural network architectures and their applications differentiable learning-to-normalize via switchable normalization the forthcoming artificial intelligence (ai) revolution: its impact on society and firms automatic detection of coronavirus disease (covid-19) using x-ray images and deep convolutional neural networks chexnet: an in-depth review deep learning covid-19 features on cxr using limited training data sets image thresholding deep anomaly detection with deviation networks covid-19 posteroanterior chest x-ray fused (cpcxr) dataset inception u-net architecture for semantic segmentation to identify nuclei in microscopy cell images covid-19 epidemic analysis using machine learning and deep learning algorithms medrxiv harmony-search and otsu based system for coronavirus disease (covid-19) detection using lung ct scan images deep learning for chest radiograph diagnosis: a retrospective comparison of the chexnext algorithm to practicing radiologists explaining the predictions of any classifier u-net: convolutional networks for biomedical image segmentation an overview of gradient descent optimization algorithms classification of images of childhood pneumonia using convolutional neural networks grad-cam: visual explanations from deep networks via gradient-based localization detection of coronavirus disease (covid-19) based on deep features deep learning in medical image analysis deep inside convolutional networks: visualising image classification models and saliency maps very deep convolutional networks for large-scale image recognition pneumonia dataset annotation methods. rsna pneumonia detection challenge discussion an efficient deep learning approach to pneumonia classification in healthcare inception-v4, inception-resnet and the impact of residual connections on learning going deeper with convolutions rethinking the inception architecture for computer vision covid-net: a tailored deep convolutional neural network design for detection of covid-19 cases from chest radiography images a deep learning algorithm using ct images to screen for corona virus disease hospital-scale chest x-ray database and benchmarks on weaklysupervised classification and localization of common thorax diseases who (2020) coronavirus disease 2019 (covid-19) situation report-81 coronaviruse/situation-reports who (2020) who timeline-covid-19 predict pneumonia with chest x-ray images based on convolutional deep neural learning networks an application of oversampling, undersampling, bagging and boosting in handling imbalanced datasets covid-19 screening on chest x-ray images using deep learning based anomaly detection learning transferable architectures for scalable image recognition we are very much obliged to our institute, indian institute of information technology allahabad (iiita), india and big data analytics (bda) lab for ensuring the needed resources and support. we also would like to extend our thanks to the colleagues for their valuable guidance and suggestions. the manuscript is also available in the pre-print arxiv with the eprint-2004.11676 under the license attribution 4.0 international (cc by 4.0). conflict of interest the authors have no conflict of interest to declare.ethical approval this article does not contain any studies with human participants or animals performed by any of the authors.publisher's note springer nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. key: cord-334495-7y1la856 authors: agricola, eustachio; beneduce, alessandro; esposito, antonio; ingallina, giacomo; palumbo, diego; palmisano, anna; ancona, francesco; baldetti, luca; pagnesi, matteo; melisurgo, giulio; zangrillo, alberto; de cobelli, francesco title: heart and lung multimodality imaging in covid-19 date: 2020-06-24 journal: jacc cardiovasc imaging doi: 10.1016/j.jcmg.2020.05.017 sha: doc_id: 334495 cord_uid: 7y1la856 abstract sars-cov-2 outbreak has rapidly reached a pandemic proportion and has become a major threaten to global health. although the predominant clinical feature of covid-19 is an acute respiratory syndrome of varying severity, ranging from mild symptomatic interstitial pneumonia to acute respiratory distress syndrome, the cardiovascular system can be involved with several facets. as many as 40% hospitalized patients presenting with covid-19 have pre-existing history of cardiovascular disease and current estimates report a proportion of myocardial injury in covid-19 patients ranging up to 12%. multiple pathways have been advocated to explain this finding and the related clinical scenarios, encompassing local and systemic inflammatory response and oxygen supply-demand imbalance. from a clinical point of view, cardiac involvement during covid-19 may present a wide spectrum of severity ranging from subclinical myocardial injury to well-defined clinical entities (myocarditis, myocardial infarction, pulmonary embolism and heart failure), whose incidence and prognostic implications are currently largely unknown due to a significant lack of imaging data. the use of integrated heart and lung multimodality imaging plays a central role in different clinical settings and is essential in diagnosis, risk stratification and management of covid-19 patients. aim of this review is to summarize imaging-oriented pathophysiological mechanisms of lung and cardiac involvement in covid-19 and to provide a guide for an integrated imaging assessment in these patients. the severe acute respiratory syndrome coronavirus 2 (sars-cov-2) outbreak arisen in central china at the end of december 2019 has rapidly reached a pandemic proportion and the associated disease (covid-19) has become a major threaten to global health (1) . as the pandemic grows, treating physicians are challenged with different and complex clinical scenarios. the most prominent feature of covid-19 is an acute respiratory syndrome of varying severity, ranging from mild symptomatic interstitial pneumonia to acute respiratory distress syndrome (ards). however, several reports have stirred the attention to possible cardiovascular involvement during sars-cov-2 infection: as many as 40% hospitalized patients presenting with covid-19 have pre-existing history of cardiovascular disease and current estimates report a proportion of myocardial injury in covid-19 patients ranging up to 12% (2) (3) (4) . identification of myocardial injury is associated to a dismal prognosis independently and on top of coexisting previous cardiovascular diseases, therefore recognition of its underlying mechanisms may offer a therapeutic opportunity (4) . in this context, the use of multiple diagnostic imaging techniques may apply to both heart and lung to provide an integrated assessment of cardiac and pulmonary function and to refine diagnosis, risk stratification and management of covid-19 patients. the pathogenesis of covid-19 is characterized by two distinctive but synergistic mechanisms, the first related to viral replication and the second to host immune response (5) . the disease primarily involves the lungs and progresses through three stages with increasing severity, corresponding to distinct histopathological, imaging and clinical findings(6-8). 1. the first stage involves incubation period, sars-cov-2 replication in the respiratory system and potential spread to target organs. during this phase alveolar and interstitial inflammation is mild, patchy and usually shows bilateral, peripheral and lower distribution, with patients presenting mild respiratory and systemic symptoms. 2. the second stage is characterized by localized lung inflammation, that shows different grades of severity, ranging from severe interstitial inflammation and thickening to air space consolidation. patients develop symptoms of viral pneumonia and eventually hypoxia, leading to clinical deterioration and need for hospitalization. 3 . in a subgroup of patients transition to the third stage occurs. this phase is dominated by widespread lung inflammation and systemic inflammatory syndrome triggered by dysregulated host immune response and cytokine storm, causing hyperinflammation, ards, shock and multi-organ damage. clinical features of covid-19 are variable. while the majority of patients present with only mild respiratory and systemic symptoms, some progress to severe forms of viral pneumonia and eventually develop severe systemic inflammatory manifestations, with an increasingly higher case-fatality rate (7) . cardiovascular adverse events may occur at different stages complicating the course of the disease and leading to unfavorable outcomes (central illustration). definition of cardiac involvement in covid-19 is challenging, as sars-cov-2 infection has multifaceted effects. from a clinical point of view, cardiac involvement during covid-19 may present a wide spectrum of severity ranging from subclinical myocardial injury to well-defined clinical entities. in a comprehensive understanding, the following clinical scenarios may be encountered: a) primary cardiac involvement; b) secondary cardiac involvement; c) worsening of previous cardiovascular diseases ( table 1) . primary cardiac involvement. this may be the consequence of viral tropism for the endothelium and (presumably) for the myocardium. a link between the respiratory syndrome and the pleomorphic cardiovascular manifestations associated with covid-19 could be identified in the angiotensin converting enzyme 2 (ace-2), a membrane-bound enzyme that serves as cell-entry receptor for the sars-cov-2(9). this receptor is expressed in a variety of tissues, including lung alveolar epithelial cells and enterocytes of the small intestine, as well as arterial smooth muscle cells and endothelial cells (9) . based on previous data from the sars-cov epidemic, myocardial infection by coronavirus is a possibility: in an autopsy series, sars-cov rna was found in 35% sampled hearts, along with macrophage infiltration and myocardial damage (10) . the extent to which these finding may also apply to sars-cov-2 is unknown. to date, no cases of sars-cov-2 nucleic acid isolation from myocardial specimens have been described. however, several cases reported on the occurrence of severe myocarditis during laboratory-proven covid-19 (11) (12) (13) (14) (15) . in all these cases myocarditis caused severe left ventricular dysfunction, but showed some degree of systolic function recovery following medical therapy, ranging from progressive improvement to complete myocardial function restoration. a single case of myo-pericarditis complicated by life-threatening cardiac tamponade has been reported, again without direct isolation of sars-cov-2 from the drained pericardial fluid (12) . in the absence of proven sars-cov-2 viral infection of the myocardium, the clinical picture overlap of these case reports with other possible differential diagnoses calls for prudence in diagnosing sars-cov-2 virus-related myocarditis. secondary cardiac involvement. this is the result of indirect myocardial damage during sars-cov-2 infection. of note, it may represent the convergence of multiple different mechanisms. in a post-mortem examination from a covid-19 patient who developed ards, interstitial mononuclear inflammatory cells were noted in heart specimens without structural damage (16) . hyperinflammatory response in advanced stage of the disease elicits a cytokine storm, chiefly mediated by il-1 and il-6 pathways closely resembling hemophagocytic lymphohistiocytosis, a life-threatening hematologic disorder characterized by uncontrolled proliferation of activated lymphocytes and macrophages, with massive release of inflammatory cytokines (9) . these cytokines have been implied in myocardial injury and adverse remodeling in clinical and experimental models of acute coronary syndromes (acs) and may exhibit direct negative inotropic and metabolic effect onto cardiomyocytes in sepsis-like settings (17) . in addition, il-1 plays a proven role in atherothrombosis and the resulting hyperinflammatory milieu may provoke atherosclerotic plaque instability and a pro-coagulant state with increased risk of arterial and venous acute thrombotic events, including type 1 myocardial infarction (mi) and pulmonary embolism (pe). indeed, there is raising concern that covid-19 patients are more prone to develop thromboembolic venous events and disseminated intravascular coagulation (18, 19) . secondary cardiac involvement may also be the consequence of hypoxia-induced myocardial damage that could lead to type 2 mi. this condition could either unmask underlaying obstructive coronary artery disease, or present as myocardial infarction with non-obstructive coronary arteries (minoca) in case of intense oxygen supply-demand imbalance (20) . moreover, altered pulmonary hemodynamics may play a role in secondary cardiac involvement. in severe covid-19 pneumonia, use of higher positive end-expiratory pressure may be associated with increased right ventricular (rv) afterload and strain due to higher pulmonary arterial pressure and pulmonary vascular resistance. pulmonary circulation hypoxic vasoconstriction and superimposed pulmonary thromboembolic events may further precipitate these effects. worsening of previously existing cardiovascular diseases. this is frequently observed during covid-19 and may explain the higher prevalence of patients with pre-existing cardiovascular comorbidities in the non-survivor cohorts (3, 4, 21) . indeed, patients with heart failure (hf) are particularly vulnerable to hemodynamic decompensation during viral infections (22) . furthermore, in predisposed patients, arrhythmias may ensue as a result of multiple mechanisms, including hypoxia, systemic inflammation and side effects of drugs used in the treatment of covid-19 (i.e. hydroxychloroquine often combined with azithromycin) (2) . chest x-ray. the recent covid-19 radiological literature has been molded by the chinese experience, with the vast majority of reports focusing on the role of chest computed tomography (ct), almost neglecting chest x-ray (cxr) contribution. on the other hand, european hospitals have drawn diagnostic algorithms in which cxr is described as a first line triage tool, mainly due to its availability and feasibility and to long reverse transcription polymerase chain reaction (rt-pcr) turnaround times. furthermore, the american college of radiology points out that ct room decontamination after scanning covid-19 patients may disrupt radiological service availability, and suggests that portable chest cxr might be considered the optimal tool to minimize the risk of cross infection (23) . as recently reported, cxr demonstrates typical radiographic features in the vast majority of covid-19 patients, including ground-glass opacities and consolidation, while pleural effusion is not common (table 2 and figure 1 ). in a retrospective cohort of 64 patients, wong et al. found that the common ct findings of bilateral involvement, peripheral distribution, and lower zone dominance can also be assessed on cxr and that severity of cxr findings peaked at 10-12 days after symptoms onset, consistently with previous ct reports (24) . despite the fact that 6 out of 64 patients demonstrated cxr abnormalities before eventually testing positive on rt-pcr, baseline cxr sensitivity resulted 69%, being significantly lower than that reported for initial rt-pcr and baseline ct (25) . moreover, differently from what has been previously reported about chest ct, radiographic and virologic recovery times were not significantly different, thus reducing the role of cxr in clinical monitoring (25) . a retrospective analysis of 9 south korean patients who underwent both chest ct and cxr further decreased the sensitivity of cxr imaging in detecting covid-19 pneumonia to 33.3% (26) . however, the significance of this result is limited by the small sample size. recently, bandirali et al. proposed a role for cxr in asymptomatic or minimally symptomatic patients in epidemic regions, which may have positive radiographic findings even after 14 days of quarantine (27) . up to date, there is no consistent report accurately depicting the course of disease on serial cxr images. chest computed tomography. chest ct is a highly accurate imaging modality for pneumonia identification and characterization. as recently reported, chest ct demonstrates typical imaging features in covid-19 patients, including bilateral ground-glass opacities (ggos), crazy paving pattern (ggos with superimposed inter/intralobular septal thickening) and/or consolidations, predominantly in subpleural locations in the lower lobes; typically, discrete pulmonary nodules, lung cavitation, pleural effusion and lymphadenopathies are not present(28,29)(table 2 and figure 2 ). pan et al. demonstrated that multiple ct scans could accurately depict the course of disease, summarized in 4 ct-based stages (28) . the typical covid-19 pneumonia often starts as small subpleural ggos, mainly affecting the lower lobes (early stage, 0-4 days after symptoms' onset), which then rapidly develops into crazy paving pattern and consolidation areas, typically affecting both lungs (progressive stage, 5-8 days after symptoms' onset). thereafter, dense consolidation become the most frequent finding (peak stage, 9-13 days after symptoms' onset). when infection resolves the consolidation areas are gradually absorbed with residual ggos and subpleural fibrotic parenchymal bands (absorption stage, >2 weeks after symptoms' onset) (figure 2) . ai et al. found that with rt-pcr as a reference, the sensitivity of chest ct imaging for covid-19 is 97% (25) . interestingly, these radiological findings are also observed in patients with clinical symptoms but negative rt-pcr results and that almost 50% and 33% of these patients have been respectively reconsidered as highly likely cases and as probable cases by a comprehensive evaluation (25) . furthermore, 60% to 93% of patients had initial positive chest ct consistent with covid-19 before the initial positive rt-pcr results (25) . finally, 42% of patients showed improvement of follow-up chest ct scans before the rt-pcr results turning negative (25) . nevertheless, it is worth emphasizing that patients with rt-pcr confirmed covid-19 infection might have normal chest ct findings at admission, when disease is still subtle (30) . additionally, chest ct can be used for characterization of covid-19 pneumonia severity. yang et al. proposed a ct-based severity score defined by summing up individual scores from 20 lung regions: the individual scores in each lung, as well as the global severity score, were found to be higher in severe covid-19 when compared with mild cases (sensitivity: 83.3%, specificity: 94%)(31). for lung evaluation presenting features that make it very attractive for assessment of patients affected by covid-19 (32) (33) (34) . lus can be performed with any two dimensional scanner, including portable ones, using linear, convex or phase array probes. specifically, highfrequency linear probe is recommended to assess the pleural line, phase array low-frequency probe is suggested to evaluate deep consolidation, while micro convex probe with small footprint is useful for evaluating posterior fields in supine patients. the entire chest can be scanned with the probe oriented longitudinally or obliquely along the intercostal spaces. the scanning protocol consists in 12-zone examination with 6 regions per hemithorax: upper and lower parts of anterior, lateral, and posterior chest wall demarcated by the anterior and posterior axillary line (32, 33) . covid-19 pneumonia is characterized by initial interstitial damage with a bilateral, peripheral and posterior distribution followed by parenchymal involvement (34) . lus effectively detects the areas affected by subpleural interstitial syndrome with the appearance of b-lines, which increase in number as the pathology spreads up covering most of the pleural line. these findings correspond to ggos and reticular pattern at ct scan ( table 2) . the characteristics of the b-lines help to distinguish within interstitial syndrome between pneumonia or ards and cardiogenic pulmonary edema. specifically, inflammatory patterns are characterized by the presence of bilateral, irregularly distributed b-lines with spared areas and coalescent b-lines mostly in posterior fields; furthermore, the pleural line appears typically thickened and irregular with reduced or absent lung sliding (32) . as the disease progresses, lung consolidations become frequent. the subpleural consolidation areas are identified as anechoic hemispheric areas close to the pleural line with a hyperechogenic base. extensive consolidation appears as non-translobar and translobar consolidation with hepatization of lung tissue and air bronchogram which distinguish them from consolidations in resorptive atelectasis (figure 3) . however, lus also presents limits since it is operator dependent and abnormalities affecting the central regions surrounded by aerated lung are not detectable. with the aim of increasing the reproducibility it would be convenient to establish a scanning model and a severity score. the lus score, validated with the chest ct comparison, provides a numerical assessment of regional loss of aeration that can be used to assess the response to treatments(33) (figure 3 ). echocardiography. even though echocardiography should not routinely be performed in patients with covid-19 and restricted to those in whom it is likely to result in a change in management, bedside echocardiography is a clinically useful tool in different clinical settings in emergency department (ed), intensive care unit (icu) and non-icu wards (35) . compact and highly mobile machines should be the ideal ultrasound system to adopt, privileging dedicated probes and machines in infected areas. a miniaturized handheld ultrasound equipment that can be easily protected and cleaned may be an alternative option (35, 36) . a pragmatic strategy based on the use of focused cardiac ultrasound (focus) seems the most reasonable approach (37) . focus should be combined with lus for the evaluation of patients with respiratory failure. the covid-19 crisis highlights the need for imagers to be cross-trained (lus and focus) and be more nimble: sonographers, cardiologists, and emergency physicians who are not familiar with lus can learn quickly with initial support of expert colleagues and web resources (38) . however, since focus is not being performed as the definitive diagnostic test, if no usable information is obtained, comprehensive echocardiogram and/or other diagnostic testing have to be considered (37) . the aim of echocardiography is to reliably identify cardiac abnormalities and coexisting heart disease in order to facilitate triage and guide patient management. echocardiography is also recommended for the evaluation of patients who develop symptoms consistent with a cardiac etiology. information must quickly include biventricular function, gross valvular abnormalities, wall motion abnormalities, pericardial effusions and surrogates of a patient's volume status, including inferior vena cava collapsibility and ventricular size (37) . transthoracic echocardiography (tte) is the standard technique, while transoesophageal echocardiography (toe) should be avoided due to the high risk of equipment and personnel contamination, unless there is a clearly defined indication that requires toe imaging or inadequate tte imaging quality due to patient-specific factors (intubated patients, poor image quality, inability to position the critically ill patient for optimal image acquisition) (35) . the most common echocardiographic abnormalities encountered in our experience on covid-19 patients in the non-icu setting are reported in table 3 . acute worsening of respiratory symptoms is a leading indication for performing echocardiography in these patients, frequently depicting a picture of acute cor pulmonale: rv dilatation, paradoxical septal motion and pulmonary hypertension. in this clinical setting pe seems relatively frequent (figure 4) . echocardiography may expedite diagnosis of this condition. ct coronary angiography is a well-established tool to effectively and safely rule-out cad in the setting of acute chest pain, thanks to its excellent negative predictive value (95-100%) (39) . of note, ct angiography can combine coronary arteries, pulmonary arteries and thoracic aorta assessment using dedicated "triple rule-out" (tro) protocols. in selected patients with variable degrees of respiratory symptoms, showing cardiac enzyme and ddimer elevation, a dedicated tro approach, with lung parenchyma instead of thoracic aorta as the third focus of the examination, may solve different clinical questions in one sitting (40) . although most of the currently available ct scanners allow to image coronary arteries with high-resolution and limited motion artifacts, clinical judgement is advised, since dedicated scanners can improve image quality. additionally, ct angiography could rule-out left atrial appendage thrombus, allowing direct-current cardioversion in patients with atrial fibrillation, thereby limiting operator exposure deriving from toe examination. moreover, cardiac ct could provide advanced diagnostic assessment through myocardial characterization (41) . indeed, ct examination can be completed with a delayed iodine-enhanced scan to identify areas of myocardial necrosis or fibrosis. this further evaluation may result especially useful in patients with minoca, allowing to differentiate myocardial infarction from stresscardiomyopathy, which is typically characterized by absence of myocardial late enhancement, and to diagnose acute myocarditis, detecting myocardial scar with typical nonischemic pattern. in this case, one can speak of "quadruple rule-out" having a single examination looking for lung involvement, coronary and pulmonary artery patency and myocardial scar (42) . however, cardiac ct remains limited in the detection of myocardial edema, which represents the hallmark of acute myocardial inflammation (41) . cardiac magnetic resonance (cmr) is the imaging of choice for the diagnosis of acute myocarditis, revealing with high sensitivity focal or diffuse myocardial edema through shorttau inversion recovery (stir) sequences and mapping techniques (t2 and native-t1), potentially associated to necrotic foci visible with late gadolinium enhancement (lge), diffuse expansion of extracellular volume fraction (ecv) and hyperemia (43, 44) (figure 5) . the recent introduction of parametric mapping enables cmr to reveal diffuse myocardial edema that can be missed by conventional sequences, increasing its accuracy in the diagnosis of inflammatory cardiomyopathies. currently, few case reports showed cmr findings consistent with acute myocarditis in patients with laboratory-proven sars-cov-2 infection (13) (14) (15) . myocardial edema was the key for cmr diagnosis in all of these cases, underscoring the importance of including mapping techniques in cmr protocols adopted in covid-19 patients with suspected myocarditis (43) . therefore, in selected covid-19 patients not requiring icu, when clinical presentation and biomarker alterations suggest acute-onset myocardial inflammation, if the diagnosis is likely to impact on management, cmr may be considered to confirm acute myocarditis, after exclusion of alternative relevant clinical conditions, including acs and hf, by means of other rapidly available imaging modalities (i.e. cardiac ct scan or tte). nuclear cardiology imaging. nuclear cardiology encompasses several non-invasive imaging modalities and techniques that can be used for myocardial perfusion and viability assessment, as well as for the diagnosis of infective endocarditis, cardiac sarcoidosis and amyloidosis. however, most of these conditions can be proficiently and safely evaluated with other imaging modalities after covid-19 clinical resolution. therefore, in covid-19 patients, the use of nuclear cardiology tests should be restricted to very specific indications when they may yield diagnosis or directly influence the clinical management and no alternative imaging modalities can be performed (i.e. suspected infective endocarditis of prosthetic valves or intracardiac devices), in order to reduce healthcare personnel exposure related to long protocols and imaging acquisition times (45) . invasive cardiac imaging. when evaluating the role of invasive cardiac imaging modalities in covid-19 patients, several aspects deserve consideration. in the complex rearrangement of the healthcare service, all the efforts should be directed to ensure the standard-of-care and timely access to the catheterization laboratory for patients with acute cardiovascular conditions, irrespectively of sars-cov-2 infection. therefore, the use of ica in covid-19 patients should be restricted to those presenting with clinical or hemodynamic instability, including acute myocardial infarction, myocarditis, cardiogenic shock or cardiac arrest (figure 6 ). in these cases an invasive strategy is pivotal to ensure diagnosis and interventional treatment (46) . in addition, ica eventually combined with coronary intravascular imaging or left ventriculography plays an important role in identification and differential diagnosis of minoca (9) . basing on our direct experience, minoca accounts for >25% of acs in covid-19 patients. notwithstanding, patient status, severity of respiratory compromise, comorbidities and the risk of futility should be carefully evaluated when considering indication to invasive strategies in covid-19 patients. some clinical and laboratory risk factors for in-hospital death have already been identified in covid-19 patients (7, 8) . the quantification of lung and cardiac involvement by multimodality imaging could effectively delineate the severity of the disease and eventually the prognosis, providing a base for further clinical decision making. quantification of lung damage using a chest ct severity score (ct-ss) has been proposed to identify patients who need hospital admission (31) . this score was defined summing up individual scores from 20 lung regions: scores of 0, 1, and 2 were respectively assigned if parenchymal opacification involved 0%, <50%, or â�¥50% of each region (ct-ss range 0-40). the individual scores for each lung as well as the total score resulted significantly higher in patients with clinically severe covid-19 as compared to mild cases. a ct-ss <19.5 was highly effective in severe covid-19 pneumonia rule-out, with a npv of 96.3%(31). in the same way lus could be effective in evaluating covid-19 pneumonia severity and monitor its modifications over time. for this purpose the numerical assessment of regional loss of aeration measured by global lus score could represent a useful tool (33) . the global lus score can be calculated as the sum of regional aeration scores attributed to each lung region during a standard 12-zone examination scanning: 0 if a-lines or <3 b-lines are visualized; 1 if â�¥3 b-lines involving â�¤50% of the pleura; 2 if b-lines becoming coalescent or involving >50% of the pleura; 3 if tissue-like pattern(33) (figure 3) . the global lus score showed good correlation with lung density as assessed by ct scan and has been applied in the icu setting to quantify and monitor lung aeration in weaning from mechanical ventilation and in ards patients on extracorporeal membrane oxygenation (ecmo) (33) . so far, the implementation of the global lus score to monitor disease evolution and to guide decision making in covid-19 patients has not been systematically investigated. similarly, despite growing evidence pointing at the negative prognostic impact of cardiovascular involvement in covid-19, no specific risk scores have been developed and validated. interestingly, although great emphasis has been posed on the link between myocardial injury and mortality, the actual incidence of specific cardiovascular clinical conditions (myocarditis, mi, pe and hf) and the respective prognostic implications in different stages of covid-19 is largely unknown due to a significant lack of imaging data (4) . a systematic approach with the use of multimodality imaging to precisely characterize covid-19-related cardiovascular manifestations should be warranted to provide clinicians with comprehensive risk stratification tools. the imaging modalities are useful in the management of covid-19 patients in different clinical settings, from triage in the ed to icu and non-icu wards (figure 7) . emergency department/triage. a rapid and efficient diagnosis of covid-19 is of paramount importance to accurately manage the high number of patients presenting to the ed with suspected sars-cov-2 infection. considering the high probability of covid-19 among patients currently accessing ed with fever and respiratory symptoms, the main goal is to stratify patients with positive sars-cov-2 rt-pcr test (or with clinically highly suspected infection despite a negative test) in order to discharge those with mild symptoms and admit to non-icu or icu departments those with severe or life-threatening infection. a simultaneous clinical evaluation and lus performed by the same visiting physician (reducing the number of operators exposed), combined with laboratory testing and cxr, allow a fast diagnosis, risk stratification and decision-making regarding patient destination. in this context, lus has the potential to rapidly discriminate initial forms of covid-19 from advanced presentations (34) . focus is an adjunct to recognize specific ultrasound signs in patients with or suspected cardiac symptoms (37) . this quick stratification could be subsequently confirmed by cxr, trying to limit the number of ct scans performed in the ed setting, reserving ct for cases with uncertain diagnosis or to rule-out other causes of illness such as pe. of note, several patients have a severe form at ed presentation, rapidly becoming non-invasive ventilation (niv)-dependent and, therefore, cannot easily undergo ct scan; in these patients, lus is of paramount importance for rapid diagnosis and stratification. despite its potential diagnostic utility, no unequivocal advantage has been demonstrated for a lus-guided strategy over standard cxr and (if appropriate) ct scan evaluation in patients with suspected or confirmed covid-19. furthermore, lus requires closer contact with the patient, potentially exposing clinicians to higher risk of aerosolized particles inhalation, mandates use of more protective personal protection equipment (ppe) and should be performed by trained personnel. in this context, lus application is a promising technique, although its role should not be overemphasized in the absence of solid evidence; on the contrary, cxr and clinical evaluation remain pivotal for initial patient assessment. beyond ed evaluation, an important approach to take care of patients and prevent transmission is the application of telemedicine (47) . telemedicine/e-visits could be combined with home triage for patients reporting worsening symptoms or self-monitored parameters, the latter being ideally performed by dedicated teams providing both clinical evaluation and lus at the patient's home, thus more accurately differentiating patients who could continue remote monitoring and medical therapy at home from those who need hospitalization. covid-19 departments is currently based on supportive care (i.e. oxygen therapy, niv if necessary) and a combination of empirically prescribed drugs (i.e. hydroxychloroquine, antibiotics, antivirals, glucocorticoids or anti-cytokine therapies). along with clinical and laboratory evaluation, imaging is fundamental to assess covid-19 evolution and response to therapy, both in daily clinical activity and in the context of controlled pharmacological/interventional trials. baseline ct scan is frequently used to confirm diagnosis and to obtain detailed information on disease extension and severity, thus becoming also a reference for subsequent imaging follow-up (28) . of note, considering its known advantages (portability, bed-side evaluation, safety), lus seems particularly useful for serial assessments during hospital stay and may be useful to determine timing of ct imaging (34) . alongside with lung imaging, focus could be useful to assess volume status and concomitant cardiac involvement, reserving cardiac ct, ica and cmr only for selected cases, including suspected concomitant mi, pe or myocarditis (37) . icu represents the most challenging setting in the management of covid-19 patients. ideally, a baseline ct scan is needed in all critically ill patients requiring icu admission, in order to precisely describe morphological lung involvement. as in the previously described clinical settings, serial lus and cxr are fundamental to monitor disease evolution in icu patients, while ct scan could be used when clinical changes are observed, substantial modifications in morphological lung damage are suspected, or ventilator-related complications need to be excluded (32) . echocardiography could be useful to rule out concomitant cardiogenic causes of respiratory manifestations (37) . furthermore, focus allows a non-invasive hemodynamic monitoring in the icu setting: assessment of biventricular function, estimated stroke volume, filling pressures, pulmonary pressures, and central venous pressure (37) . similarly, tte helps in identifying patients at high risk of ventilator weaning failure and guides tailored therapeutic strategy. finally, when mechanical respiratory and circulation support with ecmo is needed, both tte and toe are important to guide device selection (veno-venous vs. veno-arterial) based on concomitant cardiogenic cause, assist during device placement (cannulation), and monitor cardiac function and devicerelated complications during support (48) . negative rt-pcr test deserve special consideration. as medical systems are overwhelmed, accurate balance between infection prevention and adequate healthcare assistance delivery should be pursued. beside clinical disease probability assessment, while serology tests are under development, current strategies to reduce in-hospital sars-cov-2 spread from asymptomatic patients rely on rt-pcr nasopharyngeal swab test, with important limitations (49) . therefore, adherence to international guidelines recommendations, and restriction of imaging tests to those really impacting on patients' clinical management are advocated (35, 36) . triaging protocols should differentiate between patients requiring nondeferrable but schedulable imaging examinations, who can be appropriately managed after rt-pcr test result is available, and those with urgent or emergent acute cardiovascular conditions, who should be considered sars-cov-2 positive until proven otherwise. optimization of healthcare network and patient pathways is required to avoid contamination between infected individuals and sars-cov-2 negative patients, while maintaining adequate health assistance. both patients and healthcare workers should be provided with standard ppe and keep social distance when possible. basing on our experience, rt-pcr test should be performed according to local resources in selected patients requiring hospitalization or undergoing aerosol-generating high-risk procedures, after body temperature measurement and a clinical triaging questionnaire evaluating history of fever, dyspnea or cough and sars-cov-2 exposure in the last weeks (50) . current covid-19 pandemic, sharply increased the examination workload of the imaging departments. the in-hospital infection rate was about 41% in one of chinese experience: 29% hospital staff and 12.3% inpatients (2) . in italy, up to 9% of overall cases were reported among healthcare workers with an estimated in-hospital infection rate of 10.8% (51) . sars-cov-2 transmission occurs through direct inhalation of droplets but also by touching eyes, nose or mouth after hand contact with contaminated surfaces. imagers, nurses and technicians are at high risk especially due to the close patient contact performing imaging studies. in order to prevent and mitigate the transmission, preventive measures must be implemented encompassing facilities, imaging equipment, ppe and machine disinfection procedures (35) . specific in-hospital routes between imaging department and covid-19 wards should be defined. the special environment for covid-19 dedicated imaging should include a contaminated equipment area, a separated report room and a staff cleaning room. the use of mobile equipment and dedicated scanners, ultrasound probes and machines for infected patients should be encouraged (35) . staff must undergo rigorous nosocomial infection training and equipped with highquality ppe ( table 4) , balancing the risk of transmission with the potential for scarcity of ppe, considering in some cases their re-using, with adequate precautions. the use of a checklist and a step-by-step process to ensure proper wearing (donning) and removing (doffing) are recommended. imaging personnel not directly involved should avoid any contact, and the distance between the technician and patients must be, preferably, >1-2 meters. all patients should wear a surgical mask during imaging. left-lateral patient positioning with the scanner on the right side of the bench may ensure the longest distance between patient's face and the echocardiographer during tte examination. personnel involved in toe examinations should wear full ppe as this procedure is aerosol-generating. while cuffed endotracheal tube and close-circuit ventilation could reduce the risk of aerosol generation in intubated patients, niv carries a higher risk of droplets spreading. the level of protection during toe should be full both in icu or non-icu context (35) . as sars-cov-2 is sensitive to most standard viricidal disinfectant solutions, imaging machines should be thoroughly cleaned. it is recommended to use soft cloth dipped in 2000 mg/l chlorine-containing disinfectant or 75% ethanol for scanners disinfection (35) . generally, for echocardiographic probes it is advised to immerse them for â�¤1 hour without using hot steam, cold gas, or abrasive agents, as ethylene-oxide or glutaraldehyde-based methods. automated disinfection solutions should be available. air, object surfaces and floor disinfection in the covid-19 dedicated imaging department should be carried out according to the daily operation specifications. in reading rooms social distancing should be remembered and all non-essential items removed (35) . as of today, none of the healthcare workers in the cardiac imaging department of our hospital, have been infected with sars-cov-2, underscoring the relevance of adequate ppe use and adherence to a rigorous safety protocol (52) . since ppe availability could be a significant issue especially in hard-hit areas, the use of clinical judgement should be emphasized to avoid additional staff exposure deriving from performing imaging tests unlikely to yield clinically important information on covid-19 positive or suspected positive patients. thus, the need for procedures requiring stringent ppe (i.e. toe or nuclear imaging) and the possibility to perform alternative imaging modalities (i.e. cardiac ct) or no procedure at all should be thoroughly assessed in order to optimize ppe use. sars-cov-2 outbreak has rapidly reached a pandemic proportion and has become a major threaten to global health. although the predominant clinical feature of covid-19 is an acute respiratory syndrome of varying severity, the cardiovascular system can be involved with several facets. heart and lung multimodality imaging plays a central role in different clinical settings and is essential in diagnosis, risk stratification and management of covid-19 patients. in order to prevent and mitigate the transmission, key preventive measures must be adopted encompassing the equipment, the facilities, the healthcare personnel and the disinfection procedures. 61-year-old woman with sars-cov-2 positive rt-pcr swab test presenting with sudden severe dyspnoea associated with significant d-dimer increase: ct pulmonary angiography shows gross filling defect in right pulmonary artery lobar branch for right upper lobe (a); lung parenchyma windowing demonstrates bilateral, subpleural ggos and consolidation areas, typical for covid-19 pneumonia (b); tte shows rv dilatation and septal shifting, indirect signs of severe pulmonary hypertension (c-d). 58-year-old woman with sars-cov-2 positive rt-pcr swab test presenting after 1 week of fever (38.5 â°c), cough, diarrhea with recent onset of typical chest pain, elevated cardiac markers (hs-tnt 222 ng/l), st-segment depression in inferior and lateral leads at ecg, and inferior septum hypokinesia at tte. triple rule-out ct shows peripheral lung opacities (a-b) characterized by crazy paving pattern involving both the inferior lobes, with posterior distribution, suggestive for covid-19 interstitial pneumonia (boxes), and demonstrates absence of pulmonary embolism (c) or coronary disease (d). cmr shows slight diffuse myocardial hyperintensity on t2 stir image (e) consistent with a slight increase of t2 relaxation time on t2 mapping: mean value of 55 ms (normal value â�¤ 50 ms) with a peak of 61 ms in the inferior septum (g); ir images do not show significant lge foci. viral replication and host immune response synergistically determine covid-19 pathogenesis. as the disease progresses through its three stages, different chest imaging modalities (lus, cxr and ct) demonstrate worsening lung involvement. in case of severe pneumonia tte can identify increasing pulmonary hypertension and rv impairment. cardiovascular complications related to viral infection or to systemic inflammation can occur at different stages of the disease, increasing the risk of adverse outcome, and require specific multimodality imaging assessment. clinical characteristics of coronavirus disease 2019 in china clinical characteristics of 138 hospitalized patients with 2019 novel coronavirus-infected pneumonia in wuhan, china cardiovascular implications of fatal outcomes of patients with coronavirus disease 2019 (covid-19) association of coronavirus disease 2019 (covid-19) with myocardial injury and mortality covid-19 infection: the perspectives on immune responses covid-19 illness in native and immunosuppressed states: a clinical-therapeutic staging proposal clinical course and risk factors for mortality of adult inpatients with covid-19 in wuhan, china: a retrospective cohort study risk factors associated with acute respiratory distress syndrome and death in patients with coronavirus disease coronavirus disease 2019 (covid-19) and cardiovascular disease sars-coronavirus modulation of myocardial ace2 expression and inflammation in patients with sars coronavirus fulminant myocarditis saved with glucocorticoid and human immunoglobulin life-threatening cardiac tamponade complicating myo-pericarditis in covid-19 cardiac involvement in a patient with coronavirus disease 2019 (covid-19) myocarditis in a patient with covid-19: a cause of raised troponin and ecg changes acute myocarditis presenting as a reverse tako-tsubo syndrome in a patient with sars-cov-2 respiratory infection pathological findings of covid-19 associated with acute respiratory distress syndrome the inflammatory response in myocardial injury, repair, and remodelling nlrp3 inflammasome and the il-1 pathway in atherosclerosis acute pulmonary embolism and covid-19 pneumonia: a random association? esc working group position paper on myocardial infarction with non-obstructive coronary arteries association of cardiac injury with mortality in hospitalized patients with covid-19 in association of influenza-like illness activity with hospitalizations for heart failure: the atherosclerosis risk in communities study acr recommendations for the use of chest radiography and computed tomography (ct) for suspected covid-19 infection | american college of radiology frequency and distribution of chest radiographic findings in covid-19 positive patients correlation of chest ct and rt-pcr testing in coronavirus disease 2019 (covid-19) in china: a report of 1014 cases chest radiographic and ct findings of the covid-19): analysis of nine patients treated in korea chest x-ray findings in asymptomatic and minimally symptomatic quarantined patients in time course of lung changes on chest ct during recovery from 2019 novel coronavirus (covid-19) pneumonia temporal changes of ct findings in 90 patients with covid-19 pneumonia: a longitudinal study patients with rt-pcr confirmed covid-19 and normal chest ct chest ct severity score: an imaging tool for assessing severe covid-19 international evidence-based recommendations for point-of-care lung ultrasound assessment of lung aeration and recruitment by ct scan and ultrasound in acute respiratory distress syndrome patients* is there a role for lung ultrasound during the covid-19 pandemic? covid-19 pandemic and cardiac imaging: eacvi recommendations on precautions, indications, prioritization, and protection for patients and healthcare personnel ase statement on protection of patients and echocardiography service providers during the 2019 novel coronavirus outbreak focused cardiac ultrasonography echocardiography in the time of covid-19 coronary ct angiography versus standard evaluation in acute chest pain triple rule out versus coronary ct angiography in patients with acute chest pain results from the acic consortium cardiac computed tomography in troponin-positive chest pain: sometimes the answer lies in the late iodine enhancement or extracellular volume fraction map quadruple rule out" with cardiac computed tomography in covid-19 patient with equivocal acute coronary syndrome presentation cardiovascular magnetic resonance in nonischemic myocardial inflammation: expert recommendations early t1 myocardial mri mapping: value in detecting myocardial hyperemia in acute myocarditis guidance and best practices for nuclear cardiology laboratories during the coronavirus disease 2019 (covid-19) pandemic: an information statement from asnc management of acute myocardial infarction during the covid-19 pandemic virtually perfect? telemedicine for covid-19 cardiovascular considerations for patients, health care workers, and health systems during the coronavirus disease 2019 (covid-19) pandemic presymptomatic sars-cov-2 infections and transmission in a skilled nursing facility echo in pandemic: front line perspective, expanding role of ultrasound and ethics of resource allocation istituto superiore di sanitã covid-19 integrated survelliance: key national data death from covid-19 of 23 health care workers in china key: cord-294557-4h0sybiy authors: stogiannos, n.; fotopoulos, d.; woznitza, n.; malamateniou, c. title: coronavirus disease 2019 (covid-19) in the radiology department: what radiographers need to know date: 2020-06-04 journal: radiography (lond) doi: 10.1016/j.radi.2020.05.012 sha: doc_id: 294557 cord_uid: 4h0sybiy objectives: the aim is to review current literature related to the diagnosis, management, and follow-up of suspected and confirmed covid-19 cases. key findings: medical imaging plays an important auxiliary role in the diagnosis of covid-19 patients, mainly those most seriously affected. practice differs widely among different countries, mainly due to the variability of access to resources (viral testing and imaging equipment, specialised staff, protective equipment). it has been now well-documented that chest radiographs should be the first-line imaging tool and chest ct should only be reserved for critically ill patients, or when chest radiograph and clinical presentation may be inconclusive. conclusion: as radiographers work on the frontline, they should be aware of the potential risks associated with covid-19 and engage in optimal strategies to reduce these. their role in vetting, conducting and often reporting the imaging examinations is vital as well as their contribution in patient safety and care. medical imaging should be limited to critically ill patients, and where it may have an impact on the patient management plan. implications for practice: at the time of publication, this review offers the most up-to-date recommendations for clinical practitioners in radiology departments, including radiographers. radiography practice has to significantly adjust to these new requirements to support optimal and safe imaging practices for the diagnosis of covid-19. the adoption of low dose ct, rigorous infection control protocols and optimal use of personal protective equipment may reduce the potential risks of radiation exposure and infection, respectively, within radiology departments. since the outbreak of the severe acute respiratory syndrome coronavirus 2 (sars-cov-2) pandemic a few months ago, more than 4,731,000 laboratory-confirmed cases of the new virus have been reported, with over 316,000 confirmed deaths, as per may 19 , 2020. 1 the coronavirus disease 2019 (covid-19) resulting after infection from sars-cov-2 has already affected 210 countries and territories globally. 2 this virus was first recorded in china's hubei province, where the cause of the disease was initially unknown. hence, it was first classified as unknown pneumonia. 3 new information about covid-19 emerges every day as more diagnostic tests are being carried out. medical imaging has a unique place in this new evidence-base and radiographers are working on the frontline to deliver care for some of the most seriously affected patients, often facing challenging situations with staff and resource shortages. 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 the aim is to review current literature related to the diagnosis, management, and follow-up of suspected and confirmed covid-19 cases. objectives include to: i) outline pathophysiology and basic epidemiology useful for radiographers, ii) discuss the role of medical imaging in the diagnosis of covid-19, iii) summarise national and international guidelines of imaging covid-19, iv) present main clinical and imaging findings and v) summarise current safety recommendations for medical imaging practice. the search methods and keywords for this review are appended on table 1 below for clarity. it has to be noted that in this review all available information at the time of publication was included, however we appreciate this is a fast developing area of study. coronaviruses belong to a large family of single-stranded rna viruses. although they are thought to cause mainly mild symptoms, the middle east respiratory syndrome (mers-cov) and the severe acute respiratory syndrome (sars-cov) recently, caused many fatalities. 4 in 2002, the sars-cov originating in china had a 11% mortality rate, while in 2012 the mers-cov in saudi arabia had a mortality rate of 34%. both viruses originated from wild animals. 4 sars-cov-2 targets the respiratory system. after the infection there is a variable incubation period ranging between 1 and 14 days. in some studies, average incubation periods of 3 days have been reported. 5 it has been found that the majority of covid-19 patients are asymptomatic or with mild symptoms. however, for a significant minority of cases, covid-19 can present as, or progress to, severe respiratory distress. 6 the typical clinical symptoms associated with the covid-19 disease include cough, fever, fatigue and dyspnea. 7 however, many patients can initially develop nausea and diarrhea, as well as generalised muscular pain and lack of sense of smell or taste. 8e10 haemoptysis has also been reported on a less frequent basis. 7 less common non-respiratory symptoms have also been described such as headache, urticaria, or presentation of neurological clinical features prior to or following the onset of covid-19related symptoms. 7, 11, 12 at the time of writing, the united states of america have reported the highest number of confirmed covid-19 cases, as well as the highest number of deaths, followed by spain, italy, france and the uk. mortality rate is measured using the case fatality rate (cfr), a measure describing the proportion of deaths within a defined population, and this rate varies considerably among different countries and is challenging to accurately calculate, as the number of asymptomatic or mild cases may be under-reported. 13 according to the world health organization around 5% of confirmed cases are severe and require intensive care. 14 acute respiratory distress syndrome, organ failure, septic shock and severe pneumonia are the main causes of mortality. covid-19 patients with one or more comorbidities, including hypertension, diabetes, cardiovascular diseases, cerebrovascular disease, chronic obstructive pulmonary disease (copd), malignancies, chronic kidney disease and smoking, are associated with poorer clinical outcomes and higher mortality rates. 15e18 severe asthma is also listed as a risk factor for hospitalisation. 19, 20 in addition, mortality rates have been found to dramatically increase with age. a recent uk study estimates an overall covid-19 mortality rate of 0.66%, increasing to 7.8% for people over 80. 21 a study from italy reports that 83% of the deaths were people over 70 years old. 22 similarly, there is a sharp decline of death rate in children. 21 recent data shows that children with covid-19 might be largely asymptomatic (90%) or demonstrate milder clinical manifestations and lower cfr compared to adults, while only younger than 1-year of age are more susceptible to severe disease. 23e25 there are also disproportionately more deaths in men than women, with similar cases between the two genders. this was attributed to sex-based immunological differences, or sex-based behavioral differences such as prevalence of smoking. 26, 27 diagnostic investigations for covid-19 accurate detection of an active covid-19 infection is vital for case identification, disease containment and optimal management of patients. molecular techniques are the first-line method of diagnosing covid-19. most commonly they use respiratory samples, such as nasopharyngeal swabs 28 with reverse-transcription polymerase chain reaction (rt-pcr), real-time rt-pcr (rrt-pcr) and reverse transcription loop-mediated isothermal amplification (rt-lamp) being the most common methods employed. 29 viral tests are used to detect the presence of an antigen, e.g. the virus's rna, in a patient, rather than antibodies, which affirm an immune response. however, antigen tests have limitations, including the time to obtain results, the relatively high false negative and false positive rates and the intermittent shortage of test kits during the outbreak. 29e31 there is ongoing debate about the optimal testing for coronavirus, with antibody testing gaining momentum, but equally more time is required until these tests become widely available. 32 laboratory tests are widely available and cost effective, and are used in the diagnosis and management of covid-19 patients, including differentiated white cell count, c reactive protein (crp), d-dimer and erythrocyte sedimentation rate (esr). 33 lymphopenia and mildly elevated crp have been widely reported, with the degree of lymphopenia proposed as a risk factor for more severe disease. 34 the role of imaging medical imaging plays an important role in supporting clinical decision making in the diagnosis, management and treatment of covid-19 patients. medical imaging may be useful for differential diagnosis between covid-19 and other viral respiratory illnesses with similar symptoms. 35e37 chest radiographs, chest ct, lung ultrasound, as well as mri are included in the arsenal of medical imaging, each one with advantages and limitations. chest radiographs (cxr) are the most widely used imaging modality for suspected and confirmed covid-19 cases. 38 mobile radiographs are being used with increasing frequency to avoid possible transmission during patient transfer to imaging departments, as well as the traditional role in imaging critically unwell patients. 39 classical cxr patterns of covid-19 include ground-glass opacities, consolidation and bilateral interstitial opacification associated with atypical or organizing pneumonia. 39e42 pneumothorax or lung cavitation are uncommon complications. 38, 43 imaging appearances may vary with stage of the disease (days from first symptoms) and with disease severity (fig. 1 ). 44 cxrs may be normal in cases of confirmed covid-19, both in early infection and in mild disease. in severe covid-19 there is a proportionately greater lung involvement which tends to be denser peripherally and in the lower zones. 38 the role of cxrs as the initial radiological assessment of patients presenting with respiratory distress and possible covid-19 is established. however, the available data on the accuracy of cxr in covid-19 is limited, with smaller case numbers compared to chest ct research and often without the inclusion of normal or non-covid-19 cases. for example, all patients within the analysis of zhao et al. 45 had a cxr performed but the findings have not been included. sensitivity of chest radiographs is dependent on the extent of covid-19 infection. in a cohort study (n ¼ 64 patients), cxr imaging of mild to moderate covid-19 patients was found to have a sensitivity of 69% 40 and lomoro et al. 42 found cxr sensitivity of 84% (27 of 32) . of the 170 non-hospitalised patients with mild symptoms, bandirali et al. 46 found 100 (58.8%) abnormal cxrs suggestive of covid-19, however rt-pcr confirmation was not performed. a small case series with rt-pcr confirmed covid-19 (n ¼ 9 patients) reported 50% sensitivity (true positive n ¼ 3) and specificity 33% (true negative n ¼ 1) for cxr, with two false positive cxrs (breast tissue mimicking ground glass opacification and atelectasis) without ct correlation. 47 paucity of reported specificity, small sample sizes, lack of normal and non covid-19 cases emphasises the requirement for imaging to be used as part of clinical decision making rather than in isolation. these findings can be summarised in table 2 . chest computed tomography (chest ct) has a limited but important role in clinical management of covid-19 patients. ct should be reserved for seriously ill patients, with emerging awareness of high prevalence of pulmonary thrombosis. 48 in addition, it can be used in the case of inconclusive chest radiographs 49 or unavailability of pcr tests. in the case of follow-up imaging where ct is required for clinical decision making, lowdose chest ct may be considered, as it can offer up to an 8-fold dose reduction. 50 systematic reviews and meta-analyses found that the most common imaging manifestations of the disease at ct were: bilateral, basal, ground glass opacities (ggos), crazy-paving, peripheral consolidations, reverse halo ("atoll" sign) and peri-lobular patterns (fig. 2) . 51, 52 bilateral pneumonia was predominant compared to unilateral, while most patients had more than two lobes involved, more often affecting the bases of the lungs than the apices. pericardial effusion, pleural thickening or hydrothorax were uncommon ct findings. 52e54 though viral pneumonias generally show similar imaging features, there are some characteristic ct findings which may help differentiating covid-19 from influenza-related pneumonia. 55 asymptomatic patients demonstrate single or multiple groundglass opacities, air bronchogram and nodules encircled by groundglass opacities. these are the main patterns also for early symptomatic patients, with the extra possible finding of interlobular septal thickening. 36, 44 when imaging patients between 3 and 7 days after the onset of clinical symptoms, ct demonstrated fused consolidations with air bronchogram, which tended to slightly decrease in range and density when imaging was performed between 7 and 14 days after the full clinical manifestation of the disease has taken place. 56 finally, imaging in the dissipation period 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 yradi1913_proof ■ 3 june 2020 ■ 3/10 (2e3 weeks after the onset of symptoms) revealed further decrease in lesions and thickening of bronchial wall and interlobular septum. however, even though the majority of patients developed improvement after 14 days, some studies have shown increased consolidations and development of pleural effusion on follow-up cts during the latter stages of the disease. 44 finally, a significant proportion of critically unwell patients with covid-19 have pulmonary embolic disease ranging from 23 to 30% and the role of ct pulmonary angiography (ctpa) is being established. 57, 58 despite paediatric patients having less severe symptoms than elderly cases, the ct findings in each age range are similar, most frequently bilateral sub-pleural ground-glass opacities and consolidation. 59 although, when compared to adults, the ggos in children are more localised and of lower attenuation. 60 therefore, these findings were characterised as atypical. furthermore, there is higher prevalence of peri-bronchial infiltrates and bronchial wall thickening in children compared to adults. this could be related to differences in distribution of the coronavirus infection along the respiratory epithelium between the two groups or to occurrence of co-infection. 61 similarly, chest ct performed in pregnant women diagnosed with covid-19 disease, showed that the consolidation lesions were more prevalent than in the rest of the patients. 62 low-dose protocols were implemented for these patients, minimising the risks of radiation exposure. 62 imaging of pregnant women should always be performed with extreme caution, after a thorough risk-benefit analysis for mother and fetus. relative to cxr, chest ct has higher contrast resolution without superimposed anatomy, which facilitates identification of radiographically occult abnormalities, in particular early ground glass opacification. however, many diseases manifest with similar ct findings, which may explain the relatively low specificity and risk of false positive diagnoses. 63 this is not coming as a surprise as in ct the three-dimensional nature of data acquisition and presentation ensures that superimposition of anatomy and pathology is minimised and any lesions can therefore be more easily identified and characterised. these findings suggest that ct should not be used alone as a diagnostic tool, and that swab tests must always be performed for these patients. although some studies conducted in china suggested chest ct as a first-line tool, 37, 64, 65 it must be noted that chest ct should only be reserved for the critically ill patients with unexplained deterioration. in the early stages of the covid-19 pandemic, characteristic ct appearances were seen in asymptomatic patients undergoing imaging for other reasons. 66, 67 hence, radiographers working in any healthcare setting must be aware and familiar with covid-19 findings in order to eliminate the potential risks of further transmission and improve patient management. reporting radiographers in particular are expected to be familiar with covid-19 imaging findings and preliminary clinical evaluation by radiographers is an expected competency in different countries, including the united kingdom. 39 point-of-care lung ultrasound (lus) the use of lung ultrasound (lus) in covid-19 is contentious and the evidence base is still evolving. point-of-care lus may have a 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 yradi1913_proof 68, 69 some early studies show that lus has a high sensitivity (93e98%) in imaging of acute respiratory distress syndrome (ards), as well as in cases of viral infections, such as the influenza pandemic in 2009. 70 in addition, lus reported a 93% sensitivity and 95% specificity when imaging critically ill patients with pneumonia. 71 a summary of all reported diagnostic accuracy values can be found in table 3 . although its full diagnostic value in patient management is yet to be established for covid-19 patients in larger studies, lus can depict signs suggestive of alveoral damage, subpleural consolidations, white lung regions, as well as irregular b-lines. 68, 70 a recent study reported a strong correlation with chest ct findings, and lus was recommended for imaging acute respiratory failure and lung inflammation. 76 more research with larger sample sizes would be needed to establish its added value, particularly for children or pregnant patients. however high operator dependency will remain its weak point. lastly, magnetic resonance imaging (mri), although not relevant for the evaluation of lung disease, it can contribute to the diagnostic pathway of patients with symptoms from the central nervous system. these may include various neurological manifestations, such as acute stroke, skeletal muscle injuries, consciousness impairment, or acute necrotizing hemorrhagic encephalopathy. 77, 78 the role of mri currently in the diagnosis of further secondary to covid19 complications, such as cardiac complications or persistent myositis, is still being explored and it is likely the application of mri in this area will expand as we understand more about this disease. cxr imaging of suspected or confirmed covid-19 cases should be performed with portable equipment within specifically designated isolated rooms for eliminating the risks of cross-infection within the radiology department. 79 an anterioposterior (ap) chest radiograph is performed on the patient's bed, despite known limitations of this technique, such as sub-optimal evaluation of the cardiothoracic ratio. in contrast, when cxr is performed within radiology, a posterioanterior (pa) standard technique must be used, as indicated. due to known risks of cross-infection, extreme care must be taken in relation to the optimal use of personal protective equipment (ppe) and decontamination of surfaces. 79 the technical quality of chest radiographs impacts diagnosis; mobile radiographs are performed on critically unwell patients and as a mechanism to reduce possible transmission. suboptimal image quality may occur due to rotation, incorrect exposure and reduced inspiration. it is therefore always important for the radiographer to check that all the technical and image quality criteria are fulfilled for every examination, where possible, and that neither patients' nor radiographers' safety is compromised. a plethora of guidelines on radiographic imaging considerations during the covid-19 pandemic, including a 10-point check list and other helpful evaluation tools, have been produced by the international society of radiographers and radiological technologists (isrrt). 80 regarding chest ct imaging, a standard unenhanced ct protocol 81 and multidetector (mdct) ct scanners can be used; the examination is carried out during the end-inspiration phase, when patients can follow breathing instructions. reconstruction to 1.25 mm slice thickness and multi-planar reconstruction is suggested. 82, 83 in case of clinical indications of pulmonary embolism and elevated d-dimers levels, a contrast-enhanced ct should be performed. 84 low dose ct should be used in paediatric and pregnant patients, to minimise radiation. the advantages of cxr include portability, which prevents cross-infection within radiology, cost-effectiveness, and wider availability. however, the sensitivity of the method is relatively low. on the contrary, chest ct has higher sensitivity (97e98%) 73,74 but lacks specificity, it is not widely available, has a higher radiation dose compared to cxr and its use requires thorough decontamination of the scanner room impacting workflows. lus offers the advantage of portability in the intensive care units or in a prehospital setting. 70 however, its diagnostic role has yet to be established. the role of mri is only ancillary in the case of neurological complications and its use should be strongly weighted against the impact on workflows, subject to the delays caused by decontamination. after the outbreak of the covid-19 pandemic, many professional bodies and learned societies have been quick to issue official guidelines on how medical imaging should optimally be performed for early diagnosis and related management of these patients, but also how staff should be protected from cross-infection. proportionate recommendations are offered for the protection of radiographers, as frontline staff. this knowledge is necessary for any medical imaging professional. subtle differences have been noted on the suggested imaging pathways among different countries (table 4 ). this could be mainly attributed to differences on the availability of antigen testing or of imaging equipment, the variability of diagnostic methods and techniques used. there may also be disproportionate lack of the required specialised staff (i.e. radiographers) to operate the equipment, and also dissimilar policies for the management of the pandemic. however a common denominator is that in most guidelines and recommendations medical imaging investigations are reserved for those patients who are critically ill, for those with inconclusive or insufficient prior diagnostic tests but with persistent symptoms, consistent with covid-19, and for those patients where clinical management decisions need to be imminently considered. most of the societies and professional bodies suggest that chest x-ray should be reserved for critically ill patients; the bsti identifies as such those who demonstrate oxygen saturation values below 94% (90% for patients with known copd). 49 however, they conclude that chest radiographs should not be used as a first-line tool due to low sensitivity, and that they must be restricted to imaging for intensive care unit patients. 35 similarly, the professional bodies suggest that ct imaging is indicated for those patients with clinical symptoms and inconclusive or normal imaging features on cxr and that the use of ct imaging for covid-19 should be based on clinical need and the possibility to change the management plan. 85 furthermore, the table 3 summary of sensitivity and specificity of cxr, ct and lus for the diagnosis of covid-19. chest radiograph 50e84% 40, 42, 46, 47 inadequate information, see 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 intercollegiate general surgery guidance 86 proposes for ct chest to be performed alongside abdominal imaging, especially to those patients requiring emergency surgery. this may have occurred as a consequence of the many incidental findings in asymptomatic cases. in addition, gastrointestinal may be the only presenting symptoms, hence rigorous infection control protocols must be employed regardless of the lack of respiratory-related symptoms, and radiographers must be prepared to manage these patients. 87 it is also reported that ct findings are not specific enough, and they can mimic other infections. 88 hence, ct imaging is recommended only if the clinical teams decide that it will have an impact on the management of the patient. the european bodies (esr and esti) concur with the limited use of chest ct, underlining the potential risks of infection during patient transportation. in line with the above and perhaps more conservatively to their european counterparts, north american professional bodies underline that no imaging, either cxr or ct, are recommended for the diagnosis of covid-19, and that viral testing should be the first-line method for diagnosis of the disease. 89 the european societies suggest that lung ultrasound should be used at the bedside, when needed, to eliminate the risks of further infection. 35 the fleischner society, an international consortium of worldrenowned experts in lung imaging, including radiologists and pulmonologists, states that imaging is not suggested for suspected covid-19 patients with mild symptoms, but it is recommended for patients with severe symptoms and worsening respiratory status. 90 in addition, they underline that cxr is less sensitive in the early/ mild infection in contrast with chest ct, which offers much more information at this stage. however, the final decision is left to the clinicians, as the availability of these methods, expertise and resources have to be considered. consequently, there is a consensus that imaging these patients must be generally limited to those critically ill, and that the clinicians at the point-of-care must always make a thoughtful risk-benefit analysis for these procedures, taking into account the stage of the disease and patient's clinical condition. some variability exists among countries regarding the use of medical imaging, mainly due to availability of resources and equipment, but also due to new scientific data about disease progression, which become available with time. however, radiographers should always consider these guidelines to avoid unnecessary staff and patient infection and to minimise radiation dose to patients by vetting the requested examinations or minimising radiation dose accordingly. optimal infection control procedures must take place within the radiology department, to minimise the potential risks of transmission of the virus to radiographers and other healthcare staff. decontamination of the imaging equipment is vital, and it has been widely discussed within the literature. 79,80,91e93 a recent study within radiology departments suggested that imaging equipment such as ct scanner components must be disinfected every time after contamination with 2,000 mg/l chlorine-containing disinfectant and the ct gantry must be fully wiped with 75% ethanol. 94 also, after decontamination the ct room must be closed for 1 h for ventilation and air circulation. 92 these studies are in line with the recommendations issued by the car and cstr, who also suggest standardized disinfection protocols after imaging of all suspected 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 or confirmed covid-19 patients, as well as unavailability of the equipment for a period of time. 89 keyboards, viewing stations, ultrasound probes, are also suggested to be disinfected after exposure, with alcohol-containing disinfectants (fig. 3) . 92, 95 radiology departments are encouraged to contact their vendors in order to specify the optimal disinfectants for every piece of equipment. in addition, all healthcare staff associated with cleaning, must be trained in optimal decontamination strategies, and radiology managers must develop specific infection control protocols to enhance safety within the departments. 96e98 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 radiology departments must also re-organise their facilities and staff in order to enhance safety and minimise the risks of infection. a recent study suggested some effective ways to achieve this, such as segregating radiographers into teams. furthermore, additional isolation rooms need to be created near the emergency department, where mobile radiography units can be deployed to minimise transferring patients. a two-radiographer team was found to effectively reduce the potential risks of cross-infection. 99 when non-urgent imaging is reintroduced, where possible, patients should be cohorted, outpatients screened at first presentation and possible covid-19 patients imaged using dedicated equipment, appropriate radiographer ppe and decontamination procedures. 100 however, it is crucial that asymptomatic transmission is recognised early; radiographers working on the frontline could conduct a preliminary clinical evaluation, as one method to facilitate rapid identification of unsuspected covid-19. 39 personal protective equipment (ppe) personal protective equipment (ppe) is vital for radiographers and other frontline healthcare professionals, as they can help minimise the likelihood of infection. a shortage of ppe has been reported globally given the high demand. the use of ppe must be proportional to risk of exposure. 101 a typical set of ppe for healthcare professionals consists of a long-sleeved gown, gloves, eye protection and a fluid repellent surgical mask or disposable respirator (n95, ffp2 or ffp3). this is the full ppe suggested by the european centre for disease prevention and control. 102 however, the uk government suggests the use of aprons instead of a gown for non-aerosol generating procedures (fig. 4) . 103, 104 in addition, the sessional use of ppe, or reuse have been also recommended in case of extreme shortages, while the use of double gloves is not suggested. the society of radiographers (sor) suggests that no radiographer should treat a patient without the ppe identified in the protocols. 105 more guidance is also available on the isrrt website 80 in relation to this. the covid-19 pandemic is rapidly and continually evolving. what is certain is that medical imaging will continue to play a key role in supporting clinical decision-making. further research is however needed to verify the added value of the different medical imaging modalities in diagnosis and patient management. moreover, the epidemiology of the disease is constantly changing, often impacting on the imaging findings and imaging techniques required to delineate these. what is certain is that there is more to learn in the coming months, but the authors hope that this paper will be a first tool, useful to summarise current knowledge for the radiography workforce at the frontline. medical imaging has a pivotal role in the covid-19 pandemic, offering the advantage of supplementary diagnosis and follow-up of the critically ill patients. radiographers, as frontline staff, should be familiar with the main challenges and controversies related to imaging patients with covid-19 so they can fulfill their role in safeguarding patient safety, patient care, optimise image quality as a tool for more accurate diagnosis, but also to engage in their reporting roles. most importantly, they should be aware of ways to keep themselves safe and well during this unprecedented situation by correctly using the right ppe and ensuring all their workspaces are adequately decontaminated to minimise the risk of further infection . 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 world health organization. coronavirus disease 2019 (covid-19) situation report-120 reported cases and deaths by country, territory, or conveyance di napoli r. features, evaluation and treatment coronavirus covid-19): a review of clinical features, diagnosis, and treatment covid-19: what has been learned and to be learned about the novel coronavirus disease world health organization declares global emergency: a review of the 2019 novel coronavirus (covid-19) clinical features of patients infected with 2019 novel coronavirus in wuhan, china coronavirus disease 2019 (covid-19): a perspective from china olfactory and gustatory dysfunctions as a clinical presentation of mildto-moderate forms of the coronavirus disease (covid-19): a multicenter european study don't overlook digestive symptoms in patients with 2019 novel coronavirus disease (covid-2019) alert for non-respiratory symptoms of coronavirus disease 2019 (covid-19) patients in epidemic period: a case report of familial cluster with three asymptomatic covid-19 patients nervous system involvement after infection with covid-19 and other coronaviruses cross-country comparison of case fatality rates of covid-19/sars-cov-2 unknown unknows-covid-19 and potential global morality comorbidity and its impact on 1590 patients with covid-19 in china: a nationwide analysis does comorbidity increase the risk of patients with covid-19: evidence from meta-analysis prevalence of comorbidities in the novel wuhan coronavirus (covid-19) infection: a systematic review and meta-analysis prevalence of underlying diseases in hospitalized patients with covid-19: a systematic review and meta-analysis covid-19 rapid guideline: managing symptoms (including at the end of life) in the community people with moderate to severe asthma estimates of the severity of coronavirus disease 2019: a model-based analysis covid-19 spiraling of frailty in older italian patients sars-cov-2 infection in children cdc covid-19 response team. coronavirus disease 2019 in children-united states radiography xxx (xxxx) xxx novel coronavirus disease (covid-19) in children covid-19: the gendered impacts of the outbreak provisional death counts for coronavirus disease (covid-19) diagnosing covid-19: the disease and tools for detection the epidemiology, diagnosis and treatment of covid-19 stability issues of rt-pcr testing of sars-cov-2 for hospitalized patients clinically diagnosed with covid-19 dynamic profile of rt-pcr findings from 301 covid-19 patients in wuhan, china: a descriptive study antibody tests in detecting sars-cov-2 infection: a meta-analysis the critical role of laboratory medicine during coronavirus disease 2019 (covid-19) and other viral outbreaks lymphopenia predicts disease severity of covid-19: a descriptive and predictive study covid-19 patients and the radiology department-advice from the european society of radiology (esr) and the european society of thoracic imaging (esti) the role of imaging in 2019 novel coronavirus pneumonia (covid-19) ct imaging and differential diagnosis of covid-19 portable chest x-ray in coronavirus disease-19 (covid-19): a pictorial review covid-19: a case series to support radiographer preliminary clinical evaluation. radiography frequency and distribution of chest radiographic findings in covid-19 positive patients clinical, laboratory and imaging features of covid-19: a systematic review and meta-analysis covid-19 pneumonia manifestations at the admission on chest ultrasound, radiographs and ct: single-center study and comprehensive radiologic literature review progress and prospect on imaging diagnosis of covid-19 coronavirus disease 2019 (covid-19): a systematic review of imaging findings in 919 patients a comparative study on the clinical features of covid-19 pneumonia to other pneumonias chest radiograph findings in asymptomatic and minimally symptomatic quarantined patients in codogno, italy during covid-19 pandemic chest radiographic and ct findings of the 2019 novel coronavirus disease (covid-19): analysis of nine patients treated in korea covid-19 and thrombotic or thromboembolic disease: implications for prevention, antithrombotic therapy, and follow-up radiology decision tool for suspected covid-19 recommendation of low-dose ct in the detection and management of covid-19 guidance for the reporting radiologist ct imaging changes of corona virus disease 2019 (covid-19): a multi-center study in southwest china coronavirus disease 2019 (covid-19) ct findings: a systematic review and meta-analysis imaging and clinical features of patients with 2019 novel coronavirus sars-cov-2: a systematic review and meta-analysis characteristic ct findings distinguishing 2019 novel coronavirus disease (covid-19) from influenza pneumonia a rapid advice guideline for the diagnosis and treatment of 2019 novel coronavirus (2019-ncov) infected pneumonia (standard version) acute pulmonary embolism associated with covid-19 pneumonia detected by pulmonary ct angiography acute pulmonary embolism in covid-19 patients on ct angiography and relationship with d-dimer levels clinical and ct features in pediatric patients with covid-19 infection: different points of adults ct features of novel coronavirus pneumonia (covid-19) in children differences in clinical and imaging presentation of pediatric patients with covid-19 in comparison to adults clinical and ct imaging features of the covid-19 pneumonia: focus on pregnant women and children the role of ct for covid-19 patient's management remains poorly defined initial ct findings and temporal changes in patients with the novel coronavirus pneumonia (2019-ncov): a study of 63 patients in wuhan, china chest ct findings of covid-19 pneumonia by duration of symptoms covid-19 pneumonia in asymptomatic trauma patients; report of 8 cases incidental findings suggestive of covid-19 in asymptomatic patients undergoing nuclear medicine procedures in a high prevalence region point-of-care lung ultrasound findings in novel coronavirus disease-19 pnemoniae: a case report and potential applications during covid-19 outbreak point of care and intensive care lung ultrasound: a reference guide for practitioners during covid-19 is there a role for lung ultrasound during the covid-19 pandemic? role of thoracic ultrasound in diagnosis of pulmonary and pleural diseases in critically ill patients ct on the diamond princess: what might this tell us about sensitivity for covid-19? correlation of chest ct and rt-pcr testing in coronavirus disease 2019 (covid-19) in china: a report of 1014 cases sensitivity of chest ct for covid-19: comparison to rt-pcr chest ct features of covid-19 in rome can lung us help critical care clinicians in the early diagnosis of novel coronavirus (covid-19) pneumonia? radiology 2020 neurological complications of coronavirus disease (covid-19): encephalopathy covid-19-associated acute necrotizing hemorrhagic encephalopathy: ct and mri features british society of thoracic imaging. example covid-19 action card. radiology portable cxr's-'at risk' patients in ed or wards international covid-19 support for radiographers and radiologic technologists an update on covid-19 for the radiologist-a british society of thoracic imaging statement coronavirus disease 2019 (covid-19): role of chest ct in diagnosis and management ct features of coronavirus disease 2019 (covid-19) pneumonia in 62 patients in wuhan, china pulmonary embolism in patients with covid-19: time to change the paradigm of computed tomography rcr position on the role of ct in patients suspected with covid-19 infection intercollegiate general surgery guidance on covid-19 update potential implications of novel coronavirus disease (covid-19) related gastrointestinal symptoms for abdominal imaging acr recommendations for the use of chest radiography and computed tomography (ct) for suspected covid-19 infection the canadian association of radiologists (car) and the canadian society on thoracic radiology recommendations on covid-19 management in imaging departments the role of chest imaging in patient management during the covid-19 pandemic: a multinational consensus statement from the fleischner society covid-19 pneumonia: infection control protocol inside computed tomography suites covid-19) outbreak: what the department of radiology should know microbiology for radiologists: how to minimize infection transmission in the radiology department the battle against coronavirus disease 2019 (covid-19): emergency management and infection control in a radiology department management of patients with suspected or confirmed covid-19, in the radiology department covid-19 preparedness for portable xrays in an indian hospital-safety of the radiographers, the frontline warriors infection control against covid-19 in departments of radiology corona virus international public health emergencies: implications for radiology management how to safely and sustainably reorganise a large general radiography service facing the covid-19 pandemic rsna covid-19 task force: post-covid surge radiology preparedness rational use of personal protective equipment (ppe) for coronavirus disease (covid-19) european centre for disease prevention and control. guidance for wearing and removing personal protective equipment in healthcare settings for the care of patients with suspected or confirmed covid-19 covid-19: personal protective equipment (ppe) plan personal protective equipment advice for imaging departments and teams covid-19: personal protective equipment (ppe) none key: cord-347691-ia2i8svg authors: larici, anna rita; cicchetti, giuseppe; marano, riccardo; merlino, biagio; elia, lorenzo; calandriello, lucio; del ciello, annemilia; farchione, alessandra; savino, giancarlo; infante, amato; larosa, luigi; colosimo, cesare; manfredi, riccardo; natale, luigi title: multimodality imaging of covid-19 pneumonia: from diagnosis to follow-up. a comprehensive review date: 2020-08-17 journal: eur j radiol doi: 10.1016/j.ejrad.2020.109217 sha: doc_id: 347691 cord_uid: ia2i8svg due to its pandemic diffusion, sars-cov-2 (severe acute respiratory syndrome coronavirus 2) infection represents a global threat. despite a multiorgan involvement has been described, pneumonia is the most common manifestation of covid-19 (coronavirus disease 2019) and it is associated with a high morbidity and a considerable mortality. especially in the areas with high disease burden, chest imaging plays a crucial role to speed up the diagnostic process and to aid the patient management. the purpose of this comprehensive review is to understand the diagnostic capabilities and limitations of chest x-ray (cxr) and high-resolution computed tomography (hrct) in defining the common imaging features of covid-19 pneumonia and correlating them with the underlying pathogenic mechanisms. the evolution of lung abnormalities over time, the uncommon findings, the possible complications, and the main differential diagnosis occurring in the pandemic phase of sars-cov-2 infection are also discussed. as suggested in the recently published who (world health organization) advice guide for the diagnosis and management of covid-19, chest imaging should be used for diagnostic purpose in symptomatic patients if rt-pcr is not available or its results are delayed, or in case of negative result in the presence of a high clinical suspicion of covid-19 [11] . cxr and ct are the imaging procedures generally performed in the diagnosis of covid-19 pneumonia. in this clinical setting, cxr is of easy use and is usually performed in the anteroposterior (ap) projection and in the supine position, using mobile cxr units in a dedicated and isolated room to reduce the risk of infection spreading [6, 12] . chest ct is performed with the high-resolution (hrct) technique, using thin sections (<1.5 mm) and high-spatial resolution kernel to enhance lung parenchymal anatomical details, usually without contrast medium injection. however, evidences suggest a predisposition to thrombotic and thromboembolic disease in these patients [13] , with pulmonary embolism (pe) being a known epiphenomenon of covid-19 syndrome [14] . therefore, in the appropriate clinical setting, contrast medium injection is required to rule out pe. the reported sensitivity of cxr for covid-19 pneumonia is relatively low in the early phase of the disease and in mild cases and it is inferior to that of rt-pcr essay at baseline (69% versus 91%, respectively) [15] . the rate of negative cxr at baseline tends to decrease when the interval time between the onset of symptoms and the cxr increases. in a retrospective study of 240 symptomatic patients with covid-19 infection confirmed by rt-pcr, the rate of negative cxr was 36.7% at 0-2 days after the onset of symptoms and 16.1% at >9 days [16] . the diagnostic performance of cxr increases in clinical settings with high prevalence of disease as described by schiaffino et al. in their experience based on real-life reporting without independent image review during covid-19 outbreak in northern italy [12] . cxr at admission to the emergency department showed a sensitivity of 89.0% (95% ci, 85.5%-91.8%) and a specificity of 60.6% (95% ci: 51.6%-69.2%) [12] . therefore, bedside cxr -performed in the isolated rooms -can be efficiently used as first-line imaging modality in areas with high levels of contagion and high pre-test probability of disease, particularly in cases of shortage of rt-chest ct and rt-pcr were 94% (95% ci: 91-96%) and 89% (95% ci: 81-94%), respectively, for the diagnosis of covid-19 pneumonia [19] . in another meta-analysis [20] , a pooled sensitivity of 92% (95% ci: 86-96%) was reported for ct in detecting covid-19 among studies from different areas of china; the pooled sensitivity of ct was even higher (up to 99%) in the region with more severe epidemic. based on its high sensitivity, ct has been proposed as the primary diagnostic tool in epidemic areas, in order to early recognize suspicious cases and possibly limit the spread of infection. however, a more recent meta-analysis including 37 studies with 9610 patients underlined that the sensitivity of ct significantly decreases from 94% to 75% when studies with low risk of biases are included [21] . furthermore, the possibility that ct can be normal within the first 3 days from the onset of symptoms (day 0-2) in up to 56% of cases should be taken into consideration when assessing patients with suspected covid-19 pneumonia [22] . ct can also detect abnormalities in asymptomatic patients, with a rate of 54%, according to inui s. et al. [23] . the reported specificity of ct is moderate to low; it was only 37% (95% ci: 26-50%) in the meta-analysis by kim et al. [19] . the same authors observed that in regions with a prevalence of disease <10%, the ppv of rt-pcr was more than ten times higher than that of ct, meaning that applying ct as a screening tool in these areas could potentially lead to a large number of false-positive results, with an unjustified increase in medical costs [19] . due to the abovementioned limitations, some authors suggest considering ct as a supplemental diagnostic tool, especially in symptomatic patients [24] . apart from recognizing covid-19 pneumonia features, imaging -especially ct -may reveal possible alternative diagnoses (e.g. pulmonary oedema, alveolar haemorrhage, other type of lung infections) that justify patient's respiratory symptoms [25, 26] . the rule-out role of ct is also highlighted by the recent fleischner society consensus statement, particularly in patients manifesting with moderate-to-severe symptoms and a negative or ongoing rt-pcr test [6] . the who advice guide also underlined that the use of chest ct is particularly helpful in patients with known pre-existing pulmonary diseases [11] . in order to standardize the level of suspicion of covid-19 pneumonia on ct scans, a categorical assessment scheme, the covid-19 reporting and data system (co-rads) has been proposed, with levels ranging from very low (co-rads category 1) to very high suspicion (co-rads category 5), with a co-rads 6 category reserved for rt-pcr proven cases [27] . imaging also plays a role in prognostic assessment and patient stratification in covid-19 pneumonia. some cxr scoring systems have been recently developed to answer these needs in the clinical practice with interesting results [28, 29] . in particular, higher disease scores at baseline have been associated to hospitalization, requirement for mechanical ventilation [28] and in-hospital mortality [30] . similarly, categorization and quantification of hrct abnormalities in covid-19 pneumonia have been demonstrated to correlate with development of severe disease course [31, 32] , icu admission [33, 34] , and in-hospital mortality [34] . particularly, colombi et al. [35] demonstrated that patients requiring icu admission or deceased have higher parenchymal involvement (4 or more lobes) and show less aerated lung parenchyma on baseline hrct compared to the other patients. the importance of chest imaging to assess disease evolution is also unquestioned. in patients requiring hospitalization, cxr is essential in guiding clinical management, as the serial evaluation allows an adequate assessment of the evolution of findings, avoiding unnecessary radiation exposures. this is especially true in critically ill patients [36] (fig. 1) , or during regression of symptoms in favourable cases. for a late follow-up, hrct is the modality of choice, particularly in assessing eventual persistent or fibrotic lung abnormalities. lastly, imaging is essential in the early detection of complications, such as barotrauma, superimposed bacterial lung infections and empyema. ground glass (gg) opacities, a crazy paving pattern characterized by gg opacities with superimposed septal thickening, and consolidations are common hrct findings in patients with covid-19 pneumonia [37] (fig. 2) . incidence of these findings varies among different studied populations. in a recently published meta-analysis comprising 4410 patients, gg opacities, a combination of gg opacities and consolidations, and crazy paving have been reported in 50.2%, 44.4%, and 19.5% of cases, respectively [38] . the distribution of the above reported findings is usually bilateral and multilobar with a predominant involvement of subpleural/peripheral and posterior regions of the lungs [39] . similar findings of gg, interstitial opacities and areas of consolidation with a peripheral and mid-basal distribution can be observed on cxr [15] (fig. 2 ). these imaging findings are the expression of a condition of acute lung injury (ali), the main pathological pattern of the pulmonary damage caused by sars-cov-2, that presents with a wide spectrum of histologic patterns ranging from diffuse alveolar damage (dad) with hyaline membrane formation to organizing pneumonia (op) [40] [41] [42] . the predominant pathogenic mechanism of ali induced by sars-cov-2 infection and shared by other coronaviruses is the angiotensin-converting enzyme 2 (ace2) downregulation [43, 44] , which results in excessive inflammatory cytokine release ("cytokine storm") leading to apoptosis of epithelial and endothelial lung cells [45] . direct infection of the endothelial cells and diffuse endotheliitis have been also described in several organs of covid-19 patients, included the lungs, with subsequent oedema and parenchymal ischemia due to microvascular dysfunction [46] . furthermore, a generalized microthrombotic injury, mediated by activation of complement pathways, has been observed within the lung microvasculature in pathologic specimens from patients deceased with severe covid-19 pneumonia [47] . based on these observations, microthrombosis and diffuse vascular lung injury might have a role in the pathogenesis of covid-19 pneumonia, at least in severe and critically ill cases [47] . it has also been assumed that in some patients the parenchymal abnormalities seen at hrct might be associated not only to inflammation or atelectasis, but also to ischemic and/or necrotic changes caused by perfusion defects [48, 49] . the evolution of dad is characterized by three sequential phases [50] . the first phase, or exudative phase, consists of interstitial and alveolar oedema, haemorrhage, and hyaline membrane formation, with alveolar obliteration and thickening of the inter-and intralobular septa; it has a duration of approximately 7 days. the second phase, or proliferative phase, consists of fibroblast proliferation within the interstitium and the alveoli, with appearance of op foci, parenchymal remodelling and formation of reversible traction bronchiectasis and bronchiolectasis. the last fibrotic phase is characterized by collagen deposition with progressive and variable degree of fibrotic changes and usually starts 2 weeks after the lung injury [50] ; it is potentially reversible in mild to moderate cases. similarly, progressive temporal stages of hrct findings can be recognized in patients with covid-19 pneumonia. pan et al. have described 4 stages from initial diagnosis until patient recovery; these include an early stage, between 0 and 4 days from the onset of symptoms, an intermediate progressive stage (5-8 days) and a peak stage, between 9 and 13 days [51] . starting from 2 weeks after the onset of symptoms, a gradual progressive resolution of the hrct findings can be observed (absorption stage) [51] . wang et al. confirmed this evolution, even though in their experience the absorption phase was delayed, probably because patients with more severe disease were included in their study [52] . a roughly similar appearance of the lung abnormalities in the different phases of the disease is appreciable also on cxr, even though with the intrinsic limitations of this imaging modality. the early stage is generally characterized by the presence of gg opacities, with the typical bilateral and multilobar distribution [39, 53] ; consolidations are possible, but have been described only in a minority of patients. enlargement (greater than 3 mm in diameter) of subsegmental pulmonary vessels within the areas of gg has also been described in the early phase of covid-19 pneumonia in up to 89% of cases [10, 37, 54, 55] . the aetiology of this sign is not well defined yet, although it could be related to hyperaemia and vessel wall damage induced by pro-inflammatory factors [54] or small vessel thrombosis. some authors have indicated that this sign might be helpful in the differential diagnosis between covid-19 and other viral pneumonia, even though it has been described in 22% of patients with other viral non-covid-19 pneumonia [56] . with progressive replacement of gg opacities by consolidation, the caliber of the subsegmental pulmonary vessel returns normal. it is worth noting that maximum intensity projection (mip) reconstructions usually enhance and facilitate the identification of this sign and its evolution over time (fig. 3 ). the progressive stage of covid-19 pneumonia is characterized by a more extensive lung involvement and more varied imaging features [51, 53] . gg opacities increase in density and can appear diffuse or with a crazy paving pattern on hrct. consolidations can progressively develop in the areas of gg or increase in size and number respect to previous hrct scans [53] . usually, in this phase, consolidations show patchy subpleural and peribronchovascolar distribution, which is the common appearance of the op pattern [57] associated with the proliferative phase of dad [40, 58] . similar changes are detectable also on cxr, confirming its role in identifying the temporal evolution of covid-19 pneumonia till the improvement ( figures 4, 5) . the maximum disease burden is usually observed 10 days after the onset of symptoms (peak or severe stage). at this stage other well-known hrct findings, such as reversed halo sign, band-like opacities and perilobular opacities appear [59] . traction bronchiectasis and bronchiolectasis might be visible in this phase as ancillary findings within the consolidations, due to parenchymal remodelling (fig. 6 ). in patients developing acute respiratory distress syndrome (ards), covid-19 pneumonia has an unfavourable prognosis [53, 60] . in these patients, ards shows an atypical dissociation between the relatively low impairment of the lung mechanics and the severe hypoxemia [61] , probably due to the concomitant microvascular injury that characterizes covid-19 [47] . the absorption stage is characterized by a progressive clearance of the lungs over time. this stage may have quite long temporal course, with gg opacities, linear scarring and mild residual consolidations, still evident on hrct after 4 weeks from the onset of symptoms [51] ( fig. 7) . in this phase, a decrease in density of the opacities associated with a more extensive involvement of the lung has been described (the so called "tinted" sign), possibly due to the gradual resolution of the inflammation and progressive alveolar re-expansion [62] (fig. 7) . traction bronchiectasis and bronchiolectasis tend to disappear with complete resolution of the opacities. the time required for the complete clearance of lung abnormalities may reflect the severity and the extent of lung involvement. a prompt treatment of covid-19 pneumonia has been correlated with a more rapid clearance of the hrct findings [63] . however, it must be pointed out that lung abnormalities can be seen on hrct even in presence of complete clinical recovery and negative rt-pcr swab test [64] (fig. 7) . due to the young history of this outbreak, data about the morphological characteristics and potential long-term lung abnormalities in survivors from clinically significant covid-19 disease are lacking [65] . however, the evolution of dad, particularly in patients who recovered after full-blown ards [66] , may lead to lung fibrosis [57] , which has been demonstrated on covid-19 pneumonia pathological specimens [67] . therefore, a clinical and radiological follow-up might be required in order to identify and monitor potentially progressive lung fibrotic changes with the typical hrct findings of irregular interstitial thickening, traction bronchiectasis/ bronchiolectasis, coarse reticulation, and parenchymal bands [68] . the duration and timing of follow-up with ct scans have not been defined yet. although the majority of patients present with common findings and typical manifestations of covid-19 pneumonia, uncommon features and atypical presentations are possible and can potentially represent a diagnostic challenge [18] . variable rates of possible uncommon findings in different patient populations have been reported. in this section, the results of recent systematic reviews and meta-analyses are presented [37] [38] [39] . when considering the distribution of parenchymal abnormalities, unilateral lesions can be observed, especially immediately after the onset of symptoms or in pauci/asymptomatic patients [36, 69] and has been described in 18.7% of cases in a meta-analysis of 34 studies including 4121 patients [39] . in the same meta-analysis, involvement of a single lobe and predominant involvement of the anterior lung zone with relative sparing of the posterior ones have been reported in 14.9% and 20.5% of cases, respectively [39] . a peribronchovascular distribution of lung abnormalities has been found in only 3.6% of 3466 patients among 28 studies [37] . when considering the lesion shape, a multifocal nodular appearance, usually with irregular margins or surrounded by peripheral gg, has been documented in 7.8% in a meta-analysis of 45 studies including 4410 patients [38] . enlargement of mediastinal lymph nodes has been reported in 5.1-5.4% of cases [37] [38] [39] in patients with severe disease and extensive bilateral consolidative changes. the presence of lymph node enlargement has been associated with disease severity and poor prognosis. indeed, lymphadenopathies have been observed in a significantly higher number of patients (27%) with unfavorable course of disease during hospitalization than in those discharged (14%), according to a single center study on a cohort of 410 patients [70] . among the uncommon findings, the occurrence of pleural effusion has been found in about 5% of cases [37] [38] [39] , while pleural thickening adjacent to the parenchymal opacities is far more common [38] . pericardial effusion has also been rarely reported, with a pooled prevalence between 2.7% and 3.6% [37, 38] . it is worth of note that the presence of pleural and/or pericardial effusion has been related with severe clinical course and poor prognosis of covid-19 pneumonia [71] . another uncommon hrct feature related to severe disease [71] is the evidence of bronchial wall thickening within the lung opacities, described in 14.5% of cases [38] , and possibly j o u r n a l p r e -p r o o f related to inflammatory bronchial wall damage and peribronchovascular interstitial oedema. hrct findings associated with small airways disease, such as endobronchial secretions and tree-in-bud opacities, have been described in a relatively small percentage of cases (4.1%) [37] . cavitation is the least common finding of covid-19 pneumonia, with a reported pooled prevalence of only 0.1% [38] . apart from the uncommon findings associated with covid-19 pneumonia, it should be noted that pre-existing underlying pulmonary diseases, such as fibrosis and emphysema, may lead to atypical hrct presentations of covid-19 pneumonia [18] . in case of honeycombing or paraseptal emphysema, the typical subpleural distribution may be absent (fig. 8) . the occurrence of gg opacities or consolidations superimposed on extensive background emphysema may lead to a "bubble-like" appearance, which could be misinterpreted as lung cysts or cavitation and lead to erroneous ruling-out of covid-19. on the other hand, in case of typical findings of covid-19 pneumonia, detection of cystic changes within a focal peripheral gg or consolidation area on hrct should raise the possibility of a concurrent lung adenocarcinoma, especially if other red flags (e.g. spiculated margins, pleural retraction) associated with malignancies are evident (fig. 8) . severely ill patients are more prone to develop complications. pneumothorax is one of these, reported in 1% of cases according to chen et al. [72] and usually occurring in case of rupture of subpleural bullae [73] . in the clinical context of covid-19, pneumothorax may be spontaneous, with prolonged cough and respiratory distress as known risk factors, or due to barotrauma in patients requiring mechanical ventilation. pneumomediastinum may also occur as a consequence of increased intrathoracic pressure with rupture of the alveoli, followed by air dissection through the bronchovascular bundles into the mediastinum (macklin's effect) [74] . subcutaneous emphysema may also be associated. such complications must be suspected in case of sudden clinical deterioration with rapid oxygen desaturation (fig. 9) . mcguinness et al. reported an incidence of barotrauma of 15% among covid-19 patients requiring invasive mechanical ventilation, an incidence that was higher than the overall rate of 10% noted in their retrospective cohort of ards cases [75] . the occurrence of pe has been frequently demonstrated in patients affected by covid-19, particularly in those with more severe symptoms, and it has been considered as a part of the disease rather than a true complication [13] (fig. 9) , with an incidence of 21-30% according to different series [14, [76] [77] [78] [79] . based on the above, it might be potentially reasonable to perform ct scans with contrast administration in all confirmed cases of covid-19 to assess pulmonary vessels and detect eventual arterial filling defects and lung perfusion defects on iodine map, if using a dual-source ct scanner in dual-energy mode [80] . however, in the clinical practice, the current trend is to perform ct pulmonary angiogram in cases of suspected pe due to clinical worsening, not explained by an increase of parenchymal disease burden at imaging. other possible complications are bacterial infections, which should be suspected in case of appearance of multiple centrilobular nodules with or without tree-in-bud opacities, and eventual cavitation within areas of consolidation superimposed to covid-19 pneumonia lung abnormalities (fig. 9 ). although in the context of the current pandemic the imaging findings described in the previous sections of this review can be indicative of covid-19 pneumonia in areas with high prevalence of the disease, they lack in specificity and a number of disease processes both infectious and non-infectious should be considered in the differential diagnosis, as parekh et al. have described in their recently published comprehensive review [81] . in our experience at a large tertiary metropolitan hospital in rome during the outbreak in italy, 1443 patients with symptoms suggestive of covid-19 pneumonia were admitted to the emergency department between march 7 and april 27, 2020. among them, 301 presented imaging findings indeterminate for covid-19 or suggestive of an alternative diagnosis at the baseline cxr and underwent chest ct scanning; these patients also presented with two consecutive rt-pcr tests negative for sars-cov-2. in a retrospective evaluation of the final diagnosis, infectious diseases were present in 65.1% of our patients (196/301), with the majority caused by bacteria (56.6%), followed by indeterminate community-acquired pneumonia (cap) (30.1%), viruses (7.7%) and atypical pathogens (5.6%). among the remaining 105 patients with non-infectious diseases, pulmonary oedema was observed in 57 j o u r n a l p r e -p r o o f (54.2%), followed by thoracic neoplasms (20.9%) (progression and/or first diagnosis of primary lung cancer; metastases due to extrathoracic neoplasms), acute exacerbation of ild (interstitial lung disease) (9.5%), aspiration (7.6%), ascertained drug toxicity (3.8%), alveolar proteinosis (1.9%), lipoid pneumonia (0.9%) and diffuse alveolar hemorrhage (0.9%). the most relevant cxr and hcrt findings of possible differential diagnosis among infectious and non-infectious diseases at the time of the outbreak of covid-19 are summarized in the table 1 . lower respiratory tract infections and cap represent the most likely diagnosis in patients with fever, cough and dyspnoea, which are also common symptoms of covid-19 pneumonia. bacteria and viruses are the usual causative agents of cap, even if a definite microorganism is identified in only 30% to 65% of cases [82] [83] [84] . bacterial pneumonia shows three possible patterns at imaging: lobar pneumonia, bronchopneumonia and interstitial pneumonia [85] . when a lobar pneumonia pattern is identified, the diagnosis is relatively simple and covid-19 pneumonia can be reasonably excluded on the basis of radiological findings. lobar pneumonia is characterized by a homogeneous lobar or nonsegmental opacity or consolidation, with or without air bronchogram, involving predominantly or exclusively one lobe. the abnormality is commonly confined by the fissure and the lung volume is preserved (fig. 10) . streptococcus pneumoniae, legionella pneumophila and mycoplasma pneumoniae are the most common bacteria responsible of a lobar pneumonia [85] . associated findings and/or complications include parapneumonic pleural effusion, empyema, cavitation, and lung abscess formation [85, 86] . the unilobar distribution and the presence of associated findings are helpful in differentiating a bacterial pneumonia from covid-19. in case of a bronchopneumonia pattern, the causative agent leads to bronchial epithelium inflammation, with ulcerations and fibrinopurulent exudate formation and spreading through the airways' walls and adjacent pulmonary lobules [85] . cxr shows multifocal patchy nodules and confluent opacities without air bronchogram. characteristic hrct findings are patchy nodules with centrilobular distribution and tree-in-bud appearance, confluent peribronchial focal consolidation without air bronchogram and lobular gg areas, associated with bronchial wall thickening and mucoid impaction. these features, commonly associated with staphylococcus aureus and gram-negative bacteria (pseudomonas aeruginosa, klebsiella j o u r n a l p r e -p r o o f pneumoniae, haemophilus influenzae) pneumonias, are highly suggestive of aerogenous spread of infection [85, 86] . this behaviour is in contrast to covid-19 pneumonia, which shows a prevalent subpleural distribution of findings, absence of tree-in-bud opacities and it is rarely associated with airways involvement. lastly, an interstitial pneumonia pattern can be caused by mycoplasma pneumoniae and other atypical agents, which determine direct damage of the bronchioles mucosa and subsequent inflammation and oedema of the peribronchial interstitium and interlobular septa. cxr shows peribronchial thickening and bilateral interstitial and/or interstitial-alveolar opacities [85] that may mimic covid-19 pneumonia. hrct shows patchy lobular gg opacities and/or consolidation, centrilobular nodules, and thickening of the peribronchovascular interstitium. single lobe involvement, the presence of centrilobular nodules and thickening of the peribronchovascular bundles are common findings in patients with mycoplasma pneumonia and might help in the differential diagnosis with covid-19 pneumonia [85] (fig. 11) . a wide range of respiratory viruses is responsible for the development of pneumonia [87] . depending on the pathogenesis of the infection and on the causative agent, viral pneumonia can show different imaging patterns [87] . for this reason, despite a certain degree of overlap, the awareness of the underlined pathogenic mechanisms is crucial to understand imaging findings and to address differential diagnosis when possible. three main different imaging patterns have been associated with viral infections: nodular/micronodular pattern, bronchiolar pattern and interstitial pattern [87] . the nodular/micronodular pattern is characterized by bilateral scattered multifocal nodules with well-or ill-defined margins and possible gg appearance, and it is usually determined by haematogenous viral spread to the alveoli, such as in varicella-zoster virus (vzv) pneumonia. the bronchiolar pattern shows an airway-centred distribution, with centrilobular nodules, treein-bud opacities and bronchial wall thickening, with or without peribronchovascular gg opacities and small consolidations, not dissimilar from bacterial bronchopneumonia. this pattern is due to destruction of bronchial and alveolar wall, determining small airway obstruction. it is typical of respiratory syncytial virus (rsv) and human metapneumovirus (hmpv) infections (fig. 12) ; it can be also observed in adenovirus pneumonia [87] . the interstitial pattern of viral pneumonia is characterized by multifocal gg opacities, interlobular septal thickening and consolidations, expression of the development of ali with dad [87] . this j o u r n a l p r e -p r o o f is the common imaging presentation of covid-19 pneumonia, as discussed above (fig. 2) ; however, the interstitial pattern has been also associated with other coronavirus infections, such as severe acute respiratory syndrome coronavirus (sars-cov-1) and middle east respiratory syndrome coronavirus (mers-cov) [88, 89] . sars-cov-2 infection is characterized by peripheral lung involvement and gg opacities more commonly than sars-cov-1 and mers-cov pneumonia, in which consolidations were prevalent [90] . also, differently from covid-19, sars showed unifocal distribution in 54.6% cases as initial presentation [91] , remaining confined to a single lung in approximately one-quarter of patients [91, 92] . on the other hand, when compared to covid-19 pneumonia, a progressively extension from the lower lobe periphery into upper and perihilar lung zones was shown in mers [92] , together with a significantly higher prevalence of pleural effusion (33%) [93] . nevertheless, due to the shared pattern of lung damage, the differential diagnosis between covid-19 and other interstitial viral infections is challenging [87] . in one study, a high specificity in correctly differentiating ct features of covid-19 from other viral infections was reached, with peripheral distribution of the abnormalities, gg appearance and vessel enlargement as clue findings for differential [56] . however, due to some selection biases (e.g. low number of influenza-a cases), small cohort size and different reader's level of experience on reporting covid-19 cases, the observed results can be somewhat misleading [94] . influenza pneumonia can cause bilateral reticulonodular opacities, usually with lower lobes predominance on cxr, and bilateral patchy gg opacities and consolidation, with ill-defined small nodules on hrct [87] , which can be difficult to differentiate from covid-19 pneumonia. this is particularly true during the organizing phase of the disease when an op pattern, commonly observed also in h1n1 influenza pneumonia, occur [87, 95] . compared to covid-19, influenza demonstrates higher lower lobe predominance, with more frequent subpleural and peribronchovascular distribution of the abnormalities. furthermore the presence of clustered and ill-defined lesions, nodules and bronchial wall thickening is more common in influenza pneumonia than in covid-19 [96, 97] . among non-infectious diseases to be considered in the differential diagnosis with covid-19 pneumonia, heart failure with cardiogenic pulmonary oedema is one of the most important causes of acute respiratory symptoms, especially among the elderly. on cxr, particularly in the ap projection, a differentiation between the abnormalities caused by covid-19 and signs j o u r n a l p r e -p r o o f of pulmonary oedema may not be as easy as expected. in general, blurring of the vessels, proximal pulmonary vessel enlargement, peribronchial cuffing, evidence of bilateral kerley lines, peribronchovascular thickening, middle-lower distribution of the opacities, and bilateral pleural effusion are typical findings that can allow a confident diagnosis of pulmonary oedema, when associated with the adequate clinical context [98] . this is especially true in patients with history of cardiac disease or in presence of cardiac devices as pacemaker (fig. 13) . on hrct the differential diagnosis from covid-19 is straightforward, with evidence of enlargement of the pulmonary veins, gg opacities and smooth thickening of the interlobular septa and peribronchovascular bundles, mostly with a central distribution in dependent areas, associated with pleural effusion and mediastinal lymph node enlargement [98, 99] . in patients with fibrotic interstitial pneumonia, such as those affected by idiopathic pulmonary fibrosis (ipf), nonspecific interstitial pneumonia (nsip) and chronic hypersensitivity pneumonitis (chp), the onset of an acute exacerbation can mimic symptoms of covid-19 pneumonia. acute exacerbation is defined as clinical worsening of dyspnoea over the last 30 days, which can occasionally manifest with fever and flu-like symptoms [100] , and is considered expression of ali. in this context, hrct is the modality of choice. the hallmarks are the newly appearance of bilateral gg opacities and/or consolidation occurring in the nonfibrotic areas of the lungs -usually with a multifocal or diffuse distribution -and the exclusion of alternative aetiologies, such as pe and infection, especially viral pneumonia [101, 102] . comparison with previous examinations is particularly helpful in interpreting such cases (fig. 14) . drug toxicity is a frequent cause of lung disease [103] . imaging features are various, reflecting different underlying histopathology aspects. the op pattern is a frequent manifestation of drug reaction on hrct [103] and may mimic covid-19 pneumonia in the progressive phase. the typical presentation of op include bilateral, multiple, patchy consolidations with peripheral and peribronchovascular distribution and a lower lobe predominance, associated to perilobular opacities and/or reversed halo sign [57] . in this context, clinical data, pharmacological history and time elapsed between the start of treatment and the occurrence of symptoms are fundamental for the differential diagnosis. the nsip pattern is another possible manifestation of drug reaction on hrct [103] and is usually characterized by diffuse gg opacities with superimposed reticulation and mild-to-severe j o u r n a l p r e -p r o o f traction bronchiectasis/bronchiolectasis, with a predominant bilateral lower lobe distribution [104] , less likely to resemble covid-19 pneumonia. aspiration with inhalation of oropharyngeal or gastric contents into the laryngeal or lower respiratory tract is probably an under-recognized cause of cap and lung injury [105] . the risk of aspiration is increased in patients with reduced level of consciousness, abnormal cough reflex, oropharyngeal dismotility, and gastroesophageal reflux disease (gerd). the onset of symptoms can be acute or subacute, mainly depending on the aspirated content and volume [105, 106] . especially in recumbent patients, the zonal distribution can mimic that of covid-19 pneumonia, with involvement of the posterior segments of the upper lobes and the superior and basal-posterior segments of the lower lobes. either unilateral or bilateral lung involvement is possible [105, 107] . cxr can show central ill-defined alveolar opacities or segmental and lobar opacities, whereas hrct demonstrates consolidation in decumbent areas and multifocal patchy gg opacities, mainly with peribronchovascular distribution [105, 107] , associated with centrilobular nodules and/or tree-in-bud pattern [105] (fig. 15 ). despite similarities with covid-19 pneumonia, the evidence of aspirated material filling the airways as well as the presence of centrilobular nodules and/or tree-in-bud pattern are relevant clues for the differential diagnosis [105] . other possible ancillary findings that may help in ruling-out covid-19 pneumonia are abscess, cavitation, parapneumonic effusion and empyema, or the presence of pulmonary ossification in a dendriform or nodular pattern (expression of underlying chronic aspiration) [105] . in case of aspiration of lipoid material, exogenous lipoid pneumonia can occur. acute aspiration of large volumes of mineral oil (fire-eater pneumonia) [107] can have a clinical presentation similar to pneumonia, while chronic aspiration (animal fats, mineral or vegetable oils) is characterized by insidious symptoms [108] . hrct depicts centrilobular gg opacities or multiple consolidations with a peribronchovascular distribution, mainly involving the lower lobes. also, interlobular septal thickening and patchy multifocal areas of crazy paving pattern can be seen [105, 108] . moreover, the presence of consolidations or mass-like opacities with attenuation values <10 hu (hounsfied unit) is diagnostic of lipoid pneumonia and can be seen in both the acute and chronic setting [108] (fig. 15 ). despite a typical peribronchovascular distribution of the abnormalities, in the absence of fat-containing masses, a definite exclusion of covid-19 pneumonia is not always possible based on imaging features alone. diffuse alveolar haemorrhage (dah) is another potential clinical entity which we could face in the epidemic phase of sars-cov2. dah is associated with pulmonary vasculitis or connective tissue diseases among the others; haemoptysis is usually, but not always, present. cxr shows diffuse alveolar opacities, usually in the mid-zone lungs with subpleural and apical sparing, along the bronchovascular bundles. at hrct, bilateral peribronchovascular gg opacities and smooth septal thickening are visible, with possible coexistent crazy paving, ill-defined centrilobular nodules and consolidations, in case of complete alveolar filling by blood [109] ( fig. 16) . the peribronchovascular distribution makes the diagnosis of covid-19 pneumonia less likely. in the clinical practice, the differential diagnosis between covid-19 vs non-covid-19 patients on hrct remains a critical task due to the overlap of imaging findings. a recently published study demonstrated that use of a simple scoring system based on seven common hrct features (posterior part/lower lobe predilection, bilateral involvement, rounded ggo, subpleural bandlike ggo, crazy paving pattern, peripheral distribution, and ggo +/â�� consolidation) and four uncommon ones (single lobe involvement, central distribution, tree-inbud pattern, and bronchial wall thickening) might be of help in categorizing symptomatic patients as covid-19 vs non-covid-19. a high specificity (95.35%) was achieved with a score greater than 4 [110] . however, it should be noted that imaging findings cannot provide a definite diagnosis alone and that the association with clinical and microbiological data, in a multidisciplinary context, is determinant for an accurate and rapid identification of covid-19 positive cases. due to the pandemic spread of sars-cov-2 infection, it is essential to be familiar with common and uncommon imaging findings of covid-19 pneumonia and their evolution over time on cxr and hrct. cxr might be used as first-line imaging modality in the areas with high levels of contagion as well as in the serial evaluation of hospitalized and critically ill patients. on the other hand, hrct shows a low specificity in areas with low prevalence of disease, and it should be considered the modality of choice in assessing differential diagnosis with other infectious and non-infectious lung diseases and in managing patients with preexisting lung disease. the awareness of relationship between imaging findings and underlying j o u r n a l p r e -p r o o f pathogenesis helps radiologists to increase the level of confidence in diagnosing the disease at its first presentation, as well as in recognizing possible complications and differential diagnosis. the authors declared no conflicts of interest. available at clinical features of patients infected with 2019 novel coronavirus in wuhan clinical characteristics of coronavirus disease olfactory and gustatory dysfunctions as a clinical presentation of mild-to-moderate forms of the coronavirus disease (covid-19): a multicenter european study epidemiology, clinical course, and outcomes of critically ill adults with covid-19 the role of chest imaging in patient management during the covid-19 pandemic: a multinational consensus statement from the fleischner society detection of sars-cov-2 in different types of clinical specimens sensitivity of chest ct for covid-19: comparison to rt-pcr stability issues of rt-pcr testing of sars-cov-2 for hospitalized patients clinically diagnosed with covid-19 chest ct for typical 2019-ncov pneumonia: relationship to negative rt-pcr testing use of chest imaging in the diagnosis and management of covid-19: a who rapid advice guide diagnostic performance of chest x-ray for covid-19 pneumonia during the sars-cov-2 pandemic in lombardy, italy covid-19 and thrombotic or thromboembolic disease: implications for prevention, antithrombotic therapy, and follow-up acute pulmonary embolism in covid-19 patients on ct angiography and relationship to d-dimer levels frequency and distribution of chest radiographic findings in covid-19 positive patients radiographic findings in 240 patients with covid-19 pneumonia: time-dependence after the onset of symptoms chest x-ray findings in asymptomatic and minimally symptomatic quarantined patients in codogno how imaging should properly be used in covid-19 outbreak: an italian experience diagnostic performance of ct and reverse transcriptase polymerase chain reaction for coronavirus disease 2019: a meta-analysis chest ct for detecting covid-19: a systematic review and meta-analysis of diagnostic accuracy the limited sensitivity of chest computed tomography relative to reverse transcription polymerase chain reaction for severe acute respiratory syndrome coronavirus-2 infection: a systematic review covid-19): relationship to duration of infection chest ct findings in cases from the cruise ship "diamond princess diagnostic tools for coronavirus disease (covid-19): comparing ct and rt-pcr viral nucleic acid testing ct imaging and differential diagnosis of covid-19 integrated radiologic algorithm for covid-19 co-rads -a categorical ct assessment scheme for patients with suspected covid-19: definition and evaluation clinical and chest radiography features determine patient outcomes in young and middle age adults with covid-19 covid-19 outbreak in italy: experimental chest x-ray scoring system for quantifying and monitoring disease progression chest x-ray severity index as a predictor of inhospital mortality in coronavirus disease 2019: a study of 302 patients from italy association between initial chest ct or clinical features and clinical course in patients with coronavirus disease 2019 pneumonia ct quantification of pneumonia lesions in early days predicts progression to severe illness in a cohort of covid-19 patients intensive care risk estimation in covid-19 pneumonia based on clinical and imaging parameters: experiences from the munich cohort multicenter cohort study demonstrates more consolidation in upper lungs on initial ct increases the risk of adverse clinical outcome in covid-19 patients well-aerated lung on admitting chest ct to predict adverse outcome in covid-19 pneumonia covid-19 patients and the radiology departmentadvice from the european society of radiology (esr) and the european society of thoracic imaging (esti) ct in coronavirus disease 2019 (covid-19): a systematic review of chest ct findings in 4410 adult patients ct imaging features of 4121 patients with covid-19: a meta-analysis pulmonary pathology of early-phase 2019 novel coronavirus (covid-19) pneumonia in two patients with lung cancer histopathologic changes and sars-cov-2 immunostaining in the lung of a patient with covid-19 pathological findings of covid-19 associated with acute respiratory distress syndrome high expression of ace2receptor of 2019-ncov on the epithelial cells of oral mucosa angiotensin-converting enzyme 2 protects from severe acute lung failure the pathogenesis and treatment of the 'cytokine storm' in covid-19 endothelial cell infection and endotheliitis in covid-19 complement associated microvascular injury and thrombosis in the pathogenesis of severe covid-19 infection: a report of five cases difference of coagulation features between severe pneumonia induced by sars-cov2 and non-sars-cov2 abnormal coagulation parameters are associated with poor prognosis in patients with novel coronavirus pneumonia high-resolution computed tomography findings of acute respiratory distress syndrome, acute interstitial pneumonia, and acute exacerbation of idiopathic pulmonary fibrosis time course of lung changes at chest ct during recovery from 2019 novel coronavirus (covid-19) pneumonia temporal changes of ct findings in 90 patients with covid-19 pneumonia: a longitudinal study covid-19): a systematic review of imaging findings in 919 patients chest ct manifestations of new coronavirus disease 2019 (covid-19): a pictorial review chest ct features of covid-19 in performance of radiologists in differentiating covid-19 from viral pneumonia on chest ct from the radiologic pathology archives: organization and fibrosis as a response to lung injury in diffuse alveolar damage, organizing pneumonia, and acute fibrinous and organizing pneumonia longitudinal ct findings in covid-19 pneumonia: case presenting organizing pneumonia pattern fleischner society: glossary of terms for thoracic imaging acute respiratory distress syndrome: the berlin definition covid-19 does not lead to a "typical the pulmonary sequalae in discharged patients with covid-19: a short-term observational study timely diagnosis and treatment shortens the time to resolution of coronavirus disease (covid-19) pneumonia and lowers the highest and last ct scores from sequential chest ct the unsynchronized changes of ct image and nucleic acid detection in covid-19: reports the two cases from gansu, china a health system response to covid-19 in long term care and post-acute care: a three-phase approach post-covid-19 global health strategies: the need for an interdisciplinary approach postmortem examination of patients with covid-19 prediction of the development of pulmonary fibrosis using serial thin-section ct and clinical features in patients discharged after treatment for covid radiological findings from 81 patients with covid-19 pneumonia in wuhan, china: a descriptive study association of mediastinal lymphadenopathy with covid-19 prognosis the clinical and chest ct features associated with severe and critical covid-19 pneumonia epidemiological and clinical characteristics of 99 cases of 2019 novel coronavirus pneumonia in wuhan, china: a descriptive study management of spontaneous pneumothorax transport of air along sheaths of pulmonic blood vessels from alveoli to mediastinum. clinical implications high incidence of barotrauma in patients with covid-19 infection on invasive mechanical ventilation acute pulmonary embolism and covid-19 pulmonary embolism in patients with covid-19 pneumonia acute pulmonary embolism associated with covid-19 pneumonia detected by pulmonary ct angiography pulmonary embolism in covid-19 patients: awareness of an increased prevalence pulmonary vascular manifestations of covid-19 radiology: cardiothoracic imaging review of the chest ct differential diagnosis of ground-glass opacities in the covid era microbial etiology of community-acquired pneumonia in the adult population of 4 municipalities in eastern finland etiology of community-acquired pneumonia: increased microbiological yield with new diagnostic methods community-acquired pneumonia imaging of community-acquired pneumonia high resolution ct findings in communityacquired pneumonia radiographic and ct features of viral pneumonia severe acute respiratory syndrome: temporal lung changes at thin-section ct in 30 patients middle east respiratory syndrome coronavirus (mers-cov) infection: chest ct findings radiographics update: radiographic and ct features of viral pneumonia severe acute respiratory syndrome: radiographic appearances and pattern of progression in 138 patients radiology perspective of coronavirus disease 2019 (covid-19): lessons from severe acute respiratory syndrome and middle east respiratory syndrome ct correlation with outcomes in 15 patients with acute middle east respiratory syndrome coronavirus radiological society of north america expert consensus statement on reporting chest ct findings related to covid-19. endorsed by the society of thoracic radiology, the american college of radiology, and rsna swine-origin influenza a (h1n1) viral infection: radiographic and ct findings characteristic ct findings distinguishing 2019 novel coronavirus disease (covid-19) from influenza pneumonia covid-19 pneumonia: ct findings of 122 patients and differentiation from influenza pneumonia clinical and radiologic features of pulmonary edema hydrostatic pulmonary edema: highresolution ct findings recent lessons learned in the management of acute exacerbation of idiopathic pulmonary fibrosis acute exacerbation of idiopathic pulmonary fibrosis an international working group report viral pneumonia requiring differentiation from acute and progressive diffuse interstitial lung diseases pulmonary drug toxicity: radiologic and pathologic manifestations nonspecific interstitial pneumonia: radiologic, clinical, and pathologic considerations aspiration-related lung diseases aspiration and infection in the elderly: epidemiology, diagnosis and management aspiration diseases: findings, pitfalls, and differential diagnosis lipoid pneumonia: spectrum of clinical and radiologic manifestations computed tomography of diffuse pulmonary haemorrhage with pathological correlation the authors would like to thank dr. storto for her fundamental contribution to reviewing the structure and the english language of the manuscript. key: cord-296208-uy1r6lt2 authors: greenspan, hayit; san josé estépar, raúl; j. niessen, wiro; siegel, eliot; nielsen, mads title: position paper on covid-19 imaging and ai: from the clinical needs and technological challenges to initial ai solutions at the lab and national level towards a new era for ai in healthcare date: 2020-08-19 journal: med image anal doi: 10.1016/j.media.2020.101800 sha: doc_id: 296208 cord_uid: uy1r6lt2 in this position paper, we provide a collection of views on the role of ai in the covid-19 pandemic, from clinical requirements to the design of ai-based systems, to the translation of the developed tools to the clinic. we highlight key factors in designing system solutions per specific task; as well as design issues in managing the disease at the national level. we focus on three specific use-cases for which ai systems can be built: early disease detection, management in a hospital setting, and building patient-specific predictive models that require the combination of imaging with additional clinical data. infrastructure considerations and population modeling in two european countries will be described. this pandemic has made the practical and scientific challenges of making ai solutions very explicit. a discussion concludes this paper, with a list of challenges facing the community in the ai road ahead. the covid-19 pandemic surprised the world with its rapid spread and has had a major impact on the lives of billions of people. imaging is playing a role in the fight against the disease, in some countries as a key tool, from screening and diagnosis through the entire treatment process, but in other countries, as a relatively minor support tool. guidelines and diagnostic protocols are still being defined and updated in countries around the world. where enabled, computed tomography (ct) of the thorax has been shown to provide an important adjunctive role in diagnosing and tracking progress of covid-19 in comparison to other methods such as monitoring of temperature/respiratory symptoms and the current-gold standard, molecular testing, us* corresponding author: hayit@eng.tau.ac.il;hayitg@gmail.com ing sputum or nasopharyngeal swabs. several countries (including china, netherlands, russia and more) have elected to use ct as a primary imaging modality, from the initial diagnosis through the entire treatment process. other countries, such as the us and denmark as well as developing countries (southeast asia, africa) are using mostly conventional radiographic (x-ray) imaging of the chest (cxr). in addition to establishing the role of imaging, this is the first time ai, or more specifically, deep learning approaches have the opportunity to join in as tools on the frontlines of fighting an emerging pandemic. these algorithms can be used in support of emergency teams, real-time decision support, and more. in this position paper 1 , a group of researchers provide their views on the role of ai, from clinical requirements to the design of ai-based systems, to the infrastructure necessary to facilitate national-level population modeling. many studies have emerged in the last several months from the medical imaging community with many research groups as well as companies introducing deep learning based solutions to tackle the various tasks: mostly in detection of the disease (vs normal), and more recently also for staging disease severity. for a review of emerging works in this space we refer the reader to a recent review article shi et al. (2020a) that covers the first papers published, up to and including march 2020 -in the entire pipeline of medical imaging and analysis techniques involved with covid-19, including image acquisition, segmentation, diagnosis, and follow-up. we also want to point out several special issues in this space-including ieee special issue of tmi, april 2020; ieee special issue of jhbi, 2020; as well as the current special issue of media. in the current position paper, it is not our goal to provide an overview of the publications in the field, rather we present our own experiences in the space and a joint overview of challenges ahead. we start with the radiologist perspective. what are the clinical needs for which ai may provide some benefits? we follow that with an introduction to ai based solutions -the challenges and roadmap for developing ai-based systems, in general and for the covid-19 pandemic. in section 2 of this paper we focus on three specific use-cases for which ai systems can be built: detection, patient management, and predictive models in which the imaging is combined with additional clinical features. system examples will be briefly introduced. in section 3 we present a different perspective of ai in its role in the upstream and downstream management of the pandemic. specific infrastructure considerations and population modeling in two european countries will be described in section 4. a discussion concludes this paper, with a list of challenges facing the community in our road ahead. as of this writing, according to the johns hopkins resource center (https://coronavirus.jhu.edu/), there are, approximately, 12.5 million confirmed cases with 561,000 deaths throughout the world, with 32,000 deaths in new york state alone. the rate of increase in cases has continued to rise as demonstrated by the log scale plot in figure 1 . the most common symptoms of the disease, fever, fatigue, dry cough, runny nose, diarrhea and shortness of breath are non-specific and are common to many people with a variety of conditions. the mean incubation period is approximately 5 days and the virus is probably most often transmitted by asymptomatic patients. knowing who is positive for the disease has critical implications for keeping patients away from others. unfortunately, the gold standard lab test, real time reverse transcription polymerase chain reaction (rt-pcr) which detects viral nucleic acid, has not been universally available in many areas and its sensitivity varies considerably depending on how early patients are tested in the course of their disease. recent studies have suggested that rt-rpr has a sensitivity of only 61-70%. consequently, repeat testing is often required to ensure a patient is actually free of the disease. fang et al (fang et al. (2020) ) found that for the 51 patients they studied with thoracic ct and rt-pcr assay performed within 3 days of each other, the sensitivity of ct for covid-19 was 98% compared to rt-pcr sensitivity of 71% (p < .001). on cxr and ct exams of the thorax, findings are usually bilateral (72%) early in the progression of disease and even more likely bilateral (88%) in later stages , zhao et al. (2020) ). the typical presentation in icu patients is bilateral subsegmental areas of air-space consolidation. in non-icu patients, classic findings are transient subsegmental consolidation early and then bilateral ground glass opacities that are typically peripheral in the lungs. pneumothorax (collapsed lung) and pleural fluid or cavitation (due to necrosis) are usually not seen. distinctive patterns of covid-19 such as crazy paving in which ground glass opacity is combined with superimposed interlobular and intralobular septal thickening and the reverse halo sign where a ground glass region of the lung is surrounded by an irregular thick wall have been previously described in other diseases but are atypical of most pneumonias. the use of thoracic ct for both diagnosis of disease and tracking has varied tremendously from country to country. while countries such as china and iran utilize it for its very high sensitivity to disease in the diagnosis and tracking of progression of disease, the prevailing recommendation in the us and other countries is to only use lab studies for diagnosis, use chest radiography to assess severity of disease, and to hold off on performing thoracic ct except for patients with relatively severe and complicated manifestations of disease (simpson et al. (2020) ). this is due to concerns in the us about exposure of radiology staff and other patients to covid-19 patients and the thought that ct has limited incremental value over portable chest radiographs which can be performed outside the imaging department. additionally, during a surge period, the presump-tion is made that the vast majority of patients with pulmonary symptoms have the disease, rendering ct as a relatively low value addition to the clinical work-up. as a diagnostic tool, ct offers the potential to differentiate patients with covid-19 not only from normal patients, but from those with other causes of shortness of breath and cough such as tb or other bacterial or alternatively, other viral pneumonias, bronchitis, heart failure, and pulmonary embolism. as a quantitative tool, it offers the ability to determine what percentage of the lung is involved with the disease and to break this down into areas of ground glass density, consolidation, collapse, etc. this can be evaluated on serial studies which may be predictive of a patient's clinical course and may help to determine optimal clinical treatment. complications of covid-19 are not limited to acute lung parenchymal disease. these patients have coagulopathies and are at increased risk for pulmonary embolism. diffuse vascular inflammation can result in pericarditis and pericardial effusions. renal and brain manifestations have been described by many authors and are increasingly recognized clinically in covid-19 patients. long term lung manifestations will not be apparent for many months or years, but there is the potential that these patients will develop higher rates of chronic obstructive pulmonary disease (copd) such as emphysema, chronic bronchitis and asthma than the general population. objective metrics for assessment and follow-up of these complications of the disease would be very valuable from a clinical perspective. the extraordinarily rapid spread of the covid-19 pandemic has demonstrated that a new disease entity with a subset of relatively unique characteristics can pose a major new clinical challenge that requires new diagnostic tools in imaging. the typical developmental cycle and large number of studies required to develop ai algorithms for various disease entities is much too long to respond effectively to produce these software tools on demand. this is complicated by the fact that the disease can have different manifestations (perhaps due to different strains) in different regions of the world. this suggests the strong need to develop software more rapidly, perhaps using transfer learning from existing algorithms, to train on a relatively limited number of cases, and to train on multiple datasets in various locations that may not be able to be easily combined due to privacy and security issues. it also suggests that we determine how to balance regulatory requirements for adequate testing of the safety and efficacy of these algorithms against the need to have them available in a timely manner to impact clinical care. ai technology, in particular deep learning image analysis tools, can potentially be developed to support radiologists in the triage, quantification, and trend analysis of the data. ai solutions have the potential to analyze multiple cases in parallel to detect whether chest ct or chest radiography reveals any abnormalities in the lung. if the software suggests a significantly increased likelihood of disease, the case can be flagged for further review by a radiologist or clinician for possible treatment/quarantine. such systems, or variations thereof, once verified and tested can become key contributors in the detection and control of patients with the virus. another major use of ai is in predictive analytics: foreseeing events for timely intervention. predictive ai can be potentially applied at three scales: the individual scale, the hospital scale, and the societal scale. an individual may go through various transitions from healthy to potentially contaminated, symptomatic, etc. as depicted in figure 2 . at the individual level, we may use ai for computing risk of contamination based on location, risk of severe covid-19 based on co-morbidities and health records, risk of acute respiratory distress syndrome (ards) and risk of mortality to help guide testing, intervention, hospitalization and treatment. quantitative ct or chest radiographic imaging may play an important role in risk modeling for the individual, and especially in the risk of ards. at the hospital level, ai for imaging may for example be used for workflow improvement by (semi-) automating radiologist's interpretations, and by forecasting the future need for icu and ventilator capacity. at the societal level ai may be used in forecasting hospital capacity needs and may be an important measure to aid in assessing the need for lock downs and reopenings. so far, we have here concentrated on disease diagnosis and management, but imaging with ai may also have a role to play in relation to late effects like neurological, cardiovascular, and respiratory damage. before entering the discussion on specific usages of ai to ease the burden of the pandemic, we briefly describe the standard procedure of creating an ai solution in order to clarify the nomenclature. the standard way of developing deep learning algorithms and systems entails several phases (greenspan et al. (2016) , litjens et al. (2017) ) : i. data-collection, in which a large amount of data samples need to be collected from predefined categories; expert annotations are needed for groundtruthing the data; ii. training phase in which the collected data are used to train network models. each category needs to be represented well enough so that the training can generalize to new cases that will be seen by the network in the testing phase. in this learning phase, the large number of network parameters (typically on the order of millions) are automatically defined; iii. testing phase in which an additional set of cases not used in training is presented to the network and the output of the network is tested statistically to determine its success of categorization. finally, iv, the software must be validated on independent cohorts to ensure that performance characteristics generalize to unseen data from other imaging sources, demographics, and ethnicity. in the case of a new disease, such as the coronavirus, datasets are just now being identified and annotated. there are very limited data sources as well as limited expertise in labeling the data specific to this new strain of the virus in humans. accordingly, it is not clear that there are enough examples to achieve clinically meaningful learning at this early stage of data collection despite the increasingly critical importance of this software. solutions to this challenge, that may enable rapid development, include the combination of several technologies: transfer learning will utilize pretraining on other but somehow statistically similar data. in the general domain of computer vision, ima-genet has been used for this purpose (donahue et al. (2014) ). in the case of covid-19 this may be provided by existing databases of annotated images of patients with other lung infections. data augmentation is a trick used from the beginning of applying convolution neural networks (cnns) to imaging data (lecun et al. (1989) ), in which data are transformed to provide extra training data. normally rotations, reflections, scaling or even group actions beyond the affine group can be explored. other technologies include semi-supervised learning and weak learning when labels are noisy and/or missing (cheplygina et al. (2019)). thus, the underlying approach to enable rapid development of new ai-based capabilities, is to leverage the ability to modify and adapt existing ai models and combine them with initial clinical understanding to address the new challenges and new disease entities, such as the covid-19. in this section we briefly review three possible system developments: ai systems for detection and characterization of disease, ai systems for measuring disease severity and patient monitoring, and ai systems for predictive modeling. each category will be reviewed briefly and a specific system will be described with a focus on the ai based challenges and solutions. the vast majority of efforts for the diagnosis of covid-19 have been focused on detecting unique injury patterns related to the infection. automated recognition of those patterns became an ideal challenge for the use of cnns trained on the appearance of those patterns. one example of a system for covid-19 detection and analysis is shown in figure 3 , which presents an overview of the analysis conducted in gozes et al. (2020a) . in general, as is shown here, automated solutions are comprised of several components. each one is based on a network model that focuses on a specific task to solve. in the presented example, both 3d and 2d analysis are conducted, in parallel. 3d analysis of the imaging studies is utilized for detection of nodules and focal opacities using nodule-detection algorithms, with modifications to detect ground-glass (gg) opacities. a 2d analysis of each slice of the case is used to detect and localize covid-19 diffuse opacities. if we focus on the 2d analysis -we again see that multiple steps are usually defined. the first step is the extraction of the lung area as a region of interest (roi) using a lung segmentation module. the segmentation step removes image portions that are not relevant for the detection of withinlung disease. within the extracted lung region, a covid-19 detection procedure is conducted, utilizing one of a variety of possible schemes and corresponding networks. for example, this step can be a procedure for (undefined) abnormality detection, or a specific pattern learning task. in general, a classification neural network (covid-19 vs. not covid-19) is a key component of the solution. such networks, which are mostly cnn based, enable the localization of covid-19 manifestations in each 2d slice that is selected in what have become known asheat maps per 2d slice. to provide a complete review of the case, both the 2d and 3d analysis results can be merged. several quantitative measurements and output visualizations can be used, including per slice localization of opacities, as in figure 4 (a), and a 3d volumetric presentation of the opacities throughout the lungs, as shown in figure 4 (b), which presents a 3d visualization of all gg opacities. several studies have shown the ability to segment and classify the extracted lesions using neural networks to provide a diagnostic performance that matches a radiologist rating ; bai et al. (2020) ). in zhang et al. (2020) , 4,695 manually annotated ct slices were used for seven classes, including background, lung field, consolidation (cl), groundglass opacity (ggo), pulmonary fibrosis, interstitial thickening, and pleural effusion. after a comparison between different semantic segmentation approaches, they selected deeplabv3 as their segmentation detection backbone (chen et al. (2017) ). the diagnostic system was based on a neural network fed by the lung-lesion maps. the system was designed to classify normals from common pneumonia and covid-19 specific pneumonia. their results show a covid-19 diagnostic accuracy of 92.49% tested in 260 subjects. in bai et al. (2020) , a direct classification of covid-19 specific pneumonia versus other etiologies was performed using an efficientnet b5 network (tan and le (2019)) followed by a two-layer fully connected network to pool the information from multiple slices and provide a patientlevel diagnosis. this system yielded a 96% accuracy in a testing set of 119 subjects compared to an 85% average accuracy for six radiologists. these two examples exemplify the power of ai to perform at a very high level that may augment the radiologist, when designed and tested for a very narrow and specific task within a de-novo diagnostic situation. time delay in covid-19 testing using rt-pcr can be overcome with integrative solutions. augmented testing using ct, clinical symptoms, and standard white blood cell (wbc) panels has been proposed in mei et al. (2020) . the authors show their ai system that integrates both sources of information is superior to an imaging-alone cnn model as well as a machine learning model based on non-imaging information for the diagnosis of covid-19. integrative approaches can overcome the lack of diagnostic specificity of ct imaging for covid-19 (rubin et al. ( 2020)) it is well understood that chest radiographs (cxr) have lower resolution and contain much less information than their ct counterparts. for example, for covid-19 patients, the lungs may be so severely infected that they become fully opacified, obscuring details on an x-ray and making it difficult to distinguish between pneumonia, pulmonary edema, pleural effusions, alveolar hemorrhage, or lung collapse ( figure 5 ). still, many countries are using cxr information for initial decision support as well as throughout the patient hospitalization and treatment process ). deep learning pipelines for cxr opacities and infiltration scoring exist. in most publications seen to-date, researchers utilize existing public pneumonia datasets, which were available prior to the spread of coronavirus, to develop network solutions that learn to detect pneumonia on a cxr. in (selvan et al. (2020) ), an attempt to solve the issue of the compact lungs is presented using variational imputation. a deep learning pipeline based on variational autoencoders (vae) has shown in pilot studies > 90% accuracy in separating covid19 patients from other patients with lung infections, both bacterial and viral. a systematic evaluation of one of those system has demonstrated comparable performance to a chest radiologist (murphy et al. (2020) ). this demonstrates the capability of recognizing covid-19 associated patterns, using the cxr data. we view these results as preliminary, and to be confirmed with more rigorous experimental setup which includes access to covid-19 and other infections from the same sources with identical acquisition technology, time-window, ethnicity, demographics, etc. such rigorous experiments are critical in order to assess the clinical relevance of the developed technology. in this section we focus on the use of ai for hospitalized patients. image analysis tools can support measurement of the disease extent within the lungs, thus generating quantification for the disease that can serve as an image-based biomarker. such a biomarker may be used to assess relative severity of patients in the hospital wards, enable tracking of disease severity over time, and thus assist in the decision-making process of the physicians handling the case. one such biomarker, termed the corona score, was recently introduced in gozes et al. (2020a,b) . the corona score is a measure of the extent of opacities in the lungs. it can be extracted in ct and in cxr cases. figure 6 presents a plot of corona-score measurements per patient over time, in ct cases. using the measure, we can assess relative severity of the patients (left) as well as extract a model for disease burden over the course of treatment (right). additional very valuable information on characterization of disease manifestation can be extracted as well, such as locations of opacities within the lungs, opacities burden within specific lobs of the lungs (using a lungs lobe segmentation module) and analysis of the texture of the opacities using classification of patches extracted from detected covid-19 areas (using a patch-based classification module). these characteristics are important biomarkers with added value for patient monitoring over time. the clinical covid-19 lung infections are diagnosed and monitored with ct or cxr imaging where opacities, their type and extent, may be quantified. the picture of radiological findings in covid-19 patients is complex (wong et al. (2020) ) with mixed patterns: ground-glass opacities, opacities with a rounded morphology, peripheral distribution of disease, consolidation with groundglass opacities, and the so called crazy-paving pattern. first reporting of longitudinal developments monitored by cxr (shi et al. (2020b) ) indicate that cxr findings occur before the need for clinical intervention with oxygen and/or ventilation. this fosters the hypothesis that cxr imaging and quantification of findings are valuable in the risk assessment of the individual patient developing severe covid-19. in the capital region of denmark, it is standard practice to acquire a cxr for covid-19 patients. the clinical workflow during the covid-19 pandemic does not in general allow for manual quantitative scoring of radiographs for productivity reasons. making use of the cxr already recorded during real time risk assessment therefore requires automated methods for quantification of image findings. several scoring systems for the severity of covid-19 lung infection adapted from general lung infection schemes have been proposed (wong et al. (2020) , shi et al. (2020b) , cohen et al. (2020) ). above, in figure 4 , it is shown how opacities may be located in ct images. similar schemes may be used for regional opacity scoring in cxr, as shown in figure 7 . for the administration and risk profiling of the individual patient, imaging does not tell the full story. important risk factors include age, bmi, co-morbidities (especially diabetes, hypertension, asthma, chronic respiratory or heart diseases) (jordan et al. (2020) ). combining imaging with this type of information from the ehr and with data representing the trajectory of change over time enhances the ability to determine and predict the stage of disease. an early indication is that cxr's contribute significantly to the prediction of the probability for a patient to be on a ventilator. here we briefly summarise the patient trajectory prognosis setup: we have in preliminary studies from the cohort from the capital and zealand regions of denmark, combined clinical information from electronic health records (ehr) defining variables relating to vital parameters, comorbidities, and other health parameters with imaging information. modeling was performed using a simple random forest implementation in a 5-fold cross-validation fashion. in figure 8 are as illustration auc for prediction of outcome in terms of hospitalisation, requirement for ventilator, admission to intensive care unit, and death. these have been illustrated on 2,866 covid-19 positive subjects from the zealand and capital region of denmark. these are preliminary unconsolidated results for illustrative purposes. however, these support the feasibility of an algorithm to predict severity of covid-19 manifestations early in the course of the disease. the combination of cxr into these prognostic tools have been performed by including a number of quantitative features per lung region as a feature vector in the random forest described above. imaging has played a unique role in the clinical management of the covid-19 pandemic. public health authorities of many affected countries have been forced to implement severe mitigation strategies to avoid the wide community spread of the virus (parodi and liu (2020) ). mitigation strategies put forth have focused on acute disease management and the plethora of automated imaging solutions that have emerged in the wake of this crisis have been tailored toward this emergent need. until effective therapy is proven to prevent the widespread dissemination of the disease, mitigation strategies will be followed by more focused efforts and containment approaches aimed at avoiding the high societal cost of new confinement policies. in that regard, imaging augmented by ai can also play a crucial role in providing public health officials with pandemic control tools. opportunities in both upstream infection management and downstream solutions related to disease resolution, monitoring of recurrence and health security will be emerging in the months to come as economies reopen to normal life. pandemic control measurements in the pre-clinical phase of the infection may seek to identify those subjects that are more susceptible to the disease due to their underlying risk factors that lead to the acute phase of covid-19 infections. several epidemiological factors, including age, obesity, smoking, chronic lung disease hypertension, and diabetes, have been identified as risk factors (petrilli et al. (2020) ). however, there is a need to understand further risk factors that can be revealed by image-based studies. imaging has shown to be a powerful source of information to reveal latent traits that can help identify homogeneous subgroups with specific determinants of disease (young et al. (2018) ). this kind of approach could be deployed in retrospective databases of covid-19 patients with pre-infection imaging to understand why some subjects seem to be much more prone to progression of the viral infection to acute pulmonary inflammation. the identification of high-risk populations by imaging could enable targeted preventive measurements and precision medicine approaches that could catalyze the development of curative and palliative therapies. identification of molecular pathways in those patients at a higher risk may be crucial to catalyze the development of much needed host-targeting therapies. the resolution of the infection has been shown to involve recurrent pulmonary inflammation with vascular injury that has led to post-intensive care complications (ackermann et al. (2020) ). detection of micro embolisms is a crucial task that can be addressed by early diagnostic methods that monitor vascular changes related to vascular pruning or remodeling. methods developed within the context of pulmonary embolism detection, and clot burden quantification could be repurposed for this task ). another critical aspect of controlling the pandemic is the need to monitor infection recurrence as the immunity profile for sars-cov-2 is still unknown (kirkcaldy et al. (2020) ). identifying early pulmonary signs that are compatible with covid-19 infection could be an essential tool to monitor subjects that may relapse in the acute episode. ai methods have shown to be able to recognize covid-19 specific pneumonia identified on radiographic images (murphy et al. (2020) ). the accessibility and potential portability of the imaging equipment in comparison to ct images could enable early pulmonary injury screening if enough specificity can be achieved in the early phases of the disease. eventually, some of those tools might facilitate the implementation of health security screening solutions that revolve around the monitoring of individuals that might present compatible symptoms. although medical imaging solutions might have a limited role in this space, other kinds of non-clinical imaging solutions such as thermal imaging may benefit from solutions that were originally designed in the context of x-ray or ct screening. one of the fascinating aspects that has emerged around the utilization of ai-based imaging approaches to manage the covid-19 pandemic has been the speed of prototyping imaging solutions and their integration in end-to-end applications that could be easily deployed in a healthcare setting and even ad-hoc makeshift caring facilities. this pandemic has shown the ability of deep neural networks to enable the development of end-to-end products based on a model representation that can be executed in a wide range of devices. another important aspect has been the need for large-scale deployments due to the high incidence of the covid-19 infection. these deployments have been empowered by the use of cloud-based computing architectures and multi-platform web-based technologies. multiple private and open-source systems have been rapidly designed, tested, and deployed in the last few months. the requirements around the utilization of these systems in the general population for pandemic control are: • high-throughput: the system needs to have the ability to perform scanning and automated analysis within several seconds if screening is intended. • portable: the system might need to reach the community without bringing them to hospital care settings to avoid nosocomial infections. • reusable: imaging augmented with ai has emerged as a highly reusable technique with scalable utilization that can adapt to variable demand. • sensitive: the system needs to be designed with high sensitivity and specificity to detect early signs of disease. • private: systems have to protect patient privacy by minimizing the exchange of information outside of the care setting. web-based technologies that provide embedded solutions to deploy neural network systems have emerged as one of the most promising implementations that fulfill those requirements. multiple public solutions in the context of chest xray detection of early pneumonia and covid-19 compatible pneumonia have been prototyped, as shown in figure 9 . the covictory app, part of the slow-down covid project (www. slowdowncovid19.org) implements a classification neural network for the detection of mild pneumonia as an early risk detection of radiographic changes compatible with covid-19. the developers of this system based their system in a network architecture recently proposed for tuberculosis detection that has a very compact and efficient design well-suited for deployment in mobile platforms (pasa et al. (2019) ). the database that trained the network was based on imaging from three major chest x-ray databases: nih chest x-ray, chexpert, and padchest. the developers sub-classified x-ray studies labeled as pneumonia in mild versus moderate/severe pneumonia fig. 9 . illustration of a public ai systems for covid-19 compatible pneumonias on chest x-rays from two covid-19 subjects using covictory app (www.covictoryapp.org) with mild pneumonia signs (left) and more severe disease (right). by consensus of multiple readers using spark crowd, an open source system for consensus building (rodrigo et al. (2019) ). another example is the coronavirus xray app that included public-domain images from covid-19 patients to classify images into three categories: healthy, pneumonia and covid-19. both systems were implemented as a static web application in javascript using tensorflow-js. although the training was carried out using customized gpu hardware, the deployment of trained models is intrinsically multi-platform and multi-device thanks to the advancement of web-based technologies. other commercial efforts like cad4covid-xray (https://www. delft.care/cad4covid/) has leveraged prior infrastructure used for the assessment of tuberculosis on x-ray to provide a readily deployable solution. the covid-19 crisis has seen the emergence of multiple observational studies to support research into understanding disease risk, monitoring disease trajectory, and for the development of diagnostic and prognostic tools, based on a variety of data sources including clinical data, samples and imaging data. all these studies share the theme that access to high quality data is of the essence, and this access has proven to be a challenge. the causes for this challenge to observational covid-19 research are actually the same ones that have hampered large scale data-driven research in the health domain over the last years. owing to the data collection that takes place in different places and different institutes, there is fragmentation of data, images and samples. moreover there is a lack of standardization in data collection, which hampers reuse of data. consequently, the reliability, quality and re-usability of data for datadriven research, including the development and validation of ai applications, is problematic. finally, depending on the sys-tem researchers and innovators are working in, ethical and legal frameworks are often unclear and may sometimes be (interpreted as being) obstructive. a coordinated effort is required to improve the accessibility to observational data for covid-19 research. if implemented for covid-19, it can actually serve as a blueprint for large, multi-center observational studies in many domains. as such, addressing the covid-19 challenges also presents us with an opportunity, and in many places we are already observing that hurdles towards multicenter data accessibility are being addressed with more urgency. an example is the call by the european union for an action to create a pan-european cohort covid-19 including imaging data. in the netherlands, the health-ri c initiative aims to build a national health data infrastructure, to enable the re-use of data for personalized medicine, and similar initiatives exist in other countries. in light of the current pandemic, these initiatives have focused efforts on supporting observational covid-19 research, with the aim to facilitate data access to multi-center data. the underlying principle of these infrastructures is that by definition they will have to deal with the heterogeneous and distributed nature of data collection in the healthcare system. in order for such data to be re-usable, harmonisation at the source is required. this calls for local data stewardship, in which the different data types, including e.g. clinical, imaging and lab data, need to be collected in a harmonized way, adhering to international standards. here, the fair principle needs to be adopted, i.e. data needs to be stored such that they are findable, accessible, interoperable and reusable wilkinson et al. (2016) . for clinical data, it is not only important that the same data are collected (e.g. adhering to the world health organisation case report form (crf), often complemented with additional relevant data), but also that their values are unambiguously defined and are machine-readable. the use of electronic crfs (ecrfs) and accompanying software greatly supports this, and large international efforts exist to map observational data to a common data model, including e.g. the observational health data sciences and informatics (ohdsi) model. similarly the imaging and lab data should be processed following agreed standards. in the health-ri c implementation, imaging data are pseudonimized using a computational pipeline that is shared between centers. for lab data, standard ontologies such as loinc can be employed. the covid-19 observational project will not only collect fair metadata describing the content and type of the data, but also data access policies for the data that are available. this will support the data search, request, and access functionalities provided by the platform. an illustration of the data infrastructure in health-ri c is provided in fig 10. next to providing data for the development of ai algorithms, it is important to facilitate their objective validation. in the medical imaging domain, challenges have become very popular to objectively compare performance of different algorithms. in the design of challenges, part of the data needs to be kept apart. it is therefore important that, while conducting efforts to provide access to observational covid-19 data, we already plan for using part of the data for designing challenges around relevant clinical use cases. fig. 10 . design of covid-19 observational data platform. in order for hospitals to link to the data platform, they need to make their clinical, imaging and lab data fair. tools for data harmonization (fair-ification) are being shared between institutes. fair metadata (and in some cases fair data) and access policies are shared with the observational platform. this enables a search tool for researchers to determine what data resources are available at the participating hospitals. these data can subsequently be requested, and if the request is approved by a data access committee, the data will be provided, or information how the data can be accessed will be shared. in subsequent versions of the data platform, also distributed learning will be supported, so that data can stay at its location. during the pandemic, setting up such an infrastructure from scratch will not lead to timely implementation. health-ri c was already in place prior to the pandemic, and some of its infrastructures could be adjusted to start building a covid-19 observational data platform. in denmark, a similar initiative was not in place. however, in eastern denmark, the capital and zealand regions share a common data platform in all hospitals with a common ehr and a pacs at each region covering in total 18 hospitals and 2.6 million citizens making data collection and curation relatively simple. at continental scale, solutions are being created, but will likely not be in place during the first wave of the pandemic. the burdens to overcome are legal, political, and technical. access to un-consented data from patients follows different legal paths in different countries. in uk the department of social and mental care issued on march 20, 2020 a notice simplifying the legal approval of covid-19 data processing. in denmark, usual regulation and standards were maintained, but authorities made an effort to grant permission by the usual bodies in fast track. as access to patient information must be restricted, not every researcher with any research goal can be granted access. without governance in place prior to an epidemic, access will be granted on an ad hoc and first-come-first-served basis, not necessarily leading to the most efficient data analysis. finally, data are hosted in many different it systems and the two major technical challenges lie in bringing data to a common platform, and having a (in eu gdpr) compliant technical setup for collaboration. building such infrastructure with proper security and data handling agreements in place is complex and will lead to substantial delays if not in place prior to the epidemic. in the netherlands, the health-ri c platform was in place. in denmark, the efforts have been constrained to the eastern part of the country sharing common ehr and pacs and having infrastructure in place for compliant data sharing at computerome. at a european scale, the commission launched the european covid-19 data platform on april 20 building on existing hardware infrastructure. this was followed up by a call for establishing a pan-european covid-19 cohort. funding decision will be in august 2020. even though a tremendous effort has been put in place and usual approvals of access and funding have been fast-tracked, proper infrastructures have not been created in time for the first wave in europe. the current covid-19 pandemic offers us historic challenges but also opportunities. it is widely believed that a substantial percentage of the (as of this writing) 12.5 million confirmed cases and 561,000 deaths and trillions of dollars of economic losses would have been avoided with adequate identification of those with active disease and subsequent tracking of location of cases and prediction of emerging hotspots. imaging has already played a major role in diagnosis and tracking and prediction of outcomes and has the potential to play an even greater role in the future. automated computer based identification of probability of disease on chest radiographs and thoracic ct combined with tracking of disease could have been utilized early on in the development of cases, first in wuhan, then other areas of china and asia, and subsequently europe and the united states and elsewhere. this could have been utilized to inform epidemiologic policy decisions as well as hospital resource utilization and ultimately, patient care. this pandemic also represented, perhaps for the first time in history, that a disease with relatively unique imaging and clinical characteristics emerged and spread globally faster than the knowledge to recognize, diagnose and treat the disease. it also created a unique set of challenges and opportunities for the machine learning/ai community to work side by side and in parallel with clinical experts to rapidly train and deploy computer algorithms to treat an emerging disease entity. this required a combination of advanced techniques such as the use of weakly annotated schemes to train models with relatively tiny amounts of training data which has only become widely available recently, many months after the initial outbreak of disease. the imaging community as a whole has demonstrated that extremely rapidly developed ai software using existing algorithms can achieve high accuracy in detection of a novel disease process such as covid-19 as well as provide rapid quantification and tracking. the majority of research and development has focused on pulmonary disease with developers using standard chest-ct dicom imaging data as input for algorithms designed to automatically detect and measure lung abnormalities associated with covid-19. the analysis includes automatic detection of involved lung volumes, automatic measurement of disease as compared to overall lung volume and enhanced visualization techniques that rapidly depict which areas of the lungs are involved and how they change over time in an intuitive manner that can be clinically useful. a variety of manuscripts describing automated detection of covid-19 cases have been recently published. when reviewing these manuscripts one can see the following interesting trend: all are focusing on one of the several key tasks, as defined herein. each publication has a unique system design that contains a set of network models, or a comparison across models; and the results are all very strong. the compelling results, such as the ones presented herein may lead us to conclude that the task is solved; but is this the correct conclusion? it seems that the detection and quantification tasks are in fact solvable with our existing imaging analysis tools. still, there are several data-related issues which we need to be aware of. experimental evidence is presented on datasets of hundreds and we need to go to real world settings, in which we will start exploring thousands and even more cases, with large variability. our systems to date are focusing on detection of abnormal lungs in a biased scenario of the pandemic in which there is a very high prevalence of patients presenting with the disease. once the pandemic declines substantially, the shift will be immediate to the need to detect covid among a wide variety of diseases including other lung inflammatory processes, occupational exposures, drug reactions, and neoplasms. in that future in which the prevalence of disease is lower,will our solutions that work currently be sensitive enough, without introducing too many false positives? that is the crux of many of the current studies that have tested the different ai solutions within a very narrow diagnostic scope. there are many possibilities and promising directions, yet the unknown looms larger than the known. just as the current pandemic has changed the way many are thinking about distance learning, the practice of telemedicine, and overall safety in a non-socially distanced society, it seems that we are similarly setting the stage with our current on the fly efforts in algorithm development for the future development and deployment of ai. we need to update infrastructure including methods of communication and sharing cases and findings as well as reference databases and algorithms for research, locally, country-based and globally. we need to prove the strengths, build the models and make sure that the steps forward are such that we can continue and expand the use of ai, particularly just in time ai. we believe that imaging is an absolutely vital component of the medical space. for predictive modeling we need to not limit ourselves to just the pixel data but also include additional clinical, patient level information. for this, combined effort among many groups, as well as state and federal level support will result in optimal development, validation, and deployment. many argue that we were caught unaware from a communication, testing, treatment and resource perspective with the current pandemic. but deep learning-augmented imaging has emerged as a unique approach that can deliver innovative solutions from conception to deployment in extreme circumstances to address a global health crisis. the imaging community can take lessons learned from the current pandemic and use them to not only be far better prepared for recurrence of covid-19 and future pandemics and other unexpected diseases, but also use these lessons to advance the art and science of ai as applied to medical imaging in general. pulmonary vascular endothelialitis, thrombosis, and angiogenesis in covid-19 ai augmentation of radiologist performance in distinguishing covid-19 from pneumonia of chest ct findings in coronavirus disease-19 (covid-19): relationship to duration of infection rethinking atrous convolution for semantic image segmentation not-so-supervised : a survey of semi-supervised, multi-instance, and transfer learning in medical image analysis covid-19 image data collection decaf: a deep convolutional activation feature for generic visual recognition rapid ai development cycle for the coronavirus (covid-19) pandemic: initial results for automated detection and patient monitoring using deep learning ct image analysis coronavirus detection and analysis on chest ct with deep learning guest editorial deep learning in medical imaging: overview and future promise of an exciting new technique peneta scalable deeplearning model for automated diagnosis of pulmonary embolism using volumetric ct imaging covid-19: risk factors for severe disease and death covid-19 and postinfection immunity: limited evidence, many remaining questions backpropagation applied to handwritten zip code recognition a survey on deep learning in medical image analysis artificial intelligence-enabled rapid diagnosis of patients with covid-19 covid-19 on the chest radiograph: a multi-reader evaluation of an ai system from containment to mitigation of covid-19 in the us efficient deep network architectures for fast chest x-ray tuberculosis screening and visualization factors associated with hospital admission and critical illness among 5279 people with coronavirus disease spark-crowd: a spark package for learning from crowdsourced big data the role of chest imaging in patient management during the covid-19 pandemic: a multinational consensus statement from the fleischner society lung segmentation from chest x-rays using variational data imputation review of artificial intelligence techniques in imaging data acquisition, segmentation and diagnosis for covid-19 radiological findings from 81 patients with covid-19 pneumonia in wuhan, china: a descriptive study. the lancet infectious diseases radiological society of north america expert consensus statement on reporting chest ct findings related to covid-19 efficientnet: rethinking model scaling for convolutional neural networks. arxiv.org the fair guiding principles for scientific data management and stewardship frequency and distribution of chest radiographic findings in covid-19 positive patients the role of imaging in 2019 novel coronavirus pneumonia (covid-19) uncovering the heterogeneity and temporal complexity of neurodegenerative diseases with subtype and stage inference relation between chest ct findings and clinical conditions of coronavirus disease (covid-19) pneumonia: a multicenter study the authors declare the following financial interests/personal relationships which may be considered as potential competing interests: key: cord-167889-um3djluz authors: chen, jianguo; li, kenli; zhang, zhaolei; li, keqin; yu, philip s. title: a survey on applications of artificial intelligence in fighting against covid-19 date: 2020-07-04 journal: nan doi: nan sha: doc_id: 167889 cord_uid: um3djluz the covid-19 pandemic caused by the sars-cov-2 virus has spread rapidly worldwide, leading to a global outbreak. most governments, enterprises, and scientific research institutions are participating in the covid-19 struggle to curb the spread of the pandemic. as a powerful tool against covid-19, artificial intelligence (ai) technologies are widely used in combating this pandemic. in this survey, we investigate the main scope and contributions of ai in combating covid-19 from the aspects of disease detection and diagnosis, virology and pathogenesis, drug and vaccine development, and epidemic and transmission prediction. in addition, we summarize the available data and resources that can be used for ai-based covid-19 research. finally, the main challenges and potential directions of ai in fighting against covid-19 are discussed. currently, ai mainly focuses on medical image inspection, genomics, drug development, and transmission prediction, and thus ai still has great potential in this field. this survey presents medical and ai researchers with a comprehensive view of the existing and potential applications of ai technology in combating covid-19 with the goal of inspiring researches to continue to maximize the advantages of ai and big data to fight covid-19. 1 introduction figure 1 . main scope of ai in fighting against covid-19. we collected 1273 online publications related to covid-19, sars-cov-2, and 2019-ncov from databases such as nature, elsevier, google scholar, arxiv, biorxiv, and medrxiv. then, we filter out 267 papers that explicitly use ai methods. long short-term memory [66] vae variational auto-encoder [186] serological diagnosis, chest x-ray and ct image inspection, and other noninvasive methods. benefitting from the advantages of high sensitivity and specificity, real-time reverse transcriptase polymerase chain reaction (rt-pcr) is the current standard detection technology in diagnosing the sars-cov-2 virus and bacterial infections. using rt-pcr, 9 rna positives were detected from pharyngeal swabs of patients, indicating that the sars-cov-2 virus had spread in communities of wuhan, china, in early january 2020 [106] . the shedding of the sars-cov-2 virus detected in the throat, lungs, and feces suggests multiple routes of virus transmission [217, 227] . however, rt-pcr faces the limitations of a complicated sample preparation, low detection efficiency, and high false-negative rate [106, 212, 223] . isothermal nucleic acid amplification and blood testing methods are also commonly used for the rapid screening of sars-cov-2 [103, 122, 223 ]. an ml classification method was used for blood testing to extract important routine hematological and biochemical characteristics and to provide covid-19 classification. in [223] , 105 blood test reports were collected, of which, 27 were collected as positive saples from patients with confirmed covid-19, and for comparison, negative samples were collected from patients with ordinary pneumonia, tuberculosis, and lung cancer. each sample contains 49 feature variables, including routine hematological and biochemical parameters. next, the authors implemented the rf algorithm [22] on the training samples to perform feature learning and classification. based on the extracted 11 key feature variables, they built an rf classifier and tested 253 samples of 169 patients with suspected covid-19 with an accuracy of 96.97%. although ai technologies rarely participate directly in rt-pcr and blood testing, the viral load and covid-19 case data collected in these methods provide important data sources for the subsequent ai-based analysis. medical imaging inspection is another widely used clinical approach for covid-19 detection and diagnosis. covid-19 medical image inspection mainly includes chest x-ray and lung ct imaging. ai technology plays an important role in medical image inspection and has achieved significant results in image acquisition, organ recognition, infection region segmentation, and disease classification. it not only greatly shortens the time of a radiologist's image diagnosis but also improves the accuracy and performance of the diagnosis. we will discuss in detail the contributions of ai methods to chest x-ray and lung ct imaging. ct imaging provides an important basis for the early diagnosis of covid-19. the ct imaging manifestations of covid-19 are mainly ground glass opacity (ggo) in the periphery of the subpleural region, and some are consolidated. if the situation improves, the area will be absorbed and form fibrous stripes. examples of lung ct images of normal and covid-19 cases are shown in fig. 2 [42, 161, 184] . the progress of ct image inspection based on ai usually includes the following steps: region of interest (roi) segmentation, lung tissue feature extraction, candidate infection region detection, and covid-19 classification. the representative ai architecture for covid-19 ct image classification is shown in fig. 3 . the segmentation of lung organs and rois is a foundational step in ai-based image inspection. it depicts the rois in lung ct images (such as lungs, lung lobes, bronchopulmonary segments, and infected regions or lesions) for further evaluation and quantification. different dl models (such as u-net, v-net, and vb-net) have been used for ct image segmentation [32, 94, 114, 193, 226] . in [179] , shan et al. collected 549 ct images from patients with confirmed covid-19 and proposed an improved segmentation model (named vb-net) based on the v-net [117] and resnet [69] models. in [32] , chen et al. built a dl model based on the u-net++ structure [244] to extract the rois from each ct image and detect the training curve of suspicious lesions. in [226] , xu et al. used a 3d dl model to segment the infection regions from lung ct images. they then built a classification model using resnet and location-attention structures and divided the segmented area images into three categories, such as covid-19, influenza-a viral pneumonia, and normal. in [114] , li et al. used the u-net segmentation model to extract the lung organ from each lung ct image as an roi. in [94] , jin et al. proposed an ai-based covid-19 diagnostic system, which consists of a lung segmentation module and a covid-19 diagnostic module. the lung segmentation module is implemented based on deeplabv1 [33] . in [193] , tang et al. used the vb-net model [179] to accurately segment 18 lung regions and infected regions from lung ct images and further calculated 63 quantitative features. focusing on the detection and localization of candidate infection regions, different ai methods were proposed in [65, 85, 181, 212] . in [65] , gozes et al. used commercial software to identify lung nodules and small opacities within the 3d lung volume. then, they constructed a dl model consisting of u-net and resnet structures, where the u-net module was used to extract the roi regions, and the resenet model was used to detect and classify diffuse turbidity and ground glass infiltration. the authors compared the ct images of 56 patients with confirmed covid-19 and 101 noncoronavirus patients and analyzed the ct features of covid-19 in detail. in [181] , shi et al. used a cnn model based on v-net to segment lung organs and infection regions from lung ct images. then, they used the lasso method to calculate the best ct morphological features. finally, the severity of covid-19 was predicted and evaluated based on the best ct morphology and clinical features. in [212] , wang et al. collected 195 ct images from 44 patients with covid-19 and 258 ct images from 55 negative patients. they used the cnn model with the inception structure [191] to classify randomly selected roi images and predict covid-19 disease. in [85] , huang et al. used an inferreadtm ct pneumonia tool based on ai to quantitatively evaluate changes in the lung burden of patients with covid-19. the tool includes three modules: lung and lobe region extraction, pneumonia segmentation, and quantitative analysis. the ct image features of covid-19 pneumonia are divided into four types: mild, moderate, severe, and critical. based on roi segmentation and candidate infection region detection, the important features of rois and infection regions are extracted for covid-19 classification [156] . in [156] , qi et al. collected 71 ct images from 52 patients with confirmed covid-19 in 5 hospitals. they used the pyradiomics method to extract 1,218 features from each ct image and then performed lr and rf methods on these features to distinguish between short-term and long-term hospital stay. in [180] , shi et al. used the vb-net model [179] to segment the infection and lung fields from ct images and divided them based on 96 features, including 26 volume features, 31 digital features, 32 histogram features, and 7 surface features. next, they proposed an isarf method to classify features and predict covid-19 disease. comparative experiments showed that the isarf method is superior to the lr, svm, and nn methods. in [241] , zheng et al. proposed a 3d dcnn model (named decovnet) to detect covid-19 from ct images. the proposed decovnet model includes three components. the first component uses vanilla 3d convolutional layers to extract lung image features, the second component consists of two 3d residual blocks that perform element conversion on the 3d feature maps, and the third component gradually extracts the information in the 3d feature map through 3d max-pooling and outputs the probability of covid-19. in [187] , song et al. collected 1990 ct images, including 777 images from 88 patients with covid-19, 505 images from 100 patients with bacterial pneumonia, and 708 images from 86 healthy people. they proposed a dre-net dl model based on the pretrained resnet50 structure and functional pyramid networks. dre-net extracts the top-k lesion features from each ct image to predict the classification of patients with covid-19. the lack of large-scale datasets is the main challenge that hinders the implementation of ai-based ct image inspection and affects diagnostic performance. to address these challenges, strategies such as transfer learning, data augmentation, and "human-in-the-loop" were used in [94, 179, 239] . in [94] , jin et al. used the imagenet7 dataset [47] to pretrain the proposed 2d classification network. in [239] , zhao et al. provided a public covid-19 ct scan dataset, including 275 covid-19 cases and 195 non-covid-19 cases. they used data augmentation and tl methods to alleviate the shortage of training data. in terms of data augmentation, they used transformation operations to expand the training dataset, such as random transformation, cropping, and rotation. in terms of tl, they pretrained the densenet model [84] on the chest x-ray dataset [213] and then used the pretrained model to predict covid-19. in addition, a "human-in-the-loop" strategy was adopted to reduce the workload of radiologists in annotating the training samples [179] . radiologists annotate a small portion of training samples in the first batch of training. then, they manually correct the segmentation results in the second batch and used them as annotations of the images. iterative training is performed in this way to complete the annotation of all training samples. it is commendable that several works provided open-source code of the designed models and online covid-19 ct image inspection systems. for example, li [114] , jin [94] , zheng [241] , and zhao [239] published the proposed dl models on github [88] . in addition, song et al. [187] provided an online ct diagnosis service. wang et al. [212] provided a public website for ct image uploading and testing. in [32] , chen et al. developed a public online ct diagnostic system, and anyone can upload ct images for self-diagnosis. more detailed information about ai-based ct image segmentation and classification methods is provided in table 2 and fig. 4 . compared with ct images, chest x-ray (cxr) images are easier to obtain in clinical radiology inspections. although cxr image inspection is a typical imaging method used for covid-19 diagnosis, it is generally considered to be less sensitive than ct image inspection. some cxr images of patients with early covid-19 showed normal characteristics. the radiological signs of covid-19 cxr images include airspace opacity, ggo, and later mergers. in addition, the distribution of bilateral, peripheral, and lower regions is mostly observed. examples of cxr images of normal and covid-19 cases are shown in fig. 5 from [124, 160] . the cxr image inspection process based on ai techniques usually includes steps such as data preprocessing, dl model training, and covid-19 classification. the representative ai architecture for covid-19 cxr image inspection is shown in fig. 6 . unlike ct images, cxr image segmentation is more challenging because the ribs are projected onto soft tissues, which is confused with image contrast. in this way, most dl models focus on the classification of the entire cxr image, while few works focus on segmenting rois and lung organs from cxr images. in [67] , hassanien et al. used a classification method to identify and classify covid-19 on lung x-ray images through a multilevel threshold and svm. a multilevel image segmentation threshold was used to segment the lung organs from the background, and then the svm module classified the infected lungs from the uninfected lungs. focusing on covid-19 classification based on cxr images, several studies built ai-based [237] proposed a new dl model, which consists of a backbone network, a classification module, and an anomaly detection module. the backbone network extracts the features of each input cxr image. the classification module and anomaly detection module use the extracted features to generate classification scores and scalar anomaly scores, respectively. in [209] , wang et al. introduced a covid-net dcnn model to identify covid-19 cases based on cxr images. the covid-net model uses a large number of convolutional layers in a projection-expansion-projection design pattern. they collected 13,800 cxr images from 13,725 patients (including 183 covid-19 patients) to establish a cxr database (called covidx) for training covid-net. it is commendable that the authors provided an open-source of the proposed model code and the covidx database. similar to ct images, in cxr image inspection, there is also the problem of a lack of large-scale datasets for dl model training. in [121] , loey et al. used the gan model [86] to generate more cxr images, thereby extending the scale of the cxr dataset. in addition, three dl models (alexnet [109] , googlenet [190] , and resnet18 [69] ) were used to classify cxr images into four categories: covid-19, normal, pneumonia bacteria, and pneumonia virus. in [127] , maghdid et al. used cnn and alexnet models to train cxr and ct images to diagnose covid-19 cases, respectively. among them, the alexnet model was pretrained on the imagenet dataset to perform covid-19 classification on the datasets in [41, 124, 140] . unlike existing tl and image augmentation methods, afshar et al. designed a capsule network model (named covid-caps) suitable for small-scale cxr datasets [2] . each layer of the covid-caps model contains multiple capsules, and each capsule represents a specific image instance at a specific position through multiple neurons. the capsule module [75] uses protocol routing to capture alternative models of spatial information and attempts to reach a consensus on the existence of objects. in this way, the protocol uses information from instances and objects to identify the relationship between them without the need for large-scale datasets. more detailed information about ai-based cxr image classification methods for covid-19 inspection is shown in table 3 and fig. 7 . in addition to rt-pcr detection and image inspection techniques, some noninvasive measurement methods have also been used for covid-19 detection and diagnosis, including cough sound judgment and breathing pattern detection. (1) monitoring covid-19 through ai-based cough sound analysis. schuller et al. [174] discussed the potential application of computer audition (ca) and ai in the analysis of cough sounds in patients with covid-19. they first analyzed the ca's ability to automatically recognize and monitor speech and cough under different semantics, such as breathing, dry and wet coughing or sneezing, speech during colds, eating behaviors, drowsiness, or pain. then, they suggested applying the ca technology to the diagnosis and treatment of patients with covid-19. however, due to the lack of available datasets and annotation information, there is no report on the application of this technology in covid-19 diagnosis. similarly, iqbal et al. [90] also discussed an abstract framework that uses the speech recognition function of mobile applications to capture and analyze the cough sounds of suspicious persons to determine whether the user is healthy or suffers from a respiratory disease. in [214] , wang et al. analyzed the respiratory patterns of patients with covid-19 and other breathing patterns of patients with influenza and the common-cold. in addition, they proposed a respiratory simulation model (named bi-at-gru) for covid-19 diagnosis. the bi-at-gru model includes a gru neural network with a bidirectional and attention mechanism and can classify 6 types of clinical respiratory patterns, such as eupnea, tachypnea, bradypnea, biots, cheyne-stokes, and central-apnea. (2) covid-19 diagnosis based on noninvasive measurements. in [128] , maghdid et al. designed an abstract framework for covid-19 diagnosis based on smart phone sensors. in the proposed framework, smart phones can be used to collect the disease characteristics of potential patients. for example, the sensors can acquire the patient's voice through the recording function and can obtain the patient's body temperature through the fingerprint recognition function. then, the collected data are submitted to the ai-supported cloud server for disease diagnosis and analysis. the virology and pathogenesis of sars-cov-2 are one of the most important scientific studies in the fields of biology and medicine. scientists have analyzed the virus characteristics of sars-cov-2 through proteomics and genomic studies [7, 77, 123] . in the field of virology, the origin and classification of sars-cov-2, the physical and chemical properties, receptor interactions, cell entry, and the ecology and genomic variation in sars-cov-2 have been studied [123, 217, 242] . we mainly discuss the contribution of ai in the pathological research of sars-cov-2 from the perspective of proteomics and genomics. since the advent of sars-cov-2, there have been a large number of research achievements in proteomics. five types of structural proteins of sars-cov-2 were confirmed, including nucleocapsid (n) proteins, envelope (e) proteins, membrane (m) proteins, and spike (s) proteins [145, 206, 240] . in addition, other proteins translated in the host cells essential for virus replication have also attracted the attention of researchers, such as non-structural protein 5 (nsp5) and 3c-like protease (3clpro). moreover, several studies have shown that sars-cov-2 uses the human angiotensin-converting enzyme 2 (ace2) to enter the host [77, 242] . in this field, ai techniques are used to predict protein structures and analyze the interaction network between proteins and drugs. the representative ai architecture for protein structure predication is shown in fig. 8 . in [176, 177] , senior et al. used dl models to implement the alphafold system for protein structure prediction. the alphafold system uses a resnet model [69] to analyze the covariance and amino acid residue contacts in homologous gene sequences and to predict the corresponding protein structures. the alphafold system consists of a feature extraction module and a distance prediction neural network. the feature extraction module is responsible for searching for protein sequences that are similar to the input protein sequences and constructing the multiple sequence alignment (msa). the module simultaneously generates residual position and sequence contour features, and the output of 485 feature parameters are input into the distance prediction neural network. the distance prediction neural network is a two-dimensional (2d) resnet structure, which is responsible for accurately predicting the distance between all residue pairs of every two protein sequences. the authors added a one-dimensional output layer to the network to predict the accessible surface area, distance map, and secondary structure of each residue. finally, the generated potential is optimized by gradient descent to generate protein structures. based on [176, 177] , jumper et al. [96] used the alphafold system to predict the structure of sars-cov-2 membrane proteins. they published the predicted protein structures such as 3a, nsp2, nsp4, nsp6, and papain-like proteases. although the structure of these proteins has not been verified by clinical experiments, this publication allows researchers to quickly conduct sara-cov-2 studies. in [145] , ortega et al. used a computational method to detect changes in the s1 subunit of the spike receptor-binding domain and determined mutations in the sars-cov-2 spike protein sequence, which may be beneficial for studying human-to-human transmission. they collected sequences for modeling and constructed the sars-cov-2 spike protein model from the protein data bank (pdb) [19] and used swiss-model software [12] to construct the sars-cov-2 spike protein model. then, z-dock software [151] was used to dock between the spike protein and ace2, and a clustering algorithm was used to cluster the docking results. the work indicated that the sars-cov-2 spike protein has a higher affinity for human ace2 receptors. another branch of ai-assisted proteomics research involves finding new compounds and drug candidates for the treatment of covid-19 by building interactive networks and knowledge maps between proteins and drugs. please see section 4 for details. genomics is mainly used in sars-cov-2 to analyze the origin of sars-cov-2, vaccine development, and pt-pcr detection. various ai algorithms are applied for similarity comparisons of gene sequences, gene fragments, and mirna prediction [46, 164] . in [164] , randhawa et al. used different ml methods to analyze the pathogen sequences of covid-19 and identified the inherent features of the viral genomes, thereby rapidly classifying new pathogens. they collected the complete reference genome of the covid virus from ncbi [53] , the bat î²-coronavirus from gisaid [61] , and all available virus sequences from virus-host db [135] . each genomic sequence was mapped to a corresponding genomic signal in a discrete digital sequence by using chaotic game representation [93] . in addition, the amplitude spectrum of these genomic signals was calculated by using a discrete fourier transform. on this basis, they used 6 ml classification models to train the above sequence distance matrix and compared their performance. finally, they conducted the trained ml models on 29 covid-19 sequences to classify covid-19 pathogens. the results of this work support the hypothesis that covid-19 originated in bats and its classification as a î²-coronavirus. in [46] , demirci et al. performed a mirna prediction on the sars-cov-2 genome based on 3 ml methods and identified mirna-like hairpins and microrna-mediated sars-cov-2 infection interactions. they collected the complete covid-19 genome from ncbi [53] and human-mature mirna sequences from mirbase [108] . the genomic sequences are transcribed and divided into multiple overlapping fragments, which are folded into a secondary structure to extract the hairpin structure. on this basis, the authors used 3 ml methods (e.g., dt, naive bayes, and rf) to predict the category of each hairpin and determined the similarity between the hairpins and human mirna. they searched for mature mirna targets in human and sars-cov-2 genes and analyzed the potential interactions between sars-cov-2 mirnas and human genes and between human mirnas and sars-cov-2 genes. finally, the gene ontology of sars-cov-2 mirna targets in human genes were analyzed, and the similarity between sars-cov-2 mirna candidates and mature mirnas of any known organism was evaluated using the panther classification system [134] . in [133] , metsky et al. used genomic and ai technologies to rapidly design nucleic acid detection assays and improved current rt-pcr testing of sars-cov-2. they developed a crispr tool that uses enzymes to edit the genome by cutting specific genetic code chains and used different ml methods to predict the diversity of the target genome. the authors designed the rt-pcr test method through the crispr tool, and it can effectively detect 67 respiratory viruses, including sars-cov-2. in the field of drug development, ai technologies can screen existing drug candidates for covid-19 by analyzing the interaction between existing drugs and covid-19 protein targets. in addition, ai technologies can help to discover new drug-like compounds against covid-19 by constructing new molecular structures that have inhibitory effects on proteases at the molecular level. the representative ai architecture for new drug-like compound discover is shown in fig. 9 . drug development can be divided into small-molecule drug discovery and biological product development. small-molecule drug discovery mainly focuses on chemically synthesized small molecules of active substances, which can be made into small-molecule drugs through chemical reactions between different organic and inorganic compounds. one group of ai-based drug development focuses on the discovery of new drug-like compounds at the molecular level. in [18, 182] , beck et al. proposed a figure 9 . representative ai architecture for new drug-like compound discover. dl-based drug-target interaction model (mt-dti) to predict potential drug candidates for covid-19. the mt-dti model uses smiles strings and amino-acid sequences to predict target proteins with 3d crystal structures. the authors collected the amino-acid sequences of 3c-like proteases and related antiviral drugs and drug targets from the databases of ncbi [53] , drug target common (dtc) [194] , and bindingdb [120] . in addition, they used a molecular docking and virtual screening tool (autodock vina [202] ) to predict the binding affinity between 3,410 drugs and sars-cov-2 3clpro. the experimental results provided 6 potential drugs, such as remdesivir, atazanavir, efavirenz, ritonavir, dolutegravir, kaletra (lopinavir/ritonavir). note that remdesivir shows promising in clinical trial. in [136] , moskal et al. used ai methods to analyze the molecular similarity between anti-covid-19 drugs (termed "parents") and drugs involving similar indications to screen out second-generation drugs (termed "progeny") for covid-19. they first used the mol2vec [91] method to convert the molecular structure of the parent drugs into a high-dimensional vector space, treated the drug molecule as a "sentence", and mapped its molecular substructure to a "word". then, they used the vae [186] model to generate smiles strings with similar 3d shape and pharmacodynamic properties to a given seed molecule [63] . in addition, cnn, lstm, and mlp models are used to generate the corresponding smiles strings and molecules. the authors selected 71 parent drugs as seed molecules from the literature and selected 4456 drugs as candidate progeny drugs from zinc [245] and chembl [56] . in [23] , bung et al. committed to the development of new chemical entities for the sars-cov-2 3clpro based on dl technology. they constructed an rl-based rnn model to classify protease inhibitor molecules and obtained a smaller subset that favored the chemical space. then, they collected 2515 protease inhibitor molecules in smiles format from the chembl database as training data, where each smiles string is regarded as a time series, and each position or symbol is regarded as a time point. the output of small molecules was docked to the 3clpro structure with minimal energy and ranked based on the virtual screening score obtained by selecting candidates of anti-sars-cov-2 [202] . in [192] , tang et al. analyzed 3clpro with a 3d structure similar to sars-cov and evaluated it as an attractive target for anti-covid-19 drug development. they proposed an advanced deep-q learning network (called adqn-fbdd) to generate potential lead compounds of sars-cov-2 3clpro. they collected 284 reported molecules as sars-cov-2 3clpro inhibitors. these molecules were split using the improved brics algorithm [45] to obtain the target fragment library of sars-cov-2 3clpro. then, the proposed adqn-fbdd model trains each target fragment and predicts the corresponding molecules and lead compounds. through the proposed structure-based optimization policy (sbop), they finally obtained 47 derivatives with inhibitory effects on sars-cov-2 3clpro from these lead compounds, which are regarded as potential anti-sars-cov-2 drugs. another group of studies focused on screening candidate biological products for covid-19. biological products are a type of protein products with therapeutic effects, which are mainly combined with specific cell receptors involved in the disease process. biological products are prepared from microbial cells such as genetically modified bacteria, yeast, or mammalian cell strains through biotechnology processes. in [79] , hu et al. established a multitask dl model to predict the possible binding between potential drugs and sars-cov-2 protein targets, thereby selecting available drugs for sars-cov-2. they first collected 8 sars-cov-2 viral proteins from ghddi [58] as potential targets. the proposed dl model is based on the atomnet model [188, 205] and includes a shared layer to learn the joint representation of all tasks and a task processing layer for performing specific tasks. by fine-tuning the dl model using a coronavirus-specific dataset, the model can predict the possible binding between the drugs and the protein targets and output the binding affinity score. based on existing studies, rdrp, 3clpro, and papain-like protease have been confirmed as the three principal targets of sars-cov-2 [60, 146, 206] . based on the prediction results [113, 210] , the authors selected the top 10 potential drugs with a high likelihood of inhibition for each target. in [98] , kadioglu et al. used high-performance computing (hpc), virtual drug screening, molecular docking, and ml technologies to identify sars-cov-2 drug candidates. after performing virtual drug screening and molecular docking, two supervised ml models(e.g., nn and naivebayes) were used to analyze clinical drugs and test compounds to construct corresponding drug likelihood prediction models. several approved drugs, including those used for the hepatitis c virus (hcv), the enveloped ssrna virus, and other infectious diseases, were selected as sars-cov-2 drug candidates. facing the known covid-19 protease target 3clpro, zhavoronkov et al. [240] designed a small-molecule drug-discovery pipeline to produce 3clpro inhibitors, used 3clpro's crystal structure, homology modeling, and co-crystallized fragments to generate 3clpro molecules. they collected the crystal structure of covid-19 3clpro from [230] and constructed a homology model. at the same time, molecules with activity on various proteases were extracted from [56, 89] and constituted a protease peptidomimetic dataset with 5,891 compounds. then, they used 28 ml methods (such as gae, gan, and ga) and rl strategies to separately train input datasets (e.g., crystal structure, homology model, and co-crystal ligands), and generated new molecular structures with a high score. in [78] , hofmarcher et al. used a chemai dl model [154] based on the smileslstm structure [76] to test the resistance of the molecules to covid-19 proteases. they collected 3.6 million molecules from chembl [56] , zinc [245], and pubchem [104] and formed a training dataset. then, the chemai model was trained on the dataset in a multitask parallel training way, where the output neurons of the model represent the biological effects of the input molecules. the authors used the chemai model to predict the inhibitory effects of these molecules on the 3clpro and plpro proteases of covid-19. these molecules have a binding, inhibitory, and toxic effect on the targets. a list of covid-19 drug development methods based on ai is provided in table 4 . currently, there are 3 types of covid-19 vaccine candidates, such as (1) whole virus vaccines, (2) recombinant protein subunit vaccines, and (3) nucleic acid vaccines [38, 238] . ai technology has been involved in the design and development of covid-19 vaccines. compared with explicit applications in other fields, ai technology is usually used in the sub-processes of vaccine development in an implicit manner. the ai algorithms of netmhc and netmhcpan are used in the development of covid-19 vaccines for epitope prediction [74, 97, 215] . in [74] , herst et al. obtained the sars-cov-2 protein sequences from genbank and used the msa algorithm to trim the nucleocapsid phosphoprotein sequences to possible peptide sequences. on this basis, they used netmhc and netmhcpan ai algorithms to train and predict peptide sequences [8, 97] . the pan variant of netmhc integrates the in-vitro objects of 215 hlas for prediction. finally, they used the average value of the ann, svm, netmhc and netmhcpan methods to calculate the vaccine candidates. in [215] , ward et al downloaded the sars-cov-2 nucleotide sequences from the ncbi [53] and gisaid [61] databases, and generated a consensus sequence for each sars-cov-2 protein. the sequences can be used as references for prediction, specificity, and epitope mapping analysis. next, the authors used different epitope prediction tools to predict b cell epitopes and map them to the amino acid sequences of each gene. on this basis, they used the ai-based netmhcpan algorithm to predict hla-1 peptides and obtained a total of 2,915 alleles in all peptide lengths. blastp tool [6] was used to locate the short amino acid epitope sequences to the canonical sequences of sars-cov-2 proteins. finally, the author provided an online tool that provides functions of sars-cov-2 genetic variation analysis, epitope prediction, coronavirus homology analysis, and candidate proteome analysis. in [143] , ong et al. used ml and reverse vaccinology (rv) methods to predict and evaluate potential vaccines for covid-19. they used rv to analyze the bioinformatics of pathogen genomes to identify promising vaccine candidates. they obtained the sars-cov-2 sequences and all proteins of the 6 known human coronavirus strains from the ncbi [53] and uniprot [17] databases. then, they used vaxign and vaxign-ml [71, 142] to analyze the complete proteome of the coronaviruses and predicted their service biological characteristics. next, they improved the vaxign-ml model based on ml and rv using lr, svm, knn, rf, and xgboost methods and predicted the protein level of all sars-cov-2 proteins. the nsp3 protein was selected for phylogenetic analysis, and the immunogenicity of nsp3 was evaluated by predicting t cell mhc-i and mhc-ii and linear b cell epitopes. in [158] , qiao et al. used dl to predict the patient's mutated new antigen and identified the best t-cell epitope for peptide-based covid-19 vaccines. they first sequenced the diseased cells in the patient's blood and extracted 6 human leukocyte antigen (hla) types and t-cell receptor (tcr) sequences. then, they proposed the deepnovo model to train the patient's immune peptide and to identify the best t-cell epitope set based on a person's hla alleles and immune peptide group information. the deepnovo model uses lstm and rnn structures to capture sequence patterns in peptides or proteins and predicts hla peptides from conserved regions of the virus, thereby predicting new mutant antigens in patients. in addition, they used the iedb [204] tool to predict the immunogenicity of 177 peptides. they suggested designing an epitope-based covid-19 vaccine specifically for each person based on their hla alleles. the prediction of immune stimulation ability is an important part of vaccine designing [162, 166] . different ml methods and position-specific scoring matrices (pssm) are usually used to predict epitope and immune interactions, thereby predicting the generation of adaptive immunity in the target host. in [162] , rahman et al. used immuno-informatics and comparative genomic methods to design a multi-epitope peptide vaccine against sars-cov-2, which combines the epitopes of s, m, and e proteins. they used the ellipro antibody epitope prediction tool [87] to predict linear b cell epitopes on the s protein. ellipro uses multiple ml methods to predict and visualize a given protein sequence or b-cell epitope in the structure. in addition, sarkar et al. [172] studied the epitope-based vaccine design for 15/36 covid-19 and used the svm method to predict the toxicity of the selected epitopes. in [153] , prachar et al. used 19 epitope-hla combined prediction tools including iedb, ann, and pssm algorithms to predict and verify 174 sars-cov-2 epitopes. thanks to the developed information and multimedia technology, the outbreak and spread of covid-19 were reported in a timely and accurate manner. the number of suspected, confirmed, cured, and dead covid-19 cases in each country/region is announced in real time. in addition, passenger travel trajectories and related big data are shared for scientific research. based on the rich data, numerous researchers have participated in the prediction, spread, and tracking of the covid-19 outbreak. researchers collected clinical covid-19 case data and used different ai methods to extract important features and to predict the mortality and survival rate of patients with covid-19. the representative ai architecture for prediction of patient mortality and survival rate is shown in fig. 10 . figure 10 . representative ai architecture for prediction of patient mortality and survival rate. in [152] , pourhomayoun et al. used 6 ai methods to predict the mortality rate of patients with covid-19. they used public data of patients with covid-19 from 76 countries around the world [225] , and counted 112 features, including 80 medical annotations and disease features and 32 features from the patients' demographic and physiological data. based on the filtering method and wrapper method, 42 best features were extracted, such as demographic features, general medical information, and patient symptoms. on this basis, 6 ai methods (such as svm, nn, rf, dt, lr, and knn) are used to predict the mortality of patients with covid-19. in [173] , sarkar et al. used the rf model to analyze the records of 433 patients with covid-19 from kaggle [43] and identified the important features and their impact on mortality. experimental results show that patients over 62 years of age have a higher risk of death. in [228, 229] , yan et al. analyzed a blood sample dataset of 404 patients with covid-19 in wuhan, china, and used the xgboost classification method [37] to select three important biomarkers and to predict individual patient survival rates. experimental results with an accuracy of 90% indicated that higher ldh levels seem to play an important role in distinguishing the most critical covid-19 cases. bluedot [21] and metabiota [132] are two ai companies that made accurate predictions for the covid-19 outbreak. bluedot collected large-scale heterogeneous data from various sources, such as news reports, global ticketing data, animal diseases, global infectious disease alerts, and real-time climate conditions. then, it used filtering tools to narrow its focus; used various ml and natural language processing (nlp) techniques to detect, mark, and display the potential risk frequency of covid-19; and predicted the outbreak time of transmission. it is worth mentioning that 9 days before the official 16/36 announcement of the covid-19 outbreak, bluedot accurately predicted the epidemic of covid-19 and cities with a high risk of virus outbreaks. metabiota collected large-scale data from social and nonsocial sources (such as biology, socioeconomic, political, and environmental data) and used technologies such as ai, ml, big data, and nlp to accurately predict the outbreak, spread, and intervention measures of covid-19. more ai-based covid-19 outbreak and transmission prediction methods are shown in table 5 . table 5 . covid-19 outbreak and transmission prediction based on ai methods. data sources methods country/region huang [82] yang [231] , who [216] cnn, lstm, mlp, gru china hu [80, 81] the paper [148] , who [216] mae, clustering china yang [233] baidu [16] seir, lstm china fong [51, 52] nhc [139] svm, pnn china ai [3] who [54, 216] anfis, fpa china, usa rizk [168] who [216] isacl-mfnn usa, italy, spain giuliani [62] italy [144] emtmgl italy ayyoubzadeh [14] worldometer [218] , google [201] lr, lstm iran marini [129, 130] swiss population enerpol switzerland lai [110] iata [126] , worldpop [219] ml global punn [155] jhu csse [49] svr, pr, dnn, lstm, rnn lampos [111] mediacloud [131] , phe [64] , ecdc [55] transfer learning global although the source of the covid-19 epidemic has not yet been identified, it was first reported in wuhan, china. therefore, the outbreak and spread of covid-19 in china have received extensive attention. in [82] , huang et al. used 4 dl models, such as cnn, lstm, gru, and mlp to train and predict the covid-19 case data from 7 severe epidemic cities in china. the input of these dl models is the features of the covid-19 cases, including the number of confirmed cases, cured cases, and deaths. based on the input of the previous 5 days, each model can predict the number of covid-19 cases for the following few days. the architecture of the covid-19 outbreak prediction model based on ai models is shown in fig. 11 . figure 11 . architecture of covid-19 outbreak prediction model based on dl models. in [80, 81] , hu et al. used ai methods such as mae and clustering algorithms to predict the number of confirmed covid-19 cases in different provinces and cities in china. in addition, they clustered 34 provinces and cities in china into 9 clusters based on the prediction results and further predicted the spread of covid-19 among provinces and cities. in [233] , yang et al. used the seir model [101] and the lstm model to predict covid-19 in china. the population migration data and the latest covid-19 epidemiological data from baidu [16] were input into the seir model to derive the epidemic curves. in addition, they used sars data from 2003 to pretrain the lstm model to predict covid-19 for the following few days, in which epidemiological parameters, such as the transmission, incubation, recovery probability, and the number of deaths, were selected as input features. both the seir and lstm models predicted a daily infection peak of 4,000 in the first week of february. in [51, 52] , fong et al. obtained early covid-19 epidemiological data from nhc [139] . then, they used traditional time series data analysis methods (e.g., arima, exponential, and holt-winters), ml methods (e.g., kr, svm, and dt), and ai methods (e.g., pnn) to analyze and predict future outbreaks. in addition to china, the outbreak and spread of covid-19 in other countries (including the united states, italy, spain, iran, and switzerland) have also received widespread attention. in [3] , ai et al. proposed an improved anfis method [92] to predict the number of covid-19 cases. the proposed system connects fuzzy logic and neural networks and uses and enhanced flower pollination algorithm (fpa) [232] for model parameter optimization and model training. in [168] , rizk et al. proposed an improved multi-layer feed-forward neural network (isacl-mfnn) model, which uses an internal search algorithm (isa) to optimize model parameters and uses the cl strategy to enhance the isa performance. from the official covid-19 dataset reported by the who [216] , data from january 22, 2020, to april 3, 2020, in the united states, italy, and spain were collected to train the isacl-mfnn model and to predict the confirmed cases within the next 10 days. in [62] , giuliani et al. collected the number of infected people in italian provinces [144] and used the emtmgl model to simulate and predict the spatial and temporal distribution of covid-19 infection in italy. in [14] , ayyoubzadeh used real-time covid-19 epidemic data from google trends [201] and worldometer [218] to predict covid-19 cases in iran. they collected daily epidemic data and saved them as a time series data format and then used the lr and lstm models to make predictions, thereby obtaining the outbreak and spread trend of covid-19 in iran. in [129, 130] , marini et al. developed an agent-based ai platform to predict the development of covid-19 in switzerland. the system accepts the entire swiss population as input data to simulate and predict the spread of covid-19 in switzerland. it simulates the people's daily trajectories by calibrating the micro-census data and effectively predicts the individual contacts and possible transmission routes. many studies have likewise focused on the prediction of the spread of covid-19 around the world. they collected a large amount of travel data, mobile phone data, and social media data and used ai methods to accurately predict the potential transmission range and transmission route of covid-19. in [110] , lai et al. collected a large amount of travel and mobile phone data from [219] and constructed corresponding models to predict the transmission risk of covid-19 in different countries. on this basis, they established air travel network models between domestic cities and cities in other countries to predict risk cities at home and abroad. in [155] , punn et al. used 2 ml models (e.g., svr [13] and pr [44] ) and 3 dl regression models (e.g., dnn, lstm [66] , and rnn) to predict real-time covid-19 cases. in [111] , lampos et al. used an automatic crawling tool to obtain daily confirmed covid-19 case data and related articles from online media such as mediacloud [131] , public health england (phe) [64] , and european centre for disease prevention and control (ecdc) [55] . they used the tl strategy to transfer the covid-19 model of the country where the disease spread to other countries that are still in the early stage of the epidemic curve, and thus achieving the target country's epidemic prediction. in addition, companies such as microsoft bing [20] , google [201] , and baidu [16] have aggregated multiple available data sources and developed covid-19 global tracking systems to provide a visual tracking interface. in addition to ai methods, various methods based on statistics and epidemiology are used to predict the outbreak and spread of covid-19. in [70] , he et al. collected the highest viral load in the pharyngeal swabs of 94 patients with confirmed covid-19. they fitted a generalized additive model with identity links and smooth spline curves to analyze its overall trend. a gamma distribution was fitted to the transmission pair data to evaluate the serial interval distribution. the results of statistical analysis showed that the patients with confirmed covid-19 reach the peak of virus shedding before or during symptom onset, and some kinds of transmission may occur before the initial symptoms. in [208] , wang et al. determined a set of technical indicators (e.g., number of infection cases in the hospital, daily infection rate, and daily cured rate) that reflect the infection status of covid-19. next, they proposed a calculation method based on statistical theory to quantify the iconic characteristics of each period and predict the turning point in the development of the epidemic. in addition, numerous studies based on the susceptible-infected-recovered (sir) and seir models have studied the spread of covid-19 from an epidemiological perspective. please see [25, 119, 169, 185, 195, 199, 234] for more information. when covid-19 appeared, most countries in the world adopted different forms of social control, social alienation, school closures, and blockade measures to prevent the spread of the epidemic [203] . ai technologies have been widely used in epidemic control and social management, including individual temperature detection, video tracking, contact tracking, intelligent robots, etc. many countries have used smart devices equipped with ai to detect suspicious persons in public transportation places such as airports and train stations [40, 167] . for example, infrared cameras are used to scan for high temperatures in a crowd, and different ai methods perform efficient analysis to detect whether an individual is wearing a mask in real time. in addition, dl-based video tracking technology is used to detect and track suspicious covid-19 patients in public places [31] . moreover, at the entrances and exits of cities, the identity information of each passing person was collected. then, ai-based systems are used to efficiently query the travel history and trajectory of each passing individual to check whether they are from a region seriously affected by covid-19 [35, 197] . ai technologies are also used in contact tracking of patients with covid-19 [100] . for each patient with confirmed covid-19, personal data such as mobile phone positioning data, consumption records, and travel records may be integrated to identify the potential transmission trajectory [57] . in addition, when people are in social isolation, mobile phone positioning and ai frameworks can assist the government in better understanding the status of individuals [165] . moreover, intelligent robots are used to perform site disinfection and product transfer, and mobile phone positioning functions are used to detect and track the distribution and flow of personnel. another group of studies focused on the impact of various social control strategies on the spread of covid-19. in the implementation and performance improvement of ai greatly depends on the large-scale available data and resources. therefore, we compiled available public resources that can be used for covid-19 disease diagnosis, virology research, drug and vaccine development, and epidemic and transmission prediction. three types of data and resources were summarized, including medical images, biological data, and informatics resources. 19/36 we collected 17 groups of covid-19 medical images such as cxr and ct images from individual researchers and organizations. among them, the cxr image data set published by cohen et al. [41] is widely cited, which is a collection of cxr images from multiple references. in addition, many researchers uploaded cxr and ct images to kaggle [15, 112, 124, 138, 160] for covid-19 research. moreover, organizations such as the british society of thoracic imaging (bsti), eurorad, and radiopaedia also released online cxr and ct images. table 6 displays the detailed description of medical image data resources of covid-19. table 6 . medical image data resources for covid-19 research. data type cited by refs. zhao [239] ct images [239] hrct [48] ct images [94] armato [11] ct images [94] coronacases [42] ct images -medical segmentation [175] ct images -cohen [41] cxr images [2, 9, 73, 121, 127, 137, 178, 209 , 237] wang [213] cxr images [237, 239] covidx [24] cxr images [209] adrian [159] cxr images [73] covid-net [209] cxr images [209] kermany [102] cxr images [9] mendeley data [5] cxr images -kaggle [15, 112, 124, 138, 160] cxr and ct images [2, 9, 121, 127, 137, 173, 178, 209 ] bsti [140] cxr and ct images [127] sirm [184] cxr and ct images -eurorad [50] cxr and ct images -radiopaedia [161] cxr and ct images we collected 10 biological data resources, such as ncbi, protein data bank (pdb), uniprot, clarivate analytics integrity (cai), drug target common (dtc), and virus-host db (vhdb), as shown in table 7 . these data resources provide abundant biological data resources, including gene sequences, proteins, drug molecules and compounds, and mirna sequences. informatics resources such as covid-19 situation reports, dashboards, covid-19 cases, and demographic data are gathered in table 8 [135] genome sequences virus sequences [164] pdb [149] proteins 3d shapes of proteins, nucleic acids, and assemblies [19, 145] uniprot [17] proteins sars-cov-2 protein entries and receptors [143] mirbase [108] mirna sequences human mature mirna sequences [46] zinc [245] drug compounds drug compounds and molecules [78, 98, 136 ] dtc [194] drug molecules drug molecules for 3c-like proteases [18, 182] cai [89] drug discovery empowering knowledge-based drug discovery and development [240] bindingdb [120] amino-acid sequences amino-acid sequences of 3c-like proteases [18, 182] [219] demographic data spatial demographic and air travel data [110] ghddi [58] community drug discovery community [79] humdata [68] community community perceptions of covid19 we summarize the main challenges currently faced by ai against covid-19 and provide the corresponding suggestions. at present, the applications of ai in covid-19 research mainly faces four challenges: the lack of available large-scale training data, massive noisy data and rumors, the limited knowledge on the intersection of computer science and medicine, and data privacy and human rights protection. â�¢ lack of available large-scale training data. most ai methods rely on large-scale annotated training data, including medical images and various biological data. however, due to the rapid outbreak of covid-19, there are insufficient datasets available for ai. in addition, annotating training samples is very time-consuming and requires professional medical personnel. â�¢ massive noisy data and rumors. challenges arise from relying on the developed mobile internet and social media; massive noise information and fake news about covid-19 has been published on various online media without rigorous review. however, ai algorithms seem to be powerless in 21/36 judging and filtering the noise and erroneous data. this problem limits the application and performance of ai, especially in epidemic prediction and transmission analysis. â�¢ limited knowledge in the intersection of computer science and medicine. many ai scientists are from computer science, but the application of ai in the covid-19 battle requires in-depth cooperation in computer science, medical imaging, bioinformatics, virology, and many other disciplines. therefore, it is crucial to coordinate the cooperative work of researchers from different fields and integrate the knowledge of multiple subjects to jointly deal with covid-19. â�¢ data privacy and human rights protection. in the era of big data and ai, the cost of obtaining personal privacy data is very low. faced with public health issues such as covid-19, many governments want to obtain various types of personal information, including mobile phone positioning data, personal travel trajectory data, and patient disease data. how to effectively protect personal privacy and human rights during information acquisition and ai-based processing is an issue worthy of discussion and attention. in addition to the applications investigated in this paper, ai can also contribute to the battle of covid-19 from the following 10 potential directions. 1. noncontact disease detection. in cxr and ct image detection, the use of noncontact automatic image acquisition can effectively avoid the risk of infection between radiologists and patients during the covid-19 pandemic. ai can be used for patient posture positioning, standard section acquisition of cxr and ct images, and movement of camera equipment. 2. remote video diagnosis. ai and nlp technologies can be used to develop remote video diagnosis systems and chat robot systems and provide covid-19 disease consultation and preliminary diagnosis to the public. 3. patient prognosis management. ai technology (such as intelligent image and video analysis) can be used to automatically monitor patient behavior during the follow-up monitoring and prognostic management process, in addition to long-term tracking and management of patients with covid-19. 4. biological research. in the field of biological research, ai can be used to discover protein structures and features of virus through accurate analysis of biomedical information, such as large-scale protein structures, gene sequences, and viral trajectories. 5. drug and vaccine development. ai can not only be used to discover potential drugs and vaccines but also to simulate the interaction between drugs and proteins and between vaccines and receptors, thereby predicting the potential responses to the drugs and vaccines of patients with covid-19 with different constitutions. 6. identification and filtering of fake news. ai can be used to reduce and eliminate fake news and noise data on online social media platforms to provide reliable, correct, and scientific information about the covid-19 pandemic. 7. impact simulation and evaluation. various simulation models can use ai to analyze the impact of different social control strategies on disease transmission. then, they can be used to explore more effective and scientific approaches of disease prevention and social control. 8. patient contact tracking. by constructing social relationship networks and knowledge graphs, ai can identify and track the trajectories of people in close contact with patients with covid-19, thereby accurately predicting and controlling the potential spread of the disease. 22/36 9. intelligent robots. intelligent robots are expected to be used in applications such as disinfection and cleaning in public places, product distribution, and patient care. 10. intelligent internet of things. ai is expected to be combined with the internet of things to deploy in customs, airports, railway stations, bus stations, and business centers. in this case, we can quickly identify suspicious covid-19 virus and patients through intelligent monitoring of the environment and personnel. in this survey, we investigated the main scope and contributions of ai in combating covid-19. compared with the pandemic of sars-cov in 2003 and mers-cov in 2012, ai technologies have been successfully applied in almost every corner of the covid-19 battle. the application of ai in covid-19 research can be summarized in four aspects, such as disease detection and diagnosis, virology research, drug and vaccine development, and epidemic and transmission prediction. among them, medical image analysis, drug discovery, and epidemic prediction are the main battlefields of ai in the fight against covid-19. we also summarized the currently available data and resources for covid-19 research based on ai, including medical imaging data, biological data, and informatics resources. finally, we highlighted the main challenges and potential directions in this field. this survey provided medical and ai researchers with a comprehensive view of the existing and potential contributions of ai in combating covid-19, with the goal of inspiring them to continue to maximize the advantages of ai and big data to fight against this pandemic. the covid-19 pandemic calls for spatial distancing and social closeness: not for social distancing covid-caps: a capsule network-based framework for identification of covid-19 cases from x-ray images optimization method for forecasting confirmed cases of covid-19 in china artificial intelligence (ai) provided early detection of the coronavirus (covid-19) in china and will influence future urban health policy internationally augmented covid-19 x-ray images dataset. mendeley data basic local alignment search tool the proximal origin of sars-cov-2 gapped sequence alignment using artificial neural networks: application to the mhc class i system covid-19: automatic detection from x-ray images utilizing transfer learning with convolutional neural networks covid-19: a novel coronavirus and a novel challenge for critical care the lung image database consortium (lidc) and image database resource initiative (idri): a completed reference database of lung nodules on ct scans the swiss-model workspace: a web-based environment for protein structure homology modelling support vector regression predicting covid-19 incidence using google trends and data mining techniques: a pilot study in iran covid-19 x-rays. website real-time covid-19 data. website the universal protein resource (uniprot) predicting commercially available antiviral drugs that may act on the novel coronavirus (sars-cov-2) through a drug-target interaction deep learning model announcing the worldwide protein data bank covid-19 tracker). website an ai epidemiologist sent the first warnings of the wuhan virus. website random forests de novo design of new chemical entities (nces) for sars-cov-2 using artificial intelligence covid-19 chest x-ray dataset initiative. website chinese and italian covid-19 outbreaks can be correctly described by a modified sird model. medrxiv artificial intelligence applied on chest x-ray can aid in the diagnosis of covid-19 infection: a first experience from lombardy, italy. medrxiv a familial cluster of pneumonia associated with the 2019 novel coronavirus indicating person-to-person transmission: a study of a family cluster modelling transmission and control of the covid-19 pandemic in australia a comparative study of fine-tuning deep learning models for plant disease identification clinical characteristics and intrauterine vertical transmission potential of covid-19 infection in nine pregnant women: a retrospective review of medical records distributed deep learning model for intelligent video surveillance systems with edge computing deep learning-based model for detecting 2019 novel coronavirus pneumonia on high-resolution computed tomography: a prospective study. medrxiv deeplab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs keep up with the latest coronavirus research covid-19 control in china during mass population movements at new year. the lancet fangcang shelter hospitals: a novel concept for responding to public health emergencies. the lancet xgboost: a scalable tree boosting system the sars-cov-2 vaccine pipeline: an overview xception: deep learning with depthwise separable convolutions in a time of coronavirus, china's investment in ai is paying off in a big way covid-19 image data collection ct images of confirmed covid-19 cases. mendeley data novel corona virus 2019 dataset. website approximate optimal designs for multivariate polynomial regression on the art of compiling and using'drug-like'chemical fragment spaces computational analysis of microrna-mediated interactions in sars-cov-2 infection. biorxiv imagenet: a large-scale hierarchical image database building a reference multimedia database for interstitial lung diseases an interactive web-based dashboard to track covid-19 in real time images of covid-19 cases. mendeley data composite monte carlo decision making under high uncertainty of novel coronavirus epidemic using hybridized deep learning and fuzzy rule induction finding an accurate early forecasting model from small dataset: a case of 2019-ncov novel coronavirus outbreak genome sequencing data of sars-cov-2. website disease control and prevention. weekly influenza confirmed cases. website disease prevention and control. geographic distribution covid-19 cases worldwide. website chembl: a large-scale bioactivity database for drug discovery the role of close contacts tracking management in covid-19 prevention: a cluster investigation in jiaxing, china targeting covid-19: ghddi info sharing portal. website first known person-to-person transmission of severe acute respiratory syndrome coronavirus 2 (sars-cov-2) in the usa. the lancet structure of the respiratory syncytial virus polymerase complex gisaid: global initiative on sharing all influenza data. website modelling and predicting the spatio-temporal spread of coronavirus disease 2019 (covid-19) in italy automatic chemical design using a data-driven continuous representation of molecules coronavirus (covid-19) cases in the uk. website rapid ai development cycle for the coronavirus (covid-19) pandemic: initial results for automated detection and patient monitoring using deep learning ct image analysis lstm: a search space odyssey automatic x-ray covid-19 lung image classification system based on multi-level thresholding and support vector machine. medrxiv covid-19 pandemic. website deep residual learning for image recognition temporal dynamics in viral shedding and transmissibility of covid-19 vaxign: the first web-based vaccine design program for reverse vaccinology and applications for vaccine development feasibility of controlling covid-19 outbreaks by isolation of cases and contacts. the lancet global health covidx-net: a framework of deep learning classifiers to diagnose covid-19 in x-ray images an effective ctl peptide vaccine for ebola zaire based on survivors' cd8+ targeting of a particular nucleocapsid protein epitope with potential implications for covid-19 vaccine design matrix capsules with em routing long short-term memory sars-cov-2 cell entry depends on ace2 and tmprss2 and is blocked by a clinically proven protease inhibitor large-scale ligand-based virtual screening for sars-cov-2 inhibitors using deep neural networks prediction of potential commercially inhibitors against sars-cov-2 by multi-task deep model artificial intelligence forecasting of covid-19 in china evaluating the effect of public health intervention on the global-wide spread trajectory of covid-19. medrxiv multiple-input deep convolutional neural network model for covid-19 forecasting inchina. medrxiv clinical features of patients infected with 2019 novel coronavirus in wuhan, china. the lancet densely connected convolutional networks serialquantitative chest ct assessment of covid-19: deep-learning approach generative adversarial nets ellipro: an antibody epitope prediction tool. website clarivate analytics integrity. website active surveillance for covid-19 through artificial intelligence using concept of real-time speech-recognition mobile application to analyse cough sound mol2vec: unsupervised machine learning approach with chemical intuition anfis: adaptive-network-based fuzzy inference system chaos game representation of gene structure development and evaluation of an ai system for covid-19 diagnosis. medrxiv virology, epidemiology, pathogenesis, and control of covid-19 computational predictions of protein structures associated with covid-19. website netmhcpan-4.0: improved peptide-mhc class i interaction predictions integrating eluted ligand and peptide binding affinity data identification of novel compounds against three targets of sars cov-2 coronavirus by combined virtual screening and supervised machine learning health security capacities in the context of covid-19 outbreak: an analysis of international health regulations annual report data from 182 countries. the lancet the efficacy of contact tracing for the containment of the 2019 novel coronavirus(covid-19). medrxiv modeling infectious diseases in humans and animals identifying medical diagnoses and treatable diseases by image-based deep learning a simple and multiplex loop-mediated isothermal amplification assay for rapid detectionof sars-cov pubchem substance and compound databases social distancing strategies for curbing the covid-19 epidemic. medrxiv sars-cov-2 detection in patients with influenza-like illness interventions to mitigate early spread of sars-cov-2 in singapore: a modelling study mirbase: from microrna sequences to function imagenet classification with deep convolutional neural networks assessing spread risk of wuhan novel coronavirus within and beyond china tracking covid-19 using online search covid-19 x-rays. website therapeutic options for the 2019 novel coronavirus artificial intelligence distinguishes covid-19 from community acquired pneumonia on chest ct characterizing the propagation of situational information in social media during covid-19 epidemic: a case study on weibo preliminary assessment of the covid-19 outbreak using 3-staged model e-ishr volumetric medical image segmentation: a 3d deep coarse-to-fine framework and its adversarial examples hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting sars-cov-2 infection in vitro covid-19progression timeline and effectiveness of response-to-spread interventions across the united states. medrxiv bindingdb: a web-accessible database of experimentally determined protein-ligand binding affinities within the lack of covid-19 benchmark dataset: a novel gan with deep transfer learning for corona-virus detection in chest x-ray images development of anovel reverse transcription loop-mediated isothermal amplification method for rapid detection ofsars-cov-2 genomic characterisation and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding chest x-ray images (pneumonia) visualizing data using t-sne international air travel association(iata). website diagnosing covid-19 pneumonia from x-ray and ct images using deep learning and transfer learning algorithms a novel ai-enabled framework to diagnose coronavirus covid 19 using smartphone embedded sensors: design study enhancing response preparedness to influenza epidemics: agent-based study of 2050 influenza season in switzerland. simulation modelling practice and theory covid-19 epidemic in switzerland: growth prediction and containment strategy using artificial intelligence and big data. medrxiv how ai is battling the coronavirus outbreak. website crispr-based surveillance for covid-19 using kosoko-thoroddsen, tinna-solveig f-comprehensive machine learning design. biorxiv panther in 2013: modeling the evolution of gene function, and other gene attributes, in the context of phylogenetic trees linking virus genomes with host taxonomy suggestions for second-pass anti-covid-19 drugs based on the artificial intelligence measures of molecular similarity, shape and pharmacophore distribution automatic detection of coronavirus disease (covid-19) using x-ray images and deep convolutional neural networks rsna pneumonia detection challenge. website of the people's republic of china. real-time covid-19 report. website covid-19 bsti imaging database. website polynomial neural networks architecture: analysis and design vaxign-ml: supervised machine learning reverse vaccinology model for improved prediction of bacterial protective antigens covid-19 coronavirus vaccine design using reverse vaccinology and machine learning the coronavirus datasets in italy. website role of changes in sars-cov-2 spike protein in the interaction with the human ace2 receptor: nn in silico analysis characterization of spike glycoprotein of sars-cov-2 on virus entry and its immune cross-reactivity with sars-cov identification of a potential mechanism of acute kidney injury during the covid-19 outbreak: a study based on single-cell transcriptome analysis the paper news network. website covid-19/sars-cov-2 resources. website intensive care management of coronavirus disease 2019 (covid-19): challenges and recommendations. the lancet respiratory medicine zdock server: interactive docking prediction of protein-protein complexes and symmetric multimers predicting mortality risk in patients with covid-19 using artificial intelligence to help medical decision-making. medrxiv covid-19 vaccine candidates: prediction and validation of 174 sars-cov-2 epitopes. biorxiv interpretable deep learning in drug discovery covid-19 epidemic analysis using machine learning and deep learning algorithms. medrxiv machine learning-based ct radiomics model for predicting hospital stay in patients with pneumonia associated with sars-cov-2 infection: a multicenter study. medrxiv fighting against the common enemy of covid-19: a practice of building a community with a shared future for mankind personalized workflow to identify optimal t-cell epitopes for peptide-based vaccines against covid-19 detecting covid-19 in x-ray images with keras, tensorflow, and deep learning. website novel corona virus 2019 dataset. website images of covid-19 cases. mendeley data epitope-based chimeric peptide vaccine design against s, m and e proteins of sars-cov-2 etiologic agent of global pandemic covid-19: an in silico approach the role of artificial intelligence in management of critical covid-19 patients machine learning using intrinsic genomic signatures for rapid classification of novel pathogens: covid-19 case study. biorxiv identification of covid-19 can be quicker through artificial intelligence framework using a mobile phone-based survey in the populations when cities/towns are under quarantine computational immunology meets bioinformatics: the use of prediction tools for molecular binding in the simulation of the immune system drones and artificial intelligence to enforce social isolation during covid-19 outbreak covid-19 forecasting based on an improved interior search algorithm and multi-layer feed forward neural network mathematical modeling of epidemic diseases mobilenetv2: inverted residuals and linear bottlenecks ai-driven tools for coronavirus outbreak: need of active learning and cross-population train/test models on multitudinal/multimodal data the essential facts of wuhan novel coronavirus outbreak in china and epitope-based vaccine designing against 2019-ncov a machine learning model reveals older age and delayed hospitalization as predictors of mortality in patients with covid-19. medrxiv covid-19 and computer audition: an overview on what speech and sound analysis could contribute in the sars-cov-2 corona crisis covid-19 ct segmentation dataset. website improved protein structure prediction using potentials from deep learning protein structure prediction using multiple deep neural networks casp13 detection of coronavirus disease (covid-19) based on deep features lung infection quantification of covid-19 in ct images with deep learning large-scale screening of covid-19 from community acquired pneumonia using infection size-aware classification deep learning-based quantitative computed tomography model in predicting the severity of covid-19: a retrospective study in 196 patients self-attention based molecule representation for predicting drug-target interaction very deep convolutional networks for large-scale image recognition covid-19 database. website from a single host to global spread. the global mobility based modelling of the covid-19 pandemic implies higher infection and lower detection rates than current estimates. the global mobility based modelling shape-based generative modeling for de novo drug design deep learning enables accurate diagnosis of novel coronavirus (covid-19) with ct images. medrxiv development and evaluation of a deep learning model for protein-ligand binding affinity prediction inception-v4, inception-resnet and the impact of residual connections on learning going deeper with convolutions rethinking the inception architecture for computer vision ai-aided design of novel targeted covalent inhibitors against sars-cov-2. biorxiv severity assessment of coronavirus disease 2019 (covid-19) using quantitative features from chest ct images drug target commons 2.0: a community platform for systematic analysis of drug-target interaction profiles predicting the evolution of sars-covid-2 in portugal using an adapted sir model previously used in south korea for the mers outbreak breadth of concomitant immune responses prior to patient recovery: a case report of non-severe covid-19 an investigation of transmission control measures during the first 50 days of the covid-19 epidemic in china temporal profiles of viral load in posterior oropharyngeal saliva samples and serum antibody responses during infection by sars-cov-2: an observational cohort study. the lancet infectious diseases susceptible-infected-recovered (sir) dynamics of covid-19 and economic impact of chloroquine and covid-19 coronavirus search trends. website autodock vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading school closure and management practices during coronavirus outbreaks including covid-19: a rapid systematic review. the lancet child and adolescent health the immune epitope database 2.0 atomnet: a deep convolutional neural network for bioactivity prediction in structure-based drug discovery unexpected receptor functional mimicry elucidates activation of coronavirus fusion structure, function, and antigenicity of the sars-cov-2 spike glycoprotein tracking and forecasting milepost moments of the epidemic in the early-outbreak: framework and applications to the covid-19. medrxiv covid-net: a tailored deep convolutional neural network design for detection of covid-19 cases from chest radiography images remdesivir and chloroquine effectively inhibit the recently emerged novel coronavirus (2019-ncov) in vitro structural definition of a neutralization-sensitive epitope on themers-cov s1-ntd a deep learning algorithm using ct images to screen for corona virus disease (covid-19). medrxiv hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases abnormal respiratory patterns classifier may contribute to large-scale screening of people infected with covid-19 in an accurate and unobtrusive manner an integrated in silico immuno-genetic analytical platform provides insights into covid-19 serological and vaccine targets. biorxiv who. novel coronavirus 2019 (covid-19). website virological assessment of hospitalized patients with covid-2019 covid-19 coronavirus pandemic. website the statistics datasets on holidays and air travel. website a new coronavirus associated with human respiratory disease in china estimating clinical severity of covid-19 from the transmission dynamics in wuhan, china nowcasting and forecasting the potential domestic and international spread of the 2019-ncov outbreak originating in wuhan, china: a modelling study rapid and accurate identification of covid-19 infection through machine learning based on clinical available blood test results. medrxiv inhibition of sars-cov-2 (previously 2019-ncov) infection by a highly potent pan-coronavirus fusion inhibitor targeting its spike protein that harbors a high capacity to mediate membrane fusion epidemiological data from the covid-19 outbreak, real-time case information deep learning system to screen coronavirus disease 2019 pneumonia characteristics of pediatric sars-cov-2 infection and potential evidence for persistent fecal viral shedding a machine learning-based model for survival prediction in patients with severe covid-19 infection. medrxiv prediction of criticality in patients with severe covid-19 infection using three clinical features: a machine learning-based prognostic model with clinical data in wuhan. medrxiv design of wide-spectrum inhibitors targeting coronavirus main proteases clinical characteristics and imaging manifestations of the 2019 novel coronavirus disease (covid-19): a multi-center study in wenzhou city, zhejiang flower pollination algorithm for global optimization modified seir and ai prediction of the epidemics trend of covid-19 under public health interventions a model for covid-19 prediction in iran based on china parameters covid-19: review indigenous peoples' data the epidemiology, diagnosis and treatment of covid-19 covid-19 screening on chest x-ray images using deep learning based anomaly detection progress and prospects on vaccine development against sars-cov-2. vaccines potential covid-2019 3c-like protease inhibitors designed using generative deep learning approaches deep learning-based detection for covid-19 from chest ct using weak label. medrxiv a pneumonia outbreak associated with a new coronavirus of probable bat origin network-based drug repurposing for novel coronavirus 2019-ncov/sars-cov-2 unet++: a nested u-net architecture for medical image segmentation key: cord-015352-2d02eq3y authors: nan title: espr 2017 date: 2017-04-26 journal: pediatr radiol doi: 10.1007/s00247-017-3820-2 sha: doc_id: 15352 cord_uid: 2d02eq3y nan prof. michael riccabona undertook his medical school & university training at the university of innsbruck, tirol, austria, completing his md at karl franzens university in graz, austria. following an internship in neurology, surgery and internal medicine he then specialized in paediatrics at the dept. of paediatrics, university hospital in graz. there he took charge of the paediatric radiology and sonography sections at university hospital in graz, as associate professor of paediatrics. in 1993 he additionally started to specialise in radiology, becoming associate professor of radiology in 1998 -then taking charge of the subsection of paediatric sonography at the dept. of radiology, university hospital in graz, where in march 2006 he was appointed full univ. prof at the medical university graz, austria. he has a distinguished academic career and written over 200 papers, more than 50 chapters and several textbooks, and is a very popular international speaker, delivering numerous lectures at many high profile scientific meetings. he is an active member of several reputable international societies, has been chair of the paediatric ultrasound section of the austrian ultrasound society since 2003 , and president of the society of german speaking paediatric radiologists (2010) (2011) (2012) (2013) (2014) (2015) (2016) . he is a constant source of inspiration within his subspecialist areas of interest in ultrasound and abdominal radiology in children. he has been course director at several important meetings and served as president of espr in graz in 2015 and as lead of the paediatric subcommittee at ecr in 2012. he has provided inspirational leadership as chair of the espr task force on uroradiology since 2002writing state of the art guidelines and procedural recommendations to facilitate standardised best practice for imaging within paediatric uroradiology. he has been a reviewer for many international journals. he has on-going active roles in postgraduate education for medical colleagues from eastern europe, including basic ultrasound education and refresher courses, and workshops. michael riccabona is very deserving of honorary membership of espr, which reflects his seminal role within paediatric radiology in europe and his tireless dedication working for the good of children. & to discuss how to optimally adapt the various imaging techniques minimising radiation exposure and risks during diagnostic imaging in children. & to consider common restrictions, challenges, and possible solutions in paediatric radiology within the different settings in different countries, regions, continents and clinical scenarios -discussing all these aspects with colleagues, and to mingle with experts from all over the world learning from each other and fostering networking in paediatric radiology to try to grant optimal imaging for all children. an application for the 53 rd annual meeting as well as for the 39 th post graduate course has been submitted to the eaccme® for cme accreditation of this event. the eaccme is an institution of the uems (www.uems.net). the number of cme points will be announced at the espr 2017 congress website. each medical specialist should only claim those hours of credit that he/she actually spent in the educational activity. certificates of attendance will be available in the espr myuserarea after the meeting. the answers right away is: yes we can and we constantly have to. it is part of human nature to recognize problems and to find solutions. we define ideals but we also face reality. being aware of the gap in between we are constantly driven to improve. this overview will highlight some milestones and disputes throughout the 100 years of development in the use of plain x-ray imaging. it will hint on scientific literature and sources of information. and it will hint on some swiss contributions as the espr meeting 2017 will be held in davos, switzerland. fighting the glow not the fire -122 years of x-ray imaging development and improvement: in the beginning of the clinical use of x-ray imaging, there was great enthusiasm in its potential without knowing the unfavorable dangers of uncontrolled use of x-ray. the dangers were recognized and the 'beast' was tamed and domesticated. in respect to radiation protection, the most significant achievements took place in the first half of that development. throughout the recent decades, some further considerable steps in dose reduction took place mainly by the improvement of film-screen systems and the recent introduction of computed (cr) and digitized radiography (dr). even if the early computerized x-ray imaging brought a slight increase in patient doses (which was overcompensated later on by direct digital radiography) a whole new world of further advantages launched the digital era in which we live today. as we all know new dangers arose with these techniques as the uncontrolled distribution of images and thereby confidential patient data over hospital departments and across borders throughout. also, the risk of an evitable overexposure in digital radiography is a significant issue. throughout the process of taming the radiation and controlling it, today's doses are attained within the lowest range of the danger scale. this range still is perceived as a black box within which we do not exactly know which concept reflects the potential harm best. the linear no-threshold model [lnt] is acknowledged as the concept which most reliably supports the idea of radiation protection. other concepts partly oppose the linear idea and question the relevance of that dose range because there lie so much greater health benefits in the appropriate use of diagnostic x-rays. several scientists even propagate the idea that very low levels are producing health benefits instead of physical harm [hormesis model]. nevertheless, the lnt model is widely accepted as the most helpful in the context of diagnostic radiology. some recent studies were able to support the idea of potential harm at very low dose levels as they were able to prove the induction of attributable cancers in the pediatric age group. so today we are fighting the glow, not the fire. as trained medical professionals we are fully aware of the fact that there is only little potential harm to the patient by using x-rays in the current state of the art. but on the other hand, we also have to be aware of the fact that our patients and their parents still fear the fire. one of our main tasks, therefore, is to explain the risks and benefits to the patients and their advocates and to educate the public. developed straight from the first radiographic technique's digital radiography today is state-of-the-art in plain 2d-imaging. throughout the last decade, it has replaced cr and conventional radiography in many institutions. in the united states of america, one of the most developed healthcare systems, healthcare authorities propagate incentives to abolish cr and older imaging systems by making them financially unattractive. the market is fully concentrated on the spread of dr systems. momentarily there are no real milestones but many refinements of existing systems such as tomosynthesis, dual energy subtraction and advanced auto-stitching, fluoroscopy capability, basic angiography applications and 3-d cone-beam ct images are made available. combinations of these features can be found in some recently designed x-ray machines. grid-less imaging software can reduce patient dose significantly. concerning detectors, there are cr retrofit systems which will support the easy upgrade of existing systems to dr capability. wireless detectors with large internal storage, different sizes and high resolutions of 100 microns are available. dr is becoming a part of the system in the current era of full digitalization of our lives and big data, digital radiology is a cornerstone of our healthcare systems. ris and pacs as part of integrated healthcare (ihs) systems are widely disseminated. at the next step, all accessible data will be used for analyses. the major vendors of imaging systems, as well as pacs suppliers and independent companies, offer readymade software tools for reports and evaluations of all kind. the doses from different x-ray sources can be screened internally and be used for optimization purposes. they also can be sent to remote servers for dose monitoring, comparison and optimization in multi-hospital health care provider settings or to comprehensive databases like the american college of radiology dose index registry, cancer registries, or for central billing. has everything been invented? many technologies have been declared dead before a new transformation appeared. this was the case for example with single slice ct before the invention of spiral ct by willi kalender, germany and peter vock, switzerland in 1988. often the plain x-ray image was meant to be needless or redundant as newer technologies like ct or mri approached. but it still is of value because of different reasons as the low dose, high availability, well known and easy interpretation to name a few. there are some new and sophisticated techniques on the way like the "smart x-ray source" which uses coherent beams of x-rays from an array of micronsized point sources, developed by scientists at the massachusetts institute of technology (mit). the developers promise less radiation, less weight of the equipment and a far better soft tissue resolution. another promising approach is phase contrast x-ray imaging which has the potential to reduce the dose up to 1/100 of the actual value. it also has its strengths in additional soft tissue information as recent experimental publications show (paul scherrer institute switzerland) e. g. in functional evaluation of lung fluid (munich, germany). functional imaging of the lungs can also be achieved without any radiation as the development of the known concept of electrical impedance tomography highlights. this functional imaging method usually is not within the modality spectrum of radiologists. dose control and reduction -local -regional -international the most effective measures to achieve significant dose reductions in your own department are still the same strategies which are based on the "eternal rules" as we know them from our teachers: avoiding unnecessary exposures by strictly controlling the appropriateness of a referral. justification is a shared responsibility between radiologists and clinicians. there are many tools available for justification like the appropriateness criteria, guidelines or rules (like wrist or ankle rules) of several national societies and different study groups. the process of optimization is mainly in the hands of the technicians. as many studies show, the proper collimation still has the greatest effect on dose reduction. other important factors are the positioning of the patient and the shielding of radiosensitive organs which are not relevant for image interpretation so that they may be covered by lead shields. the proper use of the grid in bigger children can now partially be replaced by software solutions. in digital radiography, a profound knowledge of postprocessing possibilities is mandatory as well as the active control of the exposure indices. dose limitation procedures should be regularly checked in a team-based approach to avoid overexposure by less experienced staff or "exposure creep". existing standards should be actively used to guarantee a constant satisfactory image quality. in 2011, the image gently campaign released a safety checklist for performing digital radiography examinations on pediatric patients which is easily applicable to every radiology service. organizational improvements: at regional and national level, efforts should be made to check for best practice use in the departments and to compare and discuss imaging strategies. the establishment of national and international dose reference levels helps to keep the overall doses low and to protect the population from unnecessary overexposure. the pidrl project prepared the "european diagnostic reference levels for pediatric imaging" as part of the eurosafe project. momentarily the results of pidrl-workgroup are harmonized with international organizations. the european guidelines on drls for paediatric imaging can be accessed as a preliminary final for workshop drafts on the internet. on a worldwide basis, the world health organization has published a fundamental information brochure concerning radiation risks and the communication of health professionals and patients. health care professionals have a shared responsibility for communicating risks and benefits of imaging procedures to patients, especially in the case of pediatric patients. the document "communicating radiation risks in paediatric imaging-information to support health care discussions about benefit and risk" is intended to serve as a tool for health care providers, to communicate known or potential radiation risks associated with pediatric imaging procedures and to support risk-benefit dialogue in health care settings. as said before we are fighting the glow, not the fire. the paper of the swiss pediatric oncology group stirred a broad discussion. among other issues, there was a question if it shouldn't be a logical consequence to transfer kids from areas with higher background radiation to safer areas. the author's answers were clear: that swiss health authorities better concentrate their efforts more effectively and with greater benefit for more people by supporting prevention "toward modifiable environmental factors leading to larger numbers of deaths from several causes, such as exposure to radon, air pollution, and second-hand tobacco smoke". this leads to the conclusion that we as medical radiological professionals do have the obligation to make every effort to prevent our patients and personnel from harm of the usage or non-usage of radiation. as health specialists, we also should support the fields of prevention with broad mass effects as far as we have the opportunity. and as human beings, we are summoned to do so in respect to other beings, to our environment and to the resources we all share. radiation protection and quality improvement is just a small part of it all, but it is our field -and 'yes we can'. "communicating radiation risks in paediatric imaging. freely available at the who homepage." computed tomography: are we doing enough? e. sorantin; graz/at summary: already in 1912 the alara principle was publishedbut the implementation is still far from complete. according to the surveys of the ec tender project "pidrl -european diagnostic reference levels for paediatric imaging" the most frequent computed tomography (ct) examinations in children are, in descending order, head/neck, chest and abdomen thus counting for about 75% of all pediatric ct's. therefore it makes sense to optimize these examinations first. surveys of the "international atomic energy agency (iaea)" in 40 countries have shown, there is considerable lack of organizationeg in about 50% of facilities protocols for children were missing, indication based protocols available only in 57%, ctdi values for head and chest two to five times of those for adults. all of these simple facts indicate we are not doing enough for radiation protection in pediatric ct. actions to lower dose in ct can be categorized in organisational, optimization and alternatives. the interdisciplinary implementation of international guidelines for ct in minor head trauma with trauma surgeons could serve as an example of organisational actions. for dose optimization knowledge about dose relevant factors according the "imaging chain" is mandatory as well as adjusting kv to pediatric needs. dose influence on image quality must be known, by exploiting the fact, that, if all ct parameters are kept constant but hte slice thickness is just halve there must be an increase in noisein particular about two times more. therefore if a standard examination is reconstructed at half slice thickness and image quality is still appropriate the amount of waste radiation is in the range of 100%. therefore if the next examination will be reduced with eg 20% mas setting less will be for sure in appropriate quality and the process can be started again. after a couple of examinations the optimal dose will be reached. thus the "half slice thickness" approach is easy to do, does not need special equipment or human resources and will help to find the appropriate dose. the third point is alternatives -ultrasound and mri being the candidates in the first row. new, radiation free, techniques like electrical impedance tomography and others are already developed and can be expected to be release soon. take home points: & we are not doing enough for ct dose savingeven more than 100 years after release of alara principle & dose saving actions can be categorized inthe subtasks organisation, optimization and alternatives & "the half slice thickness approach" is an easy to do technique to elaborate the optimal dose on an particular ct machine. prenatal thoracic mr l. alamo; lausanne/ch the generalization of screening us has considerably increased the detection of congenital anomalies in utero. in the last years, important technological advances and especially, the development of fast heavily t2-weighted sequences has led to an increasing use of prenatal mri as additional diagnostic imaging method. mri is increasingly used for evaluation of thoracic pathology, including tumours and vascular malformations as well as anomalies of the diaphragm, the lungs and more recently, even of the foetal heart: -thoracic tumours and vascular malformations: the diagnosis of a congenital tumor during pregnancy involves a tremendous emotional impact for a family. the most frequently observed thoracic tumours are teratoma, myocardial rhabdomyoma and exceptionally, pleuropulmonary blastomas. mri may provide relevant additional information concerning the origin of the lesion and its real anatomical extent. -diaphragmatic pathology: congenital hernia is by far the most commonly reported foetal diaphragmatic anomaly. the large field of view and the multiplanar possibilities of mri may help to clarify the position of the herniated organs and to evaluate the severity of lung hypoplasia, considered the most important parameter for predicting outcome. other rare pathologies include eventration, paralysis and diaphragmatic lung sequestrations. -lung anomalies: congenital lung abnormalities are a heterogeneous group of pathologies consisting of isolated bronchopulmonary or vascular anomalies or a combination of both of them. congenital pulmonary airway malformation, bronchopulmonary sequestration and bronchial atresia are the most often observed pathologies but they present significant overlap imaging findings. mri allows accurate information concerning the location and extension of the lesion and the volume of the normal and abnormal lung. -heart pathology: the evaluation of the foetal heart remains extremely difficult because of its small size and high rate of battements. the unpredictable foetal motions during data acquisition and the absence of a foetal ecg signal to synchronize data acquisition are additional problems. in the last years, different approaches have been made to overcome these challenges. radiologists should know the typical imaging findings of the thoracic pathology most often observed in foetuses. prenatal mri may provide additional relevant information in a wide spectrum of congenital thoracic anomalies, but in general, it should only be performed if it is considered that additional results might influence the management of the pregnancy and/or the therapeutic approach. therefore, it is important to know the right indications for mri and to recognize the limits of the method. interruptions during embryogenesis of the muellerian or wolffian ducts result in various, potentially complex genitourinary abnormalities of a wide spectrum or combinations. multiple imaging modalities are employed to evaluate patients with these abnormalities. ultrasound is the frontline imaging modality. mr imaging is mostly reserved for complex cases and may incorporate an mr urography, too. other imaging modalities are less frequently used or provide only ancillary information. this presentation will demonstrate the utility of ultrasound and mr imaging, in particular, in the routine diagnostic imaging of patients with the wide spectrum of muellerian and wolffian duct abnormalities. & mr imaging is reserved for more complex cases. neonatal hepatic tumors and vascular malformations d. pariente, s. franchi-abella; le kremlin bicêtre/fr neonatal hepatic tumours and vascular malformations are rare but imaging plays a key role in diagnosis and treatment. the most frequent hepatic tumour is haemangioma (fig1) which often is asymptomatic but may be complicated by cardiac failure, coagulopathy or compartment syndrome. the differential diagnosis mainly includes hepatoblastoma, hematoma (fig2), abscess, mesenchymal hamartoma, choriocarcinoma in the solitary form and metastatic neuroblastoma, cirrhosis, neonatal leukemia in the multifocal form. pertinent biological data are alpha-fetoprotein (but level may be normally high in neonate), betahcg, and urinary catechol amines. hepatic vascular malformations are rarer and include intra or extra porto-systemic shunts (pss), arterio-portal fistula or complex mixed forms. intrahepatic pss may be associated with haemangioma and regress in most cases rapidly (fig3). on the contrary the extrahepatic pss which are located below the portal vein, should be urgently closed to avoid occurrence of agenesis of the portal vein. the best imaging modality is us which must be performed with high frequency probes and colour doppler to identify hepatic vessels and assess patency, direction of flow, abnormal communication. mri and ct with contrast injection may also be useful. hepatic mass in a 12 do neonate with increased crp and afp. us showing a hyperechoic mass with thrombosis of the left portal vein (black arrow) and a track (white) extending to the mass: hematoma due to malposition of an umbilical vein catheter. hemangioma of antenatal diagnosis on d1. the mass is composed of a large anterior vascular lake corresponding to a porto-systemic shunt and a tissular hyperechoic part. the infant has remained asymptomatic. intrahepatic porto-systemic shunt between the left portal branch of segment 2 (white arrow) and the left hepatic vein (black arrow) in a neonate. at 3 months of age this shunt has completely resolved. haemangioma is the most frequent hepatic tumour in the neonate and is often asymptomatic with spontaneous resolution. levels of alphafetoprotein are physiologically high in the neonate, and can be misleading. hepatic hematoma can be secondary to traumatic delivery, to coagulation disorders or to umbilical vein catheterization. intrahepatic porto-hepatic shunts are the most frequent vascular malformations and regress in most of the cases in the first year of life. us with colour doppler remains the best imaging modality in the neonatal period. imaging in crohn disease: state of the art in diagnosis, prognosis and followup n. colavolpe, a. aschero, b. bourliere-najean, c. roman, f. khachab, h. pico, m. kheiri, g. gorincour, c. desvignes, p. petit; marseille/fr summary: during the past years the inflammatory bowed diseases (ibd) have increase in frequency (1). less than twenty-five percent of them occur in children of less than 18 years (2) and crohn's disease (cd) is twice as frequent than ulcerative colitis (uc) in the pediatric age group. specific phenotypic and genotypic subtype of ibd occur in younger children. early onset (eo) pediatric ibd (before 5 years of age) represent 11% of childhood ibd (3) . uc and undetermined colitis are more frequent in this age group. eo cd showed a more frequent isolated colonic and upper gastrointestinal involvement than later-onset disease where locations are predominantly colic and terminal ileum later on childhood. some pediatric ibd specificities exist than can interfere with the imaging findings: -cd can be limited to the terminal ileum or to the colon in up to 20% of children (4) . isolated jejunal involvement is reported to occur in 5-6% of children. this location is more frequent in the youngest and is more at risk of complicated course of disease (2) . for auvin et al. (3) the small bowel is involved in 80% of cases with less involvement of the terminal ileum than in the adult population. -uc: the classical contiguous alteration of the bowel wall from the rectum to the caecum is inconstant. a macroscopic rectal sparing is reported from 5 et 30% and the absence of continuous disease from rectum to caecum (caecal patch) described in 2% of children. transmural inflammation may be present in severe form as well as terminal ileitis without granulomata (backwash ileitis) (2) . in order to assess these pathologies, and more specifically cd small bowel locations which are difficulty explored by others modalities, small bowel follow-through, barium enema, ultrasound, computed tomography and mr imaging have been widely used. among them, mr-enterography has gained worldwide acceptance due to multiple factors including: a high contrast resolution, a multiplanar ability, an absence of radiation, the possibility to explore in the same exploration the whole bowel and the extra-bowel diseases (perianal fistulae, sacroiliac joint, biliary tract), the ability to compare of side by side consecutive studies in a reproducible manner, a more easily understood exploration by the clinicians than ultrasound, and first of all for its performances. in order to technically harmonize this exploration a recent consensus statements on mre protocol has been published by the esgar and the espr societies (5) . preparation: -depending on their age children must not have solid oral intake from 2 to 6 hours prior to the examination to reduce bowel wall motility. morning mr appointment is more favorable for this purpose. no gasless fluid restriction is recommended but is reabsorbed too quickly to distend enough the small bowel. none hyperosmolar non absorbable solution is superior to another. its ingestion must start 45 to 60 minutes prior to mre. the recommended volume is 20 ml/kg with a maximum up to 25 ml/kg. explanations long before the mre concerning the importance of such absorption and the use of a refreshed product mixed with aromatized flavors will facilitate the child's participation. -the use of spasmolytic agents is optional. however, there are recommended in adults by multiple societies including esgar (5) , the society of abdominal radiology (6) and the acr (www.acr.org/ quality-safety/standards-guidelines). but, mre without antiperistaltic agents result has reached a high diagnostic confidence and excellent agreement with ct enterography for the presence of cd (7) . if used, they need to be administered immediately prior to motion sensitive sequence (t1w dynamic enhanced sequences). if the pictures obtain with these medications are of better quality, there is no evidence that they change the final diagnosis and the children's therapeutic management (8) . the use of these products increase the length of the exploration and their side effects are frequent (nausea > vomiting) which balance their visual benefice (8) . if a spasmolytic agent is used, the recommended first line spasmolytic agent is i.v. hyoscine butylbromide (0.5mg/kg i.v). the recommended second line agent is i.v. glucagon, 0.5mg (<24.9kg) and 1mg (>24.9kg), given as a slow infusion with i.v. saline at an infusion rate at 1ml/s. -no rectal enema is needed. -exploration can be performed either at 1.5tesla or 3testla. more chemical shift and susceptibility artifacts are present with the latter. prone position has been demonstrated to allow better small bowel distension than the supine one with reduction of the peristaltism but without better lesion detections (9, 10) . large multi-elements coils are needed to cover with high resolution from the perineum up to the left colonic flexure. sequences: both morphologic steady state free precession gradient echo and 2d -t2-weighted images are needed in the axial and coronal planes. fat saturation in one of this plane is recommended and maximal slice thickness of 5 mm is required. nowadays, non-enhanced then enhanced 3d t1-fat saturation weighted sequences are mandatory. slice thickness does not exceed 3 mm. enhance sequence need to be acquired at the portal phase of injection. however, in recent studies the need for gadolinium has been questioned when dwi is added to the morphologic sequences. dwi sequences have been considered optional (5,6) but we consider their place essential in pediatric practice. they must be done with high b values, from 600 up to 800 in the coronal and axial planes with 5 to 8 mm contiguous cut in free breathing. axial plane is less prone to artefact than the coronal plane. interestingly enough shenoy et al (11) report in 27 pediatric patients that dwi does not perform as well as standard mre for detection of active crohn disease but the combination of dwi and mre increases imaging accuracy for determining disease activity compared with either technique alone. seo et al (12) in 44 young adults said that dwi mre was noninferior to contrast-enhanced mre for the evaluation of inflammation in cd. based on the exploration of 130 cd adult and pediatric population, dwi proved to be efficient and would avoid gadolinium injection (14) . sirin et al. (14) report in 37 children that dwi revealed lesions that were not detectable with mre done with gadolinium injection. finally, respectively dubron et al. (15) in 48 children and neubauer et al (16) in 33 children and young adults demonstrate better performance of dwi than gadolinium enhanced imaging. like the existing mr protocols for suspected appendicitis (17) it will not be surprising to see fast mr ibd explorations becoming an alternative to emergency us as already proposed (16) . this fast mr limited to a morphologic t2 sequence in two planes associated with dwi sequences will allow a positive diagnosis and the ibd work up. apart from bowel obstruction and its spontaneous bowel distension one of the limiting factors will be the need for an oral water agent uptake in a potential surgical patient. however, it has been published in the adult literature than an oral or rectal preparation was not necessary to rule out uc (18) nor a cd (19) . the other limiting factor is the length of exploration. mre can be shorten especially if the patient's positioning is easy to do (dorsal decubitus) (20) and if there is no need for injection, either for spasmolotytic agent and for gadolinium chelates. the suppression of the iv line, the absence of potential side effects (nausea, vomiting) of paralytic agents and the decreased of repeated long apneas with no loss of significant information will be strong progresses toward the holy grail. -positive diagnosis, disease activity, prognosis and follow-up: mre has a better accuracy to detect inflammation for the small bowel than for the colon (21) . one of its goal is to try to accurately identify features of active inflammation vs fibrotic disease. this is of paramount importance since the former may respond to medical treatment and the latter may need surgical resection. however, inflammation and fibrosis are associated within the same bowel segment and progress in a parallel way making the goal difficult to reach (22) (23) (24) . in their study based on the analyze of 20 children operated for cd strictures, barkmeier et al. (24) report than strictures demonstrating >3 cm upstream dilatation with associated feces sign were highly associated with transmural fibrosis. the most severely fibrotic strictures were associated to the greatest amount of inflammation and there was no significant correlation between stricture length, mural thickness, degree of post-contrast enhancement (arterial and delayed venous phases), diffusion-weighted imaging apparent diffusion coefficient, pattern of post-contrast enhancement, or normalized t2-weighted signal intensity and histological fibrosis or inflammation scores. however, correlation with histological specimens of cd done on a other series s312 (2017) 47 (suppl 2):s297-s pediatr radiol demonstrated that the enhancement ratio of the wall is positively correlated with disease chronicity due to a possible increasing microvessels permeability and inversely correlated to acute disease (25). on the other hand, several authors have tried to correlate the adc values to cd activity. fibrotic tissue does not restrict diffusion and presents a decrease of signal at high b values and high adc values whether acute inflammation shows decrease adc values. variable thresholds from 1.6 x10 -3 mm 2 /s to 2.4 x10 -3 mm 2 /s have been proposed to separate active vs non active disease (21) . however, others authors have reported low adc value of fibrosis compare to histology (26). even if promising results have been published with high correlation with the crohn disease endoscopic index of severity (13) , adc measurements are associated with sever limitation factors including sample size overlap between the bowel wall and its atmosphere, lack of reproducibility between mri-units and mri-vendors, non-standardized sequence b-values parameters (21) . two mre scores are available to quantify the activity of cd. one is using gadolinium injection (27) and the other dwi (13) . due to the complexity of the formula, both are difficult to use in daily practice and have not been evaluated in paediatric practice. interestingly enough if a simplify mre paediatric protocol appears to become a reality, us stays a good imaging challenger and (28). in a recent meta-analysis, based on adult and pediatric series, calabrese et al (29) reported that bowel us showed 79.7% sensitivity and 96.7% specificity for the diagnosis of suspected cd, and 89% sensitivity and 94.3% specificity for initial assessment in established patients with cd. bowel us identified ileal cd with 92.7% sensitivity, 88.2% specificity, and colon cd with 81.8% sensitivity, 95.3% specificity, with lower accuracy for detecting proximal lesions. the absence of abnormal thickness wall would have a negative predictive value, high enough to exclude the need for further exploration, especially when cd is concerned (30,31). concordance between us and mre have been variably reported from excellent (32) to just correct (33). rosembaum and al (22) report that the us findings present in children operated for cd include: bowel wall thickness above 4.3 mm (mean, 6.1 mm) and an increased frequency of loss of mural stratification and fibrofatty proliferation. others us technologies are used in children to better approach the disease activity. it includes, hydrosonograpy using specific oral agents (mannitol, sorbitol, polyethylene glycol, etc…), contrast-enhanced ultrasound and dynamic contrast-enhanced ultrasound (nowadays, contrast agent is offlabel in children) (34) and elastography (35). their enthusiastic results and their efficiency to assess disease activity need to be confirmed (36). in conclusion, as we suspected 7 years ago (37), mre has dramatically modified our approach of pediatric ibd especially when considering its orientation toward a less invasive exploration and the increasing importance of dwi imaging. a cost benefice between mre and us remains to be done on this increasing disease. heterotaxy and isomerism c. lapierre; montreal/ca summary: objectives: to review the classification of visceroatrial situs to describe the associated cardiac and non-cardiac anomalies to illustrate typical findings in fetuses, neonates and children to discuss the surgical consideration and the long-term follow-up in these patients abstract: by definition, the type of situs is determined by the relationship between the atria and the adjacent organs. anatomically, the atrial chamber differentiation is based on the morphologic aspect of the atrial appendages, earlike extensions of the atria. three types of situs exist: solitus (normal), inversus (mirror image) and ambiguus. a single type of situs is present in a patient. when the situs is neither solitus nor inversus, it is referred to as situs ambiguus or heterotaxy. heterotaxy may manifest with various abnormal visceroatrial configurations that are associated with cardiac (in 90-100% of cases) and extracardiac anomalies such as splenic abnormalities, biliary atresia and intestinal malrotation. two subsets of situs ambiguus are well-recognized: right isomerism (asplenia) and left isomerism (polysplenia). in heterotaxy, the venoatrial connections are frequently abnormal. left isomerism is usually indicated by bilateral bilobed lungs, interruption of the ivc and multiple spleens. the more likely found cardiac anomalies are: pulmonary or aortic stenosis, isolated atrial and ventricular septal defects, cardiac arrhythmia due to sinus node dysfunction as well as pulmonary veins that drain into both the right and the left atria. in the presence of right isomerism, bilateral trilobed lungs, a large symmetric liver, and absence of the spleen are frequently observed. at the cardiac level, patients are more likely to have a common atrioventricular defect, a double outlet right ventricle and pulmonary stenosis. total anomaly of the pulmonary venous return and absence of coronary sinus will always be present in right isomerism. heterotaxy can be diagnosed with high accuracy by prenatal echography. a diagnosis should be suggested in the presence of congenital heart disease, visceroatrial heterotaxy and interruption of inferior vena cava with azygos continuation for left isomerism or abnormally closed juxtaposition of inferior vena cava and descending aorta in right isomerism. the mortality in fetuses is high in the presence of heart block and hydrops whereas the cardiac anomalies influence the long-term outcome. as discussed in the literature, the clinical outcomes and long-term prognosis in these patients are relatively poor when compared with non-heterotaxy patients. the risk factors are cardiac (underlying anatomy and arrhythmia risk) and non-cardiac. based on the cardiac anatomy, one of the main determinants is left versus right isomerism. with right isomerism, the cardiac malformation is more severe and an univentricular correction is more frequent. another predictor of mortality is pulmonary vein stenosis/obstruction. whatever the severity of cardiac lesions, the postoperative or discharge mortality is higher in patients with heterotaxy. prenatal diagnosis seems not improve the survival. extracardiac anomalies also contribute to the increased morbidity and mortality. three of the more challenging entities are respiratory, immunologic and gastrointestinal. recurrent respiratory infections, failed extubation or chronic respiratory failure are frequently observed in patients with heterotaxy. recent studies revealed an association between heterotaxy and primary ciliary dyskinesia which can explain the increased postoperative respiratory complications. the spleen is important for the bacterial clearance. patients with asplenia or polysplenia are thought to have "functional asplenia". so, they are at risk for sepsis and severe bacterial infection. the incidence of intestinal malrotation is high, approximately 40% to 90%. observation versus prophylactic ladd procedure and screening for asymptomatic intestinal malrotation are a growing area of debate. the trend seems to go along conservative management and surveillance of malrotation. bronchopulmonary malformations, such as congenital pulmonary airway malformation (cpam), bronchopulmonary sequestration (bps), and congenital lobar emphysema (currently known as congenital lobar overinflation [clo] ), are common congenital lung diseases. these conditions are detected prenatally, usually in the second trimester, in countries where obstetric sonography is routinely performed. the malformations are seen as hyperechoic images with respect to normal fetal lung parenchyma, with a mass effect and homogenous appearance or with coexisting cysts. the lesions usually decrease in size along gestation. a residual mass is seen on postnatal chest radiography, the first imaging technique performed, in only 40% of cases. cpam and bps are predominantly located in the posterior lower chest and can be identified postnatally on ultrasound using a small vector probe and a subcostal and subxiphoid approach. potential feeding arteries can be visualized using color or power doppler. based on clinical and sonographic findings, the differential diagnosis between congenital lung malformations and tumors such as neuroblastoma, type i pleuropulmonary blastoma, and myofibroblastic tumor will be discussed. postnatal management and imaging of newborns with congenital lung malformations is controversial, particularly in asymptomatic patients (approximately 80% of cases). chest radiography is mandatory at birth and chest ultrasound is also recommended to confirm the prenatal diagnosis. computed tomography (ct) or magnetic resonance imaging (mri) using angiographic techniques should be performed some months (8 months) after birth in asymptomatic patients. these techniques are also recommended in symptomatic newborns and before surgery to characterize the arterial supply and venous drainage in cpam and bps, as ultrasound is limited in this regard. in premature infants, sonography complements radiography in the study of prematurity-related lung diseases such as respiratory distress syndrome and its pulmonary complications (eg, pneumothorax), in predicting bronchopulmonary dysplasia, and in diagnosing transient tachypnea of the newborn when clinical and radiographic features are inconclusive. the main ultrasound finding in these conditions is visualization of numerous "b-lines", vertical narrow-based hyperechoic bands extending from the pleural surface to the end of the field of view, representing what is currently known as "sonographic interstitial syndrome". b-lines are artifacts originating from variations in the air-fluid relationship of the lung and are better seen using high-frequency linear probes . use of sonography for follow-up of these patients will reduce the number of the chest plain films performed, and therefore, the amount of radiation exposure in this vulnerable population. for proper interpretation of the sonographic findings in these conditions, the radiologist should be familiar with current related terms, such as lines a, lines b, comet tail artifact, interstitial-alveolar syndrome, septal syndrome, and white lung. trauma is the leading cause of mortality and morbidity in children after the first year of life. motor vehicle accidents are the leading cause of death from unintentional injury in children up to the age of 15. of these cases, the abdomen is the fourth most commonly injured area. in pediatric patients non-operative management of these injuries predominate, hence the importance of early radiologic assessment for appropriate clinical follow-up. anatomically, compared to adults, childrens' abdomens are more square, less muscular and with less intraperitoneal and subcutaneous fat to absorb impact. the diaphragm is more horizontal causing downward displacement of the liver and the spleen outside the protective casement of the ribs. the pelvis is smaller and hence the bladder is displaced upward, also resulting in more vulnerability to this organ. the organ surface area is larger in children and they have a smaller body mass-hence more force applied per-unit of body surface area. the ribs are flexible, and although we see fewer rib fractures, this results in more internal damage. physiologically, children maintain hemodynamic stability longer, often presenting with only mild tachycardia, even when in severe hemodynamic shock. decrease in blood pressure may not be evident before the loss of 30% blood volume. nevertheless, bleeding is less severe and operative intervention is rarely performed. mechanics of blunt abdominal trauma include organ compression from seat belt injury with the presence of erythema, ecchymossis or abrasion on the abdominal wall increasing the likelihood of internal organ injury (55% likelihood of injury). other common mechanisms include pedestrian-car collisions( 4% with intra-abdominal injuries), falls (4% with intra-abdominal injuries), or handle bar injuries (54% with intraabdominal injuries). after the child arrives in the hospital, a trauma algorithm is initiated. generally, for the unstable patient, algorithms are similar and require a rapid atls protocol, followed by a fast ultrasound to confirm free fluid prior to operation. in stable patients, institutional algorithms vary greatly between countries and in different centers. some rely solely on mechanism to determine the need for fast vs ct (not complete ultrasound), others will rely on clinical exam (in a conscious patient with reliable exam) and blood work to determine the need for imaging (ct or us) and others may chose to perform an initial us and complete the exam with a constrastenhanced us during work hours. in the literature many management prediction rules exist based on the history, physical examination, mechanism of injury and are supplemented by blood work and/or intial imaging. most are based on retrospective reviews, with only a few controlled clinical trials. however, the validity of these studies is limited because of different populations, institutional policies and variable radiological practices in terms of when imaging is performed, which modalities are most beneficial and which are less valuable, all the while, considering the utilization of the least irradiating techniques. a representative sample of such algorithms will be discussed. routine and extensive initial trauma panels are not required according to a number of studies. abdominal ultrasound and urinalysis together have been found to confirm 98% of all intra-abdominal injures, in some studies. serial haemoglobins/hematocrit is valuable for determining ongoing s315 (2017) 47 (suppl 2):s297-s pediatr radiol blood loss and assists clinical surveillance. electrolyte abnormalities are uncommon in children unless severe shock is present (metabolic acidosis). liver function tests are elevated in most cases of blunt abdominal trauma, hence, are often performed for its high sensitivity, to avoid ct if the liver panel is negative. imaging, however, is needed for grading of the potential liver injury if the liver panel is positive. abdominal xray is not useful in blunt abdominal trauma, and is usually normal. ultrasound has an important role in the pediatric community, as a sensitive and non-irradiating modality. however, this sensitivity is dependent on the type of ultrasound performed (fast vs. complete abdominal ultrasound vs. contrast-enhanced ultrasound) but also on the qualifications and experience of the performing physician. a meta-analysis of fast in pediatrics demonstrates that it has a sensitivity of 66% (grade i-ii evidence) for identifying hemoperitoneum. a negative fast is not sufficient to rule out intra-abdominal trauma. one prospective observational trial demonstrasted that 34 % of patients without free fluid on fast (performed by formally trained pediatric truama surgeons demonstrated at least grade iii liver or splenic injuries on ct). we know that pediatric ultrasound is operator-dependent, and generally an ultrasound performed by the skillful hand of a pediatric radiologist is more sensitive than that performed by surgeons or by adult radiologists. furthermore, we know that the benefits of contrast-enhanced ultrasound in pediatric trauma exist-highly accurate in visualising lesions, hence avoiding non-contributive ct imaging, however, the feasibility of providing 24-hr contrast-enhanced ultrasound by a qualified radiologist is resource intesive: both structurally and with respect to personnel. published indications for abdominal ct in stable pediatric patients included suspected mechanism of blunt abdominal trauma, significant fluid resusitation without apparant blood loss, hemoglobin <100mg/l without obvious blood loss, multisystem trauma and unreliable abdominal exam. one series with 1500 children undergoing ct for blunt abdominal trauma demonstrate postive findings in 388 (26%), of which all solid organ injuries and 96% of hollow viscus injuries were identified on ct. however, ct has its limitations: it was found to identify gastrointestinal perforation in only 47% of patients with known perforation, but with findings of free fluid, wall thickening and/or bowel dilatation. it is also less accurate in identifying pancreatic trauma, with normal scans in 33-53% of children with pancreatic trauma. again, findings of pancreatic trauma can be non-specific: free fluid or, less commonly, thickening of the gerota's fascia, presence of mesenteric fluid or of fluid between the pancreas and the superior mesenteric vein. when and where to perform ct depends on the imaging algorithms established by individual centers. generally, unstable patients with very high grade visceral injuries are taken to surgery. the stable patients are treated with non-operative management. the literature on angiographic embolization in pediatric blunt trauma is limited to case series that demonstrate a limited utility in hemodynamically stable patients with ongoing blood loss or for the definitive treatment of traumatic pseudoanevryms. a dialogue with the interventional radiologist is imperative in such cases. common imaging findings and pitfalls will be illustrated with case examples. in conclusion, a child's anatomy and physiology must be taken into account when determing the level of urgency and appropriate imaging work-up in blunt abdominal trauma. imaging of these patients cannot follow a standard algorithm as institutions vary with respect to types of personel, training, frequency of trauma, emergency department trauma protocols and availability of an in-house pediatric radiologist. ultrasound and ct have their advantages and disadvantages with associated pitfalls that the pediatric radiologist must recognize to provide an optimal diagnostic workup with minium irradiation. take home points: a child's anatomy and physiology must be taken into account when determing the level of urgency and appropriate imaging work-up in blunt abdominal trauma. imaging of pediatric abdominal trauma cannot follow a standard algorithm as institutions vary with respect to types of personel, training, frequency of trauma, emergency department trauma protocols and availability of a pediatric radiologist. ultrasound and ct have their advantages and disadvantages with associated pitfalls that the pediatric radiologist must recognize to provide an optimal diagnostic workup with minium irradiation. sport injuries d. jaramillo; miami, fl/us the growing skeleton has unique vulnerabilities to acute and chronic injuries due to sports. the practice of intensive sports during puberty and adolescence has led to a great increase in the incidence of sportsrelated injuries. during the growth spurt of early adolescence, the physis becomes weak, and is the site of fractures and avulsions (particularly in the apophyses) and of physeal widening due to repeated stresses, such as the wrist in gymnasts or the proximal humerus of baseball pitchers. both lesions can result in growth arrest. the chondro-osseous junctions of the ossifying epiphyses and apophyses are also vulnerable to avulsions, and the avulsed fragment may be entirely cartilaginous and not visible radiographically (such as in the patellar sleeve fracture). repeated trauma to epiphyses or round bones can lead to osteonecrosis (panner's disease) but more often to osteochondritis dissecans (ocd). in adolescents, ocd occurs most frequently in the medial femoral condyle, the capitellum of the elbow and the talar dome. juvenile ocd has a better prognosis than the adult form. when the skeleton begins to mature, there are fractures unique to partially closing physes such as the triplane and tillaux fracture. some sturctures have propensity to unique injuries during adolescence. a stress on the anterior cruciate ligament (acl) can lead to a tibial eminence avulsion in puberty ( figure) , an incomplete acl tear in early adolescence or a complete acl tear later. meniscal tears are almost always vertical and often involve large meniscal fragments that can flip. patellar dislocations often result in osteochondral injuries. this review will cover the main types of sports-related injuries and the imaging modalities used to diagnose them. 11 year-old with pain and popping sensation during a fall on a football game. ap radiograph is normal & it is important to take into account the specific sport in order to anticipate subtle injuries that may be difficult to detect. a. c. offiah; sheffield/uk the radiographs obtained when inflicted injury is suspected are collectively termed the "skeletal survey". a full skeletal survey should be performed in all children below 2 years of age in whom abuse is suspected. the investigation is not complete until follow-up skeletal imaging has been performed in the 11 to 14 days following the initial survey. children below one year of age should also receive a ct brain. neurological imaging in older children will depend on the clinical scenario. ct chest/abdomen is indicated when visceral injury is suspected. in terms of imaging in suspected abuse, espr has adopted the rcr guidelines. in the absence of a history of significant trauma, fractures highly specific for abuse in pre-ambulatory children include rib, metaphyseal and diaphyseal fractures. simple linear skull fractures have a relatively low specificity for abuse. the combination of subdural haemorrhage, retinal haemorrhage and diffuse cerebral oedema/encephalopathy (the so-called, "triad") suggests shaking. whereas the presence of a skull fracture implies impact. visceral injury often results from direct blunt trauma and may therefore be accompanied by anterior and/or costochondral rib fractures. the posterior rib arcs are protected by soft tissue and posterior rib fractures result from compressive/squeezing forces rather than direct trauma. the dating of fractures has a subjective element and it is more important to recognise that fractures are in different stages of healing, rather than to assign a definite age/age range to the injuries. the major differential diagnoses are accidental trauma and osteogenesis imperfecta. if rickets is the cause of the fractures, then radiology and/or biochemistry will show evidence of rickets. a low vitamin d level, in the absence of rachitic features, is not the cause of fractures. close liaison between radiologists and paediatricians is vital and any siblings/children in the same household who are below 2 years of age should also receive a skeletal survey. remember that the presence of injury does not always mean abuse and that the absence of injury does not always exclude abuse. scoliosis may be primitive, structural, particularly during adolescence; during this period, careful follow-up is mandatory, because worsening is frequent. clinical examination with evaluation of a hump (gibbosity) with a scoliometer is mandatory, with also neurological assessment. beside radiography, additional tools have been developed to avoid xray exposure: "spinal mouse", back surface topography systems, ultrasound and other computer-assisted systems. but scoliosis can also be secondary, and imaging is important to find a cause and adapt management. among the etiologies, radiologist must recognize spine malformations, dysplastic and neuromuscular scoliosis. in addition, scoliosis may also be in relation with a primitive lesion, tumor-related or not, whether the initial disease could be within the spinal canal, spinal or paravertebral. imaging studies lies first on pa and lateral full spine x-rays, if possible with a low dose device (flat panel, slot-scanning system), keeping in mind that follow-up with repetitive exposures may be necessary. reproducible measures of different curvatures help to assess the overall static spine and the importance of scoliosis with cobb angle. the assessment of axial rotation can be obtained through 3d simulations, with frontal and axial views (see figure) . morphologic evaluation of the s317 (2017) 47 (suppl 2):s297-s pediatr radiol spine is mandatory: if a secondary scoliosis is suspected, the research to etiology needs to perform ct or mri, depending on the clinical signs and the results of plain x rays evaluation. similarly, these explorations are useful in the preoperative assessment when surgical treatment is necessary. girl scoliosis, pa and lateral views with eos®, 3d simulation, coronal and axial views take home points: clinical evaluation is always the first step in subject with suspected scoliosis radiation burning is quite low with new devices, but repetitive exposures for follow-up need to carefully respect justification for x-rays exposures new tools are available to appreciate 3d spinal deformation and evaluate prognosis and surgical procedures ct and/or mri are useful in presurgical assessment and to look for etiologies in suspected secondary scoliosis malformations of the spine and spinal cord a. rossi; genoa/it summary: embryology and classification: spinal cord development occurs through three consecutive periods: (i) gastrulation (2 nd gestational week): the embryonic disk is converted from a bilaminar into a trilaminar arrangement, with formation of the intervening mesoderm; the notochord is laid down along the midline, identifying the craniocaudal embryonic axis; (ii) primary neurulation (18 th -27 th day): under the induction of the notochord, the midline ectoderm specializes into neural ectoderm. the initially flat neural plate progressively bends and folds until it fuses in the midline to form the neural tube. the primary neural tube produces the uppermost 9/10 of spinal cord; (iii) secondary neurulation (28 th -48 th day): a secondary neural tube is laid down caudad to the termination of the primary neural tube. retrogressive differentiation of the secondary neural tube results in the tip of the conus medullaris and filum terminale. defects in one of these three embryological steps produce spinal dysraphisms, characterized by anomalous differentiation and fusion of dorsal midline structures. spinal dysraphisms may be categorized clinically in two subsets: open and closed spinal dysraphisms. in open spinal dysraphisms (osd) the placode (non-neurulated neural tissue) is exposed to the environment through a cutaneous defect along the child's back. osd include myelomeningocele, myelocele, hemimyelomeningocele and hemimyelocele, and are associated with a chiari ii malformation. myelomeningocele is by far the most common of these forms; the placode protrudes through a posterior defect and is elevated above the skin surface due to concurrent dilatation of the subarachnoid spaces. closed spinal dysraphisms (csd) are covered by intact skin, although cutaneous stigmata usually indicate their presence. two subsets may be identified based on whether a subcutaneous mass is present. csd with tumefaction comprise lipomas with dural defect (lipomyelocele and lipomyelomeningocele), meningocele, and myelocystocele. lipomas with dural defect are more common; they are differentiated from one another based on the position of the cord-lipoma interface, that lies within the spinal canal in lipomyelocele, and outside the spinal canal (ie, into a meningocele) in lipomyelomeningocele. csd without tumefaction comprise complex dysraphic states (ranging from complete dorsal enteric fistula to neurenteric cysts, diastematomyelia, dermal sinuses, caudal agenesis, and spinal segmental dysgenesis), bony spina bifida, tight filum terminale, filar and intradural lipomas, and persisting terminal ventricle. the most complicated forms (complex dysraphic states), including diastematomyelia, caudal regression, and segmental spinal dysgenesis) are related to faulty gastrulation. diastematomyelia (literally, split cord) is caused by failure of midline notochordal integration, resulting into two separate hemineural plates. caudal agenesis and segmental spinal dysgenesis are related to defective notochordal formation, characterized by absence or hypoplasia of a segment of the notochord, in turn resulting into absence or hypoplasia of a corresponding segment of the spinal cord. functional neuroimaging of cns is a fast advancing field with frequent new developments in scanner's hardware, protocols, clinical indications, and post-processing techniques. for radiation safety reasons in the case of children, functional neuroimaging is mostly based on mr techniques especially designed to focus on the assessment of functional tissue characteristics, such as neuronal activity (fmri),, metabolism (mrs) and perfusion (dsc perfusion, asl). pediatric coils with multiple elements, multiple slice excitation, 3d spectroscopy, 3d asl, reduced fov (zoom) and improved motion compensation techniques are important tools available to meet the permanent challenges of pediatric mr functional imaging: fast motionless acquisitions and increased resolution. functional mri (fmri) reveals brain activation during performance of behavioral tasks, based on the blood oxygen level dependent (bold) mri signal, which is modulated by neural activity via a process of neurovascular coupling. for children, especially of younger age unable to follow a task, resting-state fmri (rfmri) can be performed and correlates brain areas with similar spontaneous fluctuations in the bold signalthereby enabling estimates of 'functional connectivity.' main clinical applications of fmri are the delineation of eloquent cortex near a space-occupying lesion and the determination of the "dominant hemisphere" for language. intense research is conducted in the areas of language organization and development, brain plasticity, and neurobehavioral disorders (e.g. adhd). magnetic resonance spectroscopy (mrs) is a noninvasive mr technique, that detects intracellular metabolites, and may provide neuroimaging biomarkers of normal biological and pathological processes or response to a therapeutic intervention. although the main field of application of mrs is the brain tumors, it has also been of particular (2017) 47 (suppl 2):s297-s pediatr radiol usefulness in assessing ischemic or traumatic brain injury and neurometabolic disorders. perfusion mr imaging methods detect signal changes that accompany the passage of a tracer through the cerebrovascular system. a less invasive approach is arterial spin labeling (asl) that uses arterial water as an endogenous tracer to measure cbf and thus it is more suitable for pediatric studies. mr perfusion is applied in the evaluation of brain tumors, neurological diseases and developmental disorders. functional neuroimaging clinical applications are expected to expand greatly in the future due to the increasing availability of their techniques, as well as the continuous advancements in the field of pediatric research. good knowledge of these techniques will become more necessary for an effective clinical practice and will enhance the role of radiology in the healthcare system. functional neuroimaging advanced techniques based on mri allow us to study complex cns processes such as cerebral perfusion (dsc, asl), metabolic activity (mrs) and brain activation (fmri). functional neuroimaging techniques already have significant clinical pediatric applications and assisted by recent advances in mr technology are expected to become even more powerful in the near future. kidney: perfusion, excretion, obstruction k. darge; philadelphia/us the functional imaging of the urinary tract entails the evaluation of the renal perfusion and excretion. in this complex process the sites of the main abnormalities could be pre-renal, renal parenchymal, renal pelvicalyceal or post-renal or even a combination at different sites. functional mr urography (fmru) is an advanced tool that not only allows the exquisite morphological depiction of the urinary tract, but also makes it possible to generate comprehensive functional data. these provide information about the function of the kidney as well as the excretion of urine from the renal parenchyma into the pelvicalyces and ureter. the functional results are mainly divided into two groups: 1. transit timesthese are recorded in minutes and a side comparison gives idea how much time it takes for the contrast to go through the renal parenchymathe longer the more abnormal in general. 2. differential renal functionsthese can be based on the enhanced renal parenchymal volume or the patalk number generated from this area and provides in percentage the split renal function. this presentation will discuss in detail the functional aspect of mr urography and demonstrate its utility in routine pediatric uroradiologic imaging. in chronic childhood lung disease (e.g. cystic fibrosis) global pulmonary function tests (pft) can be normal although lung damage is already present. moreover, in comparison to imaging, pft is challenging in young children. thus, cross-sectional imaging became more important in the past two decades. regarding morphological evaluation, multidetector computed tomography (mdct) serves as the most sensitive and reproducible modality. for functional evaluation perfusion/ventilation scintigraphy remains the reference standard. although the individual radiation burden by a single chest ct has decreased significantly in the past, radiation doses can cumulate considerably when repeated examinations are performed in a longterm follow-up. pulmonary mri exists as an alternative method, especially for paediatric patients. however, standard h+mr sequences do not demonstrate small airway disease due to inherent limitations of low signal and rapid t2* signal decay of lung tissue. for comprehensive diagnosis, functional mri offers the unique possibility to measure regional ventilation and perfusion, and mapping relaxation times and diffusion. focussing on research applications, a variety of methods are available for these purposes. in this context, ventilation imaging using inert fluorinated gas indicates to overcome the limitations of the expensive setting necessary for imaging with hyperpolarized noble gasses. regarding lung perfusion, dynamic contrast-enhanced mri (dce-mri) is the most established method in clinical practice. however, especially in children, techniques that are completely non-invasive and do not require i.v.-contrast agents administration or gas inhalation could be promising to achieve broad acceptance. concerning non-invasive methods, ventilation can be assessed by sequences with ultra-short echo times (ute), perfusion by arterial-spin-labeling (asl) and both by fourier decomposition mri (fd-mri). in conclusion, pulmonary mri offers both, the assessment of morphology and the unique possibility to measure regional ventilation and perfusion, and mapping relaxation times and diffusion. new mr techniques that are completely non-invasive are now available. however, further scientific evaluation is needed. ibd and related arthropaties d. jaramillo; miami, fl/us musculoskeletal diseases affect about 5% of patients with crohn's disease and are the most frequent extra-intestinal manifestation of inflammatory disease. the articular manifestations of inflammatory bowel disease (ibd) are one of the seronegative arthritides, although they have a lower incidence of hla -b27 than other seronegative arthritis such as ankylosing spondylitis. there are manifestations in the joints of the extremities, and findings in the pelvis, especially in the sacro-iliac joints, and spine. involvement of the extremities occurs in about 10% of patients with ibd related arthropathies, are more common with crohn's disease, and can have either manifestations related olygoarticular jia, or can have symmetrical involvement of smaller joints. the axial manifestations include ankylosing spondylitis and sacro-iliitis. sacroilliitis is typically bilateral (figure) and often has radiographic as well as mri abnormalities. enthesitis, tenosynovitis and dactylitis can occur with ibd just as they occur with other arthritides. it is important to differentiate ibd related arthritis from septic arthritis due to extension of an enteric fistula. deceased bone mineral density is a common finding in inflammatory disease. it occurs as a combination of malabsorption of vitamin d due to intestinal involvement and the effects of therapy, particularly corticosteroids. insufficiency fractures of the spine, sacrum and extremities can mimic the symptoms of arthritis. finally, ibd can be associated with chronic non-bacterial osteomyelitis, although this association is relatively rare. this review will illustrate several of the skeletal manifestations of ibd, focusing on the arthropathies. juvenile idiopathic arthritis (jia) can be defined as an arthritis of unknown cause occuring in children younger than 16 years and of at least 6 weeks duration. juvenile spondyloarthritis (jspa) is a subset of jia and is characterized by enthesitis (inflammation at the attachment of tendons, ligaments and the joint capsule), arthritis and an increased risk of axial disease. there is also a strong association with human leukocyte antigen b27. jspa accounts for approximately 10-15% of juveniles arthritis cases in europe and is the most common form of juvenile arthritis in asia. the condition is associated with significant long-term morbidity, high health-care costs and poorer outcomes compared with other forms of juvenile arthritis as well as its adult counterpart. up to 40% of patients continue to be at risk of developing ankylosing spondylitis (as) during the disease course. recognizing spondyloarthritis (spa) in children is challenging, particularly early in the course of disease, as the signs and symptoms at disease onset differ from those seen in adults. jspa typically presents with hip and lower limb arthritis in conjunction with enthesitis. inflammatory back pain as a presenting symptom is less common. as a consequence, jspa may be missed or confused with other juvenile arthritides and patients often experience prolonged delays in diagnosis. currently there is no single diagnostic or classification system that is representative of the jspa population. according to the international league of associations for rheumatology (ilar) classification system, most childhood spa's are classified as enthesitis-related arthritis (era), psoriatic arthritis or undifferentiated arthritis. recent studies indicate that there are two clinical phenotypes of era: those with early axial disease often associated with hip arthritis in addition to peripheral arthritis; and those who follow a more peripheral disease course with arthritis and enthesitis and do not develop axial disease. the ilar classification system places patients with both axial and peripheral involvement into the era subtype, and does not specifically address children who meet the criteria for as. the correct approach to the classification of era is uncertain, and this issue is confusing to both pediatric and adult rheumatologists. unlike other categories of juvenile arthritis, jspa affects boys more often than girls, and peak age of onset is early adolescence. enthesitis is a defining characteristic of jspa. it is more common and affects more sites in the paediatric population compared with the adult one. the most commonly tender entheses are the insertions of the patellar ligament at the inferior patella, plantar fascia at the calcaneus, and the achilles tendon. arthritis in jspa is typically asymmetrical, oligoarticular (< 5 joints) and involves predominantely the weightbearing joints. isolated hip joint arthritis may be the presenting feature and predicts early axial disease. involvement of the small toe joints is common in jspa but rare in other forms of jia. midfoot arthritis and tarsitis (inflammation of the intertarsal bones, overlying tendons, entheses and soft tissue) is highly suggestive of spondyloarthritis. in adults, inflammatory back pain typically heralds the onset of sacroiliitis, whereas children seldom present with symptoms of axial disease. however, according to several studies, sacroilitis can be asymptomatic in jspa and only detectable by imaging. other axial manifestations in jspa are inflammation of the lumbar apophyseal joints and interspinous ligaments, corner lesions of the spine and other sites of axial enthesitis-osteitis including the various ligamentous and muscular attachments of the pelvis. extraarticular manifestations of jspa are highly associated with axial disease and include acute anterior uveitis, bowel inflammation, psoriasis, and cardiac disease. clinical diagnosis of jspa can be difficult and the role of imaging may be more critical than in adult disease. the major goal of imaging in jspa is to identify children with early signs of axial disease, as this group is at the greatest risk for progression to as. the presence of axial disease in spa has also major implications for treatment decisions, since traditional firstline therapies appear to have minimal effectiveness in the management of axial inflammation. in addition, recent studies in adults suggest that earlier initiation of biologic agents (anti-tnfs) may slow radiographic progression. x-rays are not sensitive to acute inflammatory changes and will only show advanced disease in the sacroiliac joints. for these reasons plain radiographs are not useful in children or adolescents. ultrasound is a non-invasive, non-ionizing and relatively inexpensive technique that can be performed in a clinical setting. it is emerging as a valid diagnostic tool in spa and can be used to visualize peripheral synovitis, tendonitis and enthesitis, but the method is heavily operator-dependent and there does not yet exist a clear definition for the diagnosis and grading of enthesitis in children. secondary changes (calcifications, enthesophytes) have been observed much less in children compared with adults. there is a need for better consensus on abnormal ultrasonographic findings that define enthesitis lesions and standardization of methods. magnetic resonance imaging (mri) is a radiation free and sensitive imaging modality for detection of synovitis as well as cartilage and bone destruction. mri of the sacroiliac joints is increasingly obtained for early detection of inflammatory changes, as it shows active inflammatory (bone marrow edema, osteitis, enthesitis and capsulitis) and structural (erosions, subchondral sclerosis, subchondral fatty change and bony ankylosis) lesions of sacroiliitis long before radiographic changes become evident. in adults, mri has become the gold standard imaging modality for detecting arthritis and enthesitis. consensus definitions of lesions indicating pathology on mri are now incorporated into diagnostic criteria for adult with spa. in children and adolescents there is no gold standard mri technique and it is therefore not clearly defined whether changes s320 (2017) 47 (suppl 2):s297-s pediatr radiol seen in the sacroiliac joints are pathologic or part of normal maturation in the growing skeleton. the use of contrast enhanced imaging for the detection of active sacroiliitis on mri in jspa is a major controversy. synovial enhancement can be detected without accompanying bone marrow edema in children, and it can be argued that contrast should be administered in order not to miss the diagnosis. some authors argue that contrast administration does not change or add substantially to the mri findings made on non-enhanced scans. certainly, given the risks associated with gadolinium administration, contrast should be used with caution. perhaps the use of contrast agents should be limited to selected cases when high stir signal in the joint is the only finding in order to confirm the presence of synovitis, and when the differential diagnosis includes etiologies such as infection or tumor. the development of new mri techniques has made it possible to perform whole body mri scans (wbmri) that allow assessment of the full range of affected entheses and joints. there is limited data on the utility of wbmri in the pediatric population. it is worth noting that edema-like changes seen in the marrow of healthy children is an important potential pitfall to consider during interpretation and further studies are required in order to establish specific reference standards for mri of the pediatric skeleton. diffusion-weighted imaging (dwi) offers a new approach to detect inflammation. inflammation produces an increase in the apparent diffusion coefficient (adc) of water molecules in affected tissues. several studies in adults and a few recent studies in children have demonstrated that adc is elevated in sacroilitis versus controls and that diffusion scores correlates well with stir images. dwi is promising as a potential biomarker of disease activity in jia and presents a novel approach to contrast-free imaging of synovitis. however, further studies are needed before it can be implemented in clinical practice. jspa is distinct from adult spa and manifests more frequently as peripheral arthritis and enthesitis. symptoms involving the spine and sacroiliac joints often occures later in this population. clinical diagnosis of jspa can be difficult, and imaging therefore plays an important role in the diagnostic workup of disease. identifying early signs of axial disease has major implications for treatment decisions and mri of the sacroiliac joints is increasingly obtained for early detection of inflammatory changes. however, mri criteria for sacroilitis in children are lacking. a major controversy in imaging of sacroilitis in jspa is the use of contrast, as children can have sacroilitis without accompanying bone-marrow edema. dwi presents a novel approach to contrast-free imaging of synovitis but further studies are needed before it can be used in clinical practice.wbmri has been shown to be more sensitive than clinical examination in the assessment of both disease activity and extent, but there is limited data on wbmri in children. normal variants in the growing skeleton may mimic pathologic changes and potentially cause overdiagnosing and -staging of disease. hence, there is an urgent need to establish specific reference standards for mri of the pediatric skeleton and to develop a gold standard mri technique for the axial skeleton in children and adolescents. juvenile idiopathic arthritis o. olsen; london/uk summary: juvenile idiopathic arthritis (jia) is common (about 1:1,000 children). diagnosis and classification are based on clinical criteria. these criteria are in flux depending on 1) contemporaneous knowledge about aetiology and 2) available treatment options. radiology has currently no role in establishing the diagnosis. the clinical classification rests on whether the child has few joints affected (oligo jia), many joints (poly jia), has a condition similar to adult spondyloarthritis (entesitis-related arthritis) or other clinical presentations (systemic-onset jia, psoriatic jia, etc). radiology can potentially assess expressions of jia, such as synovitis, tenosynovitis, systemic manifestations and permanent damage caused by inflammation. it is therefore thought to play a part in gauging the disease activity. the clinical care aims at optimising the child's everyday function, reducing acute symptoms (pain, swelling, joint restriction), allowing normal growth, minimising long-term sequelae (joint deformity) and minimising adverse effects of medical treatment. medical treatment in jia is systemic (immuno-modulation) and local (steroid injection to joints and tendon sheaths). both modes of therapy may to some degree be guided by imaging. however, there currently is no evidence that any form of whole-body imaging is efficacious for guiding treatment. this means that, in principle, indication for imaging should be 1) specific clinical questions, e.g. uncertainty regarding active inflammation at specific sites, or 2) a high pre-test likelihood of inflammation at a site which is difficult to assess clinically and where imaging offers reasonable accuracy. one example of the latter are the temporo-mandibular joints where destruction is frequently seen at an early stage, often without prior symptomatic warning. there is one fundamental challenge for imaging research in jia: what is the reference standard? for lack of anything better, a standardised clinical examination is often used as 'ground truth'. the dilemma is obvious. if clinical examination is reliable and accurate, then why bother with imaging? but we think imaging offers an improvement, then we cannot use an inferior method to set the standard. this problem is not unique to jia. as is often the case, radiology in jia is all about: knowing your clinicians (i.e. the pretest likelihood for disease) being technically eloquent (e.g. using high-resolution us probes, not delaying post-contrast mri acquisitions) knowing what is normal (e.g. normal undulations in the articular surface, focal bone marrow signal variation) not being dogmatic about individual observations or measurements interpreting your findings in a clinical context the lecture will demonstrate similarities and differences among joints and modalities in children with variable-severity jia. the following points will be made: focal areas in the bone marrow with high signal (t2) and corresponding enhancement are often seen in healthy children. in isolation, these do not signify active inflammation. active synovitis in children often is not associated with (much) effusion the combination of synovial thickening with hyperaemia (us)/abnormal contrast enhancement (mri) and surrounding softtissue swelling suggests active inflammation, however there is (yet) no established system for quantifying hyperaemia/enhancement focal pits in the carpal bones do not represent erosions unless there is an associated cartilage defect radiographs are useful for detection of destructive abnormality in mri, scan fairly soon after injecting contrast. gadolinium physiologically leaks into the synovial fluid making it difficult to delineate the synovium a few differential diagnoses to keep in mind when there is mass-like swelling within or adjacent to a joint: vascular and neoplastic lesion, pigmented villonodular synovitis, synovial chondromatosis, lipoma arborescens. synovial inflammation is not always primary. even when there is an established diagnosis of jia, do consider that it may be secondary to biomechanical abnormality (erosion, osteochondral lesion, deformity). focal areas in the bone marrow with high signal (t2) and corresponding enhancement are often seen in healthy children. in isolation, these do not signify active inflammation. active synovitis in children often is not associated with (much) effusion the combination of synovial thickening with hyperaemia (us)/abnormal contrast enhancement (mri) and surrounding soft-tissue swelling suggests active inflammation, however there is (yet) no established system for quantifying hyperaemia/enhancement focal pits in the carpal s321 (2017) 47 (suppl 2):s297-s pediatr radiol bones do not represent erosions unless there is an associated cartilage defect radiographs are useful for detection of destructive abnormality in mri, scan fairly soon after injecting contrast. gadolinium physiologically leaks into the synovial fluid making it difficult to delineate the synovium pulmonary manifestation of connective tissue disorders c. m. owens; london/uk summary: connective tissue diseases are an important cause of morbidity and mortality in children with very varied presentations. nomenclature is confusing and a more appropriate descriptive term would be "multisystem inflammatory disorder +/-autoimmunity". it is important for the radiologist to be aware of the protean radiological appearances and clinical manifestations. take home points: different patterns of diffuse lung disease (eg, desquamative interstitial pneumonia, non specific interstitial pneumonia, lymphocytic interstitial pneumonia, organising pneumonia, diffuse alveolar damage) may be present in several forms of collagen vascular disease, (and indeed other rheumatological conditions such as jia) including scleroderma, systemic lupus erythematosis, juvenile dermato and polymyositis, sjogren's syndrome and mixed connective tissue disease. these will be discussed in detail with illustrations for thin section high resolution ct with histopathological correlation. the clinical presentation, prognosis and response to therapy vary depending on the histological pattern of diffuse lung disease, as well as on the underlying collagen vascular disease. whole body imaging in children: sonography, ct, mri, nuclear medicine -what and when? r. a. nievelstein; utrecht/nl there are several (benign and malignant) disease processes in children that frequently involve more than one organ system or body region. diagnostic imaging of children with such multifocal or multisystem diseases has been quite challenging, often requiring a combination of different imaging techniques for a whole body coverage. the recent technical developments in computed tomography (ct), magnetic resonance imaging (mri) and nuclear medicine (nm) have changed the role of imaging in these children revolutionary. in the past, imaging techniques have been mainly used as a tool to detect the cause of illness and to assess the extent of disease spread before, during and after therapy (i.e. structural imaging). but nowadays, it has also become possible to use imaging techniques to gain information on the biological behavior of diseases before and during therapy (i.e. functional imaging). plain radiography, ultrasonography (us) and computed tomography (ct) have been the structural imaging techniques of choice for many decades, more recently supplemented by functional imaging techniques like single-photon emission tomography (spect) and positron emission tomography (pet). a major disadvantage of most of these techniques is the use of ionizing radiation, which may be associated with induction of second cancers later during life. this small but not negligible health risk is of particular concern in children as their tissues are more radiosensitive than adults and they have more years ahead in which cancerous changes might occur. that is why there is an increasing interest in the use of alternative imaging techniques that do not use ionizing radiation. with mri it is nowadays possible to acquire images with a high spatial resolution and excellent soft tissue contrast throughout the body, which makes it an ideal radiation-free tool for the detection of pathology, especially in soft tissue, parenchymal and bone marrow locations. moreover, recent technological advances have resulted in fast diagnostic sequences for whole-body mr imaging (wb-mri), including functional techniques such as diffusion weighted imaging (dwi). as a result, wb-mri has become a clinically feasible imaging modality for diagnosis and follow-up of multifocal and multisystem diseases in children. in this scope, the recent development of integrated pet/mri systems is very interesting, combining the superior structural imaging of mri with the functional (molecular) information of both imaging techniques while decreasing the radiation dose. traditionally, whole body imaging techniques have been mainly used for oncological indications, such as staging of malignancies, and monitoring of the effectiveness of therapy. however, whole-body imaging techniques are increasingly used for the diagnostic imaging of other benign multisystem diseases and indications, including chronic recurrent multifocal osteomyelitis (crmo), rheumatological diseases, neuromuscular diseases, neurofibromatosis type 1, generalized vascular malformations, multifocal osteonecrosis after intensive chemotherapy, fever of unknown origin, and post-mortem imaging. finally, these imaging techniques may be used for the screening of children with a cancer predisposition syndrome. during this lecture, imaging protocols and indications of the different whole body imaging techniques will be discussed with a focus on their clinical application in children with benign and malignant multifocal or multisystem diseases. (2017) 47 (suppl 2):s297-s pediatr radiol appearances is important for any radiologist involved in child imaging, because we have an important role in characterizing the lesions and guiding purposeful and minimally invasive but successful diagnostic procedures. most head and neck masses in children are benign and have an inflammatory, infective, vascular or congenital cause (cf. special presentation on vascular malformations). malignant lesions are less common, however, early diagnosis is paramount as many of these cancers are readily treatable and often curable. differential diagnosis is guided by patient age, clinical presentation, tumour localisation, and imaging characteristics. while some masses such as (epi-)dermoids, fibromatosis colli and swollen lymph nodes including atypical mycobacterial infections (mott) may be readily diagnosed by clinical inspection und ultrasound, others present special diagnostic challenges. fibromatosis, for example, is a benign lesion with an often complex and potentially destructive local spread. some malignant lesions tend to be localised such as the embryonal rhabdomyosarcoma, while others may be part of a systemic disease such as lymphoma and langerhans cell histiocytosis (lchc). in case of a suspected malignancy, patients should be referred to a specialized centre which will be able to provide the full spectrum of multidisciplinary evaluation and treatment according to the guidelines of an international oncology study group. this is also important for image guided or surgical biopsies as long term outcome and survival of many of the young patients are directly associated with these initial diagnostic and therapeutic strategies. with its excellent spatial resolution in the near field, ultrasound is the method of choice for all superficial masses. an experienced paediatric radiologist will be able to identify most of the benign lesions and in other cases will be able to guide further diagnostic decisions. tumours in the midline require thorough workup to exclude an encephalocele or a dermal sinus with connection to the intracranial space. high resolution mri is required if such an extension cannot be ruled out by ultrasound or if a tumour is larger than the transducer's scan area. soft tissue tumours in the deeper parts of face and neck as well as tumours of osseous origin are also best delineated by mri. in lesions adjacent to the skull base contrast enhanced and fat saturated mr images with high spatial resolution are of utmost importance to completely depict the tumour's extension through the foramina and along the meninges (fig. 1) . ct can provide additional information on the involvement of osseous structures. embryonal rhabdomyosarcoma. high resolution mri with fat saturation after contrast injection depicts the tumour's extension through the foramen ovale (long arrow) and along the meninges (short arrows). skull base and face lesions are less frequent in children than in adults. symptoms may be subtle or unspecific. depending on their localization, clinical findings may be common (nasal obstruction, otitis…) or more disturbing (cranial nerves palsies, exophthalmos, vision loss …). clinical history and physical examination findings are important to reduce the spectrum of differential diagnosis, but imaging data are the key features to determine the nature of these lesions. ct and mri play an important role in diagnosis, treatment survey and surgery planning of skull base and face lesions. skull base and face bone lesions are either intrinsic lesions of the bone or secondary to soft-tissue tumors or pseudo tumors invasion. this lecture will focus on bone intrinsic lesions, and include soft-tissue and pseudo tumors only as differential diagnoses. computed tomography plays the role for skull base and face of plain radiograph for long bones. therefore, the same semiology may be used to determine if the lesion is slowly or rapidly growing, aggressive or looks benign. helical ct allows reconstructions with both soft-tissue and bone algorithms as well as multiplanar reformations. it gives a good visualization of the anatomy of the skull base and allows a good depiction of the bone architecture. ct is first used for the initial work up of the disease but also for surgery and therapeutic planning (endoscopic sinus surgery with navigation). however, ct analysis may be challenging in children due to growth changes: normal process of pneumatization according to age, sutures not yet fused has to be recognized. some variations in pneumatization must not be mistaken for pathology: asymmetrical pneumatization of the petrous apex and arrested pneumatization of the sphenoid mimicking intraosseous lesion are the most common. both ct and mr imaging are complementary: most preferably, contrast-enhanced mr is associated with non-contrast high resolution ct. mri allows a good delineation of bone involvement of skull base lesion due to bone marrow changes, whether ct can fail to detect subtle extension within the bone. in addition to t1 and t2 weighted sequences, the use of specific sequences and/or techniques such as fat-saturation, diffusion, dynamic-contrast-enhanced sequences, and mr angiography helps to characterize the lesions. t1 spin echo sequence is mandatory to appreciate bone marrow infiltration in adults and older children. but when red bone marrow has not yet be replaced by fatty bone marrow, in young children, this can be challenging. it is useful to know the bone marrow fatty conversion of the skull base chronology. cranial mr can also be associated to whole body mr to look for multifocal or metastatic disease. epidemiologic data concerning bone tumors of the skull base are scarce due the rarity of these lesions. they can be classified according to their location within anterior, middle or posterior cranial fossa or classified according to their origin: osteogenic (osteoblastoma, osteoma, osteosarcoma...), chondrogenic (chondroma, chondrosarcoma), fibrous ( fibrous dysplasia, fibro-s323 (2017) 47 (suppl 2):s297-s pediatr radiol osseous lesions..), notochord (chordoma), hematopoietic (leukemia, histiocytosis ), vascular (hemangioma), neuro ectodermic ( ewing sarcoma) or unknown origin (aneurysmal cyst, giant cell tumor). the aim of this presentation is to draw attention to skull base growth changes that can mimic pathology and to describe the imaging specificities of the most common bone tumors of the skull base and face in children. because conflicted nomenclature can cause confusion, accurate diagnosis and classification of these anomalies is important for proper clinical evaluation and management. many of these patients require multidisciplinary care, consequently the usage of a correct nomenclature across all disciplines is a sine qua non. the international society for the study of vascular anomalies (issva) classification, updated in 2014, offers a comprehensive classification accepted by many subspecialities. this approach/ classification has facilitated correct communication for all medical subspecialties involved in the care of these complex vascular anomalies. pediatric radiologists play a critical role in evaluating these patients since the majority present during childhood. in this presentation, we present a state of the art mri imaging protocol with exemplary cases of the most common types of vascular anomalies in the pediatric trunk and extremities, using the current issva classification. in addition, we discuss the common syndromes associated with vascular anomalies such as klippel-trenaunay and lumbar syndrome. genetic skeletal disorders (gsd's) are a heterogeneous group of syndromes characterized by an intrinsic abnormality in growth and (re-)modeling of cartilage and bone. a large sub-group of gsd's may have additional involvement of other structures/organs beside the skeleton, such as the central nervous system (cns). cns abnormalities have an important role in long-term prognosis of children with gsd's and should consequently not be missed. sensitive and specific identification of cns lesions while evaluating a child with a gsd requires a detailed knowledge of the possible associated cns abnormalities. here, we will present and discuss a pattern-recognition approach for identifying relevant neuroimaging findings in gsd's guided by the obvious skeletal manifestations of gsd. in particular, we will discuss which cns findings should be ruled out for the various gsd. to facilitate this diagnostic approach the multiple gsd are classified based on the pattern of skeletal involvement (1. abnormal metaphysis or epiphysis, 2) abnormal size/number of bones, 3) abnormal shape of bones and joints, and 4) abnormal dynamic or structural changes). skeletal involvement is defined in accordance with online mendelian inheritance in man. the spectrum of co-existing cns involvement is extracted from an extensive literature search. selected examples will be shown based on prevalence of the diseases and significance of the cns involvement. cns involvement is common in gsd's. a wide spectrum of morphological abnormalities is associated with gsd's. early diagnosis of cns involvement is important in the management of children with gsd's. this pattern-recognition approach aims to assist and guide physicians in the diagnostic work-up of cns involvement in children with gsd's and their management. not infrequently the correct radiological differentiation of skeletal and/or central nervous system findings secondary to non-accidental injury versus inherited genetic and/or metabolic disorders may be challenging. imaging findings may be non-specific, can result in incorrect diagnosis and subsequently inadequate patient management or initiation of faulty treatment. the diagnostic work-up of children suspected of non-accidental injury or genetic/metabolic disorders requires a multi-disciplinary approach involving many key players including physicans of various disciplines, nurses, psychologists, social workers and many more. a proper and detailed medical history and physical examination of the patient, collection of the relevant family history, a metabolic and genetic work up, a detailed interview of care givers, friends and family are essential for the correct and comprehensive evaluation of imaging findings. in the current session, various exemplary and possibly confusing cases will be interactively discussed with the audience by a panel of experts (susan blaser, thierry a.g.m. huisman and andrea superti-furga). goal is to offer a case based approach to challenging patients with discussion of the best diagnostic approach including differential considerations. the zikv is transmitted mainly by the bite of female aedes aegypti and aedes albopictus mosquitoes. other forms of transmission, including through sexual intercourse, blood transfusion, and neonatal, are currently under evaluation, although more elements are still needed to assess the real importance of these transmission routes 4 . the course of the zikv infection is self-limited. so far, no specific symptoms have been attributed to the disease, and a wide variety of manifestations ranging from absent to mild symptoms (in 25% of cases) have been described. when symptoms are present, they may lead to a misdiagnosis of other bacterial and viral infections, especially other arboviroses in endemic areas. the most frequently reported symptoms are mild fever, cutaneous rash, fatigue, arthralgia/myalgia, and conjunctivitis. dizziness, malaise, edema of the extremities, anorexia, retro orbital pain, photophobia, gastrointestinal disorders, sore throat, cough, sweating, and lymphadenopathy have also been reported. infection by the zikv in adults may be associated with autoimmune complications such as guillain-barré syndrome 9 . the laboratory diagnosis of zikv infection is based on the demonstration of the virus in the urine and blood using real-time reverse transcription polymerase chain reaction (rt-pcr). the main limitation of this diagnostic method is a false-negative result after the viremia is resolved. the serological diagnosis of the disease is limited due to cross-reactivity of the zikv with other viruses of the flavivirus genus, especially those causing dengue and chikungunya. physicians should be aware of this fact when the diagnosis of zikv infection relies solely on serological results. the diagnosis is also possible by igm measurement in serum, urine, or cerebrospinal fluid using enzyme-linked immunosorbent assay (elisa) 9 . the prevention against zikv infection is similar to that of other arboviroses, including vector control and mosquito bite prevention. the first major zikv epidemics were reported in the french polynesia in 2013 and 2015. at that time, some neurological changes were observed in neonates of infected pregnant women but were not associated with a maternal-fetal transmission of the virus. the growing increase in the number of cases and the severity of the infection specific to this subpopulation then led to the evidence of a congenital disease 9 . in brazil, the situation became alarming with the report of a high number of infected individuals in the second half of 2015 2, 5 . the brazilian ministry of health attributed to congenital zikv infection the 20-fold increase in cases of neonatal microcephaly in the northeastern part of the country, particularly in the state of pernambuco. this led the world health organization (who) to declare the zikv infection a "public health emergency of international concern" in february 2016 6 . the main challenge for radiologists practicing in regions of endemic zikv infection is to become familiarized with findings of congenital zikv infection in perinatal imaging studies; this is particularly important for the prenatal screening of pregnant women 3, 6 . the diagnosis of zikv infection in the fetus by neuroimaging is based on prenatal ultrasound (us), especially in the third trimester, and complemented with magnetic resonance imaging (mri). postnatal imaging was obtained by transfontanellar us, ct or mri. the main imaging findings on ct are microcephaly, an exuberant external occipital protuberance, rectification of the frontonasal angle, and a redundant scalp skin. three-dimensional (3d) reconstruction of al skull permits a better evaluation of these findings and enhances the parents' understanding of the disease. moreover, ct scan data may yield a 3d virtual physical model that can maybe obtained from ct scan data and printed onto using thermoplastic acrylonitrile butadiene styrene 7 . the aim of this study was to describe the perinatal imaging findings in cases of congenital zikv infection. we studied 18 mothers diagnosed with zikv infection from october 2015 to november 2016. they had all presented a maculopapular rash and fever during the first or second trimester of pregnancy, and their neonates presented neurological defects that were attributed to intrauterine transmission of the zikv. the maternal diagnosis of zikv infection was confirmed by serology (n=4) or rt-pcr (n=14). all patients were torch (toxoplasma, rubella, cytomegalovirus, herpes simplex) negative. prenatal us was performed every 3 weeks after the first imaging findings, and fetal mri was obtained in all cases. microcephaly was considered present when the infant's head circumference was two standard deviations below the mean value for age and sex or below the second percentile. postnatal imaging follow-up was obtained in all cases by transfontanellar us, ct or mri. we found several cns malformations, including lissencephaly, pachygyria and/or polymicrogyria, cerebral atrophy (panel 1), enlarged cisterna magna with abnormalities of the corpus callosum, ventriculomegaly, brainstem hypoplasia, malformation of the cortical development, and cortical and/or periventricular calcifications mainly in the junction between the cortical and subcortical white matter (panel 2). the skull of the infants had a collapsed appearance, with overlapping sutures and redundant skinfolds (panel 3). craniofacial disproportion was easily identifiable, and arthrogryposis was identified in one case. similar neurological findings were observed in the infected patients and seemed to differ from findings of other infectious diseases. the finding of microcephaly in neonates with congenital zikv infection seems to be only the tip of the iceberg, as several cns malformations have been identified in connection with the disease. in brazil, a spectrum of imaging findings associated with congenital zikv infection has been observed. such findings are useful in helping radiologists to identify suspected cases of the disease. panel 1: prenatal ultrasound (37 weeks) shows calcifications (arrows) and microcephaly. axial and sagittal t2 shows relative smoothness of the brain surface (arrows) and assymmetric colpocephaly. panel 2:ax t1-wi multiple cortical-subcortical fronto-parietal hyperintense foci (arrows) and markedly hypointense on swi. sagittal t2: dysgenesis of the corpus callosum, with dilation of the posterior horns of the lateral ventricles (colpocephaly). pre-and postnatal imaging in zika virus: where are we? early insights into zika's microcephaly physiopathology, from the epicenter of the outbreak: a case for teratogenic apoptosis of central nervous system. p. jungmann; recife/br early insights into zika's microcephaly physiopathology, from the epicenter of the outbreak: a case for teratogenic apoptosis of central nervous system. in mid-october 2015, intense interaction among surgical pathology and fetal medicine specialists from university of pernambuco was only focused on the dramatic and non explained ultrasonographic (us) findings and hopelessness due to lack of explanations on the odd us discoveries on the first gestational cases of zika's microcephaly. this is the field of our history of a physiopathological hypothesis on zika virus (zikv) related microcephaly when it first struck pernambuco state (pe), northeast brazil, the place that has been at the front line of the global response to the microcephaly and responsible for a large amount of data from affected children. the outbreak onset came with a sudden increase in microcephalic newborns being reported in pe state from august 2015 (panel, fig. 1) . zikv was previously thought to cause a relatively mild disease, but was recently accepted to lead to severe and diverse neurologic conditions in s329 (2017) 47 (suppl 2):s297-s pediatr radiol some children born from infected mothers and in adults 1 . the scientific community is actively trying to uncover the extent of these disorders but little has been reported on the early days of the outbreak when doctors were approaching the unknown. while evidence that zikv is related to microcephaly in newborns is accumulating, the mechanisms of how the virus affects the fetus is still uncertain. in the outbreak onset we had to face daunting challenges to search the cause of microcephaly and the emotional toll on the families. we took a very early approach from 9 microcephalic fetuses on gestation and 46 microcephalic babies on clinical follow-up from different pe areas, evaluated between 15 october and 17 december 2015 in oswaldo cruz hospital, to propose the early physiopathologic hypothesis that, a viral-related brain developmental disruption could be the basic neuropathogenesis in zikv babies instead of a direct injurious process due to viral insult followed by active inflammation. the eight pregnant women were all in the 3 rd gestational trimester and had had normal us follow-ups till week 20 th . crucially, we were facing a temporal-geographic association of cases presenting an unanticipated pattern of us alterations. because of their late alarming findings they were re-examined and the us scans revealed sudden encephalic alterations after 28 th gestational week. such devastating us clustering images were not seen here before, but are now considered as part of the "congenital zika syndrome". we observed late appearing severe dysmorphic encephalic changes in 9 out of 9 fetuses, including small skull, small brain, sub arachnoidal space enlargement, ventricular dilation, brain calcifications of varied shape and distribution, inclined frontal bone, progressive decline of head growth potential, early fontanels closure and redundant scalp (panel, figs. 2a, 2b). we had no clues on the causes and mechanisms responsible for this phenotype of severe alterations. thus, we had no explanation to offer to patients, in particular, or to the medical community. both as physicians and human beings, we were committed straightaway to continue the study of these victims of an unknown medical tragedy, engaging our expertise in fetal imaging and immunopathology. from beginning october, the first microcephalic babies were referred to the upe pediatric infectology service for initial investigation. strikingly, the newborns exhibited "healthy" appearance, excluded the microcephaly itself and motor sequels. we then looked for csf analysis of the microcephalic babies. for that, we obtained from dr. patricia travassos, a csf specialist at upe, a cohort of 120 csf samples that have been studied for signs of meningitis or encephalitis. about 87% (42 cases) of the csf analyzed looked normal for any signs of central nervous system ongoing inflammatory responses (panel, fig. 3 ). the babies had been examined by outpatient clinic dr angela rocha, from the upe hospital infectology reference center that have stated that although small, the babies were near to full term gestation (36-39 weeks gestation), had good apgar scores and variable degrees of microcephaly and neurologic impairments, i.e. contractures, spasms, irritability and in some retinal macular atrophy. during the follow-up, the babies were cared at home, breastfeeding, gaining weight and having routine vaccines. none of them expressed signs of ongoing inflammatory reaction in the cns (panel, fig. 4 ) or alterations on peripheral blood count and other routine laboratory tests up to 3 months of age. despite the striking neurological phenotype, 80% of the babies were negative for torch agents, no deaths were recorded. furthermore only in january 2016, the first evidence associating zikv to microcephaly from rt-pcr test on amniotic fluid was reported 2 . astonishingly, a particular kind of physiopathological process linked to fetal brain development was arising without clinical manifestation of inflammatory reactions or necrotic processes in these babies. unfortunately, no necroscopic samples of affected brain tissues were available to us to monitor the presence of putative neural dysgenesis and the very nature of brain calcifications background offering histological support for our hypothesis. nevertheless, with this restricted dataset we hypothesized that whatever the etiologic agent involved in these cases, its physiopathologic mechanism must trigger the cellular death programthe apoptotic process -at a particular development window on the cns, assuming clinically that the agent was not encephaalitogenic but silently tertogenic. if not, the clinical outcome of affected babies would not be so mild as far as signs of inflammation on cns was concerned. consequently, the inflammation-free clinical status of patients suggested us that a massive enhanced apoptotic cell death during the window of telencephalic expansion was the most probable physiopathologic process operating this microcephaly phenotype, with no direct direct lytic brain lesion or significant necrosis due to usual injury. furthermore, knowles and penn 3 stated that this window is very active to select the "fittest" neural cells by a constitutive apoptotic pathway. we so hypothetized that during this developmental time window, the "fit or not fit" status of the rapid, transient amplifying neural progenitors cells facing zikv, would heavily shift the selective process toward the self-elimination of virusbearing cells through apoptotic pathways. thus, zikv-enhanced constitutive apoptotic mechanisms would lead a massive loss of developing telencephalic neuronal precursors and, consequently, provoking losses of dividing cells and the arrest of further brain development. this could be particularly inferred by the absence of the characteristic morphology of late stages structures of neocortex, according with our us images of zikv microcephalics in gestation. similar processo could also be inferred to neurocrest derivatives as deformities in the viscercrany always accompany the cephalic malformation. our initial understandings based on clinical examination on the field, when no specific laboratory test, necroscopic data or experimental evidence on the disease causality were available, conducted our physiopathological approach to the "apoptosis hypothesis" for zika microcephaly that is now gaining strong support. in february, mlarkar et al 4 showed clear connection between zikv and microcephaly, presenting cns histopathologic analyses, revealing remnants of neural germinative matrix, intense gliosis, alterations in cortical ribbon, calcifications in gray and white matters without associated necrosis, encephalitis or meningitis and the presence of the virus, further supporting neurodevelopmental arrest. similar results were showed by driggers et al 5 . the ct scans from microcephalic babies from hazin at al 6 , have added details of brain development arrest with no radiological signs of brain destruction or active inflammation. finally, experimental models have provided a body of evidence f or neuroprogenitors permissiveness to zikv and viralinduced apoptotic process. tang et al 7 demonstrated by icq that the zikv infection of cortical neural progenitors attenuates their growth and increases caspase-3 activation, calling for an apoptotic process. this finding was corroborated by the up regulations of caspase-3 genes by rna sequencing. nowakowsky et al 8 demonstrated that zikv may hijack axl protein as an entryway to infection. interestingly, axl is highly abundant on the surface of neural stem cells but not on differentiated neurons in the developing brain. recently, cugola et al 9 demonstrated that zikv was able to cause cns congenital brain dysgenesis upon vertical transmission in mice. in parallel, human brain organoids infected by zikv show a reduction of proliferative zones and disrupted cortical layers, so targeting cortical progenitors and inducing apoptotic cell death with impaired development. for babies born with zikv-related microcephaly, the many expected consequences besides the evolving congenital neurosequels, are the unanticipated pattern of persistence of zikv in cns host cells, unsafe maintenance of neuron genome stability on remaining arrested populations, implying risks for brain tumors, risks for impaired adult type neuron wiring and neuron survival in an affected neuronal circuitry. in brief, evolve life with a wide vulnerable brain. the outbreak of zikv in the americas will eventually decline as herd immunity increases, but the world remains at risk of further waves of infection in affected countries and spread into new territories 1 . while experimental studies will be carried out to fully understand the pathophysiology of zikv infection in the developing fetus, our findings provide a coherent and testable physiopathological hypothesis for cns teratogenic phenotype linked to zikv congenital infection, which may be critical for the clinical care of pregnant mothers and their babies before and after birth. take home points: fetal dysmorphisms detected by ultrassonographic and mri images in congenital zika syndrome are late findings, usually after the 20 th gestational week and requires acurate analyses. clinically, zika's virus microcephaly is an infectiuos congenital condition that is not encephalitogenic but primarily teratogenic on the nervous system. the most important process leading to zika's virus microcephaly is pathologically induced apoptosis in telencephalic neuroprecursosrs cells and neurocrest precurssors cells. viral induced autophagy and low antiviral responses during the fetal period are linked to zika virus persistence in the central nervous system of affected new borns and babies a. vossough; philadelphia/us summary: susceptibility-weighted imaging (swi) has proven to be a valuable mr imaging sequence in a variety of applications. pediatric imaging has also immensely benefitted from this technique. in this presentation we will review pediatric neuroimaging applications in trauma, arterial and venous vascular disorders, hypoxic-anoxic injury, congenital malformations, congenital heart disorders, neoplasms, and pediatric degenerative disease. use of swi in pediatrics other than demonstrating hemorrhage and calcification will be reviewed. challenges in the clinical use of swi in pediatrics, interpretive pitfalls, and sources of clinical misinterpretation of swi will also be explored. we will also briefly present ongoing research and clinical use of swi in pediatrics and potentials for future collaborative investigations. & swi is highly sensitive in detection of susceptibility effects on mri. & in many cases, but not all, swi processing can differentiate between calclium and blood products. & quantitation information can also be obtained from swi with further processing. state of the art imaging of the single ventricle d.m. biko; philadelphia/us there are many congenital heart defects that result in a functional single ventricle. this may be functional or anatomical as a result of a dysfunctional valve or absent or ineffective pumping chamber. the repair of single ventricle physiology most often involves a staged reconstruction due to changing physiology ultimately resulting in a total cavopulomonary connection or fontan procedure. to appropriately image the single ventricle throughout its stages of palliation, familiarity with the physiology of the various steps in surgical palliation of the single ventricle is essential although echocardiography is a mainstay of cardiac imaging, cross sectional imaging has a vital role in the evaluation of the single ventricle. the role of ct angiography is mostly for anatomic evaluation. although it is fast and has high spatial resolution for evaluation of vasculature, ct has lower temporal resolution than mri and is unable to quantify flow. ventricular performance along with quantification of flow can be performed with mri. systemic to pulmonary collateral flow, which has been shown to result in adverse outcomes after fontan, can be quantified. valvular insufficiency and myocardial scarring can also be assessed. additionally, high anatomic vascular detail can be obtained with mri, particularly with the recent investigational use of the blood pool agent ferumoxytol. mri also has the ability to assess the lymphatics either through non-contrast t2 weighted imaging and/or dynamic contrast mr lymphangiography as lymphatic pathology may play a role in postsurgical hemodynamics in single ventricle patients. this lecture will focus on the use of ct and mri in the evaluation of the single ventricle particularly concentrating on the developing use of mri for anatomic and physiologic assessment. take home points: in single ventricle physiology, there is only one effective pumping chamber. familiarity with the physiology of the various steps in surgical palliation of the single ventricle is essential in imaging this disorder ct angiography provides high anatomic detail but limited in its assessment of physiology since it cannot quantify flow and has lower temporal resolution than mri. mri can evaluate ventricular performance, quantify flow and valvular insufficiency, and assess myocardial scarring. high anatomic vascular detail can also be obtained with mri particularly with the emerging investigational use of ferumoxytol. with non-contrast t2 weighted mri and/or dynamic contrast mr lymphangiography, lymphatic evaluation can be performed which may play a role in post-surgical hemodynamics in single ventricle patients. neuroimaging in head trauma m. argyropoulou, g. alexiou; ioannina/gr summary: objective: head trauma in children is one of the most common reasons for visiting emergency department. however, only a small portion of patients will have a traumatic brain injury. patients with moderate or severe head trauma should undergo ct scan, however, a debate exists for the indication and yield of neuroimaging for minor head trauma. we performed a systematic literature review on the accuracy of symptoms and signs in children with minor head trauma in order to identify those with severe intracranial injuries. materials: a systematic literature search of medline (2000) (2001) (2002) (2003) (2004) (2005) (2006) (2007) (2008) (2009) (2010) (2011) (2012) (2013) (2014) (2015) (2016) was performed to identify studies assessing the diagnosis of intracranial injuries in children. the authors independently performed critical appraisal and data extraction. results: we identified studies that evaluated the performance of findings for detecting intracranial injury using the reference standard of neuroimaging or follow-up examination. mechanism of injury, multiple vomiting episodes and decline in gcs score were more commonly associated with severe intracranial injury on ct. normal variations in the amount of joint fluid, ganglion cysts, bone marrow edema, and bony depressions that resemble erosions are frequent in the wrists of children. the results of a follow-up of a healthy cohort aged 6-17 will be presented. the cohort was examined twice with mr of the wrist, and the second time also with a cartilage sequence for better visualization of the bony depressions. knowledge of these normal variations is important because they can resemble disease. bone marrow edema, joint fluid more than 2 mm, and bony depressions that can resemble erosion are frequent findings in the normal wrist. take home points: bone marrow edema, joint fluid more than 2 mm, and bony depressions that can resemble erosion are frequent findings in the normal wrist. these findings can not be attributed to dissease without additional findings of synovitis. a cartilage sequence can be of use in the differentiation between true erosions and bony depressions. mri scoring of the wrist in patients with jia-current status and future perspectives c. nusman; amsterdam/nl the wrist is a frequently affected joint in patients with juvenile idiopathic arthritis. due to recent improvements in treatment strategies, permanent damage is not that common anymore. also, imaging has been playing a key role in monitoring the disease activity in the wrist of jia patients. the past years lots of efforts have been made to improve the assessment of acute and permanent changes of the jia wrist. requisites and recommendations for the mri protocol to use for of the jia wrist are available in literature. currently, the features of scoring the jia wrist are synovitis, tenosynovitis, bone marrow edema and bone erosions. the repeatability of the above-mentioned scoring features proved to be acceptable. recent studies showed that the appearance of the wrist in healthy children can mimic pathology. therefore, construct validity of the scoring features needs to be assessed by comparing wrists of healthy children with the wrists of jia patients. & construct validity of the scoring features needs to be assessed by comparing wrists of healthy children with the wrists of jia patients a novel radiographic scoring system for permanent hip involvement l. tanturri de horatio 1 , p.l. di paolo 1 , s.c. shelmerdine 2 , p. toma 1 , k. rosendahl 3 ; 1 rome/it, 2 london/uk, 3 approximately 20-50% of children with jia, particularly those with systemic onset disease, will have hip-involvement within 1-6 years after disease onset. as scoring systems for radiographic changes in children with hip involvement are lacking, we aimed to examine the reliability of potential markers and suggest a novel scoring system. a set of 75 hip-radiographs from 75 children with jia and clinical hipinvolvement: 59 seen at the outpatient clinic at great ormond street hospital (gosh), london, and 16 seen at ospedale pediatrico bambino gesù, rome, was used. all hip radiographs were scored in a blinded fashion, once by an experienced paediatric radiologist and a paediatric radiologist with minor experience in musculo-skeletal imaging in rome, and twice by an experienced radiologist and a research fellow in bergen/ london. radiographic findings suggestive of 1) destructive change (bone erosion, flattening of the femoral head, squaring of the femoral head contour, presence of sclerosis, joint space height, and 2) growth abnormality (length and width of the femoral neck, varus/valgus deformity, the ccd angle and the trochanteric-femoral head height) were assessed. assessment of erosions of the femoral head, femoral neck and the acetabulum showed moderate to good agreement for the same reader. the inter-reader agreement was, however lower. there was a high to moderate (2017) 47 (suppl 2):s297-s pediatr radiol agreement for the assessment of femoral head flattening using the mose' circle. the measurements of femoral neck length and width, the ccd and trochanteric-femoral head lengths were precise, with 95% limits of agreements within 10-15% of the mean. we have identified a set of relative robust radiographic findings suggestive of growth abnormalities and destructive change in children with hip-jia, and suggested a novel scoring system. x-ray of a 21 years-old jia patient with severe chronic hip involvement. x-ray of a 6 years-old boy with growth abnormalities on hips (bilateral coxa magna). in jia hip involvement is often a predictor of a severe disease course. radiographic findings vary according to mode of onset and age: in younger children the initial findings may be developmental rather than destructive while children with later onset jia may have destruction/narrowed joint space as the first feature. several of the commonly used radiographic findings for chronic hipchange are inaccurate. we have identified a set of relative robust radiographic findings suggestive of growth abnormalities and destructive changes in children with hip-jia, and suggested a novel scoring system. bone age assessment -statement from the msk task force k. rosendahl; bergen/no summary: age assessment is an important, yet complex and challenging issue that authorities may need to perform to determine whether an individual is an adult or a child in circumstances where their age is unknown. there is currently no method which can identify the exact age of an individual and there are concerns about the invasiveness and accuracy of the methods in use, namely analysis of documentary evidence, interviews, physical or other form of medical examination such as imaging. the main imaging methods include carpal, collar bone and dental examinations. whilst many countries make use of these methods they do not apply them in the same way and often use different combinations and/or order. one of the main reasons for this is the fact that age assessment procedures remain to a large extent determined by national legislation, with procedures evolving through national jurisprudence (ref.: european asylum support office (easo age assessment practice in europe)). the ethical and legal aspects of using bone age to determine age will be addressed in a statement from the msk task force. the ethical and legal aspects of using bone age to determine age will be addressed in a statement from the msk task force. & the application of drls should be the responsibility of all providers of x-ray imaging. this means that drls should also be applied to imaging performed outside the radiology department. & the physical quantity used to establish drls should be an easily measurable quantity, usually directly obtainable from the x-ray equipment console, obtained either by manual recording or preferably by automatic recording and analysis. organ doses and effective dose are not considered feasible as a drl quantity because these cannot be easily determined. the ultimate mission of eurosafe imaging is to support and strengthen medical radiation protection across europe following a holistic, inclusive approach. most common imaging procedures in children and their contribution to collective dose e. sorantin 1 , c. granata 2 ; 1 graz/at, 2 genova/it summary: several countries have released "diagnostic reference levels (drl)" for imaging procedures using ionizing radiation. unfortunately those drl differ in types of procedures and granularity as well as information about the proportion of pediatric patients within the different examinations are sparse. therefore an more evidence based approach seems to be feasable -meaning releasing drl first for frequent and radiation burdened examinations. therefore a survey within europe was conducted and a questionnaire was sent to key persons of the european society of pediatric radiology (www.espr.org) as well to members of a large academic, interdisciplinary, international network within the ceepus programme (central european exchange programm for university studieswww.ceepus.info). alltogether 33 centers were contacted and an response was received from 18 (54.5.%). from one center only frequencies for interventional radiology was sent. plain films: most frequent procedures are extremities (48.5%), followed by chest films (31.4%) -both account together for more than ¾. flouroscopy: voding cysto urethrography (vcu) 37.2%, followed by upper gastro intestinal (gi) series with 32.1% -again representing 2/3 of those examinations. computed tomography: head & neck 46.8%, chest 27.8%, abdomen 13.7% -together almost 90%. interventional radiology and cardiac interventions: only limited data available and procedures quite hardly standardize and comparable. it seems adviseable, that only a few procedures are suitable for drl like peripheral insertion of vascular lines, occlusion of ductus arteriosus botalli or stent implantation for coarctation. in order to estimate the contribution to the relative collective dose all values were normalized to a chest xrays (1.0) and the following numbers could be calculated: abdominal plain film 0.1, skull 0.01, ct head 2.6, ct chest 10.2, ct abdomen 4.5. the most frequent imaging procedures using on ionizing radiation are: in plain films extremities and chest xrays in flouroscopy vcu and upper gi series in ct head, chest and abdomen therefore eu wide drl should be released for those examinations. as it could be expected chest ct is the main contributor to the collective dose. since the espr abdominal (gi and gu) imaging task force has changed its name and agenda, extending from initially only genitourinary queries to also other abdominal imaging topics, new projects have been added such as for example imaging in anorectal and cloacal malformations, imaging in paediatric inflammatory bowel disease (ibd, a joint project with esgar), or paediatric abdominal ceus applications. results of these new projects will be presented in the upcoming talkshoping that again (as the last procedural recommendations and proposed imaging algorithms) our proposals and recommendations will help to standardise paediatric imaging, to reduce radiation burden, and to facilitate comparable imaging data for future research. other topics in this session are a proposal for a more standardised approach to gastrointestinal ultrasonography, and considerations on gadolinium applications in children in the light of new observations (i.e., gadolinium deposit in tissue even in children with normal renal function). the work goes ononly achievable with active participation of interested and competent members. many interesting topics for either recommendations or joint research are on the list such as addressing late decompensating pujo or specific imaging needs in ibd in early childhood; other new ones may be proposed by any task force member. thus all espr members are invited to join the group, work with us and share their expertise. (2017) 47 (suppl 2):s297-s pediatr radiol contrast enhanced us in childhood -applications in children: literature review and results from the questionnaire c. bruno; verona/it in adults, following the characterization of focal liver lesions, several applications of contrast-enhanced ultrasound (ceus) have emerged in the last two decades, since second-generation contrast agents have been introduced and approved for use in most european countries. from many points of view, children represent an ideal population for ceus, because of the absence of radiation exposure and of need of sedation. moreover, due to the small body size many anatomical targets in children can be adequately explored with high-frequency ultrasound, obtaining images with higher spatial resolution than in adults. however, to date comparatively few data on pediatric ceus are available. although very rare and usually mild, possible adverse effects of contrast agents probably limit their use in many centers. in addition, the intravenous administration of ultrasound contrast agents in children is still off-label in europe, which makes informed consent necessary in every case. finally, for unclear reasons information on this topic does not flow easily. & from the comparison between the data available, similar or better results are likely to be obtained with ceus in children than in adults, and some specific pediatric indications might be proposed. imaging in ibd-joint recommendation statement with esgar f.e. avni 1 , m. napolitano 2 , p. petit 3 ; 1 brussels/be, 2 milan/it, 3 marseille/fr the first joint esgar/espr consensus statement on the technical performance of cross-sectional small bowel and colonic imaging (1) objective: to develop guidelines describing a standardized approach to patient preparation and acquisition protocols for magnetic resonance imaging (mri), computed tomography (ct) and ultrasound (us) of the small bowel and colon, with an emphasis on imaging inflammatory bowel disease. methods: an expert consensus committee of 13 members from the european society of gastrointestinal and abdominal radiology (esgar) and european society of paediatric radiology (espr) undertook a six-stage modified delphi process, including a detailed literature review, to create a series of consensus statements concerning patient preparation, imaging hardware and image acquisition protocols in pediatric and adult patients. the delphi process is constructed as follow: step 1 questionnaire construction to includes all contents relevant to the guideline and set up of working groups; step 2 questionnaire completed by all committee member, step 3 literature search; step 4 draft consensus produced by each wg based on the literature review and questionnaire responses; step 5 committee members indicate agreement or otherwise for each individual draft consensus; step 6 acceptance of agreed statements (more than 80% of members), face to face meeting to modify statements without agreement. committee members indicate agreement or otherwise for each modified consensus statement and final consensus statements. the questionnaire was split into four broad topics, each of them treated by a subgroup including in each of them a pediatric radiologist: (1) patient preparation for mre/mr enteroclysis/cte/ct enteroclysis, (2) mre/ mr enteroclysis technique and sequence selection, (3) cte/ct enteroclysis technique, and (4) enteric us patient preparation and technique. after an extensive literature research each member were instructed to always base their statements on the retrieved literature wherever possible, and to this end graded the strength of retrieved relevant publications from i (high) to v (low) using the criteria of the oxford centre for evidence based medicine (2) during their review process. if no relevant literature was available for a particular item, members used expert opinion to construct the consensus statements. the pediatric guidelines were based on the opinion of 3 pediatric radiologists and 4 adult radiologists who have experience in pediatric practice. & it is recommended that children aged 6-9 should not eat any solid & it is recommended that the use of a spasmolytic agent is optional. unlike adult practice, the use of spasmolytic prior to mre is considered optional in paediatric patients and use is likely dependent on the age of the patient, with older children more likely to tolerate spasmolytic injection. there are data supporting the benefits of glucagon on image quality, at the expense of prolonged imaging time and precipitation of nausea in just under half of paediatric patients (3, 4) . however, high diagnostic accuracy can also be achieved without spasmolytic (5) . & it is recommended that children aged over 9years should be nil by mouth for carbonated and milk beverages for 4-6h. ingestion of still water or non-carbonated fruit juice is recommended. & it is recommended that for dedicated colonic evaluation, a standard protocol without specific modification is used. & use of a spasmolytic agent is not recommended. & the use of i.v. us contrast is not recommended. & it is recommended that scan coverage should include an abdominal and pelvic examination, including the liver. there are no specific recommendations as to the use of hydro us in the paediatric patient as practice is not well developed. if oral contrast is given prior to us, it would seem sensible to follow the recommendations for mre in the paediatric population & it is recommended that if ct scanning is used in the paediatric population, no specific preparation is usually required although administration of positive oral contrast could be considered; for example, prior to percutaneous drainage of abscesses. limitations: there is little evidence in the literature to ascertain all these proposals. the recommendations were mainly based on expert opinion. no recommendations have been proposed for children before 6 years of age. especially the benefice of mre under sedation (6) in the younger compare to us doppler need to be explored. contrast media application is essential for a number of mri studies in children. there is some evidence that gadolinium-based contrast agents (gbca) are well tolerated in infants and children. the risk of adverse reaction is no higher in children than in adults. there are only few data available about pharamakokinetics in children, especially for the use of gbca in neonates. age-adapted reference values of the glomerular filtration rate (gfr) have to be used to identify children with a potential risk. in the past few years there was some attention toward the potential cellular toxicity of gadolinium and its role in the development of nephrogenic systemic fibrosis (nsf). there were only few children identified with proof of nsf. but, particulary renal insufficiency, poor hydration, acidosis and inflammation increase the risk for nsf. because the cases of nsf have been observed with linear componds the guidelines (2017) 47 (suppl 2):s297-s pediatr radiol from the esur and the espr and others propose to avoid linear compounds and to prefer macrocyclic gbca. in the past year several studies have described observations about possible gadolinium retention in the brain; hyperintense brain structures in native t1 weighted sequences were verified -globus pallidum and dentate nucleaus -also in children undergoing multiple mri examinations with gbca application. so, repeated mr investigations within a short time should be avoided -the cumulative dose of gbca should be recorded. consider all these points, the benefit of a contrast-enhanced study should be weighted against the potential risks before administering a gbca for each child separately. but, never deny a child an indicated cemri study. use single dose application (0.05-0.1 ml/kg body weight), improve renal function and hydration, balance acidosis -and ask your pediatric nephrologist íf necessary. gadolinium-based contrast agents are safe. macrocyclic compounds should be used in children. avoid contrast media in neonates and be careful in infants. identify risk factors. avoid repetivite application. procedural recommendation: how to perform pediatric gastrointestinal us m.l. lobo 1 , m. riccabona 2 ; 1 lisbon/pt, 2 graz/at summary: ultrasound (us) is the first imaging modality applied in the investigation of abdominal complaints in children, and an increasingly valuable imaging tool in the assessment of the gastrointestinal (gi) tract in neonates, infants and children. a comprehensive us examination is a critical first-step to optimize the potential of us diagnostic yield in many paediatric gi conditions. using proper high resolution transducers and graded compression technique is an essential part of gi us examination. a methodical and systematic analysis is crucial to facilitate a thorough evaluation of the bowel segments as complete as possible: follow bowel in a cross section, complete by longitudinal and oblique views. for some bowel sections filling is essentialsuch as stomach for gastroesophageal reflux and pyloric function, and distensibility and size of the colon by enema (e.g. for query microcolon). modern us methods are valuable, but not a pre-requisite. proper documentation of abnormal size of the gi tract segments, their luminal content, peristalsis, bowel wall characteristics and its surroundings, as well as local tenderness should be noted. a proposal for recommendation on how to perform paediatric gastrointestinal us will be presented for public discussion. & careful and dedicated us examination is crucial to obtain maximum anatomic and functional information in many gastrointestinal disorders in children. & systematic and methodical analysis helps to assess the bowel as complete as possible. & satndardization of us technique is essential to optimize us diagnostic capabilities and to allow for comparable examinations wich is essential to improve future evidence-based knowledge. hominid evo-devo: reconstructing the evolution of human development c. zollikofer; zurich/ch from an evolutionary biologist's perspective, modern humans represent the only surviving species of a group of highly specialized "bipedal great apes". they evolved more than seven million years ago in africa and managed to spread over the entire globe. in this talk, i will trace the history of our species with an emphasis on key developmental innovations that underlie major evolutionary innovations. why are we born with brains that have the size of adult great ape brains? and why do we grow up so slowly and get so old? i will highlight how advanced biomedical imaging methods help addressing these questions, and show how combined fossil, clinical and great ape data yield surprising insights into the evolution of our development. to present our experience with innovative imaging in pediatric interventional radiology. imaging technologies presented will include: 1. use of bubble contrast (lumeson) for indicatons including; complex pleural effusion and abdominal collection assessment pre and post therapy, primary g tube placement, renal perfusion pre and post rena artery angioplasty, vascular patency during central venous line placement, vascular malformation therapy and biliary tube assessment. 2. intravascular us (ivus) in arterial intervention pre and post angioplasty and venous thrombolysis intervention. 3. optical coherence tomography pilot study assesssment for renal artery intervention -validation in normal subjects. currently this imagng which uses laser light technology to assess vascular mural detail at the micron level, is only validated in coronary artery intervention in adults. 4. mr overlay -a technology that fuses mr imaging with low dose fluoroscopy and can faciltate biopsy of mr positive/ct negative lesions in the ir suite. focus will be on bone lesion biopsy and vascular malformation therapy. critical structures to be avoided can be outlined on the mr and transposed onto the fluoroscopic image during biopsy. in our experience this technology has promise in the pediatric setting with significant dose reduction when compared to ct. 5. mr fusion and i guide fusion technology enables an mr positive/ct negative lesion that would require ct guided imaging to be biopsed, using low dose c -arm ct, with fusion of the ct and mri images performed using landmarks, facilitating fluoroscopically guided biopsy in the ir suite. critical landmarks/structures to be avoided can be outlined on the ct or mr and transposed onto the fluoroscopic image during biopsy path planning and orchestration. focus will be on bone lesion biopsy. 6. color parametric flow related imaging in vascular interventionthis software enables time to peak opacification of arterial or venous contrast to be color coded in time and can provide adjunctive information for assessent of perfusion change during vascular intervention such as renal artery angioplasty, dialysis access intervention and cerebral embolization. 7. mr guided intervention -this focus will be on the initial development of an mr interventional program and our initial experience with mr arthrography. discussion will also involve the use of this modality for vascular malformation sclerotherapy and other msk interventions such as biopsy and nerve injections. 8. high frequency us imaging-focus will be on the use of a 40mhz us probe in the ir suite for various indications including visualization of smaller targets such as neonatal central venous access, superficial vascular malformation therapy and thyroid fine needle biopsy. 1. participants will become more familiar with exisitng and emerging innovative imaging technologies for pediatric intervention. participants will learn about the various indications and limitations of these technologies. 3. participants will gain insight into the process of introducing new imaging modalities into their pediatric interventional practice. increasing evidence supports the notion that autism spectrum disorder is associated with anomalies of brain function and connectivity. it is also evident that there are atypicalities in development/maturation of brain systems. particular promise arises from findings of atypical electrophysiology -indexing brain neuronal activity in real time. in particular, this talk will address a characteristic electrophsyiologic signature of delayed auditory evoked response latency (at~100ms). this, and related timing anomalies, have been proposed as biomarkers for asd -with candidate use for diagnosis, prognosis, stratification and therapy monitoring. progress along each of these axes will be discussed. however, to justify the term "biomarker", we demonstrate converging evidence from spectrally-edited (megapress) mrs and diffusion-mri. mrs offers insights into neurotransmitter levels, especially gaba and glutamate, imbalance of which may be associated with anomalous electrophysiologic oscillations in the gamma band. diffusion offers insights into the white matter of the brain (auditory pathway will be illustrated) and an interpretation of diffusion parameters as an index of central conduction velocity will be offered. combining these mechanistic measures with the spectrospatio-temporal capabilities of magnetoencephalography (meg), this talk will present a state of the art review of multimodal biomarker development in asd. take home points: meg captures brain activity in space and time as well as showing sensitivity to activity at different frequencies (where, when and what) delays in cortical neuronal response latency are evidence in asd atypical coupling between diffusion evidence of conduction velocity and timing of cortical responses in shown in asd oscillatory activity is atypical in asd (elevated "noise", decreased "synchrony") diminished inhibitory neurotransmitter (gaba) levels are shown in asd disturbance of teh typical coupling between gaba and gamma-band oscillations in development leads to anomalous adult oscillatory activity (taken to index local circuit function). multimodal and longitudinal approaches may be required to tackjle the complex and heterogeneous landscape of asd the paediatric radiologist can play an important role in establishing vascular access in paediatric patients ranging from neonates to teenagers. a breadth of knowledge and skills are needed to deal with changing body morphology and varied pathology in this age range. some of the skills particular to performing and managing vascular access in children will be discussed. different devices which can be placed, their indications, advantages and disadvantages will be reviewed. choice of access vessel is important in children, because there are known long term complication such as central venous stenosis and thrombosis, which can have a huge impact for future venous procedures or potential creation of an arteriovenous fistula of the arm for dialysis. preserving venous access sites is a (2017) 47 (suppl 2):s297-s pediatr radiol key responsibility especially in children with complex medical and surgical co-morbidities. because vascular access in children has associated morbidity it's important to manage and maintain devices that are placed. the risk of infection when repairing or exchanging a broken line will be highlighted. image guided biopsy is a very frequent procedure in pediatric patients. they range from random organ parenchyma for the diagnosis of medical disease up to tumor biopsies for histopathology analysis. different imaging modalities can be used for guidance as well as different biopsy devices and needles. ultrasound guidance is the most common modality used for this purpose in the pediatric population. the success of this procedure depends on multiple factors: from pain control up to choosing the correct device and area to sample. the radiologist performing the procedure also needs to be familiar with the potential complications of the intervention, how to prevent them and how to manage them. the intention is to perform the safer procedure as possible, obtaining the best quality of sample. the goal of this lecture is to present in a didactic way technical tips to perform safe and effective image guided pediatric biopsies, which may be applicable to different groups of operators, ranging from general pediatric radiologists performing occasional biopsies up to pediatric interventional radiologists. the objectives will be: to identify the safest approach to different types of biopsies; to describe ways to obtain the better quality of sample as possible; to demonstrate the use of different approaches in challenging clinical scenarios; to illustrate new devices currently used in specific applications; to discuss potential complications and its management and to show imaging modality integration applied to biopsy planning an performance. image guided biopsy is a frequent procedure in pediatric patients. a pre-procedure planning is fundamental in the success of the intervention. the operator must be aware of the aims of the biopsy and based on this choose the best approach, device and site for sampling. preparation and competency to manage complications is mandatory. pediatric interventional oncology: big cases in little people m. heran; vancouver/ca summary: the pediatric patient presents unique challenges in diagnosis and management of oncologic disorders. interventional radiology (ir) has a prominent role in the care of these children, with improvements in imaging and equipment offering better and safer options to traditional diagnostic and therapeutic procedures. as cancer can involve any organ system, consultations to the ir service can involve any part of the body, and can be non-vascular and vascular, simple and complex. the most common ir procedures in the pediatric oncology patient are enteric tube placement/change, vascular access, and percutaneous image-guided tissue/organ biopsy. however, with the explosion of interventional oncology in the adult setting, the variety and complexity of ir in pediatric oncology has begun to increase as well. ir techniques, such as thermal ablation, transarterial pharmacotherapy, and preoperative embolization, are now increasingly discussed in multi-disciplinary conferences as complementary or primary modes of treatment of oncologic disorders or related diseases/complications. however, although the principles of these diagnostic and therapeutic ir procedures remain essentially the same in their translation from adults to children, well recognized differences in pediatric physiology and metabolism, as well as the range in weight, size, and age of children, result in a practical question of "how do we do this?" the aim of this presentation is to provide an overview of the role of ir in the pediatric oncology patient, and to highlight areas of research and innovation. vascular anomalies encompass a spectrum of disorders including vascular tumours and vascular malformations. incorrect nomenclature and misdiagnoses resulting in inappropriate treatment are commonly experienced by patients with vascular anomalies. the currently accepted method for classification of vascular anomalies is straightforward and clinically relevant. vascular malformations can be divided into high flow lesions such as arteriovenous malformation or low flow lesions such as venous or lymphatic malformations. in children, a diagnosis can often be made with the history, examination and ultrasound. the classification of vascular anomalies will be briefly reviewed with examples of commonly encountered pathologies. a multidisciplinary team approach to the management of these conditions is vital. paediatric radiologists can play a key role not only in diagnosis but also in management, principally by injection sclerotherapy of low flow lesions and embolization of the much rarer arteriovenous malformation. many sclerotherapy agents are available with sodium tetradecyl sulphate the most commonly used for venous malformations and doxycycline for lymphatic malformations. different sclerotherapy agents have different characteristics and uses which will be covered. symptomatic relief is often achieved with treatment but multiple treatment episodes may be needed to achieve the desired outcome. ensuring the child and family understand this is vital to ensure they are satisfied with the management of the condition. contrast media is commonly used during imaging in children whatever their age and whatever the pathologic conditions. still, youngest patients are vulnerable and unstable. therefore, in neonates and infants the use of s339 (2017) 47 (suppl 2):s297-s pediatr radiol contrast media should be carefully evaluated and customized putting in balance the risk versus the benefit of its use. when using contrast media in neonates and infants, several features should be highlighted: -prematures and neonates have rather immature kidneys and some contrast media might be difficult harmful -the thyroid gland in prematures may be (transitorily) depressed by iodinated contrast media -the use of high osmolar contrast may induce a fluid shift and dehydration especially in premature and neonates -most contrast media are used off label; almost none has obtained the authorization to be used in neonates. -there are very few studies evaluating the short and long term adverse reactions in neonates and infants below the age of two. fortunately these reactions seem very rare in these age groups. -using contrast extends the duration of the examination and the need for sedation different types of techniques will potentially need ingestion, instillation or injection of contrast media: 1) opacification of the entire gi tract pre-and post-operatively 2) retrograde uretro-cystography 3) contrast enhanced ct 4) contrast enhanced mr imaging 5) contrast enhanced us 6) angiography furthermore, different types of contrast media can be used to achieve these purposes 1) barium (sulfate) 2) iodinated water-soluble contrast media (hyper-, iso-or hypoosmolar) remarks regarding opacification of the upper gi tract: -the upper or lower gi tract should be opacified using water soluble contrast in the immediate postoperative period or whenever a bowel perforation is suspected. -air can be used to confirm esophageal atresia and duodenal atresia -barium should be preferred in case of t-e fistula -either barium or water soluble iodinate contrast can be used in order to opacify (sub)obstructed upper gi tract remarks regarding the opacification of the lower gi tract -iodinated iso/hypo osmolar contrast should be used to opacify the colon in case of obstruction -a higher osmolarity iodinated contrast can be used in case of suspected meconium ileus or plug; still this contrast should be used diluted and under close clinical surveillance and adequate hydration. -in some more specific cases, for instance whenever hirschprung disease or a stenosis post nectotizing enterocolitis are suspected, barium enema can be used remarks regarding ct scan -contrast enhancement may help for the global assessment of various pathologies especially in case of cardio-vascular malformations or for the evaluation of abdominal masses. any iodinate contrast among those available is acceptable in neonates. higher osmolality contrast allows to inject a lower volume -injected volumes of 1.5 ml/kg seem adequate using 22-24 gauge needles -power injectors are acceptable as long as adequate catheters can be used -allergic or side effects are very rare and should be managed similarly to adults. remarks regarding mr imaging -the use of gd chelates in neonates remains controversial as there is no data available on the long term effects of gd injected so early in life -gd should be used only when enhancement may provide additional information compared to the non-enhanced study (cns infections, tumors, cardiovascular imaging, abdominal tumors, uro-mr imaging...) -only gd with low nsf risk should be used -gd should not be used in children with renal failure remarks regarding contrast enhanced us -little is known about the use of ce-us in neonates -indications seem equal to older children -there are very few side or allergic effects -doses suggested are 0.1 ml/year of age children present varied histological types of brain tumours. it's now possible to combine different information and image techniques to improve the diagnosis of paediatric brain tumours. the multimodal approach has increased the diagnostic specificity and permits, in most cases, the pre-operative differentiation between low and grade tumours. children with low grade lesions, and in particular the less accessible tumours, would benefit the most from avoiding biopsy. in addition, preoperative spinal mri evaluation to rule out drop metastases should be performed in patients with suspected high grade tumours. in general paediatric brain tumours are less necrotic, i.e. aggressive tumours in paediatric patients tend to be more hypercellular and homogeneous. because of its ready availability and speed, computed tomography (2017) 47 (suppl 2):s297-s pediatr radiol (ct) is the first investigation generally performed for a suspected brain tumour. ct can rule out haemorrhage or calcifications, but can also be used to evaluate tumour cellularity. a hyperdense tumour on ct reflects hypercellularity and is very often high grade. medulloblastoma are, for example, typically hyperdense on ct scans and paediatric low-grade astrocytomas are almost always hypodense. mri plays a major role in the evaluation of brain tumours. in conventional mri, the "general aspect" is the single most important parameter in predicting high-grade tumours in children. the same does not hold true for low-grade tumours, of which only 67% can be predicted using the general aspect. in our previous study, hyperintensity on t2-w and the lack of diffusion changes were the most important single parameters with 83% positive prediction. embryonic tumours, such as medulloblastoma or pnet have high tumour cellularity with consequent very low adc and hypo/isointense t2 compared to the cortex. adc values derived from dwi have been shown to be decrease in highly cellular tumours. adc values cannot reliably be used in individual cases due to the substantial overlap between tumour types previously described in the literature. nevertheless, adc has a higher predictive value in children and increases the accuracy of preoperative differentiation between low grade and high grade paediatric tumours. the cut-off values for differentiating between low and high grade paediatric brain tumours are 0.7 x 10 3 mm 2 /s and 1.0 x 10 3 mm 2 /s for minimum adc and average adc values, respectively. perfusion with relative cerebral blood volume (rcbv) is considered a marker of angiogenesis and is helpful in distinguishing high and low grade tumours. however, perfusion can be difficult to perform in small children; small catheters with manual injection are therefore used in such cases (or, as an alternative, arterial labelling). it should however be taken into account that choroid plexus tumours can have high rcbvs resulting from highly leaky capillaries. mr spectroscopy (mrs) shows the metabolic profile of the tumour. high grade tumours show elevated choline (cho) -reflecting increase in cell membrane turnover -and decreased n-acetylaspartate (naa), which represents a neuronal marker. the absolute values of the mrs peaks are not used by us; we favour to normalize the signal intensities of metabolites to their values in contralateral brain tissue. mrs is helpful not only as guidance for stereotactic biopsy (cho hot spot) but also for determining whether the tumour is high or low grade. as a rule of thumb, a 200% increase of cho when compared to the contralateral brain tissue is highly suggestive of a high-grade tumour. however, in children, increased cho levels can also be found in pilocytic astrocytoma; in this case the typical aspect with cystic component and location can suggest the diagnosis, despite the mrs result. therefore, in children, high cho levels do not necessarily imply the presence of a malignant tumour. task based functional mri (fmri) can be used for pre-operative localization of the eloquent cortex together with the identification of the language and somatomotor function. in the future, small children who are unable to cooperate will probably profit from resting-state fmri. pet mri has the advantage of integrating structural mr imaging with physiologic pet. take home points: take home points although the histology of paediatric brain tumours is diverse, their general morphological aspect on mri has a very high diagnostic reliability. unlike adult grade iv brain tumours, malignant paediatric brain tumours are less necrotic, but are highly cellular with high nuclear-to-cytoplasmic ratios. adding information on signal intensities on t2w and dwi further increases the diagnostic accuracy of conventional mri. the solid areas of high-grade tumours are iso-or hypointense on t2w and hyperintense on dwi, whereas low-grade tumours show inverse signal characteristics. advanced mr techniques (perfusion and spectroscopy) provide important biological information which can be used to correctly identify grading (high vs. low) and to guide biopsy. in children high cho levels, although suggestive, do not necessarily mean a malignant tumour. experience with central review of paediatric renal tumours g. khanna; st louis/us summary: central imaging review of pediatric renal tumors has been performed in children's oncology group since 2006. to date, more than 5500 cases of pediatric renal tumors have been centrally reviewed real time by the study radiologists. the mean time for central review was <8 days. discrepancies between local and central risk stratification were identified for detection of bilateral disease and pulmonary metastasis. in addition, central archiving of images has created a rich repository of cases for future research. the role of imaging in detection of key diagnostic features in pediatric renal tumors will be reviewed. the diagnostic performance of imaging for staging, detection of vascular invasion and tumor rupture will be discussed. real time central review of imaging is feasible in pediatric oncology wilms tumor remains the most common pediatric renal malignancy, followed by renal cell carcinoma cystic nephroma typically presents as a bosniak 3 lesion, and has high association with dicer-1 mutations is there a role for dwi in nephroblastoma? a.s. littooij; utrecht/nl wilms tumour or nephroblastoma is the most common malignant renal tumour in children. ultrasound is usually the first line investigation. mri of the abdomen is often performed to further delineate the tumor and its surroundings. the addition of diffusion-weighted imaging (dwi) to the standard mri protocol may enable subtype characterisation and allows assessing treatment response beyond necrosis and volume change. overall, the survival rate in patients with nephroblastoma is relatively good and the current focus is on finding biomarkers to further improve outcomes while reducing therapy-related side effects in these children. therefore, identifying low-or high-risk type nephroblastoma might be relevant for treatment planning. diffuse anaplastic nephroblastoma and extensive blastema in residual tumour after preoperative chemotherapy may require more intensive treatment. the limited available literature suggest a linear relation between adc values and subtypes nephroblastoma at histopathology. furthermore, the addition of dwi to the standard mri protocol may detect lesions (e.g. nephrogenic rests of nephroblastomatosis) that remain undetected at post contrast t1-weighted images. unfortunately, there is a considerable heterogeneity in acquisition techniques and methods of adc measurements. nephroblastoma often contains areas of necrosis and/or hemorrhage that can demonstrate very low adc values and consequently mimic highly cellular portions of tumours. therefore these areas should be excluded from further analysis. this lecture will highlight the potential additional benefit and limitations of dwi in children presenting with renal tumour. significantly lower radiation exposure even in comparison to low-dose pet/ct, (b) the higher diagnostic accuracy as compared to pet/ct even when using diagnostic contrast-enhanced ct, (c) the unique possibility to combine distinct mr-inherent contrasts (e.g. dwi) with specific pettracers (e.g. 64cu-labeled antibody imaging) for the evaluation of novel targeted therapies, and (d) the opportunity to stage local and systemic tumour burden within a single and highly resolved examination. on the other hand, many circumstances are challenging the extensive use of pet/mri in children. in general, the availability of pet/mri systems is low, particularly for children. thus, only a few sites in europe have experience with this technique in children, and therefore the generated scientific evidence is limited. moreover, whole-body-mri is still not a broadly adopted method for the combined assessment of local disease extent and whole-body staging, potentially replacing other whole-body modalities like the bone scan. in this context, especially the detection of pulmonary metastases is biased also against pet/mri. finally harmonized sequence protocols and specific recommendations for trace dosage are not available for pet/mri. in conclusion, further efforts are needed to keep the promises of pet-mri in the daily practice. common artefacts in paediatric mri-how to recognise, avoid or take advantage of them c. kellenberger; zurich/ch summary: while mri is a robust and radiation free imaging technique for assessing anatomy and pathology of most tissues and organs throughout the body, it is inherently prone to artefacts as no other imaging modality is. mri artefacts may impair image quality potentially leading to difficulties or errors in interpretation, but in some instances can contribute diagnostic information. main sources of image degradation are motion, disturbances of the local magnetic field and other factors inherent to image acquisition. strategies to reduce effects from various kinds of motion and adjustment of sequence parameters for eliminating artefacts will be discussed. & understanding the origin and effects of artefacts encountered in paediatric mri is essential for modification of mri protocols, so that artefacts and associated errors can be avoided. & for safely and successfully imaging children with implants and devices, the composition, location and functionality of the foreign body needs to be known. injuries to the central nervous system in abusive head trauma are responsible for the primary cause of morbidity and mortality in infants. neuroradiology has an important role in diagnosis but also in depicting injury and extent of brain damage of poor outcome. computerized tomography (ct) and magnetic resonance imaging (mri) are the primary imaging techniques. ct is usually performed in the acute phase while mri is performed the following days after injury. some injuries are better identified on mri such as diffuse axonal injury and cerebral edema with susceptibility and diffusion weighted images. abusive head trauma (aht) is the primary cause of morbidity and mortality in infancy, especially during the first year of life. aht is clinically characterized by a triad consisting of subdural hematoma, retinal hemorrhage and encephalopathy caused by brain swelling (1). the most common mechanism responsible for brain damage is thought to be caused by whiplash shaking injury explaining that abusive head trauma is also referred as shaken baby syndrome. impaction, compression and penetrating injury are also possible mechanisms as well as strangulation. however because of the variability of types and severity of injury, clinical symptoms vary from subtle to severe such as alteration of consciousness or coma (2) . the most common symptoms include vomiting, seizure, lethargy, poor feeding and apnea of which vomiting and respiratory pauses are non-specific (3). poor feeding, irritability or lethargy is also nonspecific signs. however apnea and/or retinal hemorrhages seen in children with brain injury are strongly associated with inflicted trauma (4) . in contrast to acute injury some children may manifest with increased head circumference related to chronic subdural hematomas. neuroimaging is therefore playing a crucial role to assess infants and children with a suspicion of abusive head trauma. computerized tomography (ct) and magnetic resonance imaging (mri) are the primary imaging techniques. ct is performed for the initial evaluation in cases with acute symptoms to look for hemorrhagic intracranial injury as subdural hematoma. mri is more often performed in the following days to further evaluate brain injury and to look for spine and spinal cord damage (5, 6) or in the presence of normal or equivocal ct findings (7) . however brain mri may be the first option in children presenting with increased head circumference. recently the study from flom et al showed the high sensitivity of mri for intracranial hemorrhage in well appearing infants at risk for abusive head trauma suggesting mri as a screening tool with 3 pulse sequences (axial t2, axial gradient recalled echo and coronal t1 weighted inversion recovery) (8) . ct is generally performed without intravenous contrast injection with 3d volume rendering (vr) reconstructions for identification of fractures. in some cases postcontrast images are also obtained specially to rule out deep venous thrombosis especially when children present with nonspecific clinical symptoms. mri protocol should include axial t2, t2* or susceptibility weighted images, coronal t1 images, diffusion or diffusion tensor images, and postcontrast 3dt1 images including mip reconstructions to evaluate the venous structures. mr venography can also be performed. susceptibility-weighted images are usually preferred because they allow the depiction of smaller hemorrhagic dai lesions and greater number of lesions compared to gre t2 (9) . it was also reported by colbert et al (10) that the presence of micro-hemorrhages alone was useful for outcome prediction in abusive head trauma with significant poor long-term outcome. the sensitivity and specificity of microhemorrhages was also higher than the other clinical (such as retinal hemorrhages and glasgow coma scale score) and other imaging findings for prediction of outcome. diffusion tensor imaging (dti) measurements were reported in abusive head trauma by imagawa et al: decreased axial diffusivity related to axonal injury with consequent reduced mean diffusivity did correlate with poor outcomes (11) . magnetic resonance spectroscopy (mrs) is usually not part of the standard protocol. however aaen et al (12) showed that n-acetylaspartate/creatine and/or nacetylaspartate/choline ratios were decreased significantly in the corpus callosum, frontal white matter, parieto-occipital white matter, and parietooccipital gray matter in children with poor outcomes. this study mentioned above also reported that the prediction of outcome was accurate in 100% of patients by using a logistic regression model that include age, initial glasgow coma scale score, presence of retinal hemorrhage, lactate on mrs, and mean total n-acetylaspartate/creatine. functional mri, (2017) 47 (suppl 2):s297-s pediatr radiol volumetry may be performed in long-term follow up of victims of child abuse. physical abuse is associated with altered emotion with greater activation in the salience network in response to negative stimuli, that includes amygdala, thalamus, putamen and anterior insula (13) . increased responsiveness of the right amygdala to fearful and angry faces (negative stimuli) and structural changes as reduced hippocampal volume, are reported by dannlowski et al (14) . impaired attention was also reported in patients with childhood abuse (15) with reduced activation during attention tasks in the left hemispheric ventral and dorsolateral prefrontal regions. intracranial injuries include extracerebral hemorrhages and parenchymal damage as brain swelling and ischemia, venous infarction, diffuse axonal injury, contusions and intraparenchymal hematomas (7, 16) . extracerebral hemorrhages subdural hematoma is a characteristic finding of inflicted traumatic brain injury, is generally multifocal and most commonly seen along the posterior interhemispheric scissure, over de convexities at the vertex level and/ or in the posterior fossa (17) (18) (19) . subdural hematomas are most likely bilateral but may be unilateral. all locations are related to disruption of bridging veins. the identification of bridging vein rupture allows the diagnosis of traumatism in relation to acceleration/deceleration, rotational and shearing forces due to violent shaking (20) . a mixed density appearance of subdural hematomas is frequent but is also seen in accidental traumatic brain injury (21) (22) (23) . indeed this feature is often present in the very early hours following trauma and is thought to be secondary to early sedimentation of blood clots and supernatant serum. tubular high density is often seen on non-contrast ct over the convexities in abusive head trauma. this ct feature is related to a clot secondary to venous disruption (24, 25) that can end up in thrombophlebitis. this tubular high density was reported more recently as tadpole sign (26) and lollipop sign (27) respectively seen in 40 and 44% of abusive head trauma. this appearance is strongly associated to inflicted trauma and much less frequent in accidental trauma (3 out of 83 cases (3,6%) of accidental trauma in our experience). associated venous infarction is reported in 12% of cases of abusive head trauma (24) and often located in the parieto-occipital region, unilaterally at the site of venous disruption of bridging veins. subdural hemorrhages, when multiple, in the convexity and interhemispheric, or in the posterior fossa were found significantly associated with abusive head trauma in the meta-analysis reported by kemp et al (28). in addition subdural hematoma, cerebral ischemia, skull fracture, retinal hemorrhage and intracranial injury were significantly associated with abusive head trauma in the review from piteau et al (29). subarachnoid hemorrhages (sah) and epidural hematomas are also found in inflicted trauma and are not considered discriminant-imaging features. however epidural hemorrhages, isolated skull fracture and scalp swelling were reported as significantly associated with accidental traumatic brain injury (29). sah in shaking injury is usually caused by tears of the vessels within the pia and arachnoid predominantly in the interhemispheric fissure and high convexity (7). parenchymal injury parenchymal injury include brain swelling and ischemia, venous infarction (discussed above), diffuse axonal injury related to rotationallyinduced shear-strain injury with different inertia for grey and white matter due to their different specific gravities, contusions seen in deceleration trauma with friction between the skull and brain, and in blunt trauma and intraparenchymal hematomas related to lacerated vessels. brain swelling/ischemia may be related to increased blood volume (congestive swelling), increased presence of water in the nervous tissue, and the combination of both. increased water in the nervous tissue may manifest as vasogenic edema located in the white matter due to extravasation of plasma like fluid related to incompetent blood-brain-barrier and as cytotoxic edema located in the grey matter, related to ionic imbalance. cerebral edema can be recognize on ct within the 12 hours following injury as loss of gray-white matter differentiation and decreased attenuation of grey and white matter. cytotoxic and vasogenic edema are better characterized on mri with diffusion-weighted imaging. brain swelling and edema occur early after trauma with consequent underestimation of subdural hematoma. therefore imaging should be repeated (ct or mri) especially when neurologic symptoms change rapidly. brain swelling/ edema may also involve the posterior fossa and is better identified on brain mri. two frequent patterns have been reported in abusive head trauma (24). diffuse supratentorial brain swelling (infarction) involving the cortex and white matter was reported in 39% of cases and is considered as severe hypoxic-ischemic injury with poor outcome (30). watershed infarction was reported in 36% of cases and considered a less severe form of hypoxia-ischemia. apparent diffusion coefficient (adc) values are strongly associated with poor neurodevelopmental outcomes in the acute phase (within 4 days) especially basal ganglia, thalamus, brainstem, cerebral cortex, cerebellar vermis, cerebellar cortex and mean total brain (31). during the early phase up to 1 month adc values in fewer regions (basal ganglia, thalamus, brainstem and corpus callosum) were associated with poor outcome. when patients with and without parenchymal lesions are compared, the detection of diffuse lesions during the first 3 months as well as beyond 3 months is significantly associated with severe developmental outcome (32). late mri (beyond 3 months after injury) also showed that recovery depends on the extent of brain damage. patients with diffuse lesions show more severe motor and intellectual impairments and are more likely to have blindness and epilepsy than patients with focal or hemispheric lesions (32). diffuse axonal injury (dai) is related to shear-strain injury of small medullary veins and was reported in 6% of cases of abusive head trauma (24). it is encountered in trauma with sudden acceleration-deceleration associated with rotational angular forces and in shaking-impact trauma. the lesions may be hemorrhagic or non hemorrhagic (related to axonal swelling). dai is most often located in the subcortical white matter at the gray-white matter junction, corpus callosum, basal ganglia, brainstem and internal capsule. if the lesions are large enough and hemorrhagic dai may be seen on ct. however dai is usually better identified on mri with susceptibility and diffusion weighted images. the detection of changes in the basal ganglia or brainstem during the first 3 days as well as during the first month after injury is significantly associated with poor long-term outcome in survivors (30). the presence of intraparenchymal brain micro-haemorrhages detected on swi in children with abusive head trauma correlates with significantly poor long-term neurologic outcome (10) contusion is also reported in abusive head trauma and is seen in blunt trauma with impact with or without contrecoup contusion. contusions are located at the surface of the brain (crest of gyri) and may be pial and haemorrhagic (disruption of cortical arteries). they are also found in the frontal and temporal regions related to impact of the brain on the roof of the orbit, middle cranial fossa and sphenoid wing. white matter tears are also seen in the frontal and temporal area related to the vulnerability of unmyelinated and soft white matter in infants. skull fractures are seen in blunt impact and are less frequent than long bones and rib fractures in non-accidental trauma. the most common site is the parietal bone (because of bulging of parietal bones below 1 year of age). the fracture may be linear as in accidental trauma. radiologic features significant for inflicted trauma are multiple fractures, bilateral fractures and fractures that cross suture lines (28, 33). focal underlying brain damage can be seen such as subdural hematoma and hemorrhagic contusion. hypoxic-ischemic encephalopathy is seen in strangulation injury with involvement of the territories of the internal carotid artery related to their anatomic vulnerability. neuroradiology (ct and mr) is crucial for the diagnosis of trauma, to predict outcome when showing edema and hypoxic-ischemic injury. this presentation will present an update on post mortem mri (pmmr) with relevance to clinical developments over the last 2 years. in particular, reference will be made to diagnostic accuracy of pmmr across different body parts, the current limitations of post mortem mr, and protocol development at different field strengths. imaging correlates of post mortem interval are also being investigated. maceration (autolysis within intrauterine fluid) and perimortem hypoxic brain changes caused difficulties in image interpretation, which more advanced and quantitative techniques may be able to address. jawad take home points: below 500g, 1.5-t pmmr shows a significant reduction in diagnostic yield, compared with conventional autopsy, and therefore its clinical usefulness in this setting will depend on individual circumstances. 3t pmmr performs better than 1.5 t particularly <20 weeks gestation, and particularly for the chest, heart and abdomen. diffusion characteristics in different fetal brain areas are multifactorial, with maceration the strongest predictor in most areas. international pm ct protocols c.y. gerrard 1 , o.j. arthurs 2 ; 1 albuquerque, nm/us, 2 london/uk the european society of pediatric radiology (espr) taskforce and the international society of forensic radiology and imaging (isfri) pediatric working group have combined efforts to establish best practice standards for performing perinatal and pediatric post mortem computed tomography (pmct) examinations. use of pmct in the investigation of pediatric death has increased significantly in the past decade. due to quick acquisition times and the ability to acquire thin slice, high detailed images of the whole body, (2017) 47 (suppl 2):s297-s pediatr radiol many hospitals and forensic institutes have implemented pmct into daily practice. however, there lack an overall standardization of how cases are triaged and the acquisition methods when comparing institutes using pmct. in an effort to address inconsistencies in acquisition parameters, post processing, and case selection, pmct protocols were compiled from international institutes and centres currently performing pediatric imaging. this paper will describe both the uniform and divergent elements of image acquisition and procedural uses identified among the participating centres. the outcome is to provide a single source of information that can guide already established and new centres on the best practice standards for implementing pediatric pmct. take home points: describe how pediatric post mortem computed tomography (pmct) has increased in utility over the past decade. identify the differences in acquisition methods for clinical computed tomography versus post mortem computed tomography. discuss the overall consensus of case triage and scan acquisitions when comparing institutes in aggregate. provide comprehensive statement of best practice standards for pediatric pmct. post mortem imaging research: updates and future proposals o.j. arthurs; london/uk paediatric and perinatal post mortem imaging is a new and rapidly growing field, and the post mortem imaging taskforce was founded in graz at espr 2015. the pmi taskforce aims to help reach consensus and guidance regarding imaging protocols and the potential yield of post mortem ultrasound, ct and mr. the key priorities are the themes of collaboration, image acquisition, best practice guidelines, training and education, raising awareness and access to imaging. this presentation will give updates on the latest developments in perinatal and paediatric imaging, with particular focus on where the pmi taskforce can help. in particular, protocol development is underway, and the espr meeting acts as an opportunity for collaborative working and network development, to facilitate best clinical practice and welcome new members. arthurs oj et al., espr post mortem imaging task force: where we begin. pediatr radiol (2016) ; 46: 1363 -1369 take home points: post mortem imaging is an exciting sub-specialty which requires a combination of in depth fetal medicine, perinatal autopsy and pediatric imaging knowledge to help shape and grow the clinical and research arena. dedicated personnel have an opportunity to create the evidence-based behind a growing clinical service, with clear benefits to patients, families and referring clinicians. abstracts appear as submitted to the online submission system and have not been checked for correctness and completeness. sequences, are an emerging tool for evaluating intracranial vessel disease. improved survival due to emended treatment protocols results in an increasing number of long-term medulloblastoma survivors who experience delayed treatment effects. microbleedings, developement of cavernomas, vasculitis and atherosclerotic lesions are cerebrovascular structures affecting sequelae of the applied radiochemotherapy. this study evaluates radiation-induced intracranial vascular changes. twenty-two long-term pediatric medulloblastoma survivors (mean age 25.8 years, range 10-53 years; mean years after primary radiochemotherapy 16.3 years, range 1-45 years) underwent mri. the scan protocol included precontrast 3-dimensional time of flight (tof)magnetic resonance angiography (mra), precontrast 2d t1-and 2d t2-vwisequences and postcontrast 2d t1-vwi-sequences of the medium and large intracranial arteries. vessel wall thickening, contrast enhancement and luminal narrowing were analyzed. additionally precontrast t1-, t2-swi and t1-weighted images of the supra-and infratentorial brain were acquired. results: vwi-sequences: vessel wall changes could be found in 12 (54%) and 14 patients (63%) of the right and left ica, respectively. for the ba 4 (18%) patients revealed vessel wall changes; for the left and right va 2 (9%) patients were detected with vessel wall changes, respectively. in the tof angiography no alteration of the ica, ba or vas could be identified. in total vessel wall changes for the vertebrobasilar system and the icas could be found in 16 (72%) patients. swi-sequences: all patients (100%) revealed swi lesions, the smallest lesion measuring less than 2 mm, the biggest up to 5 mm. sixteen patients (72%) were presented with lesions > 4 mm, suspicious for cavernomas. to ensure quality of life in long term childhood medulloblastoma survivors, monitoring of long-term effects, like vascular changes after rct is gaining in importance. high resolution mri, including swi and vwisequences could be used here for. this study images, asymptomatic vessel wall alterations in former childhood medulloblastoma patients through vwi sequences and micro bleedings through swi sequences. vessel wall alterations, revealing rct induced arteriosclerosis, can lead to symptomatic intracranial stenosis which is associated with ischemia, furthermore micro bleedings and cavernomas can lead to intracranial hemorrhage. however further studies are needed to standardize mri sequence protocols to ensure a high standard follow up protocol, detecting clinically still asymptomatic vascular changes. fast "black-bone" mr imaging in evaluation of craniofacial abnormalities: comparison with high resolution ct z. habib, a. talib, c. parks, s. avula, l.j. abernethy; liverpool/uk to evaluate the feasibility and diagnostic value of a fast field echo, "black bone" mri sequence in children with craniofacial abnormalities. a fast "black bone" mri sequence has been used in addition to standard brain mri in 16 children (mean age 17 months, age range 3 months to 5 years and 5 months) referred to the supra-regional craniofacial surgery unit at alder hey children's hospital, liverpool, uk. a subgroup of 10 of these patients with complex craniofacial abnormalities additionally had high resolution volume ct performed at the same visit. "black bone" mr imaging was performed on philips ingenia 3t and 1.5t scanners, using a 3d fast field echo sequence (tr=8.3 ms, te=4.6 ms, flip angle 5 0 ). this sequence can be performed with an acquisition time of less than 2 minutes. the "black bone" sequences were assessed for accuracy in evaluating the patency of the sagittal, coronal and lambdoid sutures, and, where applicable, were compared with high resolution ct. the fast "black bone" mri sequence was shown to be technically feasible in all cases. the resultant images successfully demonstrated both patent sutures, which were confidently seen, and prematurely fused sutures which were confidently not seen. visualisation of patent sutures was found to be further enhanced by the use of minimum intensity projection. in the subgroup of patients with complex craniofacial abnormalities, comparison with high resolution volume ct confirmed good sensitivity for patency of cranial sutures. there was complete agreement in 50 out of 50 sutures assessed. the "black bone" mr images were also found to produce good-quality surface-rendered images and were also suitable for 3-d printing of models for pre-operative planning. fast "black-bone" mri has proven to be technically feasible and to demonstrate cranial suture patency with good agreement with high resolution ct. additionally "black-bone" mri can be used to produce good quality surface-rendered images and 3-d printed models for surgical planning. main symptom of mucopolysaccharidosis type iva (mps iva) is progressive systemic skeletal dysplasia. this is routinely monitored by cerebral and spinal mri. the vascular system is generally not in the primary focus of interest. in our population of mps iva patients we observed vessel shape alterations of the vertebrobasilar arteries, which has not been described before materials: mri-datasets of 26 patients with mps iva acquired between 2008 and 2015 were eligible for retrospective analysis of the vertebrobasilar arteries. the vessel length and angle of the basilar artery (ba) and both vertebral arteries (va) were analyzed. a deflection angle between 90°and 130°in the vessel course was defined as tortuosity, less than 90°as kinking. the results were compared to an matched control group of 23 patients not suffering from mps. the deflection angle [°] of the va and ba was significantly decreased in the majority (85%) of mps iva patients (fig. 1) mps iva is associated with significantly increased tortuosity of vertebrobasilar arteries. therefore the vascular system of mps iva patients should be monitored on routinely basis, as vessel shape alterations had been associated with dissections, leading to a higher risk of cerebrovascular events. in the pediatric population, intraspinal cysts (arachnoid or neurenteric cysts) are rare lesions mainly located in the thoracic region, whose acute onset is not well described in the literature. (2017) 47 (suppl 2):s297-s pediatr radiol we present a series of four children seen in the last two years as spinal cord emergencies and discuss the clinical aspects, imaging diagnosis, and management approaches, particularly in the emergency setting. a comparison of our cases with those reported in the literature is also provided. as in other types of spinal cord lesions, mr imaging is the diagnostic procedure of choice, because of its potential to demonstrate the exact location and extent of the cyst and its relationship to the spinal cord, valuable information for planning surgical treatment. this is a retrospective review of 4 cases of pediatric intraspinal cyst occurring in 4 boys and 1 girl, aged 2 to 6 years, treated at our institution between 2014 and 2016. onset was sudden in all cases and mimicked transverse myelitis or infarction. all our affected patients had no preceding history of trauma and presented with signs of spinal cord compression-back pain and less commonly abdominal pain-followed by weakness. all patients underwent emergent mr imaging, including t1, t2, t2*, 3d ciss, diffusion imaging and enhanced t1 sequences, mainly in the sagittal and axial planes. in each sequence, mr imaging showed a well-defined cystic lesion with signal intensity similar to cerebrospinal fluid, and secondary spinal cord compression that was severe in most cases. blood remnants were not visualized within or around the arachnoid cyst in any patient, which correlated with the absence of trauma antecedents. three of the four cysts were located in an anterior position relative to the spinal cord, and only one was located posteriorly; this latter had an associated subdural effusion. none of our patients had an associated neural tube defect. all patients were urgently treated with cyst wall fenestration or resection. the symptoms improved in all except one patient, whose symptoms did not abate, but ceased to progress. a prompt emergent diagnosis with mr imaging is important, as the symptoms can resolve if surgical treatment is performed before the spinal cord becomes irreversibly damaged. urgent surgery is essential in these cases, particularly if progressive neurological dysfunction develops over the course of spinal cord compression. the outcome following surgical fenestration or excision is excellent in most cases. nevertheless, a long-term imaging follow-up is recommended to detect possible recurrence. the objective of this study was to evaluate the usefulness of multiparametric quantitative mri model for myelination quantification in children. twenty-two children (age range: 9-5,400 days) were scanned with multiparametric quantitative mri. total volume of myelin water fraction (mwf) (msum), the percentage of msum within the whole brain parenchyma (mbpv), and the percentage of msum within intracranial volume (micv) were obtained. mwf values of brain regions were acquired by drawing regions of interests. the values were fitted to representative models of myelin maturation. spatiotemporal pattern of mwf mapping was visually assessed. values of msum, mbpv, and micv well fitted to a developmental model of myelination. mwf of brain regions well fitted to a developmental model with high r 2 values: pons (r 2 =74.6), middle cerebeller peduncle (r 2 =75.5), genu of corpus callosum (r 2 =93.6), splenium of corpus callosum (r 2 =79.7), thalamus (r 2 =85.8), frontal white matter (wm) (r 2 =95.7), parietal wm (r 2 =95.1), temporal wm (r 2 =93.7), occipital wm (r 2 =94.5), and centrum semiovale (r 2 =82.1). mwf mapping followed the known spatiotemporal pattern of myelination. multiparametric quantitative mri is a useful tool for mwf quantification in children. retinoblastoma is the most common intraocular tumour of childhood. it is a highly malignant. retinoblastoma is curable. if detected while still confined to the globe and if there are no metastatic risk factors, the child will nearly always survive following appropriate treatment. our aim is to assess diagnostic accuracy of preoperatively performed magnetic resonance (mr) imaging for detection of tumor extent in patients with histopathologically proved retinoblastoma. local ethics committee approval and informed consent were required for reviewing of patients' images and records. fifty-eight eyes in 30 girls and 27 boys with retinoblastoma (mean age at diagnosis was 23 months ±18.9) were reviewed on unenhanced t1wi, t2wi, and gadolinium-enhanced t1-weighted mri with and without fat suppression. mri parameters such as anterior chamber hyperintensity, involvement of choroid, ciliary body, optic nerve, sclera, orbital fat, and pineal gland were determined. maximum tumor diameter was measured and correlated to metastatic risk factors. imaging and pathologic findings were compared. choroidal invasion was suspected with mr imaging in 47/58 eyes; findings were false-positive in 6 eyes and false-negative in two (accuracy, 86.2%; sensitivity, 95.3%; specificity, 60%). mr imaging findings were true-positive in 10 of 17 eyes with proved prelaminar optic nerve invasion (60% sensitivity) and false-positive in 7 (82.9% specificity, 75.8% accuracy). postlaminar optic nerve invasion was correctly detected in 23 eyes; 4 eyes were false positive, in other 4 eyes, this metastatic risk factor was missed (accuracy, 86.2%; sensitivity, 85.2%; specificity, 87%). of nine eyes with histologically proven scleral invasion, 5 eyes were true positive . in the other 4 eyes, scleral involvement was missed on mri (accuracy, 93%; sensitivity, 55.6%; specificity, 100%).extraocular fat invasion was suspected on mri in 5/58 eyes. of these, findings were truly positive in 4 eyes (80%) and in 1 eye (20%) was incorrect (false positive) (accuracy, 98.3%; sensitivity, 100%; specificity, 98%).anterior chamber hyperintensity on t1-weighted mr images obtained after contrast agent administration correlated well with main mri and histolopathology findings. tumor size (assessed in our study by the maximum diameter in mm) was statistically associated with postlaminar optic nerve invasion (ρ=.002) and choroidal invasion (ρ=.007). mr imaging shows promising role for tumor staging and detection of metastatic risk factors. tumor diameter, measured with mr imaging, is associated with postlaminar optic nerve and choroidal involvement. patterns of the cortical watershed continuum of term gestation hypoxic ischaemic injurythe "wish-bone sign" a. chacko 1 , s. andronikou 2 , s. vedajallam 1 , j. thai 2 ; 1 east london/za, 2 bristol/uk objective: background partial-prolonged term hypoxic ischaemic injury (hii) involves the cortical and subcortical watershed zones of the brain, which are visually difficult to conceive. new innovative methods of demonstrating watershed cortical atrophy using flattened maps of the brain surface gives added insight into distribution of the watershed zone by demonstrating the entire brain surface. aim determining and validating patterns of hii sustained at birth in term infants using cross-sectional mri and the innovative mercator and scroll map views of cortical surface anatomy, to define the distribution of the watershed zones in children with partial-prolonged injury. one hundred paediatric mri brain scans with an mri and clinical diagnosis of chronic term hypoxic injury were read by 3 radiologists independently. all sites of abnormality were recorded and patterns classified. (2017) 47 (suppl 2):s297-s pediatr radiol patients with partial-prolonged and combined patterns were evaluated using mercator and scroll map reconstructions, generating schematics of the watershed zone. predominant patterns of disease were partial-prolonged and acuteprofound types. the watershed zone was demonstrated, on the derived maps, representing a continuum of involvement in the shape of a 'wish-bone' extending bilateral from frontal lobes to posterior parietal lobes in band-like fashion along the para-falcine cortex and intersected by another band of atrophy in the peri-rolandic regions extending along peri-sylvian cortices. this is defined in schematics as a visual aid. predominant patterns of injury in term hypoxic ischaemic injury are described and quantified, with the 'wish-bone sign' introduced to describe the typical distribution pattern of partial-prolonged hii in the watershed zone. correlation of brain edema degree and biochemical parameters in pediatric posterior reversible encephalopathy syndrome with hematologic/oncologic diseases t. akbas 1 , s. ulus 2 , b. karagun 1 , t. arpaci 1 , c. kalayci 2 , b. antmen 1 ; 1 adana/tr, 2 istanbul/tr posterior reversible encephalopathy syndrome (pres) often associated with hypertension is characterized by typical transient parietooccipital predominantly brain edema on magnetic resonance imaging (mri) with neurological symptoms such as seizures, headache and visual disturbances. even if endothelial dysfunction, increased blood-brain barrier permeability and hyper-hypoperfusion remain as controversial mechanisms to explain, the pathophysiology of pres is unremain. the aim of our study was to investigate the correlation between brain edema degree on mri and serum biochemical parameters such as lactate dehydrogenase (ldh), albumin (alb), creatinine, uric acid (ua) and urea. a total of 27 pediatric hematology and oncology patients (19 male, 8 female, aged 3-17, mean age: 11 years 6 months) diagnosed with pres during treatment and after hematopoietic stem cell transplantation (hsct) were included in this retrospective study. underlying diseases were beta thalassemia (n:14), aplastic anemia (n:4), acute lymphoblastic leukemia (n:4), acute myeloid leukemia (n:3), lymphoid leukemia (n:1) and burkitt's lymphoma (n:1). pres was seen after undergoing hsct in 21 patients. the brain edema degree according to specified anatomical regions on fluid attenuation inversion recovery (flair) mri sequence was scored by two radiologists blinded to patients' records. the levels of serum biochemical parameters at onset of symptoms were correlated with score of brain edema degree on mri. serum ldh concentration was statistically correlated with the score of brain edema degree (spearman's rho correlation, r=0.459, p=0.016). no relationship was found between other biochemical parameters and the score of brain edema degree. our results suggest that increased serum ldh as a marker of endothelial dysfunction is the main biomarker for development of brain edema in pediatric pres patients under treatment and after hsct with underlying hematologic and oncologic diseases. objective: gadolinium based contrast agents (gbcas) have been associated with increasing signal intensities in deep brain nuclei on unenhanced t1-weighted brain imaging. until now, most studies have been performed in adults, while results on pediatric patients are sparse. therefore, the aim of this study was to evaluate if there is any difference between signs of gadolinium retention in pediatric and adult patients. in this irb-approved, single center retrospective study, we extracted all patients with at least 5 contrast-enhanced mris archived on pacs between 2009-16. all patients with gadobenate dimeglumine only enhanced mris were reviewed. seventy-six pediatric patients with the most injections and 86 adult patients with the most injections were included in the final evaluation. therapies were documented. t1 signal intensity measurements for the initial and last unenhanced brain mris were performed for dentate nucleus, pons, globus pallidus and thalamus. signal intensity ratios for dentate-to-pons (dnp) and globus pallidus-to-thalamus (gpt) were calculated and correlated with number of injections and time interval as well as therapy. differences between adults and pediatrics were assessed. mean age for the pediatric group was 9.3 years compared to 47.7 years in the adults. no significant difference was found for gender distribution (47 vs. 43% females) and follow up time (3.1 vs. 3 years). there was no difference concerning the signal intensities on first and last mri in children and adults (p=0.81/0.84, respectively). for each additional year of follow-up the change in ratio increases by 0.016 for adults but only 0.002 for peds (p=0.002). comparing therapies, in children a statistically significant difference between patients with and without former radiation was found (p<0.001) while there was no difference in adult patients with and without therapy (p=0.65). children and adults show a similar increase in t1 signal in deep brain nuclei ascribed to gadolinium deposition. in children, radiation and chemotherapy) seem to have a higher influence on gadolinium deposition. this correlation cannot be found in our adult cohort, indicating therapies have no (additional) influence. kearns-sayre syndrome (kss) is a rare mitochondrial dna-deletion syndrome characterized by early onset (<20 years), progressive external ophthalmoplegia and pigmentary retinopathy, often associated with cerebellar ataxia, muscle weakness, bilateral sensorineural hearing loss and cardiomyopathy. pyramidal symptoms may be present in kss, but they are poorly reported in the literature. through this case series, we aim to evaluate the concordance with the imaging patterns proposed by literature, correlating them with clinical and laboratory data, and to investigate possible microstructural damage with diffusion tensor imaging (dti) and magnetic resonance spectroscopy (mrs). we evaluated eight patients (8-19 years of age) with genetically confirmed diagnosis of kss. all pts. were studied with 3t/1.5t mri. in 5/8 pts. the study was completed by mrs and in 7/8 by dti imaging with reconstruction of cortico-spinal tracts (cst) using a 2-rois approach. a t-test comparative study between mean fractional anisotropy (fa) of cst in the 7 kss patients with dti and a group of 4 healthy controls was performed. cst reconstruction in a patient suffering from kss (images a-c), compared to an healthy control (images d-f). the dti study showed significantly reduced fa values, pointing out a possible microstructural damage. the disease showed an mr pattern of mixed white and gray matter signal abnormality, with periventricular and/or subcortical white matter hyperintense lesions, which in 1/8 patient presented a "tigroid pattern" (fig.2 ) three patients displayed a disease extension to the cervical spinal cord. (fig.2 ) dwi images demonstrated restricted diffusivity in almost all lesions (fig.3) , with persistence of low adc values. mrs study documented a high lactate peak in 2/8 pts. and a naa reduction in 5/8 pts; an increment of gsh was noted in one patient (fig.3) . the t-test comparative study of cst showed a significant reduction of mean fa value in kss patients compared to healthy controls (p= 0,004). involvement of the spinal cord (a-c, yellow arrows). comorbidity was suspected in "a" (myelitis). below (d-f): pale nuclei (d, green arrows) and subcortical white matter (e) alterations. right image displays the "tigroid pattern" (purple arrow). mrs showing the presence of a gsh peak, which may suggest an augmented antioxidative activity within the encephalic tissue. below: dwi hyperintensity in many regions of the brain in patients suffering from kss, due to diffusion resctriction. the integration of neuroimaging with clinical data can implement the diagnosis of mitochondrial diseases such as kss. according to our experience, comorbidities can delay the achievement of a correct diagnosis. the finding of an altered signal in the spinal cord of 2/8 pts. may suggest a new possible localization of the disease, while in one patient was referable to myelitis (fig.2, a) the evidence of a "tigroid patter" in should be taken in count in the differential diagnosis with lysosomal disorders. the presence of a prominent gsh peak may represent an augmented antioxidant activity, which may correlate with a more favorable outcome. an involvment of cst can be speculated even if pyramidal symptoms are poorly represented in kss. remotely distractible, magnetically controlled growing rod (mcgr, fig. 1 ) system has been developed to allow for gradual lengthening on an outpatient basis. this allows for safe spinal lengthening with continuous neurologic monitoring and real-time feedback by the patient. this study aims to evaluate retrospectively our ultrasound (us) geometric method and his accuracy compared with the plain radiograph (gold standard) for assessing mcgr distractions. this is a retrospective study that includes patients with early-onset scoliosis undergoing multiple consecutive distractions after mcgr implant. the rods length was measured for with us, for each distraction (3-months interval), and compared with plain radiograph follow-up (1-year interval). all patients included were treated with dual-rod systems. distraction length was monitored by a senior radiologist with us at each visit, one rod at a (2017) 47 (suppl 2):s297-s pediatr radiol time, before and after magnetic lengthening, with our geometric measurement method (fig.2) . low-dose upright two-projections radiograph were taken immediately after surgery and at 1-year intervals and measured by two radiologists (1 and 10 years of experience respectfully) (fig.3) . we compared measurements with the wilcoxon signed-rank test. from january 2014 to october 2016, a total of 5 patients (4 females and 1 male), which diagnoses included mitochondrial encephalopathy syndrome (n=1), spina bifida (n=1), ataxia of unknown cause (n=1), juvenile idiopathic scoliosis (n=1) and trisomy 8 (n=1), with a mean of 10 distractions per patient (standard deviation [sd] ±1,2), were recruited. fifty distractions for each system (95 measurements in total) were performed, targeting different lengths of distraction (from -2.0 mm to +4.6 mm) on each occasion. a total of 21 sets of plain radiographs were taken. from these, 18 sets of data points were used for correlation analysis. the mean distracted length per year on plain radiographs was 10,1 mm (sd ± 3,8 mm) and the mean distracted length on us per 6-months interval was 4,5 mm (sd ± 2,9 mm). excellent correlation was observed between radiographic and ultrasound measurements. in particular, correlation between rx measurements and ultrasound was excellent both for junior (0.001
0.96 for reader 1, > 0.88 for reader 2, and 1.00 for consensus between readers. kappas for consensus reads were 1.00 on all 15 structures (p< 0.001, lower 95% confidence limit >0.97). for reader 1, kappas were 1.00 for 14/15 structures (p< 0.001) and 0.00 for pcl. for reader 2, kappas were 1.00 for 14/15 structures (p< 0.001) and 0.45 for cartilage defects. paired t-test was used to compare mean likert scores for image quality characteristics. for both readers, sms was preferred for flow artifacts whereas tse was preferred for the three remaining image quality characteristics (p< 0.05). our primary assessment suggests that sms t2 tse is comparable to standard tse in terms of diagnostic performance in the evaluation of the pediatric knee despite modest decrease in overall image quality. the 2-fold decreased acquisition time of sms is a significant advantage which is felt to offset the mild decrease in image quality, particularly as it increases the likelihood that children will tolerate the examination without motion. mri for sacroiliitis in children: panel findings and inter-observer evaluation using standardised reporting k.e. orr 1 , m.j. bramham 1 , s. andronikou 2 ; 1 plymouth/uk, 2 bristol/uk there is little evidence regarding mri for diagnosing sacroiliitis in children with juvenile idiopathic arthritis (jia). the limited literature presents varied opinions but no published recommendations for standardisation of reporting. axial disease in jia responds poorly to conventional first-line treatments but identifying these children using history and examination findings is unreliable. standardised mri reporting (2017) 47 (suppl 2):s297-s pediatr radiol may improve diagnosis and selection of patients in whom newer biologic treatments are indicated. the aim was to use a standardised reporting proforma based on published definitions for recording mri findings in suspected sacroiliitis to evaluate inter-observer agreement and determine the reliability of findings according to specific sequences. ninety-nine sacroiliac joint mris (198 joints) were included, 80 were initial examinations and 19 were follow-up mris. the age range was between 6.6 and 20.3 years (mean age 15.4 years). three readers retrospectively reported all 99 mris using the standardised proforma. 'reader 1' was the study group panel while readers 2 and 3 were specialist paediatric radiology consultants working in the united kingdom. readers were blinded to additional clinical information and other imaging. inter-reader variation was evaluated for the presence of bone marrow oedema, erosions, effusions, ankylosis, sclerosis and enhancement, as well as the presence or absence of sacroiliitis. the quality of mri examinations was evaluated, including presence and adequacy of sequences performed and alignment of the coronal/oblique studies. mri findings were correlated with clinical details and final diagnosis. there is significant variability in sacroiliac joint mri protocols. refinement of these to include only necessary sequences based on inter-reader reliability and reinforcement of good positioning will improve reporting and result in universal standardisation. there is inconsistency in current reporting practice of sacroiliac joint mri in children but increasingly, clinicians rely on imaging to select patients with sacroiliitis and guide appropriate treatment. using a standardised reporting proforma may improve the quality and consistency of reporting. ultrasound-guided steroid tendon sheath injections in juvenile idiopathic arthritis s. peters, d.a. parra; toronto/ca objective: juvenile idiopathic arthritis (jia) is the most common chronic rheumatic disease in childhood. tenosynovitis is one of the manifestations of jia, which can explain the absence of response to treatment when adjacent joints are injected. steroid injection is one of the treatment options for tenosynovitis and it has been shown to be effective in the literature. utilizing ultrasound (us) guidance for injections into tendon sheaths has shown clinical advantage to conventional blind injections in the adult rheumatoid arthritis population. the aims of this study are to: (a) identify tendon sheaths most commonly treated in our patient population with jia referred for steroid injections; (b) describe technical aspects of the procedure; (c) characterize sonographic appearance of tenosynovitis in jia; (d) assess agreement between clinical request and sites injected. this was a 10 year single-center retrospective study ( may 2006 -april 2016 in which we recruited patients with jia referred by rheumatology for us-guided tendon sheath injections. we collected patient demographics, clinical assessment information, sonographic appearance of the tendons and technical aspects of the intervention from the procedure records. we collected data from 308 visits of 244 patients (75% female, mean age 9 years 8 months) with a total of 926 injections. the ankle region was most commonly injected (85%), specifically the tendon sheaths of tibialis posterior (22%), peroneus longus (20%) and brevis (20%). 63% of the procedures were performed under general anesthesia and triamcinolone hexacetonide was used in 97% of the injections. an "out of plane" approach was used in 86% of the interventions and the 15 mhz "hockey stick" us probe was preferred for guidance (86%). we found 2 minor intra-procedure complications without sequelae. the majority of treated sites (92%) showed peritendinous fluid and sheath thickening on us. other findings were increased color-doppler signal and echogenic peritendinous fluid. a strong agreement between clinical request and sites injected was observed and most patients required one visit (78%). us-guided tendon sheath injections are used frequently to treat patients with jia. it is a safe intervention with a high technical success rate. the ankle region, specifically the medial compartment, is the area most commonly injected in this cohort of patients. the most common sonographic finding is peritendinous fluid and sheath thickening. these findings might assist radiologists and rheumatologists to characterize and more effectively manage tenosynovitis in patients with jia. to evaluate the accuracy of the software for automatic bone age (ba) estimation based on deep learning technique, and to validate the feasibility of this system in clinical practice. the software for automatic ba estimation was developed based on deep learning technique using 18,940 left hand radiographs and estimated ba of each radiograph based on greulich-pyle method. ba estimation was done for left hand radiographs of 100 consecutive patients (9 months -17 years; 42 boys and 58 girls) in three methods: (1) ai bone age (assessed by the software), (2) ai-assisted ba (assessed by two radiologists with the assistance of the software), (3) gp atlas-assisted ba (assessed by two radiologists with only gp atlas but the software). the reference ba was determined by two radiologists by consensus. the accuracy of the estimated ba by each method was assessed using concordance rate (%), pearson's correlation analysis, the root mean square error (rmse), and bland-altman plot. reading time for ba estimation by each method was evaluated. ai bone age showed 61% of concordance rate, and a significant correlation with reference ba (r2=0.986, p<0.05). the bland-altman plot of agreement between the reference ba and ai bone age showed the mean difference of -0.20 years (95% limit of agreement, ±1.22 years). rmse was 0.42 years. in reviewer 1, concordance rates were same between both gp atlasassisted ba and ai-assisted ba (72%), and rmse of ai-assisted ba (0.20) was slightly lower than that of gp atlas-assisted ba s353 (2017) 47 (suppl 2):s297-s pediatr radiol (0.23). in reviewer 2, concordance rate was slightly higher in aiassisted ba (60%) than gp atlas-assisted ba (58%), and rmse was almost the same (0.54 in ai-assisted ba, 0.55 in gp atlasassisted ba). the reading time was reduced 20.0% in reviewer 1 and 62.7% in reviewer 2. the software for automatic ba estimation based on deep learning technique showed high accuracy and may enhance work efficiency in ba estimation by allowing radiologists to save reading time and to improve accuracy. temporomandibular joint mri findings in adolescents with primary disk displacement in comparison to those in juvenile idiopathic arthritis j. bucheli, d. ettlin, c. kellenberger; zurich/ch to investigate potential differences of morphology and degree of inflammation in temporomandibular joints (tmjs) affected by primary anterior disk displacement (add) and juvenile idiopathic arthritis (jia). in 18 adolescents (15 female, age 15 ± 2 y), contrast enhanced magnetic resonance images (fig. a) of tmjs with add were retrospectively compared to those of age-and gender-matched controls with jia. morphology of articular disk and bony structures were described. osseous deformity and inflammation were qualitatively scored with progressive 4-grade scales and compared between groups with mann-whitney-u test. mandibular ramus length, measured on gradient echo minimum intensity projection images (fig. b) , was compared between groups and to normal values with independent samples t-test. in the add-group, 31/36 disks were dislocated anteriorly and showed thickening of the posterior band (27/31). in contrast, tmj disks of jia patients were mainly flattened (n=23) and/or centrally perforated (n=12) and rarely dislocated (n=1). tmjs with add showed similar overall grades of inflammation (p=0.39) and osseous deformation (p=0.53) as tmjs in the jia group. while erosions were frequent in both groups (add 25/31; jia 32/36, p=0.55), the mandibular condyle (p<0.001) and glenoid fossa (p<0.001) were less flattened in tmjs with add. in add tmjs, bone marrow oedema was less frequent (p=0.001) and grades of joint enhancement slightly lower (p=0.03), but presence of synovial thickening (p=0.43) and degree of effusion (p=0.87) were not significantly different between groups. mandibular ramus length was not significantly different (p=0.72) between groups, but in both groups clearly decreased compared to mean normal values (p<0.0001). articular disks in tmjs affected by jia are rarely dislocated. surprisingly, tmjs with primary add show considerable inflammatory change including condylar erosions. still, chronic systemic inflammation in jia joints results in considerable higher deformity of the mandibular condyle and the temporal joint surface. observation of the mostly preserved normal shape of the temporal bone may help differentiating primary add from jia. retrospective magnetic resonance imaging (mri) study of 49 consecutive jia patients (35 female, median age 14 y) with at least two consecutive tmj mri examinations ≥ 2y apart and no csi. degree of tmj inflammation was determined on t2-weighted and contrast-enhanced t1weighted fast spin echo images (fig. a) , and degree of osseous deformity on gradient echo images (fig. b) by progressive 4-grade scales (0-3). change of respective grades was assessed with wilcoxon test. mandibular growth was determined by ramus length change and compared to normal values. over a median period of 3.4 y (interquartile range, 2.4 -4.6 y), degree of tmj inflammation improved (p<0.001) with decrease in frequency of grade 3 (4.1% to 0%) and grade 2 (19.4% to 4.1%). inflammatory grades improved both in patients with (n=39, p=0.007) and without (n=10, p=0.02) systemic disease modifying medication. the degree of osseous deformation slightly improved (p=0.04), with decrease in frequency of grade 3 (5.1% to 3.1%) and grade 2 (9.2% to 6.1%), and increase of grade 0 (48% to 54.1%). overall growth rates of mandibular ramus (median, 1.3 mm/y) were not significantly different from normal growth rates (p=0.27) (fig. c) . growth rates of tmjs from patients only receiving non-steroidal anti-inflammatory drugs (median, 1.25mm/y) were not significantly different (p=0.9) compared to patients treated with systemic disease modifying drugs (median, 1.35mm/y). in patients with systemic treatment of jia, both the degree of tmj inflammation and osseous deformity as seen on mri improved at midterm follow-up. normal growth of the mandibular ramus was maintained. these results are in contrast to those from an earlier cohort treated with csi, in which on average deformities deteriorated and growth was impaired. objective: pediatric ileocolic intussusception, ici, is a common abdominal condition for which pediatric radiologists are asked to attempt emergency pneumatic reduction. because of the high success and low complication rates of pneumatic reductions, radiologists are able to make several attempts at reduction in stable patients if the initial enema attempt is unsuccessful. we have observed patients with successful reductions with rather long periods between initial symptoms of ici and performance of the air enema. we hypothesize that successful pneumatic reduction rates are independent of length of symptoms and in stable patients, repeated reduction attempts can be performed with the expectation of successful reduction. we performed an irb-approved retrospective review of all ici with a pneumatic reduction attempt between 2008-2016 at xxx. clinical, imaging and surgical data was reviewed. time to enema was defined as the time from first symptom to first air enema attempt. linear and second order polynomial statistical analysis was performed to assess the relationship between time to enema and enema outcome. results: 66 ici were identified in 61 patients. air enema was successful in 46 ici, 77%. the mean time to enema was 37.5 hours, range 4-168 hours with sd of 42.2 hours for successfully reduced ici and 35.1 hours, range 2-336 hours with sd of 53.9 hours for unsuccessfully reduced ici. surgical resection was required in 4 patients with ischemic bowel including one with an irreducible meckel's diverticulum as lead point. there was no correlation between time to enema and successful reduction, fig 1. no patient with a successful pneumatic reduction of a ici required subsequent bowel resection. conclusions: air enema for ici can be safely performed despite prolonged time to enema with the anticipation of a successful reduction. the lack of correlation of pneumatic reducibility and time to enema suggests that in surgically cleared patients with ici, the pneumatic reduction attempt may not be a true emergency and that repeated attempts at reduction are safe. additionally, though our numbers are small, they suggest that an ici is reducible or not from the beginning and do not "become irreducible" with prolongation of the time to enema. evaluation of splenic stiffness measurements for the diagnosis and the follow-up of portal stenosis after paediatric liver transplantation c. escalard 1 , a. dabadie 2 , s. chapeliere 1 , d. pariente 1 , c. adamsbaum 1 , s. franchi 1 ; 1 le kremlin-bicêtre, paris/fr, 2 la timone, marseille/fr to report our preliminary findings about the role of splenic and hepatic supersonic shear-wave elastography (sswe) in the diagnosis and followup after treatment of portal stenosis in paediatric liver graft recipients. all paediatric liver recipients with portal stenosis treated by the interventional radiology procedure, and who underwent splenic and hepatic sswe pre and post interventional procedures, were retrospectively reviewed. demographics, data about the portal stenosis (delay post transplantation, clinical presentation, initial radiological findings, hemoglobin and platelet counts), ir procedure performed, clinical and ultrasonographic follow-up and spleen stiffness pre and post ir procedure were collected. four patients were included, median age 6,5 years (range 0,9 months to 8 years) and median delay post transplantation 3,9 years (range 1 month to 4.5 years). two patients presented with anemia, associated in one case with progressive splenomegaly. one patient had liver test abnormalities, and one had decreased portal flow found on systematic doppler followup. spleen stiffness was elevated pre-procedure in all 4 patients, from 36 to 65 kpa (normal <20 kpa), and liver stiffness was normal or mildly elevated in all. portal stenosis was successfully treated by ir in 3 patients. spleen stiffness decreased rapidly, ranging from 38 to 53% (figure 1) . however, the size of the spleen remained unchanged. in the last patient, angioplasty of the portal stenosis failed leading to portal thrombosis. spleen stiffness increased on the subsequent ultrasound ( figure 2 ). mr elastography (mre) is a novel imaging technique that provides a non-invasive evaluation of liver fibrosis. the standard sequence used for this purpose on a siemens scanner has been gradient echo (gre). we also implemented echo planar imaging (epi) available as a work-in-progress (wip). our aim is to compare the liver elastogram values between gre and epi in children. after consent from both research and referred clinical subjects, a dedicated mre of the liver was performed on a 3t mr scanner (magnetom® skyra, siemens) with a pediatric mechanical driver over the right upper quadrant. an axial t2 blade with fat saturation, coronal t1 vibe dixon and axial diffusion weighted imaging (dwi) were obtained. elastograms were obtained using both standard gre and epi, in the axial plane. for the gre sequence, 5 different slices were selected and each scanned sequentially. the epi sequence incorporated 5 different slices in just one series. images were post-processed placing regions-of-interest (roi) and measuring the stiffness in kilopascals (kpa). for each sequence and each slice the mean stiffness and then the average of the means was calculated. a spleen elastogram was simultaneously generated, without changing the mechanical driver location, and the mean stiffness was also calculated. increased stiffness was defined as >2.9 kpa in the liver and >3.6 kpa in the spleen. we focused on a technical comparison between the sequences without clinical or histological correlation of findings. we included 15 subjects that had elastogram measurements of liver and 11 of them spleen stiffness on both gre and epi sequences. mean liver stiffness on gre was 2.4 (sd+/-0.71) and on epi was 2.8 (sd+/-1.04), with a pearson's correlation of r= 0.92 (p<0.001). increased liver stiffness was found in 4/15 (26.6%) of the cases in gre and 9/15 (60%) of the cases in epi. mean spleen stiffness on gre was 3.9 (sd+/-1.39) and on epi was 4.8 (sd +/-1.45) with a pearson's correlation of r=0.69 (p= 0.01). epi reported consistently higher values than gre in both liver and spleen stiffness. our preliminary data shows a moderate to high correlation between gre and epi sequences; however, the epi values were higher in both liver and spleen. in the future, larger studies are needed to validate these thresholds and patterns among different sequences. were also reviewed if done. patient's medical & surgical treatment, and clinical progress were also reviewed. active telephone follow-up 3 days after cevus was performed. results: 120 patients giving a total of 240 pelviureteric units were referred for vus study during the study period, with age ranging from 1 month to 7 years old. no contrast-related complication was encountered. except 2 cases with failed catheterization, 108 were investigations of urinary tract infection (uti), antenatal hydronephrosis and congenital anomalies etc., and remaining 10 were follow up studies of known reflux. of all cases of uti, 36 refluxing units were picked up by vus, ranging from grade i to v. of the 36 refluxing units diagnosed by cevus, 17 were missed on mcu, among which 12 were high grade refluxes (grade iii to v) requiring treatment; whereas cevus only missed one grade i refluxing unit detected by mcu. besides, one grade iv refluxing unit identified on vus was under graded by mcu to grade i. regarding patient outcomes, one patient with mcu-missed refluxing unit presented with breakthrough uti on follow up. two refluxing units that were missed on mcu but detected on cevus demonstrated scarring on dmsa. conclusion: cevus is shown to be more sensitive in detecting vesicoureteric reflux than mcu. the fact that mcu-missed refluxes detected by cevus were associated with breakthrough urinary tract infection and scarring on dmsa indicated that the extra sensitivity brought by cevus did translate to clinical significance. difficulty in visualizing low-grade reflux is a potential limitation of this technique. with favourable diagnostic performance and safety profile, cevus can be further applied in this community in the era of radiation reduction. percutaneous transbiliary needle or forceps biopsy in hepatic masses with biliary dilatation a. dabadie 1 , s. franchi 2 , d. pariente 2 ; 1 la timone, marseille/fr, 2 le kremlin-bicêtre, paris/fr hepatic masses with biliary dilatation are rare in children and mainly include rhabdomyosarcoma of the biliary ducts, but also other masses or pseudo-masses compressing the hepatic hilum. in these patients histological diagnosis of the lesion as well as temporary biliary drainage are warranted. the objective of this study is to report our experience in percutaneous transbiliary biopsy performed simultaneously and using the same access as the percutaneous biliary drainage in children with hepatic mass obstructing the biliary ducts. children presenting with a hepatic mass causing biliary obstruction, with need for biliary drainage, were considered candidates for percutaneous transbiliary biopsy of the lesion performed at the same time. the biopsy was performed under ultrasound guidance, through a sheath introduced in the dilated biliary system, using a semi-automatic 16 gauge needle or the transluminal biliary biopsy forceps set (cook medical, bloomington, usa). between 2009 and 2016, four patients were included, three females and one male, median age 5.5 years (range 2.5-11.5). all presented with jaundice and were diagnosed with a hepatic mass with secondary biliary obstruction. percutaneous transbiliary biopsy was performed in all 4 patients using the 16 gauge needle. in one patient, the biopsy did not demonstrate any tumoral cells and a second biopsy was performed using the forceps device through the same biliary access. the samples deemed adequate for analysis by the pathology department in all patients, however the samples were larger when using the needle. a retrospective -prospective study included 62 patients of both sexes (13,72 +/-3,17 y), in a two-year span. patients were divided into two groups according to the used diagnostic method (positivegroup a on us and a1 on mri, with intestine mural thickness above 3mm, and negativegroup b on us and b1 on mri, with mural thickness below 3mm). overall sensitivity and specificity of us and mri in diagnosing ibd was calculated in comparison to pathohistological (ph) findings. us examination showed an average intestinal mural thickness of 4.93 ±1.39mm and 2.7+0.18mm in group a (28 patients) and group b (34 patients) respectively. mri examination showed an average intestinal mural thickness of 6.50±1.45mm and 2,8+0,16mm in group a1 (27 patients) and group b1 (35 patients) respectively. out of 28 patients from group a, 15 (52%) had irregular mural architecture, contrary to group b in which mural architecture irregularities have not been observed. in groups a1 and b1 14 (51.9%) and 2 (5.7%) patients had irregular mural architecture respectively. average length of affected intestinal segment on us and mri was 103mm and 105mm respectively. five patients from group a and four from group a1 had signs of fibrosis. color doppler showed hyperemia in 17 and 22 patients of group a and a1 respectively. transmural signs of inflammation were found in 59% of patients on us, and 61.3% of patients on mri. average longer diameter of mesentery lymph nodes measured by us and mri was 13.29 ±3.74mm and 12.7±5.68mm, respectively. overall sensitivity of us and mri was 88.4% and 92.31% respectively. both us and mri showed a specificity of 88%. us and mri are reliable and compatible methods in diagnosing ibd, with mri being slightly more accurate. us is an extremely valuable and widely available imaging modality in every-day clinical work, both in diagnosing and follow-up of therapy effects in children with ibd. findings in percutaneous transhepatic cholecysto-cholangiography in neonates and young infants presenting with conjugated hyperbilirubinemia d.a. parra, s. peters, j. amaral; toronto/ca objective: conjugated hyperbilirubinemia is a concerning finding in neonates and young infants, biliary atresia (ba) being one of the main diagnostic considerations. ba is a rare disease characterized by fibrosis of the biliary tree. the obliteration of the biliary system leads to cholestasis and ultimately liver parenchymal injury, cirrhosis and death. an early diagnosis of ba along with a kasai portoenterostomy operation significantly improves the long-term prognosis. percutaneous transhepatic cholecysto-cholangiography (ptcc) is one of the options described in the diagnostic algorithm of ba. the aims of this study are to: (a) describe ptcc findings in patients with conjugated hyperbilirubinemia; (b) identify the abnormal patterns encountered that justify further investigations; (c) analyze technical aspects of the procedure. this is a 16 year single-center retrospective study (2000) (2001) (2002) (2003) (2004) (2005) (2006) (2007) (2008) (2009) (2010) (2011) (2012) (2013) (2014) (2015) (2016) in which we recruited patients with the diagnosis of cholestasis (less than 6 months old) referred for ptcc. we collected patient demographics, clinical information, findings in ptcc, post-procedure management and long term clinical outcome. eigthy-nine patients were referred for ptcc in the study period. the procedure was technically feasible and successfully performed in 73 patients (68% male, mean age 2.2 months). forty-one had a pre-procedure hida scan suggestive of ba. fifty-nine patients had an ultrasound-guided biopsy in conjunction with the ptcc and in all of them the cholangiography was performed through a needle placed using ultrasound guidance in the gallbladder. 53% (39) of the patients had a normal ptcc. abnormal patterns encountered were: 1) variable degrees of hypoplastic bile ducts seen in 25%; 2) atretic gallbladder without demonstration of communication with bile ducts seen in 18%; and 3) gallbladder communication with a cystic structure not communicated with the biliary ducts (cystic biliary atresia) seen in 4%. the most common diagnosis in the abnormal group was ba (71%). alagille's syndrome, alpha-1 antitrypsin deficiency and progressive familial intrahepatic cholestasis were other diagnoses in this group. no complications related to the procedure were observed. ptcc is a safe and effective option in the diagnostic algorithm of patients presenting with cholestasis early in life. visualization of the gallbladder is fundamental to perform the procedure. the majority of studies were normal in our patient population preventing further invasive investigations. three types of abnormal ptcc patters were encountered, with ba being the most common diagnosis in this group of patients. to evaluate the additive role of shear wave elastography in the sonographic distinction of biliary atresia from other causes of neonatal/ infantile cholestatic liver disease. neonates and infants with clinical and biochemical diagnosis of cholestatic jaundice were enrolled in our study after obtaining informed written consent from the parents. grey scale, doppler and shear wave elastographic findings were recorded after 4 hours of fasting using aixplorer® ultrasound system (supersonic imagine, aix en provence, france). sedation was not needed during the study. for obtaining elastographic values, linear transducer (4-15hz) was used and after image stabilization a q-box measuring 3mm was placed in the most homogenous vessel free area. the mean of three elastographic values were recorded. hida scan, liver biopsy, intra-operative cholangiogram and histopathological evaluation of resected specimens was done wherever feasible and clinically indicated. the prospectively obtained elastographic values were retrospectively evaluated. eleven of 25 patients included in our study were proven to be biliary atresia (ba) by intra operative cholangiogram and histopathological reports. the diagnosis in the remaining 14 patients included other causes of infantile cholestatic jaundice like infantile choledochal cyst, neonatal idiopathic hepatitis, progressive familial intrahepatic cholestasis, abernathy malformation, cmv hepatitis etc. the elastographic values of ba and non-ba patients were compared. six of 25 infants were younger than 60 days which included four patients with ba and their elastographic values (18.75±2.9kpa) were significantly different from that of non-biliary atresia (7±1kpa) in the same age group (p value <0.05). similarly, for patients aged >60 days also we had a significant difference (p value <0.05) in elastographic stiffness between ba (45.7±11kpa; n=7) and non-ba (19.4±3.6kpa; n=12) groups. the mean echogenic area anterior to right portal vein (earpv) was 4.45 ±0.84mm in ba and 1.47±0.36mm in non-ba group (p value <0.05). the mean gall bladder (gb) length was 1.82±0.26 mm in biliary atresia group in contrast to 3.2±0.26mm in the rest (p value <0.05). the roc plot for earpv and gb length gave a youden index cut off value of >2.7mm (sensitivity 72.7 & specificity 71.4%) and <3.32 cm (sensitivity 100 & specificity 42.9%) respectively. infants with biliary atresia have a significantly higher elastographic value when compared to age matched patients with other causes of neonatal cholestasis. we expect to validate the findings in our ongoing study with a larger sample size. to retrospectively define in a large pediatric population the association between testicular microlithiasis and testicular neoplasia. retrospective multicenter study of scrotal ultrasounds performed between january 2000 and april 2014 in subjects <18 years of age. all unique subject scrotal ultrasound reports from each institution were reviewed for mention of microlithiasis. for subjects with serial exams, the most recent exam performed was included in the analysis. all exams mentioning microlithiasis were reviewed by site-specific investigators to confirm the presence of ≥5 punctate calcifications in the testicle on a single image. the presence of testicular germ cell and stromal tumors were determined for subjects with and without microlithiasis through review of institutional pathology and imaging databases. the risk of testicular neoplasia in the context of microlithiasis was expressed in terms of odds ratios with (a-or) and without adjustment (u-or) for fixed study site (institution) effects by logistic regression. the study population included 37,863 unique subjects with confirmed microlithiasis in 1,097 (2.9%). mean subject age was 11.1±4.7 years for subjects with microlithiasis and 9.1±5.9 years for subjects without (p<0.0001). one hundred thirty-nine subjects (0. this large, multicenter study confirms that there is a significant, strong association between testicular microlithiasis and testicular neoplasia, particularly malignant germ cell tumors. children with microlithiasis have approximately 22x greater odds of having a malignant germ cell tumor than children without microlithiasis. this reinforces the need for a large prospective study assessing the risk of developing testicular neoplasia in children with incidentally identified diffuse microlithiasis. do adc-values reflect renal function or obstruction in children with uretero-pelvic-junction obstruction? p. grehten, a.c. eichenberger, c. kellenberger; zurich/ch the use of diffusion weighted imaging (dwi) in renal mri is increasing. in adults as well as in infants a positive linear correlation between adcvalues and glomerular filtration rate has been demonstrated. the aim of our study was to assess whether renal dwi can provide information on the grade of urinary tract obstruction or renal function in children with uretero-pelvic-junction (upj)-obstruction. retrospective analysis of 19 children (age 3.1+/-4.5y) with unilateral upj-obstruction who underwent pre-and postoperative mri at 1.5t and 6 normal controls (age 6.6+/-4.0y). functional mr-urography and multiple b-value dwi were part of the mr-protocol. renal adc-values were correlated to measures of obstruction and function, and compared between obstructed and non-obstructed kidneys and between pre-and postoperative studies. no correlation was found between mean parenchymal, cortical or medullary adc-values and calyceal transit time (ctt), renal transit time (rtt) and measures of differential renal function (%parenchymal s359 (2017) 47 (suppl 2):s297-s pediatr radiol volume, vdrf, pdrf). there was moderate correlation with absolute parenchymal volume and total kidney volume, and low correlation with pelvic volume. adc-values showed high correlation with age and patient's weight. adc-values normalized for age or weight showed low correlation with rtt and ctt, but no correlation with functional measures. adc-values were not significantly different between obstructed and contralateral normal kidneys (p=0.2-0.9) or between pre-and postoperative studies (p=0.3-1). renal adc is dependent on age and weight in young children and does not correlate with differential renal function. for assessing urinary tract obstruction with adc normative values need to be established. to determine the level of knowledge and awareness of medical staff, medical students and parents concerning possible risks associated with ionizing radiation. a prospective study has been conducted at children's hospital, center for adult's radiology, and medical faculty, by filling out two anonymous questionnaires (questionnaire 1medical staff and medical students, questionnaire 2parents of the children exposed to x-ray based procedures), and it included 254 participants. statistical analysis was performed using the spss 21.0. the majority of examinees assessed their knowledge about ionizing radiation as moderate. knowledge level was statistically significantly higher only in the group of medical students who passed the course of radiology, in comparison to the group of those who have not attended the course yet. only 45% of radiologists and up to 37.5% of pediatricians, pediatric surgeons and anesthesiologists are informed about "image gently" campaign. up to 80% of radiologists, and up to 22% of clinicians, both specialists and residents, are aware of alara principle. over 60% of medical doctors think that diagnostic radiology procedures are very often performed unnecessarily among children, while only 12.5% of parents share this opinion. most of the radiologists and clinicians consider it necessary to inform parents about potentially harmful effects of ionizing radiation, but even though 60-80% of clinicians claim they do inform parents in every-day clinical practice, over 70% of parents affirm that they had never been informed about effects of ionizing radiation before diagnostic procedures were performed on their children. only 26% of pediatric surgeons and pediatricians, but 72.7% of radiologist and 60% of anesthesiologists are concerned that informing parents about ionizing radiation would cause problems in every-day work. nearly 71% of parents claimed that they would not refuse to expose their child to x-ray based diagnostic procedure, after the given information about potential harmful effects. over 70% of radiologists and less than 50% of pediatric surgeons and pediatricians support the initiative to calculate the total effective dose child was exposed to during hospitalization, and place it on the discharge list. between 50% and 85% of pediatricians and pediatric surgeons greatly underestimated the effective doses in ct and fluoroscopy procedures. there are 58-100% of clinicians who are aware that ct increases the risk of carcinoma development. this study showed that general knowledge about ionizing radiation, potential risks and effective doses in pediatric population is poor, and that organized education is required. fluoroscopy in pediatric radiology -how important is an individual impact to radiation exposure of children? j. lovrenski, i. varga; novi sad/rs to determine whether there are differences between different pediatric radiologists and radiology residents in exposure of pediatric population to ionizing radiation during fluoroscopy procedures. a retrospective study has been conducted at the regional children's hospital, and included all the diagnostic fluoroscopy examinations performed within a one-year period. the fluoroscopic data along with the names of pediatric radiologists/radiology residents performing these examinations were retrieved from the evidentiary notebooks, and included: dose-area product (dap), skin dose, and fluoroscopy time. there were 4 radiologists (r1-r4), and 4 radiology residents (r5-r8) involved in fluoroscopic examinations. we found all the fluoroscopic findings in the hospital's data base, which enabled a differentiation between positive and negative findings. statistical analysis was performed using the spss 21.0. a p-value less than 0.05 was considered statistically significant. a total of 191 fluoroscopy procedures in children (mean age 4,5y, 107 males and 84 females) have been performed within a one-year period, most of which were voiding cystourethrograms (vcug) -93, and an upper gastrointestinal (gi) series -79 examinations. radiology residents and radiologists carried out 82 and 109 examinations respectively. duration of fluoroscopy procedures performed by residents (av. 32.5s) was statistically significantly shorter in comparison with duration of fluoroscopy examinations performed by radiologists (av. 53s). dap and skin dose did not show statistically significant difference between these two groups, as well as the number of positive and negative fluoroscopic findings in groups of examinations performed by radiologists and radiology residents. mean dap value ranged from 0.82μgym 2 (r2) to 7.5μgym 2 (r7) when performing vcugs, and from 1.35μgym 2 (r2) to 4.79μgym 2 (r1) for upper gi series. mean skin dose ranged from 14.33mgy (r2) to 151.48mgy (r3) for vcugs, and from 28.66mgy (r2) to 120.54mgy (r4) for upper gi series. mean fluoroscopy time ranged from 11.76s (r2) to 48.5s (r3) for vcug, and from 26.22s (r2) to 125.75s (r3) for upper gi series. statistically significant difference was shown only between radiologists r2 and r3 for dap and skin dose values in performing vcug, and for fluoroscopy time in performing an upper gi series. for all examinations dap and skin dose were statistically significantly higher in the group of positive fluoroscopic findings. this study has shown that exposure of children to ionizing radiation during fluoroscopy procedures significantly depends on radiologist/ radiology resident and the nature of fluoroscopic finding. to evaluate image quality and radiation exposure of non-contrast pediatric chest ct with automated tube voltage selection (atvs), in combination with automated tube current modulation (atcm). non-contrast chest ct scans of 160 children (91 male and 69 female; mean age, 8.7 ± 5.4 years) were analysed retrospectively with regard to radiation exposure and image quality before and after the implementation of an automated tube voltage selection. correlations of volume ct dose index (ctdi vol ) and the effective diameter (edm), before and after the implementation of atvs were compared, and confidence intervals related to the change in correlations with and without atvs were determined using fisher's z-transformation. image quality was assessed by mean signal-difference-tonoise ratios (snrs) in the aorta and in the left principal bronchus with the independent samples t-test. subjective image quality was rated by two pediatric radiologists and a general radiologist on a 10point scale. agreement between the readers was assessed using weighted kappa coefficients. a p<0.05 were considered significant. automated tube voltage selection, in combination with an automated tube current modulation, resulted in optimization of scan protocols, homogeneity of image quality, and reduction of radiation exposure for pediatric patients. advantages and disadvantages of cone beam ct for pediatric interventions l. dance, r.b. towbin, d. aria, c. schaefer, r. kaye; phoenix/us objective: illustrate the advantages and disadvantages of cone beam ct (cbct) as an alternative to conventional ct guidance and an adjunct to angiography. there is a steep learning curve to optimize utilization of cbct. we found that cbct reliably identifies high-contrast lesions. however, the lower dose and decreased penetration of cbct resulted in poorer visualization of low-contrast lesions. also cbct can be degraded by streak artifact from hardware or dense contrast. the relatively narrow field of view can be restrictive for peripherally located lesions in larger patients. however, the anatomic display is adequate for guidance in most instances. these findings are illustrated in a series of cbct-guided cases including pulmonary nodule localization, osteoid osteoma ablation, abc sclerotherapy, renal av fistula embolization, and liver lesion biopsy. the advent of cbct as an adjunct modality in the ir suite has significantly decreased the use of conventional ct guidance and significantly decreased the radiation dose in children. we have found cbct to be a practice changer. the aim of this study is to review our local drl in pediatric fluoroscopy and to compare them to values proposed by pidrl guidelines and recent international litterature. data were prospectively collected on consecutive procedures (750 total) performed from january 2016 to december 2016 on 2 different fluoroscopy units (siemens iconos r200, luminos drf). of each procedure patients data (name, weight and birth date), examination-data (kind of procedure, date, dap [cgy*cm2], total fluoroscopy time, number of images) were recorded. data from micturating-cystourethrography(mcu), barium meal/swallow(bs) and most commonly performed procedures were divided into 4 weight-groups (<10kg,10-15kg,15-30kg,30-60kg) and of each one 75th-percentile was calculated. data were compared to europeandrl and recent literature data (by age:newborn,1-,5-,10years old). weight-groups are considered a representative sample if at least 20-patients per procedure-type and per patient-group are included. our local-drl for mcu are 7(<10kg), 10(10-15kg), 24(15-30kg) and 57(30-60kg). they results to be lower than pidrls values (30, 70, 80, 75) but higher if compared to a previous local survey of 2014 (4, 9, 18, 29) . bs data are 9(<10kg), 32(10-15kg), 28(15-30kg); these data are lower than that of a previous local survey of 2014 (23, 34, 68) . the update of local-drl is helpful in daily practice to identify (and solve) critical issues such as incorrect technique or poor practice with new flat-panel equipment. pidrl guidelines: a review of local drl for pediatric head, thorax and abdomen ct in a italian referral center a. magistrelli, v. cannatà, e. genovese, m. cirillo, r. lombardi, p. toma; rome/it the aim of this study is to review our local drl in pediatric ct and to compare them to values proposed by pidrl guidelines and recent international litterature. data were prospectively collected on consecutive procedures (347 total) performed from january 2016 to june 2016 on a somatom definition flash siemens. of each procedure patients data (name, weight and birth date), examination-data (kind of procedure, clincal question, date, ctdivol16/32 and dlp16/32) were recorded. ctdivo/dlp16 from head ct were divided into 4 age-groups (<4weeks,4weeks-1y,1-6y,≥6y) and of each one 75th-percentile was calculated. ctdivol/dlp32 from thorax (chest, cardiovascular ct angiography) and abdomen+pelvis ct examination were divided into 5 weightgroups (<10kg,10-15kg,15-30kg,30-60kg,>60kg) and of each one 75th-percentile was calculated. data were compared to europeandrl and recent literature data. weight-groups are considered a representative sample if at least 20patients per procedure-type and per patient-group are included. our local drl are substantially lower than that proposed by pidrl guidelines. specifically ctdivol/dlp32 for chest ct are 1/22(<5kg), 1,52/42(5-15kg), 1,83/56(15-30kg), 2,99/113(30-60kg), 6,07/ 239(>60kg) respectively. for cardiovascular ct angiography are 0,71/ 15(<5kg), 1,01/21(5-15kg), 2,44/36(15-30kg), 2,87/90(30-60kg), 13,72/311(>60kg). while for abdomen+pelvis ct are 1,9/47(<5kg), 2,68/83(5-15kg), 3,25/131(15-30kg), 7,77/320(30-60kg), 11,21/ 532(>60kg). data for trunk sere not collected. for head ct local drl are higher in age-group 0 and 1 but lower in other age-group if compared to routine head ct pidrl ones. the update of local-drl allowed us to identify (and solve) some critical issues such as incorrect technique. drl-curve in optimization of pediatric body ct r. seuri 1 , p. laarne 2 , a. nikkola-sihto 3 , k. nygaard bolstad 4 , m.s. perhomaa 5 , a. thilander klang 6 , k. rosendahl 4 , j. ruohonen 3 , e. tyrvainen 7 ; 1 helsinki/fi, 2 tampere/fi, 3 seinäjoki/fi, 4 bergen/no, 5 oulu/fi, 6 gothenburg/se, 7 kuopio/fi objective: diagnostic reference levels (drls) in medical imaging represent valuable tools to study dose optimization in clinical practice. this is particularly important in pediatric computed tomography (ct) as the number of the examinations in many institutions is low. drls are typically given as a percentile point, usually as 75% or 3 rd quartile of the observed distribution of patient dose. in pediatric practice drls are often given for each age-or weight group separately. we present continuous drl-curve as a feasible way to compare dose levels in pediatric body ct. during 2016-2017 a selected group of nordic hospitals collected dose values (ct dose index by volume, ctdi vol , and dose-length product, dlp) from pediatric body ct examinations on children aged 5-16 years. the dose values were imported into a dynamic excel table, previously established by the radiation and nuclear safety authority in finland, stuk (fig 1) . the stuk-table includes a graphic presentation of a continuous drl-curve presented as a function of body weight, and the program automatically calculates a dose curve and compares it to the established reference level (fig 2) . the dose values were easily exported to the excel tables, and the graphic presentation and comparison with an established drl-curve was clear and readily understandable for both radiologists and radiographers. in some of the institutions included in the present study, the weight of the patient was not recorded routinely. this represents a challenge for the use of the drl-curves provided by stuk. the drl-curves provided by stuk were feasible for clinical practice. the automatic calculation of the dose curve and graphic presentation were helpful to interpret the results. the drl-curve also allows relevant comparison even with a smaller number of patients. fifty randomly selected ct chest studies performed over 10 years to assess diffuse lung disease were included in the study sample (25 females, 25 males; mean age 9.9 years + 6.6 years), comprising 9 disorders. two pediatric radiologists and a pediatric radiology fellow blinded to the results of the cts evaluated four subsets of complete chest cts (3 slices, every third slice, every other slice, and all images below the thyroid) and compared the subsets with the entire chest ct, interpreted as the control. accuracy of evaluating the primary diagnosis and determination if significant diagnoses were missed in the reduced slice ct subsets were rendered. we assume linear distribution of dose across the anatomy to estimate dose reduction on reduced slice subsets. most significant findings were present on all reduced slice ct subsets. all relevant findings were present in 100% of subthyroid, 96% of every other slice, 86% of every 3 rd slice, and 50% of 3 regional slice subsets respectively. excluded findings included small foci of ground glass opacity, consolidation, focal mosaic attenuation, and linear parenchymal bands; peribronchial thickening, dextrocardia vs dextropositioning, tree-in-bud opacities, extent of mild bronchiectasis. with the exception of consolidation in 1 of the studies, these findings were not thought to inhibit diagnostic assessment. the underlying diagnosis was correctly identified in most of the subsets: 100% subthyroid and every other slice, 90% every 3 rd slice, and 80% of 3 regional slice subsets. dose is significantly decreased by using any of these methods. while some findings are excluded with increasing gaps between slices, equivalent diagnostic information can be provided on reduced slice ct and can serve as a viable strategy to reduce lifetime radiation dose to children and young adults with diffuse lung disease imaged for routine follow-up. as findings are missed with larger gaps, this strategy should be used with caution in patients presenting with acute symptoms 3. to extrapolate the significance of early diagnosis which will compliment to treatment planning and management. case presentation: types a and b niemann-pick disease are lysosomal storage disorders that result from deficient acid sphingomyelinase activity and lead to the accumulation of sphingomyelin, primarily in tissues of the reticuloendothelial system. type b niemann-pick disease manifestations are hepatosplenomegaly, excess bleeding and bruising, growth retardation, and recurrent respiratory infections. features of hrct include thickened peribronchovascular and interlobular septal thickening, ground-glass opacities. the intermixed regions could be characterized as showing crazy paving, although this is not the predominant pattern. type b niemann-pick disease should be added to the list of clinical entities that can demonstrate crazy paving. our patient is a sevenyear old girl, presented with dry cough and fever. physical examination revealed hepato splenomegaly. radiological work up included abdominal ultrasound examination, which showed mild hepatosplenomegaly. chest radiography revealed diffuse reticulonodular infiltration in both lungs. chest hrct was done for more comprehensive evaluation which showed multilobar bilateral peribronchovascular interstitial thickening and interlobular septal thickening with ground-glass opacities and crazy paving appearance. no honeycomb pattern was seen. no sizable pulmonary nodule or sizable mediastinal lymphadenopathy was seen. no pleural effusion was seen. finding were indicating extensive pulmonary intestitial disease. a corroborative analysis along with lab tests and genetic studies revealed the diagnosis of type b niemann pick disease. 2. the spectra of hrct features including crazy paving pattern may be encountered; though not frequent. hence should be included in the differential diagnosis of crazy paving pattern. blast from the past: lemierre's syndrome in adolescents with sore throat o. kvist; stockholm/se a minor ailment such as a sore throat could prove to be a severe disorder known as lemierre's syndrome. this syndrome mostly affects previously healthy adolescents and young adults and in its classical form should meet four diagnostic criteria; primary infection of the oropharynx, septicemia, clinical-or radiographic evidence of thrombosis of the internal jugular vein (ijv) plus secondary metastatic abscesses. the infection is caused by fusobacterium necrophorum, a species of obligate anaerobe bacteria forming part of the normal human flora. the syndrome should be suspected in any patient with pharyngitis, cervicalgia and pulmonary symptoms. the incidence of lemierre's syndrome decreased dramatically after the introduction of antibiotics but has, of unknown reasons, increased over the past 15 years. we will present four patients diagnosed with lemierre's syndrome in our department during the last 8 years. the purpose of this case report is to raise awareness of this "forgotten disease". of the four patients diagnosed with lemierre's syndrome two fulfilled all 4 criteria while two fulfilled 3 out of 4. (table 1 ). the first two presented at the emergency department with one week's history of a sore throat, left sided cervical lymphadenopathy, erythematous tonsils, leukocytosis and elevated crp. in both cases the clinical condition deteriorated and they were referred to the icu. one developed ards and required initiation of ecmo. in both patients, chest ct revealed multiple pulmonary consolidations with cavitations, findings consistent with septic emboli (image 1,2 and 3). incidentally ct-neck revealed thrombosis in the left ejvand ijv (image 4). ultrasound of the neck veins confirmed the finding (image 5 and 6). blood cultures taken on admission later proved positive to f. necrophorum. the third and fourth case, with similar clinical histories but with a less aggressive development, had positive blood cultures but no thrombosis and vice versa. (table 1 and all four patients recovered and could be discharged with oral antibiotics and anticoagulants. unique teaching points: in conclusion, lemierre's syndrome is less common today thanks to antibiotics but may still occur in previously healthy adolescents and may lead to a fatal outcome. the pediatric radiologist should be aware of typical findings like septic emboli in the lungs and thrombosis in the ijv. unicameral bone cyst associated with secondary aneurysmal bone cyst of clavicle i. dasic, g.j. djuricic, s. ducic; belgrade/rs objective: aneurysmal bone cyst (abc) accounts for 2,5 % of all bone tumors. they are benign but locally destructive lesion of the bone characterized by presence of spongy or multiloculated cystic tissue filled with blood. abcs are metaphyseal, excentric, bulging, fluid-filled and multicameral, and may develop in all bones of the skeleton. most common locations include the proximal humerus, distal femur, proximal tibia, and spine. clavicle is a very rare site for aneurysmal bone cyst with only few cases reported in literature. a 10-year-old boy reported to the university children's hospital for detailed examination of swelling of right shoulder. 2-3 days before admission parents noticed tumefaction of right shoulder. there was no history of trauma or fever. physical examination revealed tumefaction of the right shoulder, in projection of acromial end of clavicle, measuring approximately 8x8cm, which was tender and fixed. the swelling was not hot to the touch, and there was no skin discoloration over that area. regional lymph nodes were not palpable. (fig. 1a) x-ray revealed osteolytic, expansible lesion in the lateral end of clavicle and there was no pathological fracture. (fig. 1b) laboratory analyzes were within normal limits. blood cultures remained sterile. chest x ray and abdominal ultrasound were normal. computed tomography (ct) revealed a thinwalled multiloculate lesion in lateral end of right clavicle. (fig. 2a) there was no extension in the soft tissues on magnetic resonance imaging (mri). mri shows the multiloculate cavities and fluid levels. (fig. 2b) . the open biopsy was done. histopathological examination confirmed the secondary aneurysmal bone cyst on the field of simple bone cyst of clavicle. the clavicle is an uncommon site for bone tumors. review of literature shows clavicle accounts for less than 1 % of all bone tumors. the patient with an aneurysmal bone cyst generally presents with pain and swelling, which may vary in duration from weeks to several years. up to 8 % of bone tumors occur in less than 20 years of age with peak incidence in second decade. radiologically, lesion is lytic and may have a soap-bubble appearance with ballooned distension of the periosteum. the differential diagnosis for aneurysmal bone cyst include giant cell tumor, chondromyxoid fibroma and telangiectatic osteosarcoma. distinction from telangiectatic osteosarcoma is difficult because the conditions have overlapping clinical and radiologic features. the differentiation is made from the histologic features. imaging of glomus tumor of liver in a child (case report) n. tewattanarat, j. srinakarind, j. wongwiwatchai, p. komvilaisak, s. areemit, p. ungarereevittaya, p. intarawichian; khonkaen/th objective: glomus tumors occur preferentially in subcutaneous tissue of fingers and toes, but extremely rare in visceral organs. most cases of the tumors are diagnosed in adults. several cases of glomus tumors in liver have been reported in adults. a literature review, no case of glomus tumor in liver in children was published. therefore, we present clinical, imaging findings of the first case of pediatric patient with glomus tumor in liver and also histopathological features. a previously healthy 11 year-old-girl was admitted with a twoweek history of progressive dyspnea on exertion and vomiting. family history was unremarkable. physical examination revealed hypertension and smooth and firm mass at epigastrium. systolic apical murmur on heart examination was noted. liver function test (2017) 47 (suppl 2):s297-s pediatr radiol showed elevated cholesterol (396 mg/dl). other laboratory tests (complete blood count, blood chemistry, renal and liver function test, coagulation test, hepatitis profiles and alpha-fetoprotein) were within normal limits. echocardiogram found mitral and tricuspid regurgitation and poor left ventricular systolic dysfunction. abdominal mri demonstrated a 12-cm well-defined exophytic hypervascular mass with intratumoral hemorrhage at segment 3/4b of the liver. there were no other suspicious lesions in other organs. the biopsy was done and revealed glomus tumor. patient underwent preoperative embolization and the liver mass revealed decreased size to 8-cm after 1-month follow up with ultrasound. after that, exploratory laparotomy with left lateral segmentectomy was performed. the pathological results showed dilated vascular channels surrounded by uniform neoplastic cells, uniform with round nuclei, fine chromatin, inconspicuous nucleoli, and pale eosinophilic cytoplasm, and well-defined cytoplasmic border. no mitotic figures and necrosis are identified. immunohistochemical (ihc) staining of tumor was positive for cd34, smooth muscle actin (sma) and h-caldesmon. others ihc including ae1/ae3, heppar1, cd31, desmin and myogenin were negative. from these findings, the tumor was finally diagnosed as glomus tumor of uncertain malignant potential due to deep location and large size. primary glomus tumor is a rare entity of liver tumor diagnosed in children. however, it should be considered in the differential diagnosis of a hypervascular liver mass. most of these tumors are benign, however tumor in liver have malignant potential due to deep seated position. therefore, tumor removal with pre-operative embolization should be considered. brain mri in a pediatric patient with linear scleroderma en coup de sabre m. mortilla, a. rosati, e. canale, c. filippi; florence/it objective: linear scleroderma "en coup de sabre" (ecds) is a rare subset of localized scleroderma. affected individuals typically have a characteristic atrophic skin lesion involving the fronto-parietal scalp. the disease usually has a benign course but rare neurologic symptoms can be seen associated: the most common described is epilepsy. intracranial mri findings described in the literature include: focal brain atrophy, calcifications and t2-hyperintense white matter lesions that may demonstrate contrast enhancement. white matter lesions and calcifications are found in the cerebral hemisphere ipsilateral to the skin abnormality. in the literature only a few pediatric cases have been described. a 4yrs. old girl was hospitalized at our institution for evaluation of a lesion of the frontal skin associated to a history of febrile seizures and mri alterations. she presented febrile seizures at the age of 2 on april 2013. on january 2014 parents noted a frontal cutaneous lesion that was defined as "linear scleroderma, port-wine stain type". on november 2014 she performed an mri at another institution showing a diffuse white matter alteration in the left emisphere with focal lesions with high susceptibility and mild contrast enhancement. she was addressed to immunosuppresive therapy with steroids and methotrexate, with steroids stopped after 6 months. a clinical cutaneous improvement was noted. on july 2016 a second mri showed a worsening of the findings. we describe a case of a little girl with ecds with no neurologic deficits or symptoms that shows extensive and progressive neuroradiologic alterations. only a few pediatric cases have been described, but it has to be known that also in absence of symptoms, patients with linear scleroderma should be screened with mri to look for cns involvement in this immune disease. brain mri can also be used to monitor the progression of the disease and the response to therapy. mals is a vascular compression syndrome which symptoms can overlap chronic functional abdominal pain. in mals the proximal part of the celiac artery is compressed by the too low located median arcuate ligament during expiration resulting in hemodynamically significant symptoms. we report two cases with mals diagnosed primarily by ultrasonography. case 1 18-year-old girl was admitted to tartu university children's clinic (tucc) due to recurrent acute epigastric pain episodes with nausea and loss of appetite during 7 years. previous analyses were normal, abdominal uss and gastroscopy did not show any abnormalities. she was referred to paediatric radiology department for doppler us (dus) which showed narrowed proximal celiac artery (ca) with turbulent flow, increased peak-systolic and end-diastolic velocities on deep inspiration and expiration, and positive ca deflexion angle on expiration. superior mesenteric artery (sma) was markedly widened, indicating possible collateral blood-supply due to severe ca stenosis. according us findings mals was suspected. abdominal mra showed proximal ca kinking, stenosis and poststenotic dilatation and confirmed diagnosis. during dsa collateral blood-supply from sma via pancreaticoduodenale arcade (pda) was seen. laparoscopic release of mal resulted in relief of patient's symptoms, she has been pain-free for two years. case 2 16-year-old girl applied to tucc due to recurrent abdominal pain episodes for 2-3 years. usually, pain occurred 2-3 times per week about 15 minutes after the start of intense cycling training or competitions, and passed about 5 minutes resting in squat position. mild mid-epigastric bruit was audible at physical examination. dus showed two-fold increase in expiratory peak-systolic and enddiastolic blood flow velocities compared to inspiratory velocities which indicated to the hemodynamically significant worsening of ca compression by mal during expiration. mra showed proximal ca compression, upward angulation and poststenotic dilatation. preoperative ct-angiography depicted collateral supply via pda. during laparoscopic surgery ca was released by transecting mal and surrounding fibrous tissue. after surgery the girl has been pain-free for one year except single pain episode during intense competition. the diagnosis of median arcuate ligament syndrome should be considered in patients with postprandial abdominal pain that does not have other clearly established etiology. colour doppler us should be the first choice imaging method. to confirm diagnosis in pediatric patients abdominal mra is preferred in our institution, but as mra may still have a tendency to movement artifacts and inadequate spatial resolution for smaller blood vessels, in these two cases mra was followed by cta or dsa. understand the unique predilection of infantile malignancies to metastasize and present as skin-based masses, most commonly lymphoma/leukemia. case presentation: an otherwise healthy 37 day old male presented to dermatology with a pedunculated, friable red glabellar mass (centered between the eyes). first noticed as a flat, bluish lesion at 10 days, its subsequent rapid growth led to an emergency department visit where dermatology diagnosed a hemangioma and initiated propranolol treatment. despite this, the mass continued to grow rapidly, encroaching upon the patient's right eye. the patient was admitted for further workup. an elevated beta hcg, anemia (7.4 mg/dl), and thrombocytopenia (92,000) suggested an alternate diagnosis. an mri and ultrasound led to a percutaneous biopsy; pathology was consistent with choriocarcinoma. pet ct found fdg-avid glabellar, liver and lung lesions. maternal and placental testing was negative for choriocarcinoma. ultrasound demonstrates a hypoechoic hypervascular mass. mri brain demonstrates cutaneous confinement of the solid avidly enhancing glabellar mass. ct shows a peripherally enhancing liver mass with a masslike area of consolidation in the right lung. initial pet/ct demonstrated fdg avid liver and lung metastases with a small focus of residual activity at the glabella consistent with incomplete resection. follow-up pet/ct showed astoundingly rapid re-growth of the glabellar mass and enlargement of the hepatic and pulmonary masses just 12 days later demonstrating the extremely aggressive nature of this cancer. 3month follow-up pet/ct showed significantly decreased size and activity of the metastases consistent with a treatment response. in a series of 208 infants with cutaneous metastases, the following diseases presented with cutaneous involvement (ordered most to least common): leukemia, langerhans cell histiocytosis, neuroblastoma, rhabdoid tumor, rhabdomyosarcoma, primitive neuroectodermal tumor, choriocarcinoma, and adrenocortical carcinoma. pathology slides (2017) 47 (suppl 2):s297-s pediatr radiol unique teaching points: considered one of the fastest growing tumors, infantile choriocarcinoma classically presents with hepatomegaly, anemia, failure to thrive, and precocious puberty between 0 days and 5 months of life. left untreated, the disease is usually fatal within 3 weeks of presentation. a solitary cutaneous metastasis can be mistaken for infantile hemangioma both clinically and radiographically. atypical mri appearance is one important clue that can suggest an alternative diagnosis. pet/ct may be useful for staging and follow-up. a rare case of ovarian juvenile granulosa cell tumor associated with ollier's disease -generalised mesodermal dysplasia p. joshi; pune/in to demonstrate a rare case of mesodermal dysplasia -association of ovarian granulosa cell tumour with enchondromatosis case presentation: two year 8 month old girl presented with precocious puberty i.e thelarche. left hand radiograph showed the radiological age corresponding to chronological age, suggestive of peripheral precious puberty. the patient subsequently underwent a sonography which revealed a pelvic mass probably arising from the right ovary ? sex cord stromal tumour. a mri of the abdomen and pelvis confirmed the pelvic mass and revealed multiple bone lesions in the right hemipelvis -on the side of the tumour she was later operated. hpe of pelvis mass revealed juvenile granulosa cell tumour. ultrasound pelvis images reveal a solid pelvic mass, probably ovarian in etiology mri pelvis also reveals multiple bone lesions unique teaching points: the aim of the poster is to create awareness about this association. the bone lesions should not be mistaken for metastasis juvenile granulosa cell tumour of the ovary (jgct) is a well-known sexcord stromal ovarian neoplasm. ollier's disease is a rare, non hereditary mesosermal dysplasia consisting of multiple enchondromas. the association of granulosa call tumour with asymmetric ipsilateral hemiskeletal distribution may indicate generalised mesodermal dysplasia as there is also association of jgct with maffucci's syndrome, other dysplastic conditions such as microcephaly, facial asymmetry,' and potter's syndrome. review of literature showed 11 previous cases of juvenile granulosa cell tumor associated with enchondromatosis, three associated with maffucci's syndrome, and the rest with ollier's disease goldbloom's syndrome is a paediatric idiopathic disease characterized by transient bone marrow oedema with recurrent crisis of bone pain, periosteal hyperostosis, fever, increased inflammatory markers and dysproteinaemia. a case series of wbmr studies in goldbloom's syndrome is reported and differential diagnosis discussed. case presentation: a 9-year-old female girl was admitted to our paediatric department because of daily crisis of bone pain of the lower limbs, associated with fever spikes, limping and nocturnal awakenings. no history of trauma was reported. laboratory tests showed mild anaemia (hb 8.2 g/dl), thrombocytosis (plt 680000/mmc), increased inflammatory markers (ers 75 mm/h, crp 7 mg/dl), high streptolysine o and dnase-b antibody levels (aso 4280 iu/ml and adn-b 6310 ui/ml, respectively). throat swab was positive for group a β-haemolytic streptococcus (gas). unusual dysproteinaemia, characterized by hypoalbuminemia (2.8 g/dl) with increased a1, a2 and g globulinaemia, was noted. x-ray examinations of both legs resulted normal. wbmri showed markedly delineated, high and homogeneous hyper-hypointensity respectively in stir/t1 of the distal tibialperoneal meta-diaphysis of both legs (fig1a,b). distal metaphysis of femur, humerus, radius-ulna and proximal tibia were also homogeneously mildly hyperintense on stir sequences bilaterally (fig1a). bone biopsy revealed signs of chronic inflammation. infectious and neoplastic diseases were ruled out and the diagnosis of gs with dysproteinaemia seemed conceivable. steroid treatment was started in association with indomethacin, leading to a prompt resolution of the clinical picture within a few days. the follow-up stir total body mri, performed after 10 months, showed the complete resolution of bone oedema. (fig2 a,b) the sock sign is a pathognomonic whole-body magnetic resonance imaging (wbmri) feature of goldbloom's syndrome (gs).it is a well marked, symmetric, homogeneous and high bone marrow hyperintensity, localized both at the distal tibial and peroneal meta-diaphysis, which looks like a pair of socks. objective: left ventricle hypoplasia is generally thought as a part of hypoplastic left ventricle syndrome or aortic hypoplasia. it is estimated that about 15-20 ml/m2 left ventricle volume is needed in order to support systemic circulation. less than that volume generally precludes biventricular repair. however conditions associated with severe preload decrease such as total anomolous pulmonary venous return (tapvr) should be considered in the differential diagnosis. tapvr presenting as hypoplastic left ventricle syndrome is presented in this study. six month old female patient admitted to emergency service with symptoms of fever, dyspnea and coughing. emergency staff started intravenous antibiotic theraphy and from medical records learned that she has been followed for partial anomolous pulmonary venous return (papvr) and atrial septal defect (asd). lung x-rays revealed pulmonary edema. echocardiography was performed and revealed very small left ventricle, papvr and 13 mm wide asd. ecg gated cardiac ct was requested with the prediagnosis of hypoplastic left ventricle syndrome. ct images revealed dilated right cavities, very small left ventricle, pulmonary edema, tapvd and peritoneal fluid plus hepatomegaly. we then retrospectively searched our archive and found she was diagnosed as papvr when she was 10 days old. all the cavities that time, were normal sized. according to these we confirmed our diagnosis as tapvr and hypoplastic appearing cavities due to reduced preload and right chamber dilatation due pulmonary overcirculation. surgical team decided to perform corrective operation and they confirmed our diagnosis unique teaching points: small left ventricle cavity in an infant need not to be due to intrinsic hypoplasia. whenever we experience such a situtation we should search for other reasons of pseudohypoplasia in order to give a chance for corrective surgery instead of palliative procesures. we present a case report of kimura disease, a rare benign chronic inflammatory disease that involves the deep subcutaneous tissues and lymph nodes of the head and neck. we report the case of a thirteen year old male who presented with a right sided facial mass which had been present for two years but had enlarged rapidly in the preceding three months. us and mr were interpreted locally as an arteriovenous malformation. review of these examinations and catheter angiography performed at this quaternary referral centre favoured a vascular tumour. subsequent percutaneous biopsy demonstrated angiolymphoid hyperplasia with eosinophilia and blood tests showed a serum eosinophilia, consistent with kimura disease. us shows a mass consisting of scattered heterogenous foci within the fat with multiple large feeding vessels. contrast enhanced mri demonstrated a solid, homogenously enhancing, mass with multiple vascular flow voids from the right external carotid artery branches. catheter angiography showed tumour blood supply from branches of the right transverse facial artery and distal right ima. the dominant supply arose superficially from the transverse facial artery. kimura disease is a rare chronic inflammatory disorder of unknown aetiology that involves the deep subcutaneous tissues and lymph nodes of the head and neck region, most common in asian men in the third decade and sporadic in the non-asian population. the histopathological and biochemical characteristics are eosinophilic lymphfolliculoid granuloma, increased eosinophils in the peripheral blood and increased ige levels. whilst ultrasound and mri are effective imaging modalities, imaging alone does not allow confident differentiation from malignant lesions and biopsy is necessitated. kimura disease has a benign indolent course with an excellent prognosis following surgical excision although local recurrence has been reported. increased naa: is it surely canavan disease? e. varga, p. barsi, g. rudas; budapest/hu leukodystrophies are a group of rare genetic, metabolic diseases that affect the central nervous system, mainly the brain. each type of them is caused by a specific gene abnormality that leads to abnormal development or destruction of the white matter of the brain. the differential diagnosis are made on the basis of clinical and neuroradiological signs. there are some diseases which show typical changes on mr spectroscopy. we present a case of a 12 year-old boy, who has been investigated due to somatomental retardation and muscle dystrophy since his six months of age. his perinatal period was normal except of a nystagmus visible from his birth. the child has muscle dystrophy, spastic quadriparesis, contractures, scoliosis, truncal hypotonia and ataxia and mental retardation. we started examinations to find out the background pathology of his idiopathic encephalo-myopathy. the brain mri showed a bilateral, symmetrical white matter signal alteration, which referred to some kind of metabolic (2017) 47 (suppl 2):s297-s pediatr radiol disease. the mr spectroscopy revealed decreased cholin and increased naa levels, which are typical of canavan disease. despite of this, the clinical aspects and the location of the involved brain areas were more typical of pelizaeus-merzbacher disease (pmd). the pmd is a genetic disorder, which is originated of the mutation of the proteolipid protein gene (plp1) located on long arm of x-chromosome (xq21-22) . this gene has an impact on growth of the myelin sheath. various types of mutations (deletion, duplication, point mutation, insertion) of plp1 gene lead to various severity of clinical picture. all form of mutations show decreased naa level on spectroscopy, except the duplication of plp1 gene. in connection with our case, we present briefly the clinical and neuroradiological differences between the two entities. magnetic resonance imaging findings in medium-chain acyl-coenzyme a dehydrogenase (mcad) deficiency l. talamanca, d. narese, m.c. rossi espagnet, l. pasquini, d. longo; rome/it we report serial brain magnetic resonance (mri) in a patient with medium-chain acyl-coenzyme a dehydrogenase (mcad) deficiency who developed acute encephalopathy. a 12-months-old girl was admitted in the emergency department of our hospital with sudden onset of acute encephalopathy with drowsiness. baseline laboratory investigations revealed severe hypoglycemia, hyperammonemia, hyperchloremic metabolic acidosis and hyperuricemia. the patient was treated with glucose solution infusion that resulted in a gradual resolution of symptoms. the first brain mri, performed within 6 hours of onset of symptoms showed bilateral symmetric restricted diffusion on diffusion-weighted imaging (dwi) in the middle cerebellar peduncle, nucleus caudatus, putamen and periventricular white matter; the adc map showed reduced diffusivity (fig 1) . the second mri, at 72 hours after the onset, revealed bilateral and symmetric hyperintensity on t2-weighted images in the middle cerebellar peduncle, nucleus caudatus, putamen and periventricular white matter. dwi showed restricted diffusion in both globus pallidus (fig 2) . a single voxel h-mrs study performed by placing a roi in the right nucleus lenticularis revealed increased values of gaba and glutamine (fig 3) . a further mri was performed 4 weeks after the first neuroimaging and indicated widespread atrophy and the appearance of a hyperintense signal in t2-wi in both globus pallidus while dwi did not reveal any remarkable signal abnormality. single-voxel mrs of the same region showed a normalization of gaba and glutamine values. brain mri showed bilateral symmetric restricted diffusion on diffusionweighted imaging (dwi) in the middle cerebellar peduncle, nucleus caudatus, putamen and periventricular white matter; the adc map showed reduced diffusivity the second mri, at 72 hours after the onset, revealed bilateral symmetric restricted diffusion on diffusion-weighted imaging (dwi) in both globus pallidus. a single voxel h-mrs study performed by placing a roi in the right nucleus lenticularis revealed increased values of gaba and glutamine. mcad is an enzyme of the mitochondrial b-oxidation of fatty acids, an essential source of energy for cells during stress. mcad deficiency is the most common genetic disorder of fatty acid oxidation. the clinical manifestation of the disorder is typically precipitated by stress due to fasting, vomiting, fever or muscular exertion and occurs in the majority of cases before the age of 2 with the onset of acute hypoketotic hypoglycemia. clinical features of this decompensated state include seizures and lethargy proceeding to coma and death in the absence of prompt treatment with intravenous dextrose infusion. mcad deficiency usually appears in an acute form and has high morbidity and mortality rates; early diagnosis is therefore extremely important in order to promptly begin treatment and obtain a complete recovery from symptoms. mr can play a significant role in the early diagnosis of the decompensated state of the disease; in our case dwi revealed the presence of lesions with a bilateral symmetric topographic distribution that strongly suggested a metabolic disease leading to acute encephalopathy. a full-term male neonate (3 days old) with external perineal anomalies was referred to our hospital. the physical perineal examination revealed a bifid scrotum containing palpable testis and a normal configured penis located at the bottom of the bifid scrotum. two soft masses of 3 and 2 cm respectively, divided from a cutaneous notch, were located below the bifid scrotum and on the right of the midline. the rear biggest mass was normal epithelized, instead the other one was a rugged pigmented mass, which resembled the scrotum (figure 1 ). there were no additional abnormalities of the external genitalia. the other peduncolar mass, located between the right scrotum and the posterior mass, had fluid content. a mild hydrocele in the right scrotum and a sliding testis on the left side were also revealed. us examination showed a hyperechoic solid tissue, corresponding to the rear biggest perineal mass. the other peduncolar mass, located between the right scrotum and the posterior mass, had fluid content (figure 2) . a mild hydrocele in the right scrotum and a sliding testis on the left side were also revealed. mri also confirmed two perineal peduncolar masses: the biggest and posterior one, was made up by homogeneous fatty matter without contrast-enhancement after intravenous gadolinium injection (figure 3 ). the patient underwent excision of perineal masses and no complications occurred in the surgery. the histopathological examination of the perineal masses revealed two areas with different histological features: the first one was characterized by the presence of smooth muscle bundles dispersed in the dermal collagen, instead the other contiguous area showed abundant mature adipose tissue in the deep dermis and hypodermis ( figure 5 ). at last the rugged swelling mass was definitively diagnosed as as without testis tissue inside, and the rear mass was diagnosed as lipoma. the physical perineal examination revealed a bifid scrotum containing palpable testis. two soft masses of 3 and 2 cm respectively, divided from a cutaneous notch, were located below the bifid scrotum. us examination showed a hyperechoic solid tissue, corresponding to the rear biggest perineal mass. the other peduncolar mass, located between the right scrotum and the posterior mass, had fluid content mri confirmed the presence of two perineal peduncolar masses: the biggest and posterior one, was made up by homogeneous fatty matter without contrast-enhancement after intravenous gadolinium injection. neonates presenting with perineal masses are uncommon. these anomalies can occur isolated or more rarely in combination with other abnormalities such as uro-genital or ano-rectal anomalies or with contiguous subcutaneous tumors. when perineal masses are found, with prenatal diagnosis or during a newborn physical examination, it is important to look for any associated congenital anomalies or subcutaneous tumors by using imaging. to describe and emphasize the significance of the "half-moon" sign in pelvic mri. a 13-year-old adolescent, karate athlete, was submitted with left hip pain, decreased range of movement and asymmetry in thigh circumference. markers for infection or inflammation were negative. frog-leg radiograph was negative for hip effusion, slipped epiphysis and equivocal for a left trochanteric abnormality. mri demonstrated a half-moon pattern of bone marrow edema at the left intertrochanteric area and at the major trochanter, surrounding an apophyseal low-intensity lesion. ap radiograph and limited ct confirmed the presence of a lytic lesion with sclerotic margins, containing calcified chondroid matrix. chondroblastoma was histologically confirmed following excision. mri, coronal stir sequence, demonstrates semilunar-shaped hyperintense area abutting the growth plate and the cortex of the femoral neck, consistent with the half-moon sign. note edema surrounding an apophyseal low-intensity lesion and soft-tissue edema. ct confirms a typical apophyseal lesion with sclerotic margins containing chondroid matrix. unique teaching points: "half-moon" sign refers to a semilunar shape of bone marrow edema at the intertrochanteric area of the hip with its base located at the cortex of the femoral neck. this distribution differs from the distribution of edema in metaphyseal and metaphyseal-equivalent osteomyelitis. "half-moon" sign has been described in patients with stress fractures and osteoid osteomas. to our knowledge, this is the first case of chondroblastoma exhibiting this sign. whenever the "half-moon" pattern of edema is identified at pelvic mri scans, a thorough search for an occult fracture line or a nidus corresponding to an osteoid osteoma or a chondroblastoma is mandatory. mr elastography (mre) is a noninvasive imaging technique that quantitatively measures liver stiffness and provides an estimate of the degree of fibrosis. our aim was to evaluate the feasibility of performing mre using both gradient echo (gre) and echo planar (epi) sequences on siemens scanners. a dedicated mre of the liver was performed on a 3t mr scanner (magnetom® skyra, siemens) with a pediatric mechanical (2017) 47 (suppl 2):s297-s pediatr radiol driver (courtesy of mayo clinic) over the right upper quadrant. an axial t2 blade with fat saturation, a coronal t1 vibe dixon and axial diffusion weighted imaging (dwi) were obtained. elastograms were obtained using both an axial standard gre and a works in-progress (wip) epi sequence. for the gre sequence, 5 different slices were selected and each scanned sequentially. the epi sequence incorporated 5 different slices in just one series. images were post-processed placing regions-of-interest (roi) and measuring the stiffness in kilopascals (kpa). for each sequence and each slice the stiffness mean was measured and then the average of the means was obtained. a spleen elastogram was simultaneously generated, without changing the mechanical driver location, and mean stiffness was also calculated. based on cutoffs in the literature, values were considered abnormal if liver stiffness >2.9 kpa and spleen stiffness >3.6 kpa. our initial experience shows that mre is feasible on siemens scanners using both gre and epi sequences. epi sequences are a promising addition to standard gre. prone versus supine ultrasound positioning for evaluation of urinary tract dilation (utd) in children c. maya 1 , y. gorfu 2 , e. dunn 1 , k. darge 1 , s. back 1 ; 1 philadelphia/us, 2 addis ababa/et objective: ultrasound (us) is used in the initial evaluation and surveillance of utd in children. utd classification systems, including the 2014 multidisciplinary consensus, assess anterior-posterior renal pelvic diameter (aprpd) and calyceal dilation. there is currently no consensus regarding optimal patient positioning-prone versus supine-during us assessment of utd. this study was performed to determine if there is a significant difference in the measurement of the aprpd, presence of calyceal dilation, or resulting utd consensus score obtained between supine and prone positions. two raters retrospectively reviewed renal bladder ultrasounds of patients with utd of one or both kidneys. technically adequate ultrasound examinations of orthotopic kidneys that were imaged in both supine and prone positions were included. those with renal anomalies or prior surgery were excluded. aprpd measurements, as well as central and peripheral calyceal dilation, were documented in both prone and supine positions. a postnatal utd consensus score was assigned to each kidney based only on these features. 146 kidneys (77 left) in 89 subjects had utd in either the supine or prone position. mean age was 0.41 years (range: 0.01 -1.97 y). female to male ratio was 1:3 (21/68). the interclass correlation (icc) of the aprpd between raters was 0.88 and 0.87 in the supine and prone positions respectively (ps<0.001). central calyceal dilation was found in 97/146 supine kidneys and 102/146 prone kidneys by rater 1 and 94/146 supine and 99/146 prone kidneys by rater 2 (kappa 0.92). peripheral calyceal dilation was found in 54/146 supine kidneys and 65/146 prone kidneys by rater 1 and 50/146 supine kidneys and 62/146 prone kidneys by rater 2 (kappa 0.89). as such the results are presented as one. the aprpd tended to be greater when prone with a strong correlation between prone and supine measurements (0.92, p<0.001). the mean difference between supine and prone aprpd was 1.1 mm (p< 2.2). in 15 kidneys, calyceal dilation was seen in the prone position and not supine while 1 kidney had central calyceal dilation only when supine. the utd score differed between supine and prone in 13/149 kidneys, with all but one higher when prone. in 10 other kidneys, the aprpd differed between positions however concurrent calyceal dilation resulted in no change in utd class. as a screening tool, performing ultrasounds in the prone position may help identify more kidneys with utd. further research is needed to determine if these differences are clinically significant. during the evaluation of magnetic resonance enterography (mre), diffusion restriction (dr) has been utilized as a marker for bowel inflammation, but in our practice we commonly see dr in otherwise normal segments of jejunum. the purpose of this article is to assess the dr in normal loops of jejunum on mre and to determine if there is a correlation between dr and luminal distention, age, magnet field strength, and bowel segment location. a retrospective analysis of subjects with a normal mre and normal clinical work up (based on available clinical history, endoscopy reports, serum white blood cell count and inflammatory markers, and stool samples) was performed. the abdomen was divided into 4 quadrants. if available, 2 loops of jejunum were randomly chosen in each quadrant. two radiologists independently evaluated these same loops of jejunum for the following: luminal distension, wall thickness, and enhancement pattern. additionally, the loops were then evaluated for the presence or absence of dr. inter-rater reliability was determined. disagreement was resolved by consensus. presence or absence of dr was correlated with luminal distension, age, magnet field strength (1.5 versus 3 tesla), and abdominal quadrant. one hundred ninety-seven loops of jejunum were evaluated in 39 patients. not all subjects had jejunal loops in all quadrants. sixteen subjects (41%) had jejunal loops with dr for a total of 29 loops. one loop had increased wall thickness and another increased enhancement but both did not demonstrate dr. no other loops demonstrate increased enhancement or wall thickening. for the presence or absence of dr, inter-rater reliability was fair (kappa=0.39). there was no correlation between the presence/ absence of dr in relation to luminal distension, age, magnet field strength, or quadrant location. of the 16 subjects who had a single loop with dr, a 2 nd loop with dr was found in 50%. 14 year old who presented with nausea. mr enterography demonstrates no bowel thickening or abnormal enhancement. a. coronal haste demonstrates the craniocaudal position of the axial diffusion sequence for reference (line). 14 year old who presented with nausea. mr enterography demonstrates no bowel thickening or abnormal enhancement. b. axial diffusion weighted seqeunce (b=800) shows diffusion restriction within loops of jejunum (arrow) within the anterior abdomen. 14 year old who presented with nausea. mr enterography demonstrates no bowel thickening or abnormal enhancement. c. corresponding adc map demonstrates low signal within the jejunal wall consistent with diffusion restriction (arrow). diffusion restriction in normal loops of jejunum on mre was present in 41% of patients. if dr is seen in an otherwise normal segment of jejunum, this can be considered non-pathologic. a patient with a loop of jejunum with dr is likely to have an additional loop of jejunum demonstrating dr. there is no correlation with dr of normal jejunum with luminal distension, magnet field strength, or patient age. our data may help reduce overestimation of disease burden when clinically applied. imaging findings in the newborn with meconium peritonitis that require surgery p. caro dominguez 1 , a. zani 2 , a. daneman 2 ; 1 cordoba/es, 2 toronto/ca objective: meconium peritonitis is a rare condition caused by an in-utero bowel perforation resulting in spillage of meconium into the peritoneal cavity and subsequent calcification. the role of prenatal and postnatal imaging is to identify infants who require surgery. the aim of this study was to evaluate the role of postnatal imaging in meconium peritonitis and to correlate the radiologic and sonographic patterns with the need for surgery. imaging studies in infants with meconium peritonitis performed between 1999 and 2014 at our institution were reviewed separately by a pediatric radiologist, a pediatric radiology fellow and a pediatric surgeon. patients were divided in a surgical and a non-surgical group. clinical, surgical and pathology reports were reviewed to validate the diagnosis. statistical analysis: comparisons between sonographic and radiographic findings and patterns in the surgical and non-surgical groups were performed using unpaired t-test and chi-square. during the study period, there were 37 infants with meconium peritonitis managed at our institution. in the 23 (62%) who needed surgery, the most frequent surgical findings were idiopathic perforation, jejunal and ileal atresia. ultrasound identified more cases with hepatic calcifications, meconium pseudocyst, ascites and pneumoperitoneum than radiography and radiography more cases of small bowel obstruction. ascites identified with ultrasound (p=0.01) [fig 1] and bowel obstruction [fig 2] diagnosed either with ultrasound (p=0.04) or radiography (p=0.01) were associated with the need for surgical intervention. one third of children with meconium pseudocysts (4/12) [fig 3] , did not require surgery. diffuse peritoneal or hepatic calcifications as an isolated postnatal finding were not associated with the need for surgery. both radiography and ultrasonography give valuable information to the surgeon to take the decision for surgery. dilatation of bowel loops and ascites detected postnatally with radiography and/or ultrasound require surgical intervention in children with meconium peritonitis. interestingly, a large proportion of infants with meconium peritonitis can be managed conservatively. 0.17 -17.4) . those included had complete fmru analysis, dti (b=0 and b=400, 20 directions), and upjo configuration in at least 1 kidney. cases with motion artifact (n=9), post-pyeloplasty (n=3) or duplex collecting systems (n=3) were excluded. pelvicalyceal dilation grade (pcd), corticomedullary differentiation (cmd), and functional parameters were included. pyeloplasty following fmru was recorded. dti tractography was reconstructed using a fractional anisotropy (fa) and an angle threshold of 0.10 and 55°, respectively (figure 1) . user-defined regions-of-interest (roi) of the renal parenchyma, excluding the collecting system, were drawn to quantify dti parameters: mean fa, apparent diffusion coefficient (adc), tract length and tract volume. the relationships between dti quantitative parameters and fmru parameters were analyzed. age and adc (roi) (p<0.01, r 2 =0.36), tract volume (p<0.01, r 2 =0.77) and tract length (p<0.01, r 2 =0.60) were positively correlated. age and fa (roi) (p<0.01, r 2 =0.46) were negatively correlated. there was a correlation between fmru parenchymal volume and tract volume (p<0.01, r 2 =0.80), but median volumes were higher on dti (tractography=98.5 cm 3 vs. fmru=60.52 cm 3 ; p<0.01). of the 16 children, 12 had pyeloplasty, 1 had nephrectomy, 2 were managed conservatively and 1 was lost to follow-up. fa was significantly lower in kidneys that went on to have pyeloplasty in comparison to those without pyeloplasty, but the 95%ci and the iqr overlapped (table 1) . the adc, tract length and tract volume were similar between these groups (table 1) . there was no difference between the adc of fa values in kidneys with and without pcd or cmd (p>0.11). linear hierarchical regressions controlling the age did not show a significant relation between adc and cortical or renal transit times (p>0.68), but lower fa values were related to a higher renal transit time (p<0.01, r 2 =0.103). table 1 . quantitative dti parameters between kidneys with and without pyeloplasty following fmru. renal adc, fa, tract volume and tract length change with age but tractography overestimates renal parenchymal volume. there was a tendency towards a lower fa in kidneys that went on to pyeloplasty. otherwise, none of the quantitative parameters evaluated in this study differentiated degrees of upjo. echo-enhanced voiding urosonography (eevus) has become an important imaging tool in urodiagnostics; however, it has been observed that during eevus the premature destruction of ultrasound contrast agent microbubbles might occur. the purpose of this study was to evaluate the possible causes of contrast vanishing during investigations and propose the protocol to avoid false negative results. eevus was performed in 163 children from april to december 2016. sonovue mixed with saline solution in a plastic bottle is applied by continuous flow through the urine catheter. the collected data according to the protocol in this prospective study was completed in 105 children, aged from 2 weeks to 10.25 years. the protocol included general patient information, indication for eevus, duration of eevus in minutes, and the presence of vesicoureteric reflux. extensive data about sonovue were recorded: charge number, expiration date, time since opening, amount of initially administered contrast (ml sonovue/ml saline solution), grading of the initial contrast opacification of the bladder, the need for immediate readministration of contrast (dose), grading of contrast opacification during examination, and the need for readministration of contrast later in the course of the examination (dose). in addition, the data regarding bladder (ratio real/predicted bladder volume, wall thickness, ureter dilatation), saline solution, the size of urine catheter (french), and the type of antibiotic prophylaxis were collected. child observation included grading of crying and muscle stiffness. normal contrast opacification of urinary tract during examination was found in 87/105 children, while in 18/105 (17.2%) the contrast opacification was insufficient. in 12/18 (72.2%) microbubble destruction occurred during the first minute, in 4 (22.2%) in 5 minutes, and in 1 in 13 minutes after the beginning of contrast administration. the reason for unsatisfactory contrast opacification at the beginning of the eevus is probably due to small urine catheter size (25% of children with fr6 catheter had insufficient opacification compared to 13.3% with fr8 in whole cohort), time since the contrast is opened (more than 3 hours in 4 children), and insufficient bladder emptying at the beginning of the procedure. the reason for microbubble destruction later in the course of the examination is bladder overfilling in combination with increased muscle stiffness and strong crying, which led to increased bladder pressure. there was no correlation between the type of antibiotics and microbubble destruction. we should be aware of possible false negative vur results during eevus caused by premature microbubble destruction. patients with fontan-type palliation of univentricular congenital heart disease have elevated central venous pressure due to their passive pulmonary flow. the altered circulation has a negative impact on several visceral organs, and these patients have chronic liver congestion. they are at risk of developing hepatic fibrosis and cirrhosis with potential malignant transformation. these changes can occur from only a few years after fontan palliation, making early detection and grading of major importance. the patchy pattern of hepatic changes makes liver biopsy an unreliable diagnostic tool. magnetic resonance imaging (mri) t1 mapping has been suggested as a technique for non-invasive assessment and quantification of hepatic fibrosis/cirrhosis. the aim of this study was to compare two different t1 mapping sequences of the liver in adolescents with fontan palliation, and in healthy controls. materials: 15 adolescents (15-17 y) with fontan circulation and 7 young healthy adults (18-24 y) were included as a part of an ongoing national population-based study. all underwent mri (1.5 tesla) pre-and post-gadolinium contrast, including two types of t1 mapping of the liver. a 3d t1 volumetric interpolated breath-hold examination (3d vibe) sequence with dual flips with b1 correction and a modified look-locker inversion recovery (molli) sequence. t1 relaxation times (ms) were measured by placing five standardized circular regions of interest (roi) in the mid-section of the liver and one in the spleen (fig 1) . statistical analysis was performed comparing measurements pre-and post-contrast, between sequences, and patients and controls. there was a significant difference in the measurements between molli and 3d vibe with increased values for the latter. within each sequence there were small, but significant regional differences in relaxation times (table 1). the same pattern was seen in pre-and post-contrast images in both groups. there were significantly increased native t1 times on both sequences in all regions in the fontan group as compared to the controls, but not post contrast. t1 relaxation times differ between the t1 mapping sequences, molli and 3d vibe, pre-and post-contrast. t1 mapping of the liver revealed significantly increased native t1 times in adolescents with fontan palliation compared to healthy slightly older controls. these findings suggest hepatic fibrosis/cirrhosis, but may also represent a component of congestion. diagnostic accuracy of ultrasound, computed tomography and wedge portography in the work-up for mesenterico-rex bypass in children with extrahepatic portal hypertension s. toso, r. breguet, m. annoshiravani, s. terraz; geneva/ch to identify the diagnostic accuracy of ultrasound (us), computed tomography (ct) scan and portography (wedge hepatic vein portography or direct portography) in the pre-operative work-up of mesenterico-rex bypass performed for extrahepatic portal hypertension in children. we conducted a retrospective analysis of pre-operative imaging for mesenterico-rex bypass in our tertiary hospital over the last 12 years. we analyzed all patients between the ages of 0-16 years, with extrahepatic portal hypertension necessitating surgical treatment that underwent us, ct and portography. three reviewers independently analysed the patency of the left portal vein, mesenteric vein, splenic vein and the presence of communication between the left and right portal vein on preoperative imaging with correlation to surgical findings. statistical analysis of diagnostic accuracy was performed. eleven patients underwent mesenterico-rex bypass for portal hypertension secondary to portal vein thrombosis. two patients had partial liver transplant. ct with ultrasound correlation was sufficient in responding to the preoperative criteria in 72% (8/11) cases. portography was useful in the 27% (3/11) cases where ct could not respond to preoperative criteria, in particular the presence of left-right communication. there was good inter-rater correlation for each modality and good correlation of findings between modalities. in the majority of cases the use of ultrasound and ct is sufficient for preoperative planning for mesentrico-rex bypass. portography is mandatory in cases with large intra-hepatic cavernoma, where the left-right communication could not be confirmed on ct. contemporaneous clinical data was reviewed where available, and a clinical decision on disease severity and activity on a likert scale made with and without imaging. fifty-three patients underwent mre and bowel us in the specified timeframe (29 male; median age 13.02 years, range 4-16 years). twenty patients had sufficient contemporaneous clinical information to be analysed. inter-observer variability for the imaging scores was assessed using bland-altman plots. where variability was beyond pre-determined limits, the studies were consensus reviewed. mean scores were used for the studies within accepted limits of variability. there was no significant difference between total mre and us scores (wilcoxon signed-rank test z=1.13, p=0.26). at the bowel segment level, there was no significant difference between the mre and us segment scores for the ileum and terminal ileum (wilcoxon-signed rank test, z=0.72, p= 0.472), but significant differences were present between the imaging scores for other bowel segments, with mre identifying more abnormalities. there is a significant positive correlation between mre and clinical consensus scores (spearman's rho=0.598, p=0.0053) and between us and clinical consensus scores (spearman's rho =0.657, p=0.0016). imaging caused a refinement to the original clinical assessment in 8 of the 20 cases, with jejunal and ileal disease the most common reason for 'upgrading' a score and absence of any detectable abnormality on us and mre the most common reason for 'downgrading' a score. we found good agreement between mre and us for total patient imaging scores, ileal and terminal ileal scores. both mre and us scores correlated well with the gold standard clinical consensus, with imaging altering the original clinical decision in 40% of cases. although us detected fewer abnormalities than mr, it correlates marginally better with the clinical consensus, suggesting it is at least equally clinically valuable. background: differentiating between acute osteomyelitis (om) and acute bone infarct (bi) in children with sickle cell disease (scd) is a challenge for clinicians and radiologists, particularly when blood cultures are negative. although bone aspiration is the gold standard test for om diagnosis, it is an invasive procedure and infrequently performed. magnetic resonance imaging (mri) has shown a potential role in differentiating between acute bi and acute om. the goal of this case series is to evaluate the utility of fluid signal on unenhanced fat-suppressed (fs) t1-weighted mr sequence in distinguishing acute bi and om in children with scd. methods: we reviewed a total of 22 children with scd admitted with long bone pain during the one -year study period 2015-2016 attributed to either an acute bi or an acute om. twelve of 22 patients with available bone aspiration, blood culture, and mri data were evaluated for fluid signal, marrow signal and other criteria. of 12 patients, nine patients were diagnosed as acute bi and two patients had acute om and one with coexisting bi and om. the diagnosis was based on the fluid signal on t1 unenhanced t1 fs mr images as compared to aspiration cytology in which eight of nine patients with bi had hyperintense fluid signal on non-contrast t1 fs mr images while one of two patients with om demonstrated hypointense fluid signal. the last patient was diagnosed as a probable coexisting lesion (om&bi) based on a giant well demarcated hypointense marrow signal with an extraosseous hyperintense fluid signal. in acute bi, an abnormal hyperintense subperiosteal or paraosteal fluid signal is frequently observed on unenhanced t1-fs weighted images. this finding was present in the majority of cases in our study population regardless of age, sex or site in the appendicular skeleton. mri fluid signal characteristic on unenhanced t1 fs shows promise as a criterion to differentiate between acute bi and om. role of mri to assess skeletal age in pediatric celiac disease s. bernardo, m. martino, a. laghi, e. tomei; rome/it objective: coeliac children are often subject to weight loss and lower somatic growth rate, compared to healthy children of the same age. the purpose of this study was to asses the feasibility of magnetic resonance imaging (mri) of the hand and the wrist to assess skeletal age and growth delay. we enrolled in our study 39 coeliac children (13 males and 26 females) affected by histological proven coeliac disease, with a chronological age ranged between 5 years and 1 month and 16 years and 4 months (mean age of 10years, +/3 years and 8 months standard deviation). a single mri sequence (t13d se, acquisition time: 1 minute 31 seconds) of the hand and wrist in coronal plane was performed of each patient to estimate the skeletal age. patients' data were compared with a population of normal subjects. the preliminary results showed a delay in skeletal age in children affected by coeliac disease in 85,7% of the simple study, with a delay of maturity of 0.83 years (+/-2,2 years of sd). only 3 children showed advance mri skeletal age when compared to normal subjects. mri of hand/wrist to assess skeletal age may be considered as a reliable indicator of somatic growth. mri, without radiation exposure, can be an used as a diagnostic tool in skeletal delay. it could play an important role in the follow up of coeliac children, after glutenfree diet. the prevalence of metaphyseal injury and its mimickers in otherwise healthy children under two years of age p. eide, å. djuve, r.e. gjøsaeter, k.f. forseth, a. nøttveit, c. brudvik, k. rosendahl; bergen/no objective: metaphyseal lesions in infants and toddlers are believed to have a high specificity for inflicted injury, however, normal metaphyseal irregularities may mimic pathology and lead to overdiagnosis. during the period 2010-2015 all children between 0 and 2 years, seen at the a&e department in bergen (bergen legevakt) due to an injury, and who had radiographs taken, were included. data on previous injury, age, sex and injury mechanism were drawn from the medical notes and pacs archive. all radiographs were reviewed by two researchers and an experienced paediatric radiologist, registering the following: number, site and type of fractures, signs of healing (yes, no), bone structure (normal, pathological) and metaphyseal appearances (shape (normal, metaphyseal collar, metaphyseal irregularity), injury). the study was approved by the institutional review board. six hundred one children (293 girls) between 2 and 24 months of age (mean 17.8 months) were included, of whom 218 (109 girls) had a total of 275 fractures. one hundred eight of the fractures (39.3%) involved the forearm, followed by leg-fractures (59/275, 21.5%) and fractures to the clavicle (37/275, 13.5%). one epiphyseal separation and one metaphyseal lesion (without a history of trauma) were seen. one thousand three hundred twenty metaphysis were analysed, of which 212 (16.1 %) were defined as either irregular (105/1320, 8.0 %) or demonstrating a metaphyseal collar (107/1320, 8.1 %). metaphyseal lesions with a history of trauma did not occur in otherwise healthy neonates and infants under 2 years of age, indicating that this type of fracture has a particular mechanism. metaphyseal irregularities are frequent, particularly around the knee, and should not be mistaken for clms to evaluate whether mri might be used for age estimation, based on greulich-pyle (gp) atlas criteria. 1.5tesla mri of the left hand was conducted in 60 adolescents, and subjectively evaluated by two blinded radiologists. for sequence optimization, coronal mri sequences (t1 tirm, t1 vibe-3d-we, and t1 se) and a left hand x-ray were compared in ten patients (eight male, two female; mean age, 13.5 years). the ages of 50 healthy volunteers (17 s375 (2017) 47 (suppl 2):s297-s pediatr radiol male, 33 female; mean age, 15 years) were assessed from coronal t1 vibe-3d-we. bland-altman plots, intraclass correlation coefficients (icc), and logistic regression models were calculated. coronal t1 vibe-3d-we achieved the best image quality. the correlation between estimated patients' ages on x-ray and mri was high. icc showed high inter-observer agreement (0.95 for x-ray, 0.97 for mri). the estimated age of the healthy volunteers tended to be older than their chronological age. the probability of overestimation was higher in girls than in boys. coronal t1 vibe-3d-we of the left hand is feasible for skeletal age estimation by gp criteria with a high readers' agreement. the likelihood of overestimation of healthy children makes it necessary to develop a new hand atlas representing changes since the 1950s. to assess the relationship between the radiographic findings of metabolic bone disease (mbd) and serum biochemical markers in preterm infants. preterm infants in our neonatal intensive care unit between january 2014 and september 2016 were included. two readers retrospectively reviewed the wrist radiography for grading according to mbd severity. we recorded the levels of alkaline phosphatase (alp) and phosphorous (p) immediately after birth, on the same day of the first wrist radiography (alp-s, p-s), the highest alp levels before the first wrist radiography (alp-hb) and during follow-up (alp-h), and the lowest p levels before the first wrist radiography (p-lb) and during follow-up (p-l). patients were subdivided into four groups according to mbd severity determined by wrist radiography for the first analysis, and were divided into two groups according to mbd presence for the second analysis. one-way analysis of variance with a tukey multiple comparison and the student's t-test were used for statistical comparisons in the two analyses, respectively. a receiver operator characteristic (roc) curve was constructed to determine the optimal cut-off values of the biochemical markers for the radiological prediction of mbd. of the 159 patients, 94, 39, 19, and 7 infants were classified into grades 0 1, 2, and 3, respectively. in the first analysis, alp-s, alp-hb, and alp-h were significantly different between grades 0-1 and 2-3 (all p<0.001). plb was significantly different between grades 0 and 2 (p=0.001) and p-l was significantly different between grades 0 and 2 or 3 (p<0.001 or p=0.001). in the second analysis, alp-s, alp-hb, alp-h, p-s, p-lb, and p-l were all significantly different between the two groups (p<0.001). the roc curve of alp-h showed the largest area under the curve values (0.752, 95% confidence interval=0.676-0.828; p=0.039) for detection of a radiographic change. the optimal cut-off value of alp-h was 473.5 u/l, and the sensitivity and specificity were 81.5% and 47.9%, respectively. the first wrist radiography was obtained at 8.3 ± 5.1 weeks after birth, and alp-h was measured at 6.9 ± 5.3 weeks after birth. the cut-off value of alp for the prediction of abnormal radiological changes in wrist radiography was determined to be was 473.5 u/l. our findings indicate that the highest alp level at around 6.9 weeks after birth could be a valuable predictor of radiological mbd in preterm infants, including those with very low and extremely low birth weights. quantitative grading of tmj synovitis in children with jia-influence of mr-coil, timing after contrast-injection and location of measurements on joint-to-muscle enhancement ratio a. hamardzumyan schmid, c. kellenberger; zurich/ch objective: assessment of signal intensity ratio between joint space and longus capitis muscle on contrast-enhanced mri has been proposed as reliable method across different mr-scanners and protocols for grading temporomandibular joint (tmj) arthritis in juvenile idiopathic arthritis (jia) with a cut-off of 1.55 for diagnosing synovitis. the aim of this study was to investigate potential influences on such enhancement ratios (er). retrospective evaluation of 14 contrast-enhanced mr-studies of 28 tmjs in 7 girls with jia (age 11.3±4.3y) obtained at two occasions with two different coils on a 1.5t scanner. joint-to-muscle er were calculated from signal intensity measurements in different joint compartments, muscles and sequences obtained with varying delay after contrast-injection, and compared with paired sample t-test. er of tmjs without synovitis (n=10) and tmjs with synovitis (n=18), determined by qualitative criteria, were compared to er reported in the literature. superior and inferior joint space to longus capitis muscle er for normal tmjs (2.2±0.9; 2.6±0.7 respectively) exceeded 1.55 in all but one case (figure) and for tmjs with synovitis (3.3±1.1, 3.8 ±1.1) were significantly higher than in 211 cases with synovitis from the literature (2.5±0.8, p≤0001). the same er were higher when obtained with dual-ring coil (3.7±1.1; 4.1±0.9) than with multichannel surface coil (2.2±0.7; 2.6±0.8; p≤0.0002). while er to longus capitis muscle were higher than those to pterygoideus muscle for both coils (p≤0.008), er to pterygoideus muscle did not differ between coils (p>0.2). not considering the timing of the scan, er to pterygoideus muscle were highest in the inferior joint space (1.65 ±0.63), followed by the anterior joint recess (1.52±0.44) and superior joint space (1.49±0.62). comparing images acquired immediately after contrast injection to later images (median delay 11min, range 5-15min), pterygoideus muscle er in the superior (1.2±0.4 to 1.8 ±0.7) and inferior (1.2±0.3 to 2.0±0.6) joint space increased substantially (p<0.0001), while er in anterior recess showed no significant increase (1.5±0.4 to 1.6±0.5, p=0.18). conclusion: joint-to-muscle er are clearly dependent on 1) the signal profile of the mr coil with muscles located further away from the coil providing higher er, 2) the time of image acquisition after contrast-injection with later obtained images providing higher er, and 3) where the joint signal intensity is measured. as these factors need to be accounted for, the described normal and pathologic ranges of joint to longus capitis muscle er cannot be generalised for every mr-system and imaging protocol. integration of 3d c-arm ct images with navigational software provides real-time fluoroscopic guidance during percutaneous interventions in the interventional radiology (ir) suite. a trajectory, drawn from skin entry point to the target lesion on the 3d c-arm ct data, is overlaid on intraprocedural fluoroscopy for real-time needle guidance. this study describes our experience with syngo iguide (siemens) needle guidance software in a range of clinical applications in the pediatric ir suite, including technical success, radiation dose and procedure time. in this irb approved study, all percutaneous interventions performed in the ir suite using syngo iguide over a 3-year period were included. cases were classified by procedure type; for each type, mean effective radiation dose (msv) was estimated using pcxmc program (v2.0.1.3, stuk) and procedure times were evaluated. forty-five patients (25 male, 20 female; mean age: 11±6 years) underwent iguide-assisted interventions including: bone biopsies -35/45 (22 pelvic, 7 lumbar, and 6 lower extremity), intra-articular steroid injections -5/45 (4 sacroiliac, and 1 temporomandibular joint), lumbar punctures -2/45, percutaneous catheter placements -2/45 (cecostomy, and chest tube placement) and bone biopsy with radiofrequency (rf) ablation -1/45. iguide was used in particular for the cecostomy procedure due to high sub-hepatic cecal pole position, and in the chest tube procedure due to the presence of loculated pneumothoraces. all procedures were technically successful. the diagnostic bone biopsy rate was 97.7%. the mean estimated dose and procedure times for each procedure type are listed in table 1 . sonography of neonatal spine (sus) is a simple, non-invasive, quick, relatively inexpensive method to evaluate lumbar spine anomalies in infants less than 4 months of age. unossified posterior neural arches allow beam penetration to obtain high-resolution images of the intra-spinal contents. sus is carried out at the bedside, does not utilize radiation & requires no sedation. linear array transducers with extended field-of-view permit diagnostic sensitivity equal to mri. factors affecting mri resolution like patient movement, pulsation & vascular flow do not affect sus. we use sus as first-line screening test in neonates with lumbosacral cutaneous stigmata & spinal dysraphism (sd) associated syndromes. this was a prospective study approved by the institutional ethics committee. thirty five children (age range of 5 to 15 years) with clinically suspected and complicated pulmonary tb were enrolled in the study. lung mri and ct scan was performed in all the patients. the sensitivity, specificity, positive predictive value (ppv), and negative predictive value (npv) of lung mri in detection of radiological findings that were considered highly suggestive or diagnostic for tb, were calculated, with ct as the standard of reference. lung mri performed equivalent to ct in detection of pleural effusion, mediastinal/hilar lymphadenopathy and lung cavitation with sensitivity and specificity of 100%. agreement between ct and mri in detection of each finding was almost perfect (k: 0.8-1). lung mri was found to be comparable to ct scan for detecting various radiological abnormalities which were highly suggestive for tuberculosis. being a radiation free imaging modality, it has the potential, particularly in children, to replace chest radiographs and ct scan in the coming years. to evaluate differences of myocardial strain assessed by feature tracking using ssfp cardiac mri sequences between pectus excavatum (pe) patients and healthy volunteers. in this prospective study, cardiac mri was performed in 10 pe patients (with a pathologic haller-index above 3.25) and 10 healthy volunteers (5 males and females, respectively; age range 13-30 years) including short-and long-axis cine-ssfp sequences on a 3t scanner. post-examination analysis included standard cardiac volumetry with measurements of the biventricular ejection fractions (ef). additionally, manual biventricular contouring by an experienced radiologist, and subsequent automated strain assessment using dedicated software (circle cvi 42 ®) was performed. longitudinal, radial, and circumferential peak systolic strain and strain rates were analyzed for both ventricles. left-ventricular ef was normal in all patients. five pe patients had a normal right-ventricular ef, in 5 pe patients rvef was slightly impaired (40-44%), all healthy volunteers had a normal rvef. compared with healthy volunteers, pe patients showed a significantly higher apical left-ventricular strain (radial: 53±16.8 vs. 26±8%, p<0.001; circumferential: -23.7±4.8 vs. -15.3±3%, p=0.001) and strain rate (radial: 3.76±1.05 vs 1.71±0.33s -1 , p<0.001; circumferential: -1.92±0.51 vs. -1.08±0.48s -1 , p=0.002). mid right-ventricular strain (radial: 14.4±6.5 vs. 8.2±2.6%, p=0.019; circumferential: -9.7±3.3 vs. -6.4±1.5%, p=0.019) and strain rate (radial: 1.07±0.45 vs. 0.63±0.24s -1 , p=0.015; circumferential: -0.79±0.27 vs. -0.5 ±0.14s -1 , p=0.011), as well as apical right-ventricular strain (radial: 25.6±8.6 vs. 16.5±8.6%, p=0.009; circumferential: -15.9±3.8 vs. -11.1±4%, p=0.019) and circumferential strain rate (-1±0.26 vs. -0.72±0.24s -1 , p=0.029) were also significantly higher in pe patients than in healthy volunteers. left-and especially right-ventricular radial and circumferential strain and strain rate increased from the bases to apices in pe patients. longitudinal strain and strain rate did not differ significantly between pe patients and healthy volunteers. myocardial strain assessed by cardiac mri differs significantly between pe patients and healthy volunteers. as the chest wall deformity usually leads to a compression of the basal parts of the ventricles, higher values of myocardial strain in the mid and apical ventricles in pe patients might indicate a compensation mechanism to enhance especially right ventricular output against sternal compression. to determine the normal range of the haller index (hi) value, and its dependence on the age, sex, and respiratory phase. evaluate the possibility of reduction of the effective dose (ed) of ionizing radiation using a single-slice ct scan technique. the retrospective-prospective study included 165 patients (av. 12y, sd 5y). it consisted of 3 parts. the prospective study included evaluation of ct scans performed by single-slice technique in 30 patients with pectus excavatum both in inspiratory and expiratory phase, without topogram. hi was measured in each patient in both respiratory phases. in retrospective study, 100 ct scans of the chest in children without pectus excavatum were analyzed to determine normal range of hi values depending on the age (0-5y, 5-10y, 10-15y, 15-18y) and gender. the retrospective study also included the analysis of another 30 ct scans in patients who were operated or diagnosed with pectus excavatum. in the latter group of patients the average value of ed of ionizing radiation was calculated, and the values were compared with the average ed obtained using low-dose ct examinations applied in the new protocol (single-slice technique). the normal value of hi was 2.23±0.32. a significant positive correlation between age and value of hi was found. older patients had higher hi (0-5y: 2.01±0.29, 15-18y: 2.35±0.33). results of mann-whitney test did not demonstrate any difference between gender in the observed group, however girls had generally higher hi in all age groups. in the group of patients who were operated/diagnosed with pectus excavatum, hi was 3.34±0.88. the average value of hi in inspirium in children with diagnosed deformity was 2.69±0.76, while in expirium it was 3.49±1.19. only 3/32 (9%) patients had hi value over 3.25 (a boundary value for surgical treatment) during inspirium, while 13/32 (41%) patients had it in expirium, which showed statistically significant difference (p=0.012). single-slice ct technique during the inspiratory and expiratory phase showed average ed of 0.02msv, which is an equivalent of 1 chest xray. it reduced ed more than 20 times in comparison with low-dose whole chest ct. the value of haller index increases with the age and in expiratory phase. we propose the single-slice ct technique without topogram in expiratory phase, as a sufficient and reliable technique in evaluation of haller index and preoperative preparation. mps iva is a lysosomal storage disorder caused by a deficiency of nacetylgalactosamine-sulfatase. main symptom is a systemic skeletal dysplasia. affection of the vascular system has not been described yet. our goal is the analysis of the vascular system in patients with mps iva, based on the example of the aorta. in a retrospective study, 32 patients with mps iva were included. the aorta in its course from 4 th thoracic vertebrae to 10 th was analyzed on the basis of 49 craniospinal mr and 4 ct examinations. to describe the course of the aorta, the area around the vertebral body was devided into 5 equal parts (fig.1) . high buckled arteries in relation to the length of the affected aortal part were indicated as aortal kinking, and a moderate twist in relation to the length of the affected aortal part as aortal tortuosity. results: twelve of 32 patients had an aortal kinking, 10 of 32 patients an aortal tortuosity, 4 of these had moderate and 3 strongly tortuous aortae. seven patients had a normal aortal course, 4 couldn't be analyzed. one patient revealed both, aortal kinking and tortuosity. this study reveals the occurrence of aortic tortuosity in patients with mps iva. we suggest that this complication could be due to glycosaminoglycane deposition in the aortic intima, which may be s379 (2017) 47 (suppl 2):s297-s pediatr radiol associated with an increased vulnerability of the vascular wall. we conclude that the examination of the vascular system should be included in regular follow-up protocols. lung ultrasound in the diagnosis and follow-up of pneumonia in children -is it really as reliable as chest x-ray? s. balj-barbir, j. lovrenski, s. petrović; novi sad/rs to investigate the role of lung ultrasound (lus) both in the diagnosis and follow-up of pneumonia in children. a prospective study was carried out in the regional children's hospital, and included 130 children (av. 2.9y, sd 2.93y) with clinically suspected pneumonia, in whom initial lus and subsequent chest x-ray (cxr) were performed within 24h. the final diagnosis of pneumonia at discharge was used as a reference test to determine the reliability of lus, cxr, clinical and laboratory findings in the diagnosis of pneumonia. children with pneumonia formed a study group, while the control group consisted of children without diagnosed pneumonia. lus finding of subpleural lung consolidation was considered a diagnostic sign for pneumonia. the children with lus signs of pneumonia were followed-up until complete resolution of the lus findings. there were from one to five follow-up lus examinations performed. a final diagnosis of pneumonia was confirmed in 105/130 (80.8%) patients, and 77/105 (73.33%) were hospitalized (av. 10.84, sd 6.19 hospital days). in diagnosis of pneumonia lus, cxr, auscultation, elevated crp, and tachypnea showed sensitivity of 94.3%, 93.3%, 79%, 80% and 32.38%, and specificity of 100%, 92%, 56%, 48% and 100% respectively. lus detected lung consolidations in 99 of 105 children with final diagnosis of pneumonia, and in 84/99 patients lus showed air-bronchogram (figures 1, 2) . lus was superior to cxr in the detection of lung consolidations smaller than 15mm. interstitial lung changes were detected by lus in 50/105 (47.62%) patients, and by cxr in 21/105 (20%). lus and cxr detected pleural effusion in 24/105 (22.86%) and 14/105 (13.33%) patients respectively. mcnemar's test showed no statistically significant difference, and cohen's kappa coefficient showed almost perfect agreement (0.864) between us diagnosis of pneumonia and final diagnosis of pneumonia. during the follow-ups, moderate to substantial agreement between lus and clinical evaluation of the course of the disease was obtained (k=0.406-0.621). in children with complete clinical and incomplete us regression of pneumonia, consolidations of less than 15mm were the most prevalent finding. the average time period until complete resolution of the lus findings was 16.3±10.24 days. children with us detected pulmonary consolidations larger than 50mm were statistically significantly longer hospitalized than others. lung ultrasound in the diagnosis of pneumonia in children is just as reliable as radiography, and should be included in the standard diagnostic protocol. the latest uk nice guidelines for childhood tb contact screening require that a chest x-ray (cxr) be requested only when mantoux or igra testing is positive or if there is a documented reason e.g. clinical concern. nice clarifies the role of cxr in determining treatment choice. we aimed to review cxr referral and treatment in the current climate of european migrant screening. retrospective review of 148 paediatric referrals to the infectious diseases clinic for tb contact screening of whom 125 had cxrs, from october 2009 to august 2015 and correlation with the medical notes. a panel of 3 paediatric radiologists independently interpreted radiographs in the clinical context of tb contact screening and a majority decision was reached. of 148 patients referred to the infectious diseases unit, 125 underwent cxr in addition to a mantoux and igra test. of those 125 cxr's, 20 were reported as having features of pulmonary tb but only 16/20 (80%) were treated as active tb. eighteen of the 105 (17%) cxr's which were reported as having no features of pulmonary tb, were treated as active tb. of those 18, only 6/18 (33.3%) had a clearly documented reason. review of the 125 radiographs (mean age 14 years) by the panel of radiologists noted that all were of readable quality, 20 radiographs were in keeping with a diagnosis of tb, 6 were inconclusive and 99 were normal. the diagnosis of tb was based on lymphadenopathy in 19 and (2017) 47 (suppl 2):s297-s pediatr radiol milliary nodules in 1. parenchymal abnormality was seen in 8 patients [one was the milliary] and effusion was seen in 3. this correlated well with the initial radiology reports of duty radiologists. of the 125 who underwent cxr, referral information was available in 120.109 (90.8%) of these had been appropriately referred because of a +ve mantoux/igra. only 8 out of 11 (72.7%) of those who had cxr despite a -ve mantoux or igra, had a documented reason. according to nice guidelines, 20% of cxr reported positive for tb were not treated for active tb. this may represent a lack of clarity regarding the definition of 'latent tb'. furthermore, only a third of the 17% of patients who received treatment despite negative radiographs had a documented reason. the current migrant crisis requires clarity of x-ray definitions of latent tb to avoid the 20% under-treatment and 17% overtreatment identified in our population. is there really no cardiac problem for performing sports ö.i̇. koska, p. bayindir, h. alper; izmir/tr objective: sudden death in young is a rare condition excluding known anomalies and sudden infant death sydrome; but its consequences are devastating because they are so unexpected. %25 of them occur in a context of sports event. everyday parents of millions of children admit hospitals in order to get permission for participating in sports events. and after physical examination and ecg, physicians are expected to decide such an important issue. however there are a number of silent reasons that may lead to child to sudden death. altough we don't perform ct scans for such indications, we have encountered several cases with abnormalities that can lead to sudden child death and while reporting an examination, awareness of these devastating conditions may be usefull. we searched our database from 01.01.2016 to 01.01.2017 in order to see how often we diagnosed such a silent reason from the ct images that are performed for some reasons. as our center is a tertiary center we have performed 452 cardiac ct examinations in that period that are mainly for excluding or defining complex cardiac anomalies. in order to prepare a pictorial review of unexpected but ct detectable sudden cardiac death reasons, we excluded congenital heart diseases (namely obstructive, shunting or complex anomalies) and ecg detectable arrhytmic diseases. the non arrhytmic, non traumatic reasons for sudden cardiac death excluding congenital heart diseases in the papers are: hypertrophic cardiomyopathy (cmp) (%36), some coronary artery path and origin anomalies (mainly abnormal left coronary artery from pulmonary artery (alcapa), and interarterial path)(%24), increased cardiac mass (%10), dilated cmp (%9), marfan disease (%6), myocarditis (%3), ischemic heart disease (%2). we detected 18 examinations according to our inclusion criteria and selected one cases of each; rca path anomaly, alcapa, dilated cmp, hypertrophic cmp, subaortic discrete membrane and increased cardiac mass for presentation although sudden cardiac death is rare in young children it is a so devastating condition that it must be taken into account for every situation. awareness of silent conditions and active search of them may protect professionals from medicolegal issues and unpleasent results. to describe the spectrum of chest ct scan findings of pulmonary involvement in childhood langerhans cell histiocytosis (lch) and propose a simple scoring system to evaluate the profusion and distribution of the main lung lesions. one hundred forty-six chest ct scans of the 48 pediatric patients with pulmonary lch enrolled in the french national database for lch until april 2016, could be retrospectively and independently reviewed by 3 pediatric radiologists. for each ct scan a semi-quantitative analysis was performed for nodular opacities and cystic abnormalities. the chest was divided in 6 fields (upper, medium and lower field of the left and the right lung) and for each field, both for the nodules and the cysts the score was 0=no lesion, 1= lesions involving up to 25% of the parenchyma, 2=25-50%, 3= 50-75% and 4=more than 75%. of 29 patients evaluated at diagnosis, 18 patients (62%) presented with nodules, 13 patients (45%) presented with cysts and 6 patients (21%) presented a combination of both nodular and cystic lesions. on the initial ct scan, median nodules total score was 1 and median cysts total score was 0. during subsequent ct scans almost the same percentage of patients with nodules (30 patients, 62 %) was found but we observed an increase number of patients with cysts (30 patients, 62 %), median nodules total score was 1 and median cysts total score was 2. the distribution of nodules and cysts was symmetric in the upper, medium and lower fields with an involvement of costo-phrenic angles in 68% of the cases. patients with pneumothorax (13 patients, 27%) had a higher cysts median score (10) than patients without pneumothorax (1). we found alveolar condensation in 12 patients (25%). none of them showed signs of infection at bal examination or any improvement after a treatment with a standard antibiotic therapy while they did show regression under the lch standard regimen of chemotherapy. we proposed a score for semiquantitative analysis of distribution and profusion of nodular and cystic lesions on chest ct scans that can be a useful tool in pediatric population to monitor lung involvement. we found a significant correlation between pneumothorax and a high cysts median score. alveolar condensation could be considered as a possible manifestation of plch in children. lung bases involvement was found in 68% of the cases, representing an important different imaging feature from adult plch. high resolution computed tomography for chronic small airway disease in hiv infected adolescents a.-m. du plessis 1 , s. andronikou 2 , h. zar 1 ; 1 cape town/za, 2 bristol/uk early treatment with antiretroviral therapy (art) and decline in infected infants due to prevention of mother-to-child transmission has resulted in an increase in the population of hiv-infected adolescents. pulmonary disease is common among them. cxr is considered insensitive and terminology inconsistent. therefore, despite concerns related to radiation dose in paediatric patients, high resolution computed tomography (hrct) is the modality of choice for the evaluation of small and large airway pathology, prominent in chronic pulmonary disease. hrct findings are used for prognosis, treatment decisions and defining anatomic extent of bronchiectasis for surgical intervention. the aim of this paper is to demonstrate the spectrum, frequency and extent of airway pathology in hiv-infected adolescents using hrct. a nested sub study was undertaken within the cape town adolescent antiretroviral cohort (ctaac); a prospective, descriptive cohort study of 520 hiv-infected adolescents on art and 120 age matched hiv-s381 (2017) 47 (suppl 2):s297-s pediatr radiol negative controls. hrct was performed on 100 patients who demonstrated abnormal lung function (defined by forced expiratory volume in 1 second (fev1) of <85% and/or low lung diffusion capacity (dlco)). single phase, contrasted multi-detector volume acquisitions were performed from the thoracic inlet to the diaphragms at full inspiration and image data postprocessed to yield thin (0,6mm) and thicker slice images (5mm). three 1mm slices at 3 cm intervals were performed in full expiration. three radiologists interpreted the c t scans independently, with strict imaging definitions, and a majority decision was generated for each finding. ages of patients ranged from between 10 to 17 years with a mean of 14,2. there were 53 females and 47 males with a ratio of 1:0,8. bronchiolitis obliterans was seen in 70% of patients and bronchiectasis was demonstrated in 39%, 12% of which was classified as severe (involving either an entire lobe or more than 50% of at least 2 lobes). there was an absence of lymphadenopathy (a sign of primary tuberculosis (tb)), lymphocytic interstitial pneumonitis (lip) and post tuberculous apical architechtural distortion. miliary tb was identified in a single patient. ground glass was seen in 10% and consolidation in 13%. the majority of hiv infected adolescents with poor lung function demonstrated bronchiolitis obliterans strongly emphasizing the use of hrct for confirming small airways disease. hrct was also useful for demonstrating extent of associated bronchiectasis in 39%. hrct allows classification of patients into those with diffuse small airways disease requiring medical management and those with local disease requiring surgery. background: chronic recurrent multifocal osteomyelitis (crmo) is an autoinflammatory paediatric non-infectious bone disease. presenting symptoms are non-specific, prolonging diagnosis, and leading to deformity, morbidity and unnecessary procedures. imaging is critical to diagnosis, with whole-body mri (wbmri) commonly used in all stages of care. in our institution, a baseline whole-body coronal stir sequence is routinely obtained. aim: to determine lesion distribution and extent on baseline wbmri via retrospective panel review of all patients clinically diagnosed with crmo, and to determine any patterns of involvement that could facilitate earlier radiological diagnosis. method: all patients diagnosed with crmo since december 2009 using published bristol criteria were identified and baseline whole body mris reviewed. the reviewing radiologists were blinded to the original report and previous investigations. each mri was reviewed for focal lesions consistent with crmo. the extent of metaphyseal and epiphyseal lesions was categorized into involvement of thirds of the width of the structure. the wbmri of forty children (28 girls, 12 boys), averaging 12 years (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) were reviewed by the panel using a majority decision rule. three hundred three lesions were recorded, averaging 7.6 lesions (0-26). the tibia was most affected (83 lesions), most commonly the distal tibial metaphysis (25 lesions in 16 patients, 9 bilateral). rib, metatarsals and distal femoral epiphyseals lesions were common. humeral, hand and skull lesions were few. complete metaphyseal involvement, the 'smouldering physis', was most prevalent within the proximal and distal tibial metaphyses. although ranked seventh, the reportedly more common clavicular lesions were the site of the most florid lesions, demonstrating bone expansion and periosteal reactions. two clear patterns of involvement emerged. in patients with clavicular lesions, fewer overall lesions were observed (average 6.4), mainly affecting the axial skeleton and feet. patients with tibial involvement had a higher number of overall lesions (average 9.3), but few lesions outside the lower limbs. only four patients had a both clavicular and tibial lesions. twelve vertebral lesions and four cases of spodylo-discitis were identified; two at t5/6 level, one at t6/7 level, and one involving both t5/6 and t6/7. our series of 40 cases of crmo with baseline wbmri, one of the largest in the published literature, identifies the common sites that should be interrogated for involvement. this study also demonstrates potential as-yet undescribed patterns of skeletal involvement that can be used to aid radiological diagnosis and highlights a non-infective cause for spondylo-discitis. hepatic hemangiomas (hh) are the most common benign vascular tumors encountered in the pediatric population. two main types have been described congenital and infantile, which both display distinct clinical courses and biological features. hemangioendothelioma in differential diagnosis of hh is controversy. recent literature suggests that congenital hepatic hemangiomas present in a focal form, whereas infantile hepatic hemangiomas present in either a multifocal or diffuse form. the goal of this study is to evaluate the features associated with focal and multifocal hh. the records of 45 patients diagnosed with a hepatic hemangioma at a tertiary pediatric hospital from 1994 to 2015 were reviewed. we divided our series into 2 groups: focal or multifocal including diffuse form. clinical endpoints are: age of diagnosis, presence of cutaneous hemangioma, alpha-fetoprotein, thrombocytopenia, cardiac insufficiency. imaging endpoints were echogenicity on us (hypoechoic, hyperechoic or mixed), vessels density on color doppler (<2; 2-5, >5 cm 2 ). presence of calcifications, venous lakes, visible vessels and aortic tapering were assessed on us, ct-scan, mr and angio. treatment and outcome were analyzed. univariate and bivariate analysis were done. this study included 24 focal (9m, 15f) and 21 multifocal (17m, 28f) hh. antenatal diagnosis was done in 6 focal and 1 multifocal hh. focal lesions were associated with the presence of cutaneous hemangiomas (p<0.001) and calcifications (p<0.05). no other variable was significant. conservative management was decided in 12 focal and 11 multifocal hh. steroids (focal: 6, multifocal: 3), steroid-interferon (focal: 1, multifocal: 3), propranolol-steroid (focal:4, multifocal: 4) and surgery in one focal form. complete regression was observed in most lesions (focal: n=19, multifocal n=20), whereas incomplete regression <50% was observed in 4 patients (focal: n=3, multifocal: n=1) and 3 patients in the focal group with the pathology diagnosis of hemangioendothelioma. hepatic hemangiomas demonstrate a wide range of radiological features, with important overlaps in focal, multifocal or diffuse forms. focal and multifocal hh can be seen in congenital hemangioma, infantile hemangioma or hemangioendothelioma. the association of cutaneous infantile hemangioma in the focal group confirmed that the focal lesion can be seen in infantile hemangioma. however, calcifications are more frequent in focal hh which is described in congenital hemangioma or hemangioendothelioma. (2017) 47 (suppl 2):s297-s pediatr radiol symptomatic and asymptomatic meckel's diverticulum in the pediatric population -a retrospective analysis of imaging findings with histopathologic correlation n. abu ata, r. cytter-kuint, j. bar-ziv, i. hadas-halpern; jerusalem/il objective: though meckel's diverticulum (md) is a relatively common gastrointestinal anomaly, many of the symptomatic and most of the asymptomatic md's are often missed on abdominal imaging. our purpose is to describe the radiologic appearance of symptomatic and asymptomatic md in the pediatric population and to correlate the radiologic findings with histopathology. a retrospective analysis of all children diagnosed with md between 1/2004-10/2016 and had relevant imaging (n=38) was done. imaging studies-ultrasound (us), computed tomography (ct) and magnetic resonance imaging (mri) were retrospectively reviewed and evaluated for visualization of md in symptomatic and asymptomatic patients. findings were compared with the preoperative radiology report and the pathology specimen. symptomatic group (n=24): mean age 8.6±6 years, nineteen males. md presented with abdominal pain in 9 patients, small bowel obstruction (sbo) in 6 patients, gastrointestinal bleeding or anemia in 5 patients and intussusception in 2 cases. md was identified in 13 preoperative reports (54.16%) and retrospectively identified in 4 more cases (overall 17 patients, 70.1%). in 7 cases, an inflamed or perforated md were found. in 4 cases, mucosal lining resembling gastric folds was seen. inverted meckel and prominent tissue surrounding the diverticula were seen in 2 patients. in a single case md was misdiagnosed as a duplication cyst. asymptomatic group (n=14): mean age 10.9±6, eight males. md was not mentioned in any of the original reports and only 3 md's were identified retrospectively (21.4%). no mucosal abnormality or irregularity were noted. histopathology: ectopic gastric mucosa was found in 16/24 (66.67%) of the symptomatic patients vs. 2/14 (14.29%) in the asymptomatic group. all 4 patients with sonographic appearance of gastric mucosa had gastric mucosa in pathology (specificity-100%, positive predictive value-100%). md has a variety of radiologic appearances. it can be detected in most of the symptomatic patients but is almost undetectable in asymptomatic patients. heterotopic gastric mucosa is more common in the symptomatic group. inflamed gastric mucosa may have a typical appearance resembling a small stomach, a sign that was not described before and has both high specificity and high positive predictive value for gastric mucosa within a meckel diverticulum. preliminary results on dna damage from ct irradiation in pediatric patients i. dilevska, e. nagy, w. schwinger, e. sorantin; graz/at the increased radiation sensitivity in children, compared to adults, is a well-recognised fact in the pediatric radiology community. the high-dose irradiation induced dna damage has been well established, however the dosages that are clinically used in everyday ct procedures are so low, that it remains unclear how this severely affects the dna and can induce cancer in the long run. the aim of this study is to assess the effects of lowdose ionizing radiation from ct in children by establishing a standardization curve ranging from the high to the low, medically significant ctdi values. this is done by measuring the phosphorylation of the h2ax histone (γh2ax), which is considered a biomarker for quantification of radiation induced dna double-strand breaks (ddsbs). the detection of the γh2ax histone was done by two methods: immunofluorescence microscopy (im), which is an established method for detection and quantification of this histone and the new, promising flow cytometry technique (facs). for this study, 35 leucocyte samples ("buffy coats") were provided by the local transfusion department and these samples were irradiated with a clinical ct scanner (aqilionone, tmse) with values ranging from 2153,6 to 0,16 mgy ctdi. afterwards the samples were processed with both methods. for the immunofluorescence, twostep immunostaining was used with two different antibodies and the cell and foci counting were done on an olympus xc10 microscope, while the facs staining was done with a one-step antibody and the samples were measured on a navios flow cytometer (beckman coulter). comparable results were detected with both methods, with a good correlation between the facs and im, with a linear incline (r 2 >0.95) in the high and in the low dosages from 0,16 to 5.0 mgy ctdi. however, in the samples irradiated with doses below 5.0 mgy ctdi, there seems to be less phosphorylated h2ax than in the native samples. two possible explanations arise: a) low dose irradiation initiates repair that extends to ddsbs occurring naturally or b) low dose irradiation doesn´t cause phosphorylation of this histone, but affects other dna damage and repair pathways. the preliminary findings show that the facs analysis can be used as a valid replacement method for the labor-intensive if method in the higher dosages. however more analysis should be done to establish its accuracy in the lower regions since underlying mechanisms are not clear yet. a. turkaj 1 , g. cicero 2 , e. sorantin 1 , r. coroiu 3 ; 1 graz/at, 2 mesina/it, 3 cluj-napoca/ro there is only little information available regarding imaging procedures in the trauma setting of pediatric patients. such data can serve as a rational basis for running pediatric trauma units. the purpose of the paper is to investigate the number, types and distribution of imaging procedures in a tertiary pediatric trauma center serving children of about 1.2 million inhabitants with approximately 225.000 children. all trauma-caused admission to the emergency room and their imaging procedures were analyzed retrospectively occurring within a period of 15 months. a cohort of 263 patients (m:f=1.9:1) were analyzed. patients were grouped according to age into the following categories: neonates, infants, middle childhood, early adolescence, late adolescence. imaging procedures were classified into plain films, us, ct and mri. furthermore, the time of admission was noted and categorized in time slots 07:00-15:00, 15:00-20:00, 20:00-24:00 and 24:00-07:00. referral cause was divided in domestic accidents, motor-scooter-bicycle accidents, car accidents, sport injuries, falls from height and others. the average age was 9.5 ± 5.4 years, aligned in the following age-groups neonates 13 (5%), infants 68 (26%), middle childhood 42 (16%), early adolescence 58 (22%), late adolescence 82 (31%). a total of 364 imaging procedures were performed, of which 210 plain films (58%), us 81 (22%), ct 69 (19%) and mri 4 (1%). there was a statistically significant difference of imaging procedures due to age in particular in us and ct. regarding the timeslots: 07:00-15:00 ;111 patients (42%), 15:00-20:00; 118 patients (45%), 20:00-00:00; 31 patients (12%), 00:00-07:00;3 patients (1%). domestic accidents were the leading referral cause with 82 cases (31%) prevailed age groups were infants and middle childhoods corresponding for more than 75%. motorscooter/bicycle accidents accounted for 51 cases (19%) of which most were early and late adolescence (more than 85%), s383 (2017) 47 (suppl 2):s297-s pediatr radiol sport's accidents 38 (14.4%) equally shared among middle childhood, early and late adolescences. car accidents 37 (14.0%) cases and fall from height 22 (8,3%) did not show any prevalence according to the age groups. for the first time detailed data about imaging procedures at the emergency room for pediatric patients are now available. over half of the admissions (55%) occur outside regular work hours thus representing a challenge for the staff in duty and this fact should be considered in working schedules. due to strict interdisciplinary developed diagnostic pathways the number of ct examinations was reasonable low. head ct in a regional children's hospital without mri -effective doses and justification of clinical indications j. lovrenski, n. milenković; novi sad/rs to calculate effective doses (ed) for pediatric head computed tomography (ct), to determine the most common referral diagnoses, and the share of normal and pathological ct findings. a retrospectiveprospective study comprised all the children with performed ct examination (16-slice scanner) within a one-year period. pediatric ct protocols were used. the values of ed for head cts were calculated based on the two different models (shrimpton's and icrp publication 103). average ed for different age groups was expressed as the number of chest x-rays (cxrs) (1cxr 0.02msv). the most common non-traumatic referral diagnoses for head cts were determined, as well as percentage and type of pathological ct findings. a share of pathological ct findings was also determined for trauma as a referral diagnosis. head cts were represented with 649 (71%) in total number of 924 ct examinations within a one-year period. the different calculation models have shown the difference in ed values of up to 20.5%. eds for head cts were equivalent of 66 (15 years of age and older) to 140 (younger than 3 months of age) cxrs for one sequence of scanning. the most common non-traumatic referral diagnoses for head cts were: loss of consciousness, epilepsy, headache, convulsions, and vertigo. in this group of 273 patients, 86% of completely normal ct findings were found. pathological findings in this group consisted of the patients with the most common non-traumatic referral diagnoses were as follows: cortical atrophy (15 patients), arachnoid cyst (10), ischemia (6), porencephalic cyst (3), agenesis of the corpus callosum (1), chiari malformation -type i (1), open-lip schizencephaly (1), and tumor of the posterior cranial fossa (1). most common incidental, extracerebral pathology discovered included sinusitis and otomastoiditis. in patients with trauma as referral diagnosis, the share of pathological ct findings was 59.5%. it is necessary to get clinicians familiar with the extent of ionizing radiation that children are exposed to during the head ct examinations. a more careful selection of children for head cts is necessary in an every-day clinical practice, especially for patients with non-traumatic referral diagnoses. diffuse and symmetric diffusion restriction involving the white matter of the brain in patients with neonatal seizures j.-y. hwang 1 , y.j. lee 2 , y.-w. kim 2 ; 1 yangsan-si, gyeongsangnam-do/ kr, 2 yangsan-si/kr this study aimed to evaluate magnetic resonance (mr) imaging findings in patients with neonatal seizures focused on the diffuse white matter lesions on diffusion weighted image (dwi) in addition to clinical manifestations. a total of 55 neonates aged less than 1-week old underwent brain mr imaging for evaluation of neonatal seizures between november 2008 and december 2016. among them, 12 patients showed diffuse and symmetric pattern of high signal intensity on dwi. clinical, laboratory, and mr images were analyzed retrospectively. nine patients were males and three patients were females. patient age was 5.2 ± 0.8 days (range, 3-6 days). all the patients were born at full term. the most frequent month of the hospital visit was march (n=4) and january (n=3). eight patients showed generalized clonic seizure and four patients showed partial clonic seizure. stool viral test was performed in nine patients. among them, five patients were positive for the rotavirus and one patient was positive for the astrovirus. nine patients underwent cerebrospinal fluid analysis, however, all showed negative results. mr imaging was performed at 2.2 ± 1.6 days after onset of seizures. diffuse and symmetric diffusion restriction were distributed along the cerebral white matter tracts and both thalami (fig 1) accompanied with high signal intensity on either t2-weighted images or on the fluid-attenuated inversion recovery (flair) sequence. multiple foci of high signal intensity on t1-weighted images at the centrum semiovale that was affected on dwi were also observed. follow-up period was 11.3 ± 9.6 months (range, 1.8 -30.9 months) and developmental delay was encountered in three patients. six patients underwent follow-up mr imaging at the age of 8.4 ± 4.2 months (median, 8.5 months; range, 3.7-13 months). five patients showed volume loss in cerebral white matter on both sides of the brain and four patients showed high signal intensity of the periventricular white matter on either t2weighted images or flair sequences (fig 2) . myelination delay was not observed in follow-up mr images. diffuse and symmetric diffusion restriction involving the cerebral white matters can be seen in patients with neonatal seizures on mr imaging. our study shows that rotavirus is commonly encountered, but not exclusively detected in these patients. nevertheless, viral infection-associated encephalopathy should be considered when a patient is presented with characteristic clinical and mr findings. whole body mri on diagnosis and follow-up of neurofibromatosis type 1 d. grassi, v. tostes, e. caran, h.m. lederman; sao paulo/br demonstrate that whole body mri is effective on showing neurofibromatosis type 1 involvement of different regions of the body not known by the clinicians. review of 41 patients with neurofibromatosis type 1 (nf1) who underwent whole body mri throughout their follow-up with the majority of them had only brain and spine imaging studies. it was possible to demonstrate that whole body mri provides an overview of nf1 systemic manifestations and neurofibroma's extension beyond the clinic expectation. despite being rare, sarcomatous degeneration was suspected when there was any difference on the characteristics of the neurofibromas. whole body overview where its possible to see the neurofibroma's extension in right cervical region, scoliosis and multiple plexiform neurofibromas. only the biggest neurofibroma was detected by clinical exam. however it is possible to identify two others neurofibromas. whole body view of multiple plexiform neurofibromas. whole body mri is a radiation-free exam and it is useful on the diagnosis of nf1 and on patient's follow-up. it provides an overview of the systemic s385 (2017) 47 (suppl 2):s297-s pediatr radiol involvement and neurofibroma's extension beyond the clinical expectation. during patients follow up, it could also show tumor's characteristics modification, which was considered as a possible sarcomatous degeneration. accuracy of non-radiologists and lay-persons for identifying children with cerebral cortical atrophy from 'mercator map' curved reconstructions of the brain s. vedajallam 1 , a. chacko 1 , s. andronikou 2 , e. simpson 2 , j. thai 2 ; 1 east london/za, 2 bristol/uk objective: background: communication of cortical brain atrophy in children with term hypoxic ischaemic injury (hii) to parents and the legal fraternity contesting compensation rights can be very difficult using text and standard cross-sectional images. when demonstrating the cortex in hii, a single image of the brain surface, much like the way a map of the earth is derived from a globe, can be generated from curved reconstruction of coronal magnetic resonance imaging (mri) scans i.e. a mercator map. lay people's ability to identify abnormal scans from such maps without prior training requires evaluation before routine use. aim: to determine the sensitivity and specificity of lay people in detecting abnormal brain scans through review of mercator flat-earth maps of the brain, without prior training. ten mercator map images were provided to 100 participants with a distribution of 5 hii, 1 cortical dysplasia and 4 reported normal. participants were required to identify abnormal scans. sensitivity and specificity overall and for sub-groups were derived by averaging true positives and negatives; false positives and negatives. the results show a strong ability for lay-people to identify normal versus abnormal mri brain studies using mercator maps. the sensitivity and specificity in this group is 67% and 73% respectively. non-radiologist physicians and radiographers performed slightly better than lay people as expected. radiologists of course had very high sensitivity and specificity of 86% and 100%. the mercator map is therefore a viable tool in the communication of complex mr imaging to the lay-person. safety and efficacy of sphenopalatine ganglion blockade in childreninitial experience l. dance, c. schaefer, d. aria, r. kaye, r.b. towbin; phoenix/us objective: sphenopalatine ganglion (spg) blockade is known to be a safe and effective migraine headache treatment among adults. this paper will report the initial experience in the pediatric population with spg blockade. one hundred thirty-three procedures were performed in 85 patients ages 7 to 18 from february through november 2015. pre-intervention headache scores were recorded on a scale of 1 to 10. the procedure was performed supine with neck in hyperextension. anesthesia of the bilateral nares was accomplished with lidocaine spray and gel. contrast was injected using a sphenocath confirming catheter position. 4% lidocaine was injected. patients remained supine with neck in hyperextension for 10 minutes. post-intervention headache scores were recorded. mean pre-treatment score of 5.55 decreased to 3.28 post-treatment (δ2.27, 95% ci 1.34-3.20, p<0.0001). there were no complications. spg blockade is a safe and effective treatment for migraine headaches in children which results in decreased reliance on intravenous drug therapy. orbital masses represent a spectrum of benign and malignant lesions in children that can be challenging to diagnose and treat. imaging plays an important role in diagnosis, due to a potentially limited clinical examination and risks associated with biopsy. mr imaging is a powerful tool for imaging the orbit, due to the excellent tissue contrast it provides. yet conventional mri has a limitation in discriminate the benign from malignant lesions. diffusion-weighted imaging (dwi) is non-invasive rapid technique uses the water diffusibility to produce contrast among different kinds of tissues. our propose was to assess the role of dwi and calculated apparent diffusion coefficient (adc) values in characterization of the pediatric orbital masses regarding benignancy or malignancy. one hundred and thirty patients with recently diagnosed orbital masses and who underwent preoperative conventional mri and dwi were included in this study. the orbit was divided into six compartments: the eye globe, retroocular fat, optic nerve, lacrimal system, bony boundaries and extra-ocular muscles. the average adc obtained from each tumor was compared with the histopathological diagnosis determined from subsequent surgical sample. seventy girls and sixty boys with orbital masses were included in this study. their age was ranged from 1 month to 18 years. the globe is the seat of lesions in 43/130 cases, optic nerve in 27/130 case. seven cases have lesions in the lacrimal gland. forty-five of cases was diagnosed as having benign masses & 85 of cases have malignant lesions. there is a statistically significant difference between the mean adc value of the benign lesions (1.39±0.52 x10 -3 mm 2 /s) and the mean adc value of the malignant lesions (0.69±0.22 x10 -3 mm 2 /s) (p<0.001). the optimal adc cutoff value that was determined for discrimination between these lesions is: 1.075 x10 -3 mm 2 /s), with sensitivity of 97.14% and specificity of 75%. using conventional mri alone in predicting benign and malignant lesions has the sensitivity of 76% and specificity of 91% with 94% positive predictive value and 67% negative predicative value. combining dwi and conventional mri has increased accuracy, as the sensitivity and specificity were 95%, 86% respectively with 93% positive predictive value and 90% negative predicative value. (2017) 47 (suppl 2):s297-s pediatr radiol adc values provide an accurate, sensitive, fast, and non-invasive mean of characterization of pediatric orbital tumors. a priori tumor characterization is useful in timing and treatment planning for orbital tumors. utiliy of resting state fmri in children for preoperative language mapping l.-m. leiber, m. delion, a. ter minassian; angers/fr to assess if resting state fmri is able to detect language eloquent areas in childrens. six children, from 8 to 15 years old suffering from brain lesions were enrolled in this study. they underwent mri with one 3dt1 morphology session and three 10 minutes fmri sessions, including one resting state fmri and two language task induced activity fmri sessions. analysis was performed using a generalized linear model for the first one and a spatial independent component analysis approach for the two others. language maps were compared with cortical mapping obtained by intraoperative direct stimulation. language network was identified systematically by resting state session but not by task induced activity sessions. moreover, in two of the six patients, resting state fmri was able to detect eloquent areas found during intraoperative cortical mapping that were not present in task induced activity sessions. resting state fmri appears superior to task induced activity fmri in detecting language eloquent areas. is sclerotherapy an effective treatment option for ranulas or thyroglossal duct cysts in children? d. aria, l. dance, c. schaefer, r. kaye, r.b. towbin; phoenix/us to assess the utility of sclerotherapy in the treatment of ranulas and thyroglossal duct cysts materials: from 2015-2016, 8 patients varying in age from 16 months to 32 years were referred to the ir department for sclerotherapy. of the 8 patients, 6 had a diagnosis of ranula while 2 had the diagnosis of thyroglossal duct cyst by either mr, ct, or us. sclerotherapy treatments were performed with standard sclerosing agents, i.e. sotradecol 3% foam, absolute ethanol, and bleomycin. in the subset of patients with ranulas, sclerotherapy was commonly performed in accordance with salivary (submandibular and/or sublingual) gland botox injection or ethanol ablation. 22-gauge or 5f sheathed needles were used for us-guided access to the lesions, with ranula sclerotherapy being performed after placement of side-hole drainage catheters (5-10f) due to their increased viscosity. the preferred sclerosing agent was injected with dwell times ranging from 15 mins to 3 hours. salivary gland injection/ablation was performed under usguidance using a 22-gauge needle with volume injection targeted centrally within the gland or in the portion of the gland abutting the ranula. after treatment, all patients were scheduled for follow-up ultrasounds at a minimum of 8 weeks to assess lesion response or residual disease. a total of 23 sclerotherapy treatments were performed. of the 8 patients, 2 were lost to follow-up after single sessions for ranula and thyroglossal duct cyst. the other 6 patients all had follow-up ultrasounds after each of the remaining 21 sclerotherapy sessions. four of these patients showed initial improvement with either decreased size of lesion or lesion resolution while the other 2 showed no improvement with either stable or increased size on initial follow-up. the 4 patients who initially showed promising response unfortunately had recurrence on follow-up imaging and ultimately, demonstrated no favorable response to sclerotherapy after subsequent treatments regardless of whether treatment was combined with ethanol/botox salivary gland injection. in summary, all 6 patients who were successfully followed show no appreciable response to treatment for ranula or thyroglossal duct cyst. despite the emergence of clinical requests for sclerotherapy of ranulas and thyroglossal duct cysts, in our case series, sclerotherapy has not proven to be an effective treatment option using our current drug regimen. role of the susceptibility-weighted imaging (swi) in the neuroimaging of term newborns g. rudas, e. varga, p. barsi, l. kozák, ü. méder; budapest/hu objective: susceptibility-weighted imaging (swi) was introduced in the neonatal neuroimaging only a few years ago. we can find only a few publications about its advantages and disadvantages. according to our experience, swi is extremely useful not only for detecting bleedings but for the diagnosis of other diseases as well. during the last year we had 164 mri examinations on term newborns (0-14 days of life) and the swi gave additional information in 54 cases. we used a 3t philips insignia scanner. in the case of the questionable hypoxic-ischemic-encephalopathy (19 cases) and the metabolic diseases (5 cases) we could find increased signal intensity in the cortex; in the case of stroke we could find the thrombus itself in 6 cases; the avm were much clearer using the swi in 3 cases; at the pvl in 3 cases we could visualize the cysts better using swi; in the case of congenital heart disease (11 cases) and in the case of sinus thrombosis (4) we could find microbleedings and/or dilated veins; in 3 cases the position of the lateral ventricle drain or shunt was much clearer using the swi. the swi gave important additional information in 54/164 (33%). the swi is a strongly recommended new sequence at the mri examination of the term newborns' brain. a disadvantage of swi is that it requires ca. three minutes examination time (in contrast to t2* which is only 1 minute long). mechanical birth-related trauma: imaging of the "accidents of birth" a. chaturvedi, j.g. blickman; rochester/us objective: 1. to discuss definition, incidence and risk factors leading to "mechanical birth-related trauma" and compare these with existing literature. offer an organ-system based classification scheme encompassing the varied manifestations of birth-related trauma and describing the implications on care decisions. materials: the hospital imaging department database was searched for neonates who presented with history of difficult/traumatic birth at our obstetric center between january 1, 2013-june 30, 2016. search software used was primordial customised radiology solutions, san mateo, ca. the search terms used were "macrosomia", "shoulder dystocia", "instrumental delivery", "malpresentation", "cephalopelvic disproportion", "forceps" and "vacuum". initial and follow-up imaging and clinical data on these neonates was reviewed and compiled by two board-certified pediatric radiologists. the relevant literature was reviewed and findings compared. organ-system based classification scheme for birth-realted trauma. in our study, mechanical trauma of birth was seen to manifest within different organ systems, which have been listed below in the order of occurrence within our sample. injuries to the skull (sutural overlap, dents and fractures), scalp hemorrhages (subgaleal hematoma, cephalhematoma, caput). intracranial intraand extra-axial hemorrhages (subdural, subarachnoid, epidural, intraparenchymal). clavicle fractures neonatal brachial plexus injury. sternocleidomastoid hematomas. adrenal hemorrhages. cervical spinal cord contusions. schematic diagram depicting intra-and extracranial hemorrhages by location. 6-year-old with history of calvarial fracture at birth-fracture did not heal but enlarged secondary to leptomeningeal entrapment at the fracture sitean entity called "growing fracture" or "leptomeningeal cyst". multiple newborn organ systems can be injured from mechanical trauma of birth. our numbers compare favourably with the existing literature. mechanical birth-related trauma can occur simultaneously with hypoxic-ischemic birth injury. although most of these injuries spontaneously and completely resolve, long-term complications can be seen in some cases. few of these injuries are life-threatening. imaging plays a crucial role in diagnosis and follow-up, and can assist in decision making as well as in counselling the parents. ewing sarcoma of tibia in an infant girl a. seehofnerova, j. skotáková, i. červinková; brno/cz objective: ewing sarcoma (es) is the second most common primary bone malignancy in children. it histologically originates from neuroectodermal tissue and consists of small round blue cells. ewing tumour family is very close to primitive neuroectoderm tumour (pnet) family with diverse stage of differentiation, ewing sarcoma being less differentiated. approximately 50 % of the cases occur between ten and twenty years of age with slightly higher prevalence in male gender. nine-month-old caucasian girl presented to local surgery department after she had wedged her lower leg in a bed. the right lower leg was swollen and painful. she was initially diagnosed with a ligament injury and underwent standard treatment. oedema gradually disappeared, but swelling and pain increased after three weeks. she also suffered from a fever of 38.1°c (100.58 o f). at that point x-ray of her right lower leg was performed with report describing pathologically changed structure of tibia and she was referred to our university centre. (2017) 47 (suppl 2):s297-s pediatr radiol we made a second reading of the plain film, reporting sclerotic heterogeneous bone structure of the right tibial diaphysis and distal metaphysis, onion-like periosteal reaction with sunburst spiculation and cortical bone destruction. her laboratory results were: crp 7.2 mg/l, ld 6.2 μkat/l, nse 44.3 μg/l, ferritin 18 μg/l. crp has been raising for a week to 25 mg/ l, then decreased to normal level. differential diagnosis was established as a primary bone malignancy (especially es) or, less likely, an osteomyelitis. mri revealed pathological signal of bone marrow of diaphysis of the whole tibia with cortical scalloping and periosteal spiculated apposition. epiphyses were spared. dorsal cortical bone was interrupted with extraosseal spread of the process. intraosseal part enhanced heterogeneously, whereas extraosseal component enhanced almost homogeneously after contrast medium administration. total size of the tumour was assessed as 97x23.5x20 mm (22.8 ml) . adjacent muscles were oedematous with post-contrast enhancement. there were also few enlarged lymph nodes in popliteal region. results from the biopsy confirmed ews with positive ews/fli1 gene. tumour was assessed as a localized disease, enneking iib. patient underwent chemotherapy according to aews1031doc protocol and a knee-exarticulation with no traces of tumour in resection lines. nowadays she is in the first complete remission. x-ray: ap view mri: etw1_tse postcontrast, sagittal view, pre-treatment mri: etw1_tse postcontrast, sagittal view, after initial treatment unique teaching points: ewing sarcoma belongs to common primary bone tumours in children but is a very rare unit in infants. despite the age predilection it is necessary to consider this diagnosis even in children younger than one year of age. scimitar syndrome together with pulmonary sequestration and horseshoe lung: congenital pulmonary venolobar syndrome b.e. derinkuyu, h.n. özcan, y. tasci-yildiz, h g. cınar, u.a. orun; ankara/tr objective: congenital pulmonary venolobar syndrome (cpvls) comprises of a spectrum of pulmonary developmental anomalies. the main components of cpvls are hypogenetic lung partial anomalous pulmonary venous return, absence of pulmonary artery, pulmonary sequestration, systemic arterialization of lung, absence of inferior vena cava. minor components of cpvls include tracheal trifurcation, eventration and partial absence of the diaphragm, phrenic cyst, horseshoe lung, esophageal and gastric lung, anomalous superior vena cava, and absence of the pericardium. in this case presentation, we present a baby with scimitar syndrome, pulmonary sequestration, horseshoe lung and right aberran subclavian artery. a 3 month-old girl was admitted to our hospital with the suspicion of scimitar syndrome from a different hospital. she did not have any symptoms. the physical examination was unremarkable. on plain radiograph, the baby had dextrocardia. there was a doubtful tubular structure with the shape of scimitar and a nodular radioopacity behind the heart (figure 1 ). transthoracic echocardiography demonstrated the dextrocardia, atrial septal defect and the right pulmonary artery hypoplasia. afterwards, the ct angiography was done for confirmation of scimitar syndrome and other accompanying abnormalities. on the ct angiography, there was a partial anomalous pulmonary venous return to the suprahepatic inferior vena cava known as scimitar syndrome. besides this, there was a right pulmonary extralobar sequestration in the lung base. the arterial supply was arising from the celiac trunk, while the venous drainage was going directly to the inferior vena cava. therefore, the right lung was hypoplastic of which the tongue of the right pulmonary parenchyma passing between the aorta and heart, appearing confluent with the left lung in a horseshoe configuration. there was dextrocardia and right aberran s389 (2017) 47 (suppl 2):s297-s pediatr radiol subclavian artery. the patient was subjected to catheterization and angiography for treatment. on plain radiograph, the baby had dextrocardia. there was a doubtful tubular structure with the shape of scimitar and a nodular radioopacity behind the heart unique teaching points: the term cpvls is an umbrella to a group of pulmonary parenchymal and vascular anomalies that may present in combination. mdct is a helpful diagnostic tool in the preoperative evaluation for delineation of the components of this syndrome. congenital pulmonary venolobar syndrome refers to a wide spectrum of pulmonary developmental anomalies that may appear single or in combination. the main components of congenital pulmonary venolobar syndrome are hypogenetic lung (including lobar agenesis, aplasia, or hypoplasia), partial anomalous pulmonary venous return, absence of pulmonary artery, pulmonary sequestration, systemic arterialization of lung, absence of inferior vena cava, and accessory diaphragm. in this case presentation, we describe a child with scimitar syndrome, bilateral sequestration, hypogenetic lung (single lobed left lung) and right aberran subclavian artery. an 8 year-old syrian girl was admitted to our hospital with the history of heart defect. she did not have syncope or ciyanosis whereas she has easy fatigue and palpitation. on plain radiograph the anomalous draining vein was seen as a tubular structure paralleling the right heart border in the shape of a turkish sword ("scimitar") ( figure 1) . transthoracic echocardiography demonstrated the scimitar vein as well as large patent ductus arteriosus (pda), atrial septal defect and left pulmonary hypoplasia. afterwards, the ct angiography was done for confirmation of scimitar syndrome and other accompanying abnormalities. on the ct angiography, there was a partial anomalous pulmonary venous return to the suprahepatic inferior vena cava known as scimitar syndrome. besides this, there was a bilateral intralobar pulmonary sequestration in the lung bases. the arterial supply of the right side was arising from the celiac trunk, while the left side feeding artery was originating directly from the descending aorta. therefore, the left lung had a single lobe with single pulmonary vein draining to left atrium. there was a large pda and right aberran subclavian artery. the patient was subjected to catheterization and angiography for treatment. the right sided anomalous draining pulmonary vein and the feeding artery of the right sequestration were closed in the first session. the procedure was completed without any complication. afterwards, the closure of the feeding artery of the left pulmonary sequestration and the pda were planned in the next sessions. on plain radiograph the anomalous draining vein was seen as a tubular structure paralleling the right heart border in the shape of a turkish sword ("scimitar") unique teaching points: congenital pulmonary venolobar syndrome comprises a heterogeneous group of uncommon abnormalities that may occur in combination. diagnosis of congenital pulmonary venolobar syndrome can be confirmed by ct angiography that allows detailed evaluation of vascular, tracheobronchial, and pulmonary parenchymal abnormalities with a single short, noninvasive procedure. neck infection disclosing diagnosis of congenital fourth branchial arc anomaly in a girl h.n. özcan, z. aycan, b. ardıclı, m. haliloglu; ankara/tr objective: congenital branchial arc anomalies are rare entities. herein, we describe the imaging findings of acute suppurative infection of the neck caused by fourth branchial fistula in a child. case presentation: an 11-year-old girl presented to our pediatric emergency department with fever, left sided neck swelling and redness. her complaints were started five days ago. on her physical examination, there was a 4x4 cm, stiff, painful mass lesion with redness on the left side of the neck. blood count and thyroid function tests were in normal range; however, c-reactive protein level and erythrocyte sedimentation rate were elevated. neck ultrasonography revealed diffuse soft tissue swelling, a hypoechoic mass consistent with abscess in the left thyroid lobe and perithyroid tissue. the left lobe of the thyroid gland had poorly defined margin and a focus of air. contrast-enhanced neck mr imaging demonstrated an abscess in the left thyroid and perithyroid tissue ( figure 1) and enhancement of the soft tissue plane around the left pyriform fossa (figure 2 ). barium swallow revealed the sinus tract originating from the left pyriform sinus apex. the patient was operated after antibiotic treatment and sinus tract was surgically excised. the aim of this report is to describe three cases of right kidney wilms' tumor with cavoatrial tumor extension, referred to our institution between january and september 2016. case presentation: three children, two girls (3 and 7 years old) and one boy (2 years old) were admitted at the emergency service with cardiac failure symptoms; the latter had also liver failure. echocardiography showed right atrial thrombus in all three patients, as an extension of massive obstructive thrombosis of the inferior vena cava (ivc). abdominal ultrasonography revealed in all patients a right renal mass, associated to right renal vein thrombosis that extended to the ivc and to the right hepatic vein. contrast enhenced computed tomography confirmed findings. patients were treated primarily with chemotherapy before surgery, with partial regression of the thrombus in two patients and no response in one. (2017) 47 (suppl 2):s297-s pediatr radiol unique teaching points: wilms' tumor is the most common renal malignancy in children and its intravascular extension is a well-recognized event. incidence of tumor extension to inferior vena cava (ivc) is reported to be of 2-5% and intra-atrial extension of 0,2-1,2%. it occurs most commonly in tumors located in the right kidney (probably due to the shorter path of the right renal vein compared to the left one). this complication does not directly influence the prognosis of malignancy, but the degree of intravascular extension determines technical surgical strategy and increases difficulty of the surgical procedure, especially when there is intracardiac involvement, which increases morbidity. several classifications have been proposed in the adult age group, but due to the similarity of the degree of intraoperative difficulty, the same classifications are used in children. pritchett et al. (1986) described the relation between thrombus and hepatic vessels: level i -intrahepatic intravascular extension; level iiintrahepatic extension; and level iii -suprahepatic or atrial extension. staehler et al. (1987) proposed a different classification that was posteriorly modified and detailed by daum (1994) : stage i -small extension (thrombus size within ivc <5 cm); stage iilarge thrombus (> 5 cm within the ivc), but still below the hepatic vessels; stage iii -thrombus extending to the level and above the hepatic vessels; stage iv -intra-atrial thrombus. a 4 year old boy presented with a soft tissue mass in his forearm which appeared to have grown quickly in size over a period of three to four months. physical examination demonstrated a welldefined mass in the dorsal aspect of the forearm with no pulsatile bruit. intial differentials included a vascular anomaly or a sarcomatous lesion. the patient proceeded to have an ultrasound examination which revealed a very well-defined heterogenous subcutaneous mass, mostly solid in substance. the lesion measured 3.5 cm x 1.6 cm x 3.5 cm (transverse x length x depth). there was no evidence of muscle invasion. prominent internal arterial vascularisation was demonstrated and the mass was classed as inderminate in nature. subsequent mr findings demonstrated a mass with t1 signal isointense to muscle, hyperintense t2 signal and marked homogenous enhancement. small foci of intralesional t1 hyperintensity and larger areas of t2* gradient hypointensity were noted, in keeping with small areas of intralesional blood. vessels were seen to extend from the subcutaneous fat into the lesion. the mass slightly distorted the underlying extensor muscles and tendons of the forearm but there was no deep extension across the fascia. findings deemed the lesion to be more malignant in nature. the patient underwent incisional biopsy and histological findings confirmed a diagnosis of angiomatous fibrous histiocytoma. these tumours are rare soft tissue tumours which most commonly occur in children, adolescents and young adults. while it is rare, there is a potential for local recurrence and metastasis. therefore, it is essential to identify these tumours where possible or at least consider them as a differential for a soft tissue mass in a child. the surgeon commented that the imaging findings and report were essential in making the initial decision about whether to perform an incisional or excisional biopsy as the best treatment for the tumour is wide surgical excision with clearance of margins. unique teaching points: angiomatous fibrous histiocytomas are rare lesions with potential for recurrence and metastasis and therefore should be identified and managed appropriately as a malignant tumour. they are often confused as soft tissue haemangiomas or complex haematomas. it is very important to be aware of the presentation and imaging findings, remembering this form of tumour as a key differential for a soft tissue mass. nasopharyngeal anlage tumor in a neonate with the initial presentation of respiratory difficulty: correlation between imaging and clinicopathologic findings p.-s. tsai, d.-c. lin, s.-l. shih; taipei/tw the etiologies of nasal or nasopharygeal obstruction are variable in neonates. the respiratory symptoms are varied in these cases. mass lesions in nasal cavity or nasopharynx are extremely rare during the neonatal period. however, we must keep it in mind when respiratory problems occur in the neonatal period. here, we report a case presenting with sleep apnea resulting from nasal obstruction by a rare benign salivary anlage tumor in nasopharynx and discuss the imaging findings as well as clinicopathologic characteristics. the 28-day-old female infant had loud breathing sound, slow feeding and sleep apnea since birth. nasal endoscope and laryngoscope disclosed a polypoid tumor in nasopharyngeal cavity with a stalk connecting with posterior nasal septum. further magnetic resonance imaging (mri) revealed a lobulated mass about 1.6 cm in greatest diameter occupying posterior nasal cavity to the nasopharynx that was intermediate signal intensity on t1-weighted/t2-weighted images and heterogeneous gadolinium enhancement. the patient then received endoscopic resection. the tumor was shown locating in nasopharyngeal cavity and having a stalk from posterior nasal septum, partially occluding the choanae as well. resected tissue fragments displayed tan and whitish in color grossly. microscopic examination demonstrated duct-like structures and mesenchymal elements in a nodular pattern which are typical features of salivary gland anlage tumor. until now, there is no tumor recurrence for four years. unique teaching points: "salivary gland anlage tumor (sgat)" was firstly introduced in a report by dehner et al in 1994 . the tumor that has histologic resemblance to the developing salivary gland, is believed to be a hamartoma originating from minor salivary gland rather than a true neoplasm. congenital sgat displays male predilection and is a rare cause of neonatal airway obstruction. the mass is usual in the midline and attached to the posterior nasal septum or posterior nasopharygeal wall by a delicate pedicle. favorable results with simple excision are obtained. once massrelated airway obstruction is established, further examination with computed tomography (ct) or mri is helpful in anatomic evaluation, size measurement, characteristics definition and intracranial involvement. if mass induced airway obstruction is suspected in a neonate and sgat is considered based on imaging studies, invasive procedure should be careful due to the potential of tumor dislodgement from its fine pedicle resulting in complete airway obstruction. the association of intussusception with malrotation is referred to as waugh syndrome. [1] malrotation occurs in approximately 1 in 500 live births. [2] the incidence of malrotation amongchildren with intussusception is 40%. we hereby present a case report of waugh's syndrome associated with midgut volvulus. case presentation: a 5 month old male child reported to the emergency department with the clinical history of vomiting, abdominal distension, bloody mucoid stools and incessant cry. routine blood examination revealed hb: 12.9 gm%, tlc: 14500/cu mm, plt-2. 9lac/cu. mm. ultrasound (us) examination was performed and it revealed dilated fluid-filled small bowel loops with moderate amount of free fluid, right iliac fossa showed bowel within bowel appearance suggestive of target/pseudo kidney sign of bowel intussusception. no pathologic lead point was identified. transverse ultrasound image through the upper abdomen showed superior mesenteric vein noted to the left of the superior mesenteric artery hence malrotation should be considered. in view of surgical emergency non contrast enhanced ct was done and axial image showed target/sausage shaped soft tissue density mass it had alternating areas of low and high attenuation due to bowel wall and mesentry. on emergency laparotomy patient was found to have intestinal malrotation with duodenojejunal junction on the right of the midline and 180 mid gut volvulus in clockwise direction. intussusception with terminal ileum (gangrenous), caecum, appendix, whole of ascending colon, transverse colon were telescoping into descending and sigmoid colon. the volvulus was derotated and the in tussusceptum was reduced. the gangrenous terminal ileum and appendix was resected and ladd's procedure was done, a diverting ileostomy was created. the patient recovered uneventfully after which an ileo-colonic anastomosis was created transverse ultrasound shows a mass with a swirled appearance of alternating hypoechoic and hyperechoic "bowel-within-bowel" appearance (target sign) unique teaching points: on ultrasonography multiple, concentric, target like appearance of wall layers of invaginated segments (target sign) on axial scan, as well as pseudokidney sign (sandwich sign) on longitudinal scans were accepted as diagnostic criteria for intussusception. [8] it can assess the relative positions of the smv and sma which are mostly abnormal in malrotation. upper gastrointestinal contrast study is the imaging reference standard for diagnosis of malrotation with or without volvulus. abnormal position of the duodeno-jejunal junction. spiral, "corkscrew" or z-shaped course of the distal duodenum and proximal jejunum, and location of the proximal jejunum in the right abdomen. [2] a high degree of clinical suspicion and radiologist's awareness of this entity is helpful in guiding the surgeons towards diagnosis and prevention of morbidity and mortality. a rare case of epidermal naevus syndrome p. joshi; pune/in to acquaint the radiologists with the entity of epidermal nevus syndromes (enss) which are a group of rare complex disorders characterized by the presence of skin lesions known as epidermal nevi associated with additional extra-cutaneous abnormalities, most often affecting the brain, eye and skeletal systems case presentation: this one and a half year old child was referred to us for neuroimaging. he had multiple hairy naevi over his face, limbs including the palms, since birth, associated with blackish discolouration of his entire trunk. unique teaching points: epidermal nevi are overgrowths of structures and tissue of the epidermis, the outermost layer of the skin. the different types of epidermal nevi can vary in size, number, location, distribution and appearance. neurological abnormalities that can be associated with enss can include seizures, cognitive impairment, developmental delays and paralysis of one side of the body (hemiparesis). skeletal abnormalities can include abnormal curvature of the spine, the term "epidermal nevus syndrome" has generated significant controversy and confusion in the medical literature. originally, the term was used to denote a disorder that was actually several different disorders erroneously grouped together. in the recent past, the term was used to denote a specific disorder now known as schimmelpenning syndrome. however, the term epidermal nevus syndrome could be correctly applied to several different disorders. therefore, the umbrella term "epidermal nevus syndromes" now represents a group of distinct disorders that have in common the presence of one of the various types of epidermal nevi. however, there is so far no general agreement how to classify the types of this diverse group of disorders, adding to the confusion within the medical literature. these disorders are quite different from one another and are not "variants" of each other as is sometimes mistakenly stated in the medical literature. in the future, as the genetic molecular basis of these disorders is better understood, the classification may change or expand. bilateral axillary lump in a newborn diagnosed as hematoma h.n. özcan, u. aydingoz, m. haliloglu; ankara/tr objective: most birth traumas are self-limiting and have a favorable outcome. injuries to the infant that result from mechanical forces during the birth process are not uncommon. they occur most commonly on head and neck after vaginal breech delivery. however, soft tissue hematomas can be rarely seen after caesarian section (c/s). herein, we describe imaging findings of a newborn with bilateral axillary lump diagnosed as hematoma. case presentation: a 31-year-old woman was admitted to an outside hospital at 38 weeks' gestation for c/s due to prior caesarean operation. it was her fourth pregnancy (g4 p2). the pregnancy was unremarkable and she had normal ultrasounds at gestation. there was no history of trauma or fall during antenatal period. according to the c/s reports, the process of operation was uneventful any undue prolongation and without having used any other instrumentation. the weight of the female baby was 3.1kg at birth. on the 3 rd postnatal day, her mother noticed a left axillary swelling, then admitted to a tertiary children's hospital. her physical examination revealed, bilateral axillary asymmetry with a fluctuant, nontender swellings. there was no redness or discoloration of the skin. there was no clinical feature suggestive of trauma or bleeding diathesis. a superficial ultrasonography showed solid heterogeneous, hyperechogenic masses 38x35 mm in the left axillary region and 15x13 mm in the right side. doppler study did not reveal any flow in the masses. contrast enhanced mr imaging demonstrated, bilateral axillary mass lesions with fluid levels and smooth contours (figure 1 and 2) . t1w images demonstrated hyperintense component suggesting hemorrhage. after the administration of the gadolinium-based contrast material, lesions showed peripheral enhancement (figure 3) . a diagnosis of hematoma was entertained. the child was managed non-operatively. she was monitored clinically and radiologically. follow-up ultrasounds scan revealed significant regression of the swellings. unique teaching points: soft tissue hematomas can be rarely seen in newborns. the formation of axillary hematoma on the background of c/s is a rare complication, which, to the best of our knowledge, has not been previously reported. ultrasonography and mr imaging readily depicts hematoma and aids in the differential diagnosis. colorectal carcinoma (crc) is extremely rare in pediatric age, with an estimated annual incidence of approximately 1 case per million individuals. the majority of reported cases occur in adolescence, while the incidence is further lower for children under 10 years. the distribution between males and females is not equal, with higher prevalence in males (ratio of 2:1). the etiology in children is unclear as these tumors are often sporadic and not linked to a preexisting adenomatous polyp, unlike adults. predisposing factors such as familial polyposis of the colon, other polyposis syndromes, ulcerative colitis and familial multiple cancer syndromes were reported in 10 % of cases. advanced stage at diagnosis, aggressive histologic subtypes (poorly differentiated, signet ring and mucinous adenocarcinoma) and poor survival are the hallmarks of pediatric crc. case presentation: a 15-year-old male presented with a history of dyspeptic symptoms (recurrent epigastric-right flank colic pain and heartburn) for the last eight months, without evidence of irregular bowel function. after a prior diagnosis of esophagitis secondary to a gastroesophageal reflux disease, physical and laboratory examinations revealed anorexia, progressive body weight loss, microcytic iron deficiency anemia and positive fecal occult blood test. during an emergency access, abdominal ultrasound identified rounded target liver lesions and circumferential heterogeneous mural thickening of the ascending colon. contrast-enhanced computed tomography scan (cect) demonstrated a marked circumferential wall thickening of the ascending colon and cecum with a longitudinal extension of 85 mm and thickness of 22 mm; the mass contained lowdensity areas and calcifications. furthermore 5 hypovascular hepatic lesions along with lymph node metastases containing calcifications were identified. no lung metastases were found. histopathological analysis confirmed the diagnosis of metastatic colon adenocarcinoma. after chemo-and radio-therapy, only the hepatic lesions showed reduction in size and number. the patient subsequently underwent right hemicolectomy. one month after surgery he is in a rigorous follow-up through ultrasonographic evaluation of pleural effusion and ascites and cect. unique teaching points: crc, although rare, should be suspected in children presenting with unexplained persistent abdominal pain, progressive body weight loss and positive fecal occult blood test. ultrasound imaging can be appropriate in the preliminary detection of abnormal bowel wall thickening, lymph node and liver metastases; cect is mandatory to confirm the radiological diagnosis and complete the staging. to increase awareness of this rare syndrome and its varied presentation in order to facilitate its early diagnosis and treatment to prevent poor prognostic outcomes. case presentation: lemierre syndrome is a rare disease characterized by an initial infection of the head and neck leading to the development of a septic thrombophlebitis which has a propensity to spread and involve the jugular and facial veins. this progressive infection then leads to the development of metastatic septic emboli to the respiratory tract. we present the case of a 7 year old boy who attended with a 1 week history of fever and a cough. initial imaging on admission demonstrated a large left sided hydropneumothorax with multiple cavitating lesions throughout the lung parenchyma in addition to thrombosis of some of the segmental pulmonary veins. the hydropneumothorax was surgically drained and the patient was transferred to the paediatric intensive care unit after further deterioration with the development of a broncho-pleural fistula. following a short course of antibiotics there was no clinical or radiological improvement and sputum cultures grew coliform organisms which raised suspicion for a more distant source. when pus was noted to be discharging from the left ear, a contrast enhanced ct of the head and neck revealed a left mastoiditis with multiple cerebral abscesses and occlusive thrombi in the left jugular vein, transverse venous sinus, sagittal and straight sinuses. following this diagnosis antibiotic therapy was modified and targeted at anaerobes, which was vital in assisting the patients recovery and successful discharge home. unique teaching points: clasically the majority of lemierres syndrome begins in the oropharynxinvolving the palantine tonsils and peritonsillar tissue often presenting with fever, sore throat and neck pain. our case demonstrates an atypical presentation with sepsis and respiratory symptoms as a result of the septic emboli which delayed diagnosis. we have learnt from this case the importance of considering lemierres syndrome in patients presenting with signs of a respiratory infectionin particular cavitating pulmonary lesions-that have not improved with conventional therapy and to have a low threshold to investigate the head and neck as a potential source of infection. when the working hypothesis of meningitis could not help e. kovacs 1 , n. pinter 2 , g. balázs 1 , a. machovitsch 1 , a. arany 1 , z. liptai 1 , l. fonyad 1 , p. benke 1 ; 1 budapest/hu, 2 amherst/us objective: neuroinfection still represents a diagnostic challenge in the everyday practice, where clinical evaluation, imaging, laboratory and pathological workup and treatment goes hand in hand under the pressure of time. we summarized a case in which, despite the extensive multilateral collaboration the battle was lost, to bring attention to the possible causes. a two year old, previously healthy female was taken to the emergency department for altered state of consciousness and fever. she also suffered from gingivitis. the unconscious child underwent an emergency ct scan: hydrocephalus with signs of raised intraventricular pressure was detected. subsequently mri of the head and spine was performed, and showed signs of diffuse leptomeningeal enhancement with basal predominance. multiple dwi restricted parenchymal lesions with basal predominance were also found. repeated csf and blood tests did not reveal any causative organism, although the gradually increasing crp suggested infection. two weeks after the onset of symptoms a follow up mri study showed extensive cerebral and spinal swelling with no focal lesion. the child passed away three days later due to cardiac failure. autopsy and neuropathological evaluation could not reveal the cause of the disease, which was identified only weeks after the child died, by culturing sputum and csf. unique teaching points: an overview of the clinical and radiological presentation of meningitis basilaris is carried out. attention is given to the circumstances, when tuberculotic infection should be suspected, and antituberculotic treatment should be started, even before the confirmation of the presence of mycobacteria can be obtained. to describe the clinical, laboratory and mri findings of chronic nonbacterial osteomyelitis(cno) in a patient with a negative radiograph and emphasize useful imaging findings, including an unusual radial pattern of edema in both femoral heads. case presentation: a 14-year-old adolescent, was referred with progressive debilitating hip pain and inability to walk since 5 days, that was unsuccessfully treated with non-steroidal anti-inflammatory drugs. during hospitalization he developed fever up to 38 ο with normal full blood count and smear, elevated esr (95mm/h) and crp (12.8 mg/dl), positive serologic markers for streptococcus (asto) and ebv and received antibiotics with relative good response. blood cultures did not grow any pathogens, the rest of serology was negative for acute infection, tuberculin skin test was negative and immunological investigation was unremarkable. pelvic radiographs were negative. mri showed a symmetric pattern of bone marrow involvement around both triradiate cartillages, at both femoral heads and (2017) 47 (suppl 2):s297-s pediatr radiol major trochanters. complementary evaluation of tibial areas with a limited protocol disclosed asymptomatic involvement of tibial epiphyses and apophyses. a radial pattern of edema was seen at the femoral heads with alternating stripes of involved and uninvolved areas. clinical course and imaging appearances were highly suggestive of cno. rapid clinical improvement occurred during hospitalization while a repeat mri 6 months later showed complete resolution of hip findings and the patient was free of any symptoms or signs. coronal stir sequence at presentation showing the radial pattern of bone marrow edema (arrowheads) alternating with stripes of normal marrow (*) at both femoral epiphyses. note hyperinense edema (arrows) around triradial cartillages. coronal stir sequence showing the predilection of bone marrow edema symmetrically around triradial cartillages (arrows) and at major and minor trochanters (arrowheads). coronal stir sequence at 6-months follow-up shows resolution of edema. unique teaching points: cno is a not well known chronic autoinflammatory bone disorder affecting primarily children and adolescents. positive serology for streptococcus or other infectious agents has been previously reported as in our case and may be a triggering factor. striking mri findings with a negative radiograph may occur at initial stages. symmetrical distribution of non-specific bone marrow edema around epiphyses and apophyses is highly suggestive of the diagnosis in the appropriate clinical setting and following exclusion of suppurative bone infections as well as bone or hematologic malignancies. the radial pattern of edema in our patient is unusual and considered to comply with the direction of main trabecular systems in femoral heads. in 26/31 chest cts, 103 nodules (median size 2.0 mm) were detected. display mode a with 5 mm mip yielded the best interreader variability (κ=0.294) and the highest sensitivity (73.9%) compared to mode b and c (κ=0.218, sensitivity 61.2% and κ=0.238, sensitivity 59.3%, respectively). perifissural nodules were detected in all subgroups. conclusion: mip improves the detection of pulmonary nodules in chest cts of young children, but overall interreader agreement is only fair. nodules, including perifissural nodules, occur in children with and without malignancy. images were subsequently read and interpreted by board-certified radiologists and nuclear medicine physicians in communal reading. in case of identifying suspicious lesions in cect additional imaging (mri) or biopsy was performed. compared to pet/ct employing only low dose ct (ldct), the use of cect resulted in the identification of 19 additional suspicious lesions in 13 patients. furthermore the use of cect allowed us to qualify 3 lesions as benign/ physiologic which in pet/ldct were identified as suspicious and 16 lesions suspect for metastases or tumor. in those 37 patients who received combined integrated fdg pet/ct including both ldct and cect the ctdi ranged in between 1,3-15.5 mgy (n= 10.2 mgy) and the dose length product (dlp) ranged in between 82.4-2464 mgy *cm (n= 812.6 mgy *cm) specificity was significantly higher combining pet and ct compared to stand-alone ct and pet. our study showed that the acquisition of cect in combined integrated pet/ct leads to an increased specificity and thus represents an essential component of a good fdg pet/ct in pediatric oncology. in assessment of lymph nodes, inflammatory foci and liver lesions diagnostic contrast enhanced ct is essential. comparison of the detectability of ubos in neurofibromatosis type i patients with proton density-weighted and flair sequences in 3t mri l. porto, s. lescher, n. hillenbrand; frankfurt/de objective: neurofibromatosis type 1 (nf1) is an autosomal-dominant congenital disease. in nf1 patients, significant numbers of so-called unidentified bright objects (ubos) can be found in brain imaging, with predilection sites at the basal ganglia and the dentate nucleus. ubos seem to develop at a very early age, contrary to other criteria leading to diagnosis. the detection of ubos might therefore prove helpful in the early diagnosis of nf1, complementing the clinical diagnosis based on criteria of the "national institutes of health consensus development conference". the aim of the study was to investigate whether the detectability of ubos increases at 3t by comparing proton density-weighted images (pdw) with fluid-attenuated inversion recovery (flair) sequences. a total of 14 nf1 patients (7 male, 7 female, between 8 and 26 years old, mean age 15.4 years) were examined by a 3t magnetic resonance scanner. the presence of ubos was evaluated on pd-w and flair images by 4 evaluators (2 experienced neuroradiologists, 1 junior radiologist and 1 student in his final year). detectability was rated by a three-point scoring system for dedicated regions: lesions which were "well defined/detectable", "suspicious" or "detected after a second look". the wilcoxon signed-rank test was used for comparisons between the raters. the level of significance was p<0.05. significantly more lesions were marked as "well defined/detectable" in the pd-w sequence compared to flair (p<0,001 for all four evaluators together, as well as for each evaluator separately). in particular, pd-w proved to be superior for detecting ubos located in the medulla oblongata (p=0,001) dentate nucleus (p=0,002) and hippocampal region (p=0,007), regardless of the level of the raters' experience. this is the first study that compares flair and pd-w at t3 for the diagnosis of ubos in nf1. significantly more ubos are detected in the pd-w compared to flair sequences, especially for the infratentorial regions. as ubos occur at very early stages of the disease in patients with suspected nf1, pd-w might aid an early diagnosis in these patients. assessment of radiation doses from diagnostic imaging in the followup of paediatric oncology patients p. logan 1 , r. harbron 2 , k. mchugh 1 ; 1 london/uk, 2 newcastle-upon-tyne/uk objective: previous literature (1,2) has suggested paediatric oncology patients accumulate a large radiation burden as a consequence of routine diagnostic imaging examinations during therapy. we retrospectively looked at the effective doses from routine ct and nuclear medicine in three cohorts of children, namely patients with hepatoblastoma, wilms' tumours and rhabdomyosarcoma (rms). of note, in our centre we rely on repeated mris of the primary site for many tumours. effective doses (e), in millisieverts (msv), were estimated using the ncict dose estimation tool (lee et al 2015) , based on details specific to each procedure: patient age, scan region, scanner type and ct dose index (ctdi -an indicator of radiation exposure recorded at the time of each scan). doses for general radiography were estimated using pcxmc v2.0 monte carlo simulations, assuming standard exposure factors and field size. there were 59 patients in total (18 hepatoblastoma, 21 wilms', 20 rms). there were 33 boys. the mean age was 3 years 2 months (ranging from 14 days -11 years 10 months). the mean and median cumulative effective doses from ct for the whole cohort were 7.88 msv and 4.24 msv respectively. four patients in the wilms' cohort had a dmsa nuclear scintigram (0.7 -1.0 msv), no hepatoblastoma patient had any nuclear medicine imaging, and 16 patients with rms received a bone scan (3 -3.5 msv) or a pet scan (approximately 8msv). cumulative radiation doses from routine radiological investigations in paediatric oncology can be kept in a much lower range than reported in the literature (1,2). in our institution, the followup of solid intra-abdominal tumours with mri, with additional ct or nuclear medicine only when clinically justified, has resulted in a significantly low radiation exposure in these patient cohorts. mri of the primary tumour site should be implemented as a replacement for ct imaging when there is no significant detriment to the diagnostic information obtained. (2017) 47 (suppl 2):s297-s pediatr radiol mri-based evaluation of multiorgan iron overload is a predictor of adverse outcomes in pediatric patients undergoing allogeneic hematopoietic stem cell transplantation f. zennaro 1 , d. zanon 2 , r. simeone 2 , g. boz 2 , f. degrassi 2 , m. gregori 2 , g. schillani 2 , c. boyer 1 , n. maximova 2 ; 1 nice/fr, 2 trieste/it objective: iron overload is associated with poor clinical outcomes in patients undergoing allogeneic hematopoietic stem cell transplantation (hsct). although the effects of hepatic and cardiac siderosis on patient outcomes have been extensively studied, less is known about the effects of siderosis in other organs. the medical records of 44 consecutive pediatric patients who underwent allogeneic hsct in our institute from 2011 to 2015 were retrospectively reviewed. mri was used to measure iron concentrations in the liver, spleen, pancreas and bone. these patients were divided into two groups, 18 with non-elevated (<100 μmol/g; group 1) and 26 with elevated (>100 μmol/g; group 2) liver iron concentration (lic) at baseline. in group 1, only two patients had normal iron concentrations in all organs. none of the patients of group 2 presented with pathological iron concentrations in only two organs. comparisons of baseline data with results of the first follow-up mri performed 1-6 months after hsct, showed a general worsening of iron accumulation. in group 1, none of the patients showed complete absence of iron overload in a single organ. in group 2, none of the patients showed a total absence of siderosis involving fewer than three organs. this study confirms the correlations between iron overload and the risks of transplant-related complications, such as transplant related mortality, sinusoidal obstruction syndrome, infections, pancreatic insufficiency, and metabolic syndrome, in transplant recipients with systemic siderosis. another important finding of this study was the close correlations between pre-transplant bic and times to neutrophil and platelet engraftment (p<0.001 each). (15), ganglioneuromas (gn, 2) and ganglioneuroblastoma (gnb, 1), examined by 3t mri were retrospectively grouped according to tumor entity, risk factors (bone marrow metastasis, mycn amplification or 1p36 deletion) and therapeutic regime (observation versus chemotherapy). dw (b values 0, 400 and 800) and conventional mri images (t2, t1 pre and post contrast) were analyzed for tumor size, relative si-and absolute adc-values at baseline (base; no therapy), and after 3 (fu1) and 12 (fu2) months. adc values in nb were lower than in gnb and gn (0.75*10 -3 mm 2 /s versus 1.28*10 -3 mm 2 /s; p<0.05). there was a tendency towards lower adc values in tumors with risk factors (n=6) versus no risk factors (n=7) at baseline, which did not reach statistical significance (p=0.08). during follow-up shrinkage of tumor volume was noted (baseline 1206 ml, fu1 159 ml, fu2 51 ml; p<0.05 baseline vs. fu1; p=0.08 baseline vs. fu2). in the observation group, tumor adc values rose without relapse (0.89*10 -3 to 1.07 *10 -3 mm 2 /s). only in eventually relapsing tumors adc values tended to decrease further (0.95*10 -3 to 0.71*10 -3 mm 2 /s, p=0.17), despite initial reduction in tumor size. to establish inter and intra-observer variability in the radiological detection and assessment of pulmonary nodules at diagnosis in children with wilms tumours. a test set of ct thoraxes at diagnosis from 15 patients enrolled in the multicentre 'improving population outcomes of renal tumours of childhood' (import) study in the uk were assessed. five radiologists (3 chest, 2 paediatric) from 5 different centres (4 uk, 1 netherlands) completed a scoring sheet for nodule assessment on the same studies on two occasions, 6 months apart. the readers were blinded to patient respiratory symptoms, the original radiology reports and also that they were scoring identical cases. descriptive statistics, modified bland altman graph and fleiss kappa scores were used for statistical assessment. in total, 93 different pulmonary nodules across the 15 ct thoraces at both rounds were scored by at least one reader. 81 (87%) were seen by at least one reader in round 1 and 81 (87%) in round 2, 69 (85.2%) nodules were seen by at least one reader in both rounds. only 20 (21%) nodules were scored by all 5 readers in round 1, 16 (17%) by all 5 readers in round 2, and 14 (7%) nodules by all 5 readers in both rounds. of the 20 nodules seen in the first round, 11 were measured to be >5mm in at least one dimension and of these, 7 were classified as malignant by all 5 readers. the limits of agreement for mean difference in nodule size in anterior-posterior, transverse and longitudinal measurements were ±1.85mm, ±1.76mm and ±1.88mm respectively. the fleiss kappa scores ranked from poor to fair agreement for nodule border smoothness (0.03), nodule shape (0.2), solidity (0.2) and impression of malignancy (0.4). within the same readers for both rounds, nodule detection rates of agreement were between 60.7-82.9%. the average intra-reader percentage of observed agreements for nodule border smoothness, shape, solidity and impression of malignancy were 84.2%,63.9%, 90.6% and 84.8% respectively. conclusion: detection and characterisation of pulmonary nodules on ct thorax shows both intra-and inter-observer variability. this has important implications for the interpretation of metastatic disease at presentation. fever without a focus is defined as febrile illness without an initial obvious cause or localizing signs. our aim is to assess the diagnostic value of whole-body mri (wb-mri) in the diagnostic work-up of children with fever without a focus. we retrospectively searched for subjects who underwent wb-mri for fever without a focus. a total of 29 children (m=17, f=12), mean age 6.1 years (range: 0.08 -17.61) were included. 8/29 (27.6%) subjects were immunosuppressed and 6/29 (20.7%) subjects were hospitalized at onset of fever. the reference standard was based on positive cultures, biopsy or surgery. when this was not possible, a probable diagnosis was made based on clinical follow-up or serology. initially, the wb-mri images were reviewed independently by 2 pediatric radiologists blinded to all clinical information. at the end of each case the final diagnosis and the diagnostic category (5 categories: a. normal, b. infection, c. oncologic, d. rheumatologic, e. miscellaneous) was recorded. this was followed by a consensus read for comparison with the reference standard. for statistical analyses all subjects were treated as fever without a focus. results: reference standard: the diagnostic category of the reference standard was as follows: infectious 12/29 (44.8%), oncologic 1/29 (3.4%), rheumatologic 8/29 (27.6%), miscellaneous. 1/29 (3.4%). even after extensive work-up in 7/29 (20.7%) no clear cause for the fever was found table 1 . wb-mri: wb-mri diagnosed the cause of fever without a focus in 7/29 subjects (24.1%) ( table 1 ). in 2 subjects (6.8%) wb-mri results were falsely positive (1 jia and 1 myositis), and in the remaining 20 subjects no imaging findings compatible a cause of febrile disease were found. interobserver agreement was fair (kappa 0.53). in children with fever without a focus wb-mri provided the diagnosis in in almost a quarter of the cases. given the multiplicity of causes of fever without a focus, some of them not possible to visualize on mr imaging, wb-mri may be considered in routine imaging practice when evaluating pediatric patients with fever without a focus. to compare linear measurement/volume to direct volumetric measurements using 3 dimensional(3d) post-processing software. for this irb approved study initial diagnostic ct or mr exams in 100 patients(11mo-20yr) with solid tumors were reviewed by 11 radiologists and 3 technologists. radiologists recorded measurements in 3 axes in their routine method, described tumor shape (sphere, ellipse, cone) and surface texture (smooth, almost smooth, or mildly, moderately, markedly irregular). three technologists individually, and 3 radiologists by consensus, used 3d processing software (intellispace portal, philips, cleveland, oh) to directly measure tumor volume. tumor volume (v) was calculated from linear measuments using the following equations: sphere v=4/3πr 3 , ellipsoid v=8πr 2 or 16πr 2 , conicalv=2πr 2 or 4πr 2 , and cuboid v=(xyz). inter-reader variability in tumor measurement in all tumors and for tumors divided by surface characteristics was assessed amongst radiologists and technologists, and radiologist consensus using coefficient of variation (cov). tumor shape analysis was reported as 14 sphere, 84 ellipse, 2 cone, and surface texture 8 smooth, 10 almost smooth, 32 mildly irregular, 31 moderately irregular, 19 markedly irregular. inter-reader variability of as much as 1,119cc above to 383cc below the mean tumor volume was found when using radiologist determined linear measurements, with standard deviation (sd), range 0.65-413. inter-reader variability amongst technologist derived volumes was considerably less, range 102cc above to 90cc below the mean, with sd, range 0.03-97. cov analysis shows a greater degree of variation in tumor volume calculated from linear measurements [smooth(7%), almost smooth(12%), mildly(20%), moderately(18%), markedly(27%) irregular] than direct volume determination [smooth(5%), almost smooth(6%), mildly(3%), moderately(3%), markedly(5%) irregular]. variation was significant only for tumor with irregular surface texture [smooth (p=0.26), almost smooth (p=0.23), mildly (p=0.003), moderately (p=0.001), or markedly (p=0.002) irregular]. variation in linear/volume measurements in very irregular tumors. light blue=middle 50% tumor volume measurements by pediatric radiologists. whiskers mark limits of range. ▲♦ • markers =measurements by technologists. note broad degree of variation. (2017) 47 (suppl 2):s297-s pediatr radiol variation in linear/volume measurements in almost smooth tumors. light blue=middle 50% tumor volume measurements by pediatric radiologists. whiskers mark limits of range. ▲♦ • markers =measurements by technologists. note narrow degree of variation. both graphs show the same informationthe % relative variation in tumor volume measurements determined by 3 dimensional linear measurements (11 pediatric radiologists) v. volumetric processing (technologists & consensus group). radiologist generated measurements are subjective and unreliable. variation in measurement technique leads to differences in calculated tumor volume which significantly over or under estimate volume in tumors with irregular texture and is not significant in smooth tumors. quail-quantitative mri-based evaluation of pancreatic iron overload in pediatric patients undergoing allogeneic hematopoietic stem cell transplantation f. zennaro 1 , m. gregori 1 , f. degrassi 1 , e. cattaruzzi 2 , y. diascorn 3 , c. boyer 3 , n. maximova 1 ; 1 trieste/it, 2 muggia/it, 3 nice/fr objective: iron overload (io) is a relatively common but often neglected transplantrelated complication and has been associated with poor prognosis in patients undergoing allogeneic hsct for hemato-oncological disease. pancreatic io is frequent among patients with transfusion-dependent anemias, but is uncommon among patients with hematologic malignancies. the causes of pancreatic io and the potential effects of pancreas iron deposits on transplant outcomes or on the risk of developing significant late effects in long-term hsct survivors have not been yet determined. our institute routinely uses magnetic resonance imaging (mri) with various gradient-recalled-echo (gre) sequences to quantitatively measure the iron concentration in abdominal parenchymal organs in all pediatric patients before and after allogeneic hsct. this study retrospectively analyzes the correlations of pancreas io with the type of conditioning regimen and pretransplant liver iron concentration (lic) in 50 pediatric patients who underwent allogeneic hsct in our transplant unit over the last 5 years. we enrolled 50 patients, age 1-17 years. pre-transplant mean lic was 147,55 μmol/g (normal values 36 μmol/g). 3 (6%) patients had mild liver io and 25 (50%) had moderate or severe io. pretransplant mean pancreatic iron concentration (pic) was 29,59 μmol/g, whose only 4 (8%) had mild pancreatic io and none had severe io. post-transplant mean lic was 162,6 μmol/g, only one patient had mild liver io but 36 patients (72%) had moderate or severe io. post-transplant mean pic was 66,22 μmol/g, 12 (22%) patients had moderate or severe io. mean pre-transplant pancreatic volume was 33,41 cm 3 , while mean post-transplant pancreatic volume (evaluated 30 days after transplantation) was 25,41 cm 3 . 11 (91,7%, p<0,001) patients with post-transplant moderate or severe pancreatic io underwent tbi-based conditioning. mean reduction of pancreas volume in tbi group was 12,02 cm 3 (p< 0,001). no pancreatic volume reduction was observed in chemotherapy-based group. all patients with pancreatic io have had exocrine pancreatic insufficiency and 9 (83,3%) patients have had metabolic syndrome. volume reduction well correlate (mean 49,8%, p<0,0001) with pancreatic io. this study confirms that pancreatic iron overload is not so rare in patients with hematologic malignancy underwent allogeneic hsct, with increased risk of metabolic syndrome and total deficit of exocrine pancreatic activity, but not of endocrine activity. iron overload monitoring allows for chelation therapy optimization. mr is fast, reproducible and more reliable compared to serum ferritin and transfusional history and allows a multi organ evaluation. pulmonary tb is common in south africa, with many children affected. diagnosis can be challenging and chest x-ray remains fundamental for diagnosis. interpretation is difficult and shown to have wide inter-reader variability. no study however has compared cxr findings and interreader agreement between ambulatory and hospitalised patients. this study compares the frequency of cxr changes, as well as interreader agreement in ambulatory compared to hospitalised children with suspected tb. from nolungile clinic and red cross children's hospital respectively was done. each sample contained 50% proven tb and 50% negative controls. two paediatric radiologists and one paediatrician served as blinded, independent readers for the database using standardised ticksheets. our study demonstrated no significant difference in lymphadenopathy, but an increase in parenchymal change in the hospitalised group. we otherwise showed similar results to literature regarding finding frequency, but poor inter-observer agreement. if the least expert reader were removed, results were comparable with available literature. this highlights the need for development and study of explicit cxr criteria for lymphadenopathy to improve the value of cxr for paediatric tb in all settings. lung ultrasound in pediatric pneumonia -why is it necessary to use the additional trans-abdominal approach? j. lovrenski; novi sad/rs objective: to emphasize the need of lung ultrasound (lus) technique modification, which enables detection of pneumonia in children not visualized by using solely the standard trans-thoracic approach. a prospective study was carried out in the regional children's hospital, and comprised a 2-year period. the inclusion criterion was us finding of pneumonia detected by trans-abdominal, and not with trans-thoracic approach. lus examinations were performed using a combined, trans-abdominal and trans-thoracic approach. longitudinal, transversal (intercostal), and oblique sections were used. trans-abdominal examination included transhepatic and trans-splenic approach. the ultrasound probe was angulated from the most anterior to the most posterior sections while examining the lung bases by trans-abdominal approach. a pneumonia-positive lus finding included subpleural consolidation with air-bronchogram, or with an adjacent area of interstitial/ alveolar-interstitial edema. lus was always performed before the other diagnostic modalities (chest x-ray (cxr) and computed tomography (ct)), if they were indicated by pediatrician or radiologist. within a 2-year period in 14 children (mean age 3.9y, sd 2.3y) the pneumonic focus was discovered using the trans-abdominal approach, while the trans-thoracic approach showed a normal lus pattern. all the children had the clinical symptoms of pneumonia (fever and cough, with or without dyspnea/tachypnea). the auscultatory finding was positive in 4 children. cxr was performed in three children, showed a right-sided pneumonia in two, and was negative in one patient. one child had a contrastenhanced chest ct, which confirmed a left-sided pulmonary base abscess detected during lus examination by trans-splenic approach only (figures 1, 2) . apart from pulmonary symptoms, there has not been any other associated diseases found, apart from otitis media in two children. each child responded to the antibiotics treatment with resolution of infection and us signs of pneumonia. in this oral presentation we will explain and give anatomical and technical reasons for pneumonia-positive us findings within lung bases, that remained undetected by the trans-thoracic approach. left-sided abscess abutted on the spleen (s), and was detected by trans-splenic us approach. it did not contact the pleural surface approachable by trans-thoracic ultrasound (black semi-lunar mark). l-liver. conclusion: trans-abdominal (trans-hepatic and trans-splenic) approach should become an inseparable part of each lus examination, along with a standard trans-thoracic approach. this modification of technique is expected to result in a further increase of lus sensitivity in diagnosing pneumonia. is thoracic ultrasound really competitive to computed tomography in children -a two-year retrospective study j. lovrenski, k. antolović; novi sad/rs to compare diagnostic accuracy of thoracic ultrasound (us) and computed tomography (ct) in children. a retrospective study was conducted in the regional children's hospital, and comprised a 2-year period. the inclusion criteria were: chest ct performed within 24h after the us examination of thorax, and us and ct examinations in the same patient performed by different pediatric radiologists. all us examinations were performed using a combined transabdominal-transthoracic approach. ct examinations were done (2017) 47 (suppl 2):s297-s pediatr radiol according to the body mass based pediatric ct protocols. each hemithorax was analyzed separately in terms of comparison between us and ct findings. statistical analysis included the calculation of sensitivity, specificity, positive predictive value (ppv) and negative predictive value (npv) of ultrasound in diagnosis of pulmonary pathological entities. out of 176 children with chest ct, 21 of them (mean age 8,2y, sd 5,6y) fulfilled the criteria to enter the study group. lung us showed sensitivity, specificity, ppv and npv in diagnosis of pleural effusion: 100%, 96,4%, 91%, 100%; lung consolidation: 100%, 100%, 100%, 100%; lung abscess: 75%, 100%, 100%, 97%; and interstitial lung disease: 57%, 100%, 100%, 91%, respectively. within 3 hemithoraces multiseptation of pleural effusion was observed by us only. air bronchogram within lung consolidation was observed in 5 hemithoraces both by us and ct examinations. necrotic areas within pulmonary consolidations were detected by us in 4 hemithoraces, which was later confirmed by ct examination. lung abscesses were diagnosed in 3 hemithoraces by both us and ct. two small lung abscesses filled with air (1 hemithorax) and bronchiectasis (2 hemithoraces) were detected only by ct examinations. other pathological findings detected both by us and ct examinations were: congenital pulmonary airway malformation (cpam) (1 hemithorax), pulmonary sequestration (1 hemithorax), partial pneumothorax (2 hemithoraces), hidropneumotorax (1 hemithorax), inflamed pneumatocele (1 hemithorax), hydatid cyst (1 hemithorax), pericardial effusion (2 patients), soft tissue masses of thoracic wall with initial bone destruction (2 patients), and lymphomas (2 patients) (figures 1-3) . in one patient us and ct revealed cysts and an extremely dilated bronchus within lung consolidation (pathohistological finding: cpam type 2 combined with subsegmental bronchial atresia, and extensive bronchopneumonia). us examination, unlike ct, could not differentiate between eventration of the left hemidiaphragm and diaphragmatic hernia in one patient. to determine and compare the accuracy of frontal cxrs alone and 'combination frontal-lateral' set of cxrs for diagnosing lymphadenopathy in children with tb using patients with confirmed tb and controls without tb, and to compare findings in hiv-infected and hiv-uninfected children. a total of 172 children (ie: 88 children with gene xpert confirmed tb and 84 control patients admitted with lower respiratory tract infections), which were part of a larger south african study, who had both frontal and lateral cxrs, were included. three qualified radiologists read the cxrs in 2 separate sittings one month apart (one for the frontal x-ray alone and one for the 'combination frontal-lateral' cxrs) for the presence of lymphadenopathy. odds ratios and 95% confidence intervals were calculated to determine the presence of lymphadenopathy using a consensus reading on the frontal cxr and frontal-lateral cxr combination according to the final diagnosis of tb. inter reader agreement was determined using the kappa statistic. lymphadenopathy was reported in 86 (50%) patients on the frontal cxr alone and in 143 (83%) patients on the frontal-lateral cxr combination. 52 (60%) of the 86 patients with lymphadenopathy on the frontal cxr alone were gene xpert positive versus 72 (50%) of the 143 patients with lymphadenopathy on the frontal-lateral cxr combination. in all patients, the consensus reading using a frontal-lateral cxr combination resulted in a 5-fold increase (or 4,9; 95% ci 2, 4) in calling lymphadenopathy compared to using a frontal cxr only in the gene xpert positive group, the consensus reading using a frontal and lateral cxr combination resulted in a 3 fold increase (or 3,1; 95% ci 1,5-6,6) in calling lymphadenopathy compared to a frontal cxr only. overall inter reader agreement for all 3 readers when evaluating for lymphadenopathy was 'fair' on both the frontal cxr (k= 0,2088) and the frontal-lateral cxr combination (k=0,273). the addition of a lateral view to the standard frontal cxr increased the rate of calling lymphadenopathy. however, the accuracy of diagnosing lymphadenopathy on chest x-ray as a marker for tb was poor. this poor accuracy was further hampered by only 'fair' inter reader agreement for the presence of lymphadenopathy on chest x-ray. dynamic 4d ct imaging in children has significant advantages over routine ct scanning, bronchography and bronchoscopy for diagnosing trachebronchomalacia because it can be performed during free breathing without anaesthesia or invasive airwayaccess.itcanalsodemonstratevascularcausesoftracheo-bronchomalaciain the same sitting. the technique is currently performed in 1 paediatric center in the uk. we aimed to report pitfalls encountered while setting up a dynamic 4d ct imaging service for children and report the findings of studies performed. materials: dynamic 4d ctscanning was introduced after installation of a large array (320 slice) ct scanner, applications specialist training and review of the literature. imaging parameters in use by greenberg and colleagues (arkansas children's hospital, usa) were applied. referral indications, pitfalls encountered, quality of scanning and imaging findings/diagnoses were reviewed and enumerated. results: nineteen paediatric dynamic 4d ct scans (9 females, 10 males; 11 days -3 years 9 months; mean 12 months) were performed over 15 months. the first 4 studies were performed without ivi contrast due to lack of experience and 13 subsequent studies were performed with contrast ( figure 1 major pitfalls included initial failure to perform contrasted studies for simultaneous evaluation of vessels, initial failure to withdraw the endotracheal tube, patient motion under care of nurses and clinicians, failure to appreciate the value of imaging the full lung volume while trying to keep dose length product to a minimum and failure to appreciate that collapse of the airway is often in the ap plane and not appreciated on coronal slab projections -rotating 3d volume rendered images is a requirement ( figure 3 ). additional obstacles were initial clinician and radiologist lack of support after early failures and colleague concerns regarding the radiation dose. objective: diagnosis of pulmonary tuberculosis (ptb) in children relies on chest radiography, however there is wide inter-observer agreement in detecting lymphadenopathy, the hallmark of ptb. paediatric airways are pliable, thus detection of airway compression may be a more objective criterion for the presence of lymphadenopathy. thus the objctive was to assess the usefulness of airway compression on chest radiographs for diagnosis of ptb in children. chest radiographs of children admitted to red cross children's hospital with suspected ptb were read by two readers according to a standardised format and a 3 rd when there was disagreement. radiographs of children with definite ptb were compared to those with lower respiratory tract infection (lrti) from another cause. the prevalence and location of airway compression was evaluated. findings were correlated with hiv status and age. inter-observer agreement was assessed using kappa statistic. 13 .9% in older children (or 1.7; 95%ci 1.00-3.00). no association with airway compression and hiv infection was found. inter-observer agreement ranged from 0.0-0.4. eighteen-month-old male patient diagnosed with ptb; hiv negative. majority agreement of airway compression at lmb indicative of lymphadenopathy. left upper lobe opacity is in keeping with a ghon focus. there is a strong association between airway compression on chest radiographs and confirmed ptb, particularly in infants, irrespective of hiv status. however, clinical use is limited by poor inter-observer agreement. paediatric ultrasound-guided biopsies in a tertiary oncology centre: five years experience n. parvizi, m. smedley, s. chakraborty; oxford/uk objective: histological diagnosis is almost always essential to guide appropriate therapy for children diagnosed with cancer. tissue can either be obtained by surgical/open or image-guided percutaneous biopsy. the aim of this study is to assess the safety and diagnostic accuracy of ultrasound-guided biopsies in a tertiary oncology referral centre. a retrospective analysis of clinical data, imaging findings and histological diagnosis of patients aged 0 to 18 years between january 2010 and december 2015 was carried out. a total of 113 ultrasound-guided biopsies were performed in our institution on 110 patients. most of the biopsies were performed in theatre with the patient under general anesthetic and with an 18-gauge spring-loaded core biopsy needle with a minimum of two cores per patient. in 99% of lesions the needle biopsy was diagnostic. the single nondiagnostic case did not have sufficient material to make a full diagnosis and a surgical biopsy was required. eighty-two of the biopsied lesions were malignant and 31 were benign. in no cases was a repeat biopsy required. the vast majority of the biopsies were performed within one week of request with over half performed within 3 days. all biopsies were performed without complication and in the majority of cases the patients were discharged the same day or following an overnight stay. ultrasound-guided percutaneous biopsy is an accurate and safe technique in order to acquire tissue from suspected malignant lesions in children. these can be performed instead of or in addition to open biopsy and will often result in a shorter hospital admission and recovery time. the role of imaging in the diagnosis of thymoma in paediatric patients with myasthenia gravis j. adu, t. a. watson; london/uk thymomas are exceedingly rare tumours in the paediatric age group, with only very few cases having been reported in the literature. thymomas are commonly associated with myasthenia gravis (mg), with thymectomy being potentially curative. ct is the mainstay imaging modality for thymoma diagnosis in the adult population. while, chemical shift mr imaging can be helpful to distinguish thymoma from other anterior mediastinal abnormalities. currently, there is no consensus on the imaging pathway for children with mg with suspected thymoma. our aim is to review the imaging of patients who were referred to our institution for management of mg, and suggest an imaging pathway in cases where thymoma is suspected. we performed a retrospective search of the local pacs system of cases between 2000 and 2016 using the search terms "thymoma" and "myasthenia gravis" in the clinical indication for the study and the body of the final report. forty-three cases were identified using the search criteria. eight cases were excluded owing to an absence of cardiothoracic imaging. 20/35 of all cases (57%) had chest x-rays (cxr's), of these 17/20 (85%) were normal. the three remaining patients who had abnormal cxr's went on to have ct scans, which confirmed an anterior mediastinal mass (amm) in all three cases. 26/35 of all cases (74%) had cross-sectional imaging (mri 15/26 cases, ct 11/26 cases). of those, 21/26 of cases (81%) had normal studies. specifically, all 15 mri studies (100% of cases) were normal, while only 5/15 ct scans (66%) demonstrated an anterior mediastinal abnormality. 12/35 of all cases (34%) had both cxr and cross sectional studies. 9/12 of these cases (75%) had a normal ct or mri. in the remaining three cases, the amm was clearly demonstrated on both cxr and the crosssectional imaging. in our series, radiography, ct and mri studies were normal in the vast majority of cases. however, given that thymectomy is potentially s405 (2017) 47 (suppl 2):s297-s pediatr radiol curative, it is appreciated that clinicians may still be keen to radiologically investigate paediatric patients with myasthenia gravis. cxr is not an efficacious imaging modality in this context, as patients with a normal cxr may be falsely negative, and patients with an abnormal cxr may undergo cross-sectional imaging regardless. we propose that mri should be used as first line investigation for patients in this population. this approach will negate the need for ionizing radiation, maximize diagnostic yield, and facilitate surgical planning if deemed clinically appropriate. increased risk of venous thrombosis of the arm with multiple peripherally inserted central catheters insertion in paediatric patients r. gnannt 1 , n. waespe 2 , j. donnellan 2 , k. liu 2 , l. brandao 2 , b. connolly 2 ; 1 zurich/ch, 2 toronto/ca objective: peripherally inserted central catheters (piccs) are associated with superficial and deep venous thrombosis of the arm. the impact on the incidence of developing a thrombosis of the arm when inserting a subsequent picc remains unclear. the purpose of this study was to analyze the incidence of deep, upper limb thrombosis of repeated upper limb piccs in children. the study population included all patients who underwent their first successful arm picc insertion between january 2010 and december 2015. subsequent ipsilateral arm piccs were included in the analysis. patients were followed until march 2016 or until any alternative central venous line insertion (jugular, femoral, saphenous or umbilical vein lines -because of their thrombogenic effect). for each picc insertion the following data were collected: date of insertion and removal, weight of the patient, type of picc (1.9fr, 2.6fr, 3fr, 4fr, 5fr), left or right arm, and vein cannulated (basilic, brachial, cephalic). all symptomatic deep and superficial thrombosis of the arm were correlated with the picc database. four thousand one hundred thirty-eight piccs were inserted. applying inclusion and exclusion criteria, 1955 piccs remained for analysis. first, 2 nd , 3 rd , and 4 th picc insertions in the same arm were identified in 1773, 146, 30 and 6 patients, respectively. in total there were 57 upper body deep symptomatic thrombotic events diagnosed with ultrasound. a 1 st , 2 nd , 3 rd , and 4 th picc insertion was associated with 46/1773 (incidence 2.6%), 6/146 (4.1%), 4/30 (13.3%), and 1/6 (16.7%) thrombotic events, respectively. an increasing hazard ratio was seen with higher numbers of picc insertions, which was significant when comparing the 1 st with the 3 rd picc insertion in the same arm (hr 3.9, ci95%1.3-11.4, p=0.01). after excluding any confounder, double lumen piccs were associated with a significantly higher risk of thrombosis than single lumen (or 5.3, ci 1.2 -23.4, p=0.026). repetitive picc insertions in the same arm are associated with an increased risk of thrombosis. double lumen piccs are associated with a higher risk of thrombosis compared to single lumen lines. diagnostic performance of lung ultrasound for the detection of community acquired pneumonia in children j.a.m. stadler 1 , s. androunikou 2 , h. zar 3 ; 1 paarl/za, 2 bristol/uk, 3 cape town/za objective: chest radiographs (cxr) are considered the first line imaging modality when investigating cases of suspected community acquired pneumonia (cap) in children. however, cxr interpretation is limited by moderate sensitivity and specificity and poor inter-and intra-rater reliability and expose children to potentially harmful ionizing radiation. point-of-care lung ultrasound (lus) has been proposed as alternative to cxr for diagnosis of pneumonia in children and some published data suggest accuracy and reliability as good as or better than cxr. most of these studies however, were performed in in-hospital settings creating a bias for selceting more severe disease and consequently more overt radiological findings. the mean age of children in most of these studies were also well above one year, while the highest incidence and risk of complicated pneumonia occurs during the first year of life. the purpose of our study was to assess the diagnostic performance of lus for the diagnosis of pneumonia in both hospitalised and non-hospitalised children in an age group representative of the most at risk segment of the population. we performed a lus on 147 children who presented with clinical signs consistent with the who case definition for childhood pneumonia. one hundred of these patients also had chest radiographs performed as part of routine clinical care. inter-rater reliability (irr) between a general practitioner and an expert paediatric radiologist were assessed for the interpretation of lus findings consistent with pneumonia. where radiographs were available concordance between lus and cxr findings of pneumonia were also assessed. results: seventy-four hospitalised and 73 non-hospitalised clinically defined pneumonia cases were included with a median age of 10.3 years. our general practitioner reported lus findings consistent with pneumonia in 77/147 (52%) compared with 66/147 (45%) by the paediatric radiologist. substantial overall agreement between the reporters was found with an overall agreement proportion of 0.79 and kappa=0.62. agreement for the presence of lung consolidation or for a normal scan was also substantial with kappa of 0.67 and 0.68 respectively. agreement on the finding of interstitial syndrome was moderate with kappa=0.45. agreement was higher in hospitalised than in non-hospitalised cases with kappa of 0.62 and 0.56 for the respective categories. results showing concordance between lus and cxr findings are pending. conclusion: lus shows substantial irr for the diagnosis of pneumonia in children. irr are higher for the detection of consolidation or for no pathology than for interstitial syndrome. irr also appears to be lower in clinically less severe disease. 'white-out' on plain chest radiograph-a late presentation of congenital diaphragmatic hernia a. fagan 1 , c. stewart 2 , k. halliday 3 , s. rao 2 , d.t. chang kwok 2 ; 1 peterborough/uk, 2 lincoln/uk, 3 objective: awareness of the limitations of plain radiograph and computed tomography in diagnosis of late presentation of congenital diaphragmatic hernia. case presentation: a 2 year old boy presented with a 6 day history of pyrexia, vomiting and respiratory distress. he was haemodynamically stable, and had no audible air entry over his upper left thorax with occasional wheeze over the left base. he had bronchiolitis previously but did not require ventilatory support. he was otherwise well with unremarkable antenatal scans. initial chest x-ray showed a large air collection with fluid or soft tissue density within the left hemi-thorax and mediastinal shift to the right. repeat x-ray (figure 1 ) demonstrated the nasogastric tube below the diaphragm. complicated pneumonia was suspected but as the findings were atypical a non-contrast ct was performed. this was interpreted as showing a large hydropneumothorax. (figure 2) . a chest drain was inserted which drained only a small volume of fluid, and a repeat chest film showed no change. ct chest and abdomen with oral and intravenous contrast revealed a bochdalek diaphragmatic hernia (figure 3) . fortunately the chest drain had not entered the herniated stomach. the hernia was surgically corrected and the child recovered well. (2017) 47 (suppl 2):s297-s pediatr radiol bochdalek is the most common congenital diaphragmatic hernia (cdh). it is often diagnosed on prenatal ultrasound, with mri used for confirmation. cdh which is not diagnosed in the perinatal period may be asymptomatic and imaging findings can be confusing. postnatal x-ray typically shows an opacified hemi-thorax with or without gas bubbles. there can be mass effect with mediastinal shift. the position of an ng tube can be helpful in localising the stomach, but in this case the infradiaphragmatic position of the tube gave false reassurance. in neonates, the position of an umbilical venous catheter may demonstrate abnormal location of the liver. computed tomography generally demonstrates a posterolateral defect (foramen of bochdalek), which is located on the left in 80% of cases. ct is useful for excluding lung masses or bronchopulmonary foregut malformations, which may appear similar to cdh on x-ray. ct can also identify anatomical abnormalities associated with cdh. late presenting cdh is often misdiagnosed as pleural effusion or pneumothorax. there are other case reports published where chest drains were inserted before cdh was diagnosed. it is important to keep cdh in mind as a potential cause of unilateral hemithorax opacification, even in previously asymptomatic older children. ct with oral contrast can be useful in diagnosis. ovarian tuberculosis with peritoneal dissemination mimicking ovarian tumor with peritoneal seeding d. grassi, v. tostes, a. duarte, s. abib, h.m. lederman; sao paulo/br consider tuberculosis (tb) as a differential diagnosis whenever the case enrolls in an endemic region. case presentation: female, 13 years old adolescent, who presents with abdominal pain and weight loss. abdominal sonography was performed in a public family practice location and bilateral ovarian masses were detected. she was referred to an oncology pediatric facility for further investigation. abdominal mri and chest ct were performed where dissemination through the peritoneal and mesenteric lymph nodes could be detected; chest ct was normal. the patient underwent surgical intervention for diagnosis and on pathology the findings in the bilateral ovarian masses were secondary to tb involvement. sonography showing bilateral pelvic masses. t2-weighted coronal overview bilateral ovarian masses. unique teaching points: whenever a case enrolls in an endemic region of tuberculosis, it is important to consider it as a possible differential diagnosis. in this case, the initial presentation mimicked ovarian tumor with mesenteric seeding. however, only after surgical approach was possible to diagnose ovarian tuberculosis with mesenteric lymph nodes and peritoneal involvement. retrospectively, patient's uncle was discovered as having pulmonary tb. langerhans'-cell histiocytosis with thoracic involvement in infant and young child: ct findings s.-l. shih 1 , k. tsai 1 , w. huang 2 , f.-s. yang 1 ; 1 taipei/tw, 2 taitung/tw the purpose of the study was to evaluate the ct changes of thorax in the patients with langerhans'-cell histiocytosis. the 3-month-old female infant presented with generalized hemorrhagic macular rash over the skin for 2 months. the laboratory findings showed hemoglobin 7.8gm/dl (normal 11.0~16.0gm/dl). the chest radiograph showed bilateral reticulonodular infiltration. high-resolution computed tomography (hrct) of chest showed multiple cystic-like lesions (1-6mm) in the right middle and bilateral lower lobes. the pathological report was langerhans'-cell histiocytosis after skin biopsy from upper chest. then she was on scheduled chemotherapy. she was in remission after one-year treatment. the 1y10m-old girl presented with fever for 3 months. the physical examination revealed hemorrhagic-macular rash over the skin in the anterior chest wall and hepatosplenomegaly. the laboratory findings revealed albumin 2.5g/dl (normal 3.8-4.7g/dl) and hemoglobin 7.0g/dl (normal 11.0-16.0g/dl). hrct of chest showed multiple cystic-like lesions (1-6mm) in the bilateral lower lobes with left pleural effusion as well as multiple osteolytic lesions in the vertebral bodies of t7, t8, t11 and t12. the pathological report was langerhans'-cell histiocytosis after skin biopy from anterior chest wall. then she was on scheduled chemotherapy. she was doing well 10 years after treatment. the 1y6m-old girl presented with yellowish discoloration of skin for one month. the laboratory findings revealed direct/total bilirubin 4.4/7.9mg/dl (normal 0.1-0.5/0.3-1.2mg/dl), got 223iu/l (24-46iu/l) and gpt 220iu/l (12-27iu/l). the chest radiograph revealed enlargement of upper mediastinum. the ct scan of chest and upper abdomen showed punctuate calcification with heterogeneous enhancement in the upper mediastinum and several minute cysts in the lower lobes of lung (hrct) as well as dilatation of bilateral intrahepatic bile ducts in the liver. the pathological report was langerhans'-cell histiocytosis after biopsy from thymus and liver. then she was on scheduled chemotherapy and got initial response. unique teaching points: langerhans'-cell histiocytosis affecting the lungs and thymus may be in isolation or as part of a multiorgan disease. the pulmonary changes on ct scan may not have corresponding respiratory symptoms. ct scan of thorax may have multiple minute cysts (1-6mm) in the lungs, pleural effusion, calcification in the thymus and osteolytic lesions in the thoracic spine. case of fungal infection of the soft tissue in a child with acute myeloid leukemia (ultrasound aspects of diagnosis) i. begun, s. kondaurova; minsk region/by objective: early diagnosis of fungal infections of the tissues is essential for a successful and complete recovery. we describe a clinical case of fungal infection of the soft tissue in a child with acute myeloid leukemia (aml). ultrasound were made for the characteristics of the structural changes in the area of interest to perform biopsies followed by bacteriological culture studies. case presentation: patient k., 13 years old, diagnosed with aml, from which after a course of induction chemotherapy with neutropenia about 3 weeks on the skin of the foreskin appeared removable hard white coating. cultures of plaque it possible to establish the presence of fungi of the genus trichosporon spp. after 4 days, there were hyperemia, compaction and ulceration of the glans penis, which led to extensive tissue defects. with help ultrasound were determined the structural deformation of the glans penis with the pronounced around changed tissues vascularization. after 7 days in the rear surface projection of the left thigh and the lateral surface of the left calf were defined erythematous papules which progressed to ulceration with central black scab. by standard ultrasound were visualized: subcutaneous nodal education oval 1,0х0,4 sm on hip and echogenic skin thickened portion having an average degree of severity of dorsal acoustic shadow on the lower leg (weakening of the signal behind scab). in cultures of biopsies of subcutaneous foci were revealed fungi of the genus trichosporon spp too. the patient received the combination treatment (intravenous liposomal amphotericin b and surgical rehabilitation of lesions of glans and corpus cavernosum of penis). after the stabilization of patient state the treatment of the underlying disease was continued. unique teaching points: for some patients, lesions of superficial tissues may be the only sign of systemic fungal infections, and rapid recognition of these lesions may contribute to early diagnosis and treatment. ultrasound examination in such a situation naturally becomes an main imaging tool and by choice method. the scanning high-resolution of foci of the thigh of the patient k. in grayscale made possibility to determine the configuration consisting of the central echogenic focus surrounded by a hypoechoic rim (fig. 1) with peripheral changes in the type of "infiltrative" according by the active fungal infection at the exit of cytopenia. duplex and triplex ultrasound scanning were indicating to the perifocal vascularization with low level vascular resistance around of the affected area (see fig.2-3) . to increase knowledge and awareness of rare cases and diseases in order to be able to better manage and treat patients in the future. case presentation: an 3-month-old female was presented to our hospital with abdominal distention that increased in the past 3 months associated with low-grade fever, loss of weight and mild respiratory distress. abdominal ultrasonography revealed an enlarged liver with multifocal hypoechoic lesions scattered all over the liver (fig 1) . a ct scan with iv contrast (mri was not available at that time in our district) revealed severe hepatomegaly with the presence of multiple, variable in size, hepatic hypodense lesions which had peripheral (ring) enhancement after contrast injection in the arterial phase (fig 2) . progressive centripetal filling in portal phase is seen and in the delayed images many of the lesions were completely filled (fig 3) . reduction in the aortic caliber (mid-aortic syndrome) below the level of celiac branch was noted. a diagnosis of hemangioendothelioma was made although liver biopsy was not done due to fear of hemorrhage. alternative diagnosis to infantile hemangioendothelioma in this age group is hepatoblastoma, mesenchymal hamartoma and metastatic neuroblastoma. the patient was transferred to another city to a hospital with pediatric oncology department for follow up and treatment. unfortunately the lack of experience and knowledge of such rare cases led to mismanagement and delayed treatment and after less than 2 months the patient was brought back to our hospital to the pediatric icu due to deterioration of her status due to congestive heart failure. unfortunately the patient died shortly afterwards. hemangioendothelioma is twice as common in girls and can have complications due to high output chf secondary to arteriovenous shunting hemangioendotheliomas tend to involute spontaneously without therapy over a course of months to years. they are followed with sequential ultrasounds. medical therapy is reserved for severely symptomatic lesions (e.g. anemia, consumptive coagulopathy, high-output chf) and includes high-dose steroids or alpha-interferon. in cases of failed medical management, surgical resection should be performed. if partial hepatectomy is not technically achievable, transarterial embolization should be used either as definitive therapy or as a temporizing measure until liver transplantation is possible. the sad outcome of this case was mainly due to mismanagment due to lack of medical experience and knowledge of such rare cases so we suggest that such rare cases should be catalogued in a national data bank for future consultation and teaching purposes. fatal outcome of acute gastric dilatation causing acute abdomen compartment syndrome in a child: a case review c.s. yoon; seoul/kr to describe and review presumed acute abdominal compartment syndrome in a child. case presentation: a 3 years and 5months old boy was admitted to emergency room due to abdominal distention. he suffered abdominal pain and vomited since yesterday after lunch. on physical examination, his abdomen was rigid and distended. body temperature is 37.2°c. the white cell count was increased (24,140/μl). esr is 25 mm/hr and c-reactive protein was 123.4 mg/l. creatinine was increased (1.24mg/dl). amylase and lipase were increased (164 u/l and 104 u/l respectively). prothrombin time was prolonged (15.5 sec). plain abdomen radiograph shows markedly distended stomach with air-fluid level (fig.1) . first trial of nasogastric tube insertion was failed due to kinking of tube at gastroesophageal junction. contrast-enhanced abdomen ctscan shows marked distensionofstomachwithlargeamountoffoodmaterialsandintraluminalairwith prominent external compression in the duodenal 3rd-4th junction. esophageal air distention is also markedly noted with l-tube insertion. no opacification of large vessel with contrast media, without contrast enhancement of spleen, pancreas and left kidney is noted (fig.2) . prob. markedly compressed and poorly defined lower abdominal aorta with faintly visible both common iliac arteries and femoral arteries. after ctscan, nasogastric tube exchange was performed due to poor drainage of gastric fluid. about 700cc of gastric fluid was drained. however, sudden cardiac arrest of the patient was developed. although vigorous cardiopulmonary resuscitation was performed, the patient was died. (2017) 47 (suppl 2):s297-s pediatr radiol unique teaching points: acute abdomen compartment syndrome is a very serious and lifethreatening disease. as soon as possible, rapid diagnosis and adequate treatment are necessary for good prognosis. delayed diagnosis and treatment may result in fatal outcome. pleuroperitoneal fistula in a pediatric patient with primary hyperoxaluria type 1 w.p. chu; hang hau/hk to illustrate the imaging features of pleuroperitoneal fistula in a pediatric patient suffering from primary hyperoxaluria type 1 case presentation: an 11-year-old girl with the history of primary hyperoxaluria type 1 was repeatedly admitted to the hospital for recurrent right pleural effusion despite chest drain insertion. the right pleural fluid was transudative in nature and the microbiological cultures for bacteria and mycobacterial species were negative. the radiographic examination [ figure 1 ] showed moderate right pleural effusion a n d f e a t u r e s o f o x a l o s i s i n c l u d i n g b i l a t e r a l c o r t i c a l nephrocalcinosis and generalized increased in bone sclerosis. delayed planar images of the peritoneal scinitigraphy [ figure 2 ] obtained 3 and 5 hours after injection of technetium-99 m suphlur colloid found diffuse tracer activity at the right hemithorax, suggestive of pleuro-peritoneal fistula. the patient subsequently required thoracoscopy and surgical decortication at the right hemithorax and renal transplantation. primary hyperoxaluria is due to defective glyoxylate metabolism and results in increased synthesis of oxalic acid. cortical nephrocalcinosis and diffusely increased bone sclerosis are characteristic radiographic features. pleuroperitoneal fistula is unusual in patients without peritoneal dialysis. possible cause in this patient is increased intra-abdominal pressure related to portal hypertension and cirrhosis. osteosarcoma with pulmonary intra-arterial tumor embolism metastasis a. alzaher, f. alzaher; dammam/sa objective: osteosarcoma rarely invade the veins and small number of cases has been reported with venous invasion at the presentation. however, to our knowledge, no case has been reported with venous invasion and isolated distal metastasis as intra-arterial pulmonary embolisms. we are presenting a case of pediatric pelvic osteosarcoma with venous invasion and pulmonary arterial tumor embolisms as isolated distant metastasis at the presentation. the purpose of this case report is to describe the rare presentation of distant metastasis as isolated pulmonary arterial embolism that might be overlooked radiological. additionally, such tumor embolism might cause respiratory symptoms and differentiating tumor emblism from pulmonary thromboembolism is crucial to avoid the unnecessary anticoagulation. case presentation: fourteen year old boy who presented with 3 months history of right hip and lower limb pain after trauma. this was associated with lower limb swelling. the plain radiography showed right pelvic iliac bone aggressive mass, along with lobulated, soft-tissue components, extensive areas of osseous matrix, and malignant periosteal reaction. the patient could not tolerate the mri and ct scan was performed and it showed that the mass was invading the right external and internal iliac vein with imaging appearance was most consistent with osteosarcoma. patient staging was then carried on with mri under anesthesia and chest, abdomen and pelvic ct scan. the unenhanced and iv contrast enhanced chest ct scan showed multiple beaded expansion of sub segmental pulmonary arteries with soft tissue destinies and calcification suggestive of intra-arterial pulmonary tumor embolisms. there was no isolated pulmonary nodule or any other site of distant metastasis. unique teaching points: we present this case to increase the awareness of isolated intra-arterial pulmonary tumor embolisms as osteosarcoma metastasis especially with the present of venous invasion. additionally, such condition might be with respiratory symptoms and differentiating the tumor embolism from pulmonary thromboembolism is crucial to avoid the unnecessary anticoagulation. case presentation: a 14-year old boy with acute myelodysplastic syndrome presented with recurrent, acute severe anemia (hemoglobin 68 g/dl) and melena. his past history was significant for bone marrow transplant twice followed by graft-versus-host-disease of intestines, bilateral lung transplants for bronchiolitis obliterans, renal failure, scleroderma and acute pancreatitis. ct angiography performed previously did not identify active extravasation. several days before, upper gi endoscopy had demonstrated ulceration of the greater curvature of the gastric wall that was initially treated with epinephrine injection and surgical clip placement. at the time of referral, endoscopic interventions were unsuccessful leading to progressive clinical deterioration. a decision was taken to proceed to angiography to isolate the arterial source of hemorrhage, with an intention to embolize, if feasible. catheter angiography via transfemoral 4fr access revealed a left gastric artery pseudoaneurysm with active extravasation into the gastric lumen through the ulcer. after selecting the feeding pedicle of the left gastric artery with a microcatheter, the pseudoaneurysm was embolized using 40% nbca in lipiodol, resulting in complete angiographic obliteration of the bleeding source. on repeat cbc 6 hours post-procedure, the hemoglobin had increased from 80 to 115 g/dl. the patient remained hemodynamically stable in the intensive care unit. there is no evidence of bleeding recurrence 25 days later. unique teaching points: catheter angiography can define the bleeding source with greater accuracy than cta in children. there should be a low threshold to perform catheter angiography, with an intention to proceed to treatment. nbca embolization is a feasible and effective option for treatment of acute gi bleeding in children. case presentation: an infant born by cesarean section at 38 weeks of gestation, after nonreassuring cardiotocoghraphy, with meconium aspiration at birth, severe hepatocellular failure with hyperbilirubinemia, signs of hemorrhage, edema, ascites, hypoglycemia, increased ferritin values, and lactic acidosis was referred for ultrasound and magnetic resonance. both examinations showed signs of liver cirrhosis with portal hypertension; in addition, on t2-weighted images and gradient-echo images, the signal intensity of the liver and the pancreas was lower than that of the spleen and skeletal muscle, a finding consistent with abnormal iron deposition in those organs. a biopsy of the lower lip confirmed the diagnosis of neonatal hemochromatosis. unique teaching points: although the diagnosis may be suspected clinically, it must be confirmed by demonstrating the generalized iron overload affecting, among other organs, the salivary glands, liver and pancreas, with sparing of the reticuloendothelial system. the underlying cause may be associated with an an alloimmune mechanism; thus, intravenous immunoglobulin during gestation is administered in selected cases to prevent the severity of neonatal hemochromatosis. diagnosis is then crucial not only for management of the affected infant, but also for prevention in the future offspring. fishing for the answer -a rare case of paediatric exogenous lipoid pneumonia secondary to fish oil aspiration h. moodley, d. white, g.d. baker; johannesburg/za objective: lipoid pneumonia is a rare condition caused by the intrapulmonary accumulation of endogenous or exogenous fat containing substances. in the acute exogenous form secondary to aspiration of oil, it is important to make the diagnosis and remove the causative agent to prevent or arrest the progression of pulmonary fibrosis. radiopathological findings usually prompt the diagnosis, as aspiration of mineral oils is usually unnoticed due to the lack of reactive airway symptoms and patients present with vague chronic respiratory symptoms. case presentation: we present the clinical, radiological and pathological correlation of exogenous lipoid pneumonia in a 4-month-old male patient with recurrent respiratory tract infections. a ct chest demonstrated an extensive crazy paving pattern of the dependent lung segments bilaterally. the lung biopsy findings of occasional intra -alveolar macrophages with larger (2017) 47 (suppl 2):s297-s pediatr radiol foamy cytoplasmic vacuoles, raised the possibility of an exogenous lipoid pneumonia secondary to aspiration. on further history, the patient was found to have been fed fish oil by his mother, confirming the diagnosis. unique teaching points: the rare diagnosis of exogenous lipoid pneumonia can be confirmed on ct chest by measuring the hounsfield units in the most hyperdense components of consolidation (typically -150 to -30 hu). histopathological confirmation can be obtained provided that the specimens are not embedded in paraffin. the possible role of visual evaluation of dwibs in childhood renal masses based on our five cases e. varga, g. rudas; budapest/hu objective: nowadays, the diffusion-weighted mri has a great importance not only in the differential diagnosis and follow-ups of childhood renal tumors, but in the early detection of recurrence of the disease as well. the dwibs with appropriate b-values and the adc calculation can be helpful in distinguishing between benign and malignant processes. however, the adc calculation is a time consuming method and in addition, there are cases when we cannot use this technique, but we can still apply the visual evaluation of diffusion. case presentation: between 2013-2016, we had 5 cases in which the visual assessment of dwibs was the best method which helped to make the appropriate therapeutic decisions. left kidney of an infant with nephroblastomatosis was removed because of an arising wilms tumor. 2,5 years later, in the contralateral kidney, a small area of diffusion restriction appeared on the dwibs in one of the cystic residual lesions, but the anatomic sequences haven't showed any changes comparing with the previous examinations. in another patient with beckwidt-wiedemann syndrome, the follow-up ultrasound examination showed a little bulging of the surface of the left kidney. accordingly, the mri showed a barely distinguishable nodule, but the dwibs referred to a wilms tumor. in a 17-month-old child, more nodules were visible in both kidney on the dwibs than on other sequences. with the help of visual evaluation of dwibs, we were able to detect the malignant lesion easily and quickly, among a lot of cystic and solid nodules of the kidneys in a seven years old patient with sclerosis tuberosa. an 8-month-old infant was followed with a benign cystic renal disease and a new small solid nodule was found on the last ultrasound examination. instead, the visual assessment of dwibs indicated a multilocular cystic wilms' tumor. unique teaching points: the diffusion-weighted mri is suitable for differentiate benign and malignant renal lesions in children. the dwibs (with appropriate b-values) and the adc calculation are very sensitive methods in pediatric oncoradiology. the adc calculation is a long process andas our cases demonstrated -we cannot apply in every cases. the visual evaluation of dwibs is a time saving method which is spared from limitations of adc histogram-based assessment, so it may become very useful in the everyday practice. we can use it in the differential diagnosis and follow-ups of childhood renal tumors and we can detect the recurrence of the malgnancy very early and easily. mr urography in a 9-years-old female with unusual urinary dribbling m.c. terranova, c. tudisca, d. narese, g. li voti, s. salerno; palermo/it objective: congenital anomalies of kidney and urinary tract (cakut) occurs in up to 3.2% of infants, and clinically they can range from asymptomatic patients, in which anomaly is detected incidentally even in adulthood, to ante-natal or post-natal mortality due to bilateral kidney agenesis or acute renal failure. dmsa renal scintigraphy is considered gold standard, for evaluation of those cases electable for surgery, in order to assess renal function, depict and locate ectopic kidney and guide surgical management, but has the important limit of radiation exposure and may undetect poorly functional renal moieties. the advent of modern magnetic resonance technics proven to be able to assess anatomical malformations and renal function, overcoming the limits of dmsa scintigraphy, may be used as a valid alternative, especially in vulnerable pediatric population. we herein describe a case of a young girl with small left renal bud and ectopic ureter, draining in vagina, discovered by mr and undetected by previous dsma scintigraphy. case presentation: a 9 years old girl was referred for continuous urinary dribbling, after starting toilet training, with normal bladder voiding pattern, unrelated to any physical and psychological events, and no history of urinary tract infections. physical examination revealed vaginal septa and micturition training was practiced, with no symptoms improvement. abdominal us study was performed, reporting empty left renal fossa and hypertrophic right kidney; no ectopic kidney nor sign of urine stasis or other urogenital anomalies were detected, and dmsa renal scintigraphy was planned. it depicted only normal right kidney radionuclide uptake but no evidence of left renal or ectopic renal tissues activity. patient then underwent mr evaluation for suspected genito urinary malformation, that revealed a small cystic formation, with a slight cortex, at the level of the iv lumbar vertebra -that represented the left immature renal bud -supplied by a short fluid-filled tubular structure, located postero-medially to the bladder -that configured the left ectopic ureter, draining in left vaginal wall. bladder was normal, and regularly connected with the right orthotopic ureter (fig 1) . pre-surgical cystoscopy and vaginoscopy, followed by left ascending urethrogram were performed, confirmed previous mr findings, and patient underwent successfull laparoscopic left nephron-ureterectomy. unique teaching points: mr urography has proven to be a rapid, safe, radiation free, systematic diagnostic tool especially in the evaluation of poorly functioning renal systems, and of collecting system, bladder and ureteral abnormalities, overcoming the limits of conventional imaging technics agenesis of the dorsal pancreas: case report c. lanza, g. pieroni, l. amici, a. giovagnoni; ancona/it objective: agenesis of the dorsal pancreas (adp) is a rare malformation. since 1911 and until 2008, 53 cases have been reported. majority of the patients with this anomaly are asymptomatic or associated with abdominal pain, hyperglycemia, diabetes mellitus, and acute or chronic pancreatitis. case presentation: we present a case report of a 11-year-old girl with adp, diagnosed incidentally during radiological evaluation for abdominal pain. she was hospitalized in the pediatric department for recurrent abdominal pain for the past 10 months. there was no history of nausea, vomiting or trauma. biochemical investigations showed a normal random serum glucose, serum amylase levels slightly increased (149 u/l; reference value 28-100 u/l) and slightly elevated serum pancreatic lipase levels (138 u/l; reference value 22-51 u/l). the day after serum amylase levels decresed up to 80 u/l and lipase levels to 78 u/l. us revealed increased -size pancreatic head with normal contour and echotexture with no parenchymal calcification or duct dilatation. the body and the tail of the pancreas were poorly visualized. mr imaging examinations revealed only a partial visualization of the pancreas: the pancreatic head and the uncinate process were visualized with defined margins with peripancreatic fat stranding, but the distal neck, body, and tail of the pancreas were absent. on mrcp, the dorsal pancreatic duct of santorini and the minor duodenal papilla could not be visualized. the ventral pancreatic duct of wirsung and the common bile duct were normal and clearly visualized. these findings were compatible with complete adp, eliminating the need for ercp. unique teaching points: the clinical presentation of dpa varies greatly ranging from incidental detection on x-ray, surgery or autopsy through to the development of a ductal adenocarcinoma of the pancreas. abdominal pain and diabetes are the most frequent clinical manifestations reflecting exocrine and endocrine insufficiency as most of the islands of langerhans are located in the tail of the pancreas. there have also been reports of an increase in the size of the remnant pancreas and recurrent acute pancreatitis as a form of presentation. diagnosis requires confirmation of the absence of the neck, body and tail of the pancreas and duct of wirsung using endoscopic retrograde cholangiopancreatography (ercp) or mrcp. one hundred four mr images of foetal cns with a us suspicion of acc were retrospectively reviewed. foetal mri was performed at 1.5 t magnetom avanto (siemens, erlangen, germany) without motherfoetal sedation. polymicrogyria, lissencephaly, schizencephaly, subependymal heterotopias and migration disorders were evaluated. cortical findings were compared to three types of acc (complete agenesis, partial agenesis and hypoplasia). genetic tests were collected. postnatal mri or foetopsy for diagnostic confirmation were collected. on 104 foetuses, fetal mri was able to detect cortical malformations in 32 cases even in early gestational ages (<24gw). the mean gestational weeks (gw) at mr diagnosis was 26 (range: 22-36gw). mr imaging found 13/32 polymicrogyria, 7/32 lissencephaly, 5/32 schizencephaly, 4/32 subependymal heterotopias and 3/32 neuronal migration disorders. 22/32 had complete acc, 4/32 had partial acc and 6/32 had cc hypoplasia. statistically significant correlations (p<0.005) between complete acc, focal polymicrogyria and cortical dysmorphism affecting frontal lobes were found. fetal cns mri can detect cortical development malformations in complex acc, providing further information for the clinician to assess the severity of perinatal outcome. mri is a useful tool in improving obstetrical genetic prenatal counselling to predict pregnancy and foetal prognosis. clinical signs of the neonatal lymphatic flow disorder (nlfd) are a combination of the congenital chylothorax, chylous ascites and body edema. it can present as neonatal chylothorax (nc), neonatal chylous ascites, or congenital lymphatic dysplasia (cld). the prenatal appearance of lymphangiectasia has been described as nutmeg lung. the purpose of this study is to describe prenatal and postnatal imaging features and outcomes of neonates with nlfd. materials: this is a retrospective case series of neonates in our institution that had pre-and postnatal lymphatic imaging and nlfd. all patients had prenatal imaging (fetal mri and us) and underwent postnatal dynamic contrast mr lymphangiography (dcmrl) with a three-dimensional (3d) t2 space. conventional lymphangiography (cl) when performed was also reviewed. six patients with nlfd were identified (3 with nc and 3 with cld). one patient had congenital heart disease. nutmeg lung was seen in all patients on fetal mri and 4 patients on fetal us. 5/6 patients had pleural effusions, 2/6 had ascites and 1/6 had body wall edema prenatally. postnatal mri with 3d t2 space revealed soft tissue edema in the upper chest and neck (5/6 patients), mediastinal edema (5/6 patients), interstitial lung edema (6/6 patients), retroperitoneal edema (5/6 patients), and ascites (6/6 patients). dcmrl demonstrated lymphatic flow to the pleural space (5/6 patients) and to the abdominal cavity (1/6 patients) and dermal backflow (2/6 patients). cl was performed in 4 patients, all of which had collateral lymphatic flow to the lung. lymphatic intervention was performed in 3 patients, lipiodol injection for 2 patients with nc and thoracic duct embolization (tde) for 1 patient with cld. mean hospital duration in the first 4 months of life was 51 days (range 5-113) for nc and 105 days (range 75-120) for cld. all 3 patients with cld died after 4 months of age due to respiratory distress including the patient that had tde and both with findings of dermal backflow. the pleural effusions in the 2 patients with nc resolved post lipiodol injection and in the other patient with nc it resolved with conservative therapy. conclusion: nlfd is a disorder that can be recognized on prenatal and postnatal imaging. in this small series, nutmeg lung was present in all patients with nlfd and may be easier to recognize with fetal mr than us. dermal backflow on dcmrl suggests a poor prognosis. both prenatal and postnatal imaging may guide treatment and interventions in nlfd. fetal mri and postnatal ct scans of prenatally diagnosed bpms from 10 patients with available histology were analyzed retrospectively. the fetal mri and ct images were reviewed by two radiologists blinded to histological findings. specific diagnosis was assigned based on predetermined criteria. the accuracy of fetal mri was evaluated. the agreement rate in fetal mri diagnosis between two radiologists was 100 %. an overlap of 80% in fetal mri and histopathological diagnosis was reached. when comparing fetal mri and postnatal ct examinations, the agreement of the results was also 80%. the least matching histological diagnosis was bronchopulmonary sequestration (bps). fetal mri is very accurate in characterizing the bpm spectrum and provides important information on lesion type and structure when compared with histology. with relatively small number of patients high correlation between prenatal mri and postnatal ct was reached. therefore, further investigation with more patients is needed. we hypotethise that fetal mri in late pregnancy could in the future replace early (neonatal) ct examinations if fetal mri provides sufficient inforfmation for clinical management. real time virtual sonography: a new integrated approach for the evaluation of fetal cerebral pathologies? s. bernardo, a. antonelli, v. vinci, m. saldari, c. catalano, l. manganaro; rome/it objective: real-time virtual sonography (rvs) is a new technique that uses magnetic navigation and computer software for the synchronized display of real-time us and multiplanar reconstruction mri images. the purpose of this study was to evaluate the feasibility and ability of rvs to assess the main cerebral pathologies in fetuses with suspected us anomalies. materials: this is a prospective study. fusion imaging (hitachi hi vision ascendus) was offered to 35 patients undergone fetal mri for a us suspicion of cerebral pathology. the mri image dataset acquired was loaded into the fusion system using a cd support and displayed together with the us image. both sets of images were then manually synchronized and images were registered. the possibility to record the images in a video format allowed, however, the possibility to re-evaluated the examination. results: rvs was technically possible in all cases. data registration, matching and fusion imaging were performed in 25 minutes at the beginning and in less than 15-20 minutes after practice. the ability of rvs imaging to assess the main anatomical sites and fetal anomalies was evaluated and compared with standard us and mri images. the principal application of rvs was the study of midline, cerebral gyration and vascular malformations because it also allowed adding a real time doppler signal on mri images. fusion imaging helped the diagnosis in 25%. in the 25/ 35 cases of encephalic pathology, fusion imaging improved the diagnosis; in the other cases mri was superior to us even using the rvs. this is a preliminary study on the feasibility and practical use of a fetal mri-us real-time fusion imaging. both techniques are complementary but still independent and the retrospective synthesis of these exams allows optimal analysis of fetal cerebral anomalies. this technique has many advantages especially on the pedagogic plan. however, rvs is currently limited to the research area. role of foetal mri in the evaluation of ischaemic-haemorrhagic lesions of the foetal brain s. bernardo, a. antonelli, v. vinci, m. saldari, c. catalano, l. manganaro; rome/it the aim of this study was to define the role of fetal magnetic resonance imaging in the evaluation of cerebral ischaemic-haemorrhagic lesions and the extension of parenchymal injuries. from september 2010 to december 2016 we performed 271 fetal mri of cerebral region in foetuses with suspected abnormalities on ultrasound or cmv infection and toxoplasma serum conversion. fetal mri was performed with a 1.5-t magnet system without materno-fetal sedation. fetal mri detected ischaemic-haemorrhagic lesions in 14/271 fetuses, revealing a 5% pathology incidence. mri confirmed the diagnosis in 3/14 cases with us suspect of ischaemic-haemorrhagic lesions associated with ventriculomegaly. in 1/14 cases with us findings of cerebellar haemorrhage, mri confirmed and provided additional information regarding the parenchymal ischaemic injury. in 8/14 cases with us suspect of ventriculomegaly (n=3), corpus callosum agenesis (2), cerebellar vermis hypoplasia (1), holoprosencephaly (1), spina bifida (1) mri detected ischaemic and haemorrhagic lesions unidentified at us examination. in 2/14 fetuses with us suspicion of intracerebral tissue space-occupying lesion, mri modified the diagnosis to extra-axial haematoma associated with dural sinus malformation. results were compared to fetopsy or after-birth follow up. fetal mri is an additional imaging modality in the diagnosis of cerebral ischaemic-haemorrhagic lesions and it is useful in providing further information on the extension of parenchyma injury and associated abnormalities and in improving delivery management. the contribution of mid-trimester virtual autopsy with mr imaging a. d'hondt, n. d'haene, j. rommens, m. cassart, f.e. avni; brussels/be the aim of the study was to assess the potential contribution of fetal virtopsy (post-mortem mr imaging (pm-mri)) in the second trimester of pregnancy. during a one-year period, post-mortem mr imaging (pm-mri) was performed in all fetuses who died in utero or whose pregnancy was interrupted due to major malformations. the study was performed in agreement with the local ethical committee. fetuses of <26 weeks that underwent obstetrical ultrasound and pm-mr were included. mr imaging examination was performed on a 1.5 tesla magnet with a standardized protocol. the findings on pm-mri were compared to obstetrical sonographic findings (and to pathology when available). we have analyzed separately the findings in the central nervous system and those in the rest of the fetus (chest, abdomen and skeleton). the results were classified in three categories according to the diagnostic accuracy: ultrasound>pm-mri, ultrasound=pm-mri and pm-mri>ultrasound. the us and pm-mri data of ten fetuses were analyzed. their gestational age ranged from 17.6-26 weeks and their bodyweight ranged from 160-930g. for the cns malformation: pm-mri offered a better diagnostic accuracy than us in 7 cases (70%) (e.g. agenesis of the corpus callosum and holoprosencephaly). in 3 cases (30%) us offered the same information than pm-mri. there was no case where us was more accurate than pm-mri. for the rest of the body malformations: pm-mri offered a better diagnostic accuracy in 5 cases (50%) (e.g. heterotaxy anomalies or vertebral segmentation anomalies). in 3 cases (30%), us offered the same information as pm-mri. there were 2 cases (20%) where us showed major malformations that were not diagnosed on the pm-mri (two cases of cardiac malformation). post mortem mr imaging is more accurate than obstetrical ultrasound in detecting major malformations in the cns as well as in the rest of the body. the present exceptions are cardiac malformations. the examination offers an easy evaluation of the deceased fetus. it provides, in most cases, important additional information. diffusion coefficient and perfusion fraction parameters correlate with gestational age in normal human in vivo placenta: a preliminary study a. antonelli, m. guerreri, s. bernardo, s. capuani, c. catalano, l. manganaro; rome/it to investigate the potential of diffusion parameters derived from a biexponential analysis as marker to evaluate the perfusion quality of normal in vivo placenta. eighteen normal pregnancies, fulfilling the study inclusion criteria, have been analysed at 1.5 t magnetom avanto (siemens, erlangen, germany) without mother-foetal sedation. dw imaging was collected using seven b values: 0, 50, 100, 150, 400, 700, 1000 (s/mm 2 ). three regions of interest (rois) have been considered -central (c), peripheral (p) and umbilical (u) regions. a bi-exponential model was used to obtain perfusion fraction (f), pseudo-perfusion (d*) and apparent diffusion (d) coefficients. pearson test was performed to investigate correlation between diffusion parameters and gestation weeks (gw), body mass index (bmi) and basal glycaemia (bg). the average values on all rois were d=1.41±0.16•10 -3 (mm 2 /s), d*=1.81 ±1.28•10 -2 (mm 2 /s), f=3.28±0.18•10 -1 , in good agreement with the literature. in the c roi, a positive correlation (p<0.04) was observed between f and gw. after 30 gw in the p roi a positive correlation between f and gw (p<0.05) and a negative correlation between d and gw (p<0.0001) were found. no correlation was found between d, d*, f, bmi and bg. conclusion: the f increase reflects normal placenta perfusion physiology. on the other hand, the decrease of d highlights placental parenchyma maturation becoming more fibrotic during late gestational age. bi-exponential model provides more and useful information about placental morphological changes compared to mono-exponential diffusion model. to demonstrate the diagnostic value of fetal mri in the detection of fetal central nervous system (cns) impairment in prenatally echocardiographic diagnosed congenital heart diseases. we retrospectively examined 24 fetuses between 19 gestational weeks and 33 gestational weeks who performed a fetal mri in our institution after a second-line ultrasonography, between april 2010 and october 2015. fetal heart and cns studies were performed with a 1.5 tesla magnet (siemens magnetm avanto) without maternal sedation. prenatal findings were compared to fetopsy results, fetal mri after 30 gw or postnatal mri. in our sample of 24 cases, 7/24 had interatrial septal defect (iasd),intervertricular septal defect (ivsd), and atrioventricular canal defect (cavc), 6/24 had cardiac rhabdomyomas, 3/24 had hypoplastic left heart syndrome and hypoplastic aorta, 2/24 had transposition of the great vessels, 2/24 had fallot tetralogy, 2/24 had aorta coartation and 2/24 had intracardiac masses of uncertain significance. magnetic resonance imaging was able to detect cns impairment: we recognize 11/24 corpus callosum (cc) dysgenesis (4/13 cc hypoplasia, 4/13 complete cc agenesis, 3/13 partial cc agenesis), 7/24 ventriculomegalies or hydrocephalus, 3/24 subtentorial anomalies (dandy-walker, vermian hypoplasia and vermian malrotation) and 3/24 gyration anomalies. due to the high risk of cns involvement in prenatal congenital heart diseases, it is essential to suggest an mri study of the evolving fetal brain especially in complexes forms that suggest a syndromic background. fetal mri of the cns is mandatory in the study of congenital heart disease due to the high rate of encephalic anomalies associated, particularly in iasd, ivsd and cavc. first experiences and diagnostic utility of micro-ct for fetal autopsy j.c. hutchinson 1 , x. kang 2 , s.c. shelmerdine 3 , m. cannie 2 , v. segers 2 , n. sebire 3 , j. jani 2 , o.j. arthurs 3 ; 1 newcastle upon tyne/uk, 2 brussels/ be, 3 london/uk perinatal autopsy remains poorly accepted by parents, despite yielding information that affects the management of future pregnancies in around 30% of cases. microcomputed tomography (micro-ct) has shown promising results in the examination of ex-vivo fetal organs, and may provide diagnostic imaging in cases where traditional autopsy is challenging, and s416 (2017) 47 (suppl 2):s297-s pediatr radiol existing post mortem imaging techniques (ct and mri) provide insufficient diagnostic resolution. our objective was to examine whole fetuses non-invasively using micro-ct, and compare the findings with standard autopsy as the gold standard. in this ethically approved double blinded study, terminated fetuses or miscarriages underwent iodinated micro-ct examination followed by conventional autopsy. images were acquired using a nikon xth225st microfocus-ct scanner with individual specimen image optimisation. forty indices normally assessed at perinatal autopsy were evaluated for each imaging dataset by two independent reporters and a consensus report produced. autopsies were performed blinded to the imaging findings by one of two perinatal pathologists. we examined 8 fetuses, with a gestational age range of 11-16 gestational weeks. 36/320 indices were non-diagnostic (11%), but there was agreement for 271/284 diagnostic indices (overall concordance of 95.4% (95% ci 92.3, 97.3%). in seven out of eight fetuses (87.5%), the same final diagnosis was made following micro-ct examination and autopsy examination; in one case, micro-ct was non-diagnostic. ten false negatives indices included a vsd, laryngeal anomaly, ambiguous genitalia and incomplete bowel rotation, none of which changed the overall diagnosis. three apparent false positives on micro ct were a cloacal anomaly, incidental cystic neck lesion and thymic atrophy, which were not detected at autopsy. micro-ct of early gestation whole fetuses can provide highly accurate datasets with three-dimensional renderings of complex disease processes. this approach confirms the potential of this technology for non-invasive examination of small fetuses. investigation of perinatal body organ diffusion-weighted post mortem mri s.c. shelmerdine 1 , m. cheryl 2 , j.c. hutchinson 1 , n. sebire 1 , o.j. arthurs 1 ; 1 london/uk, 2 southampton/uk objective: diffusion weighted magnetic resonance imaging (dwi) uses water molecule diffusion to generate mr contrast images, and can reveal microstructural or functional changes in tissues, quantified by measuring the apparent diffusion coefficient (adc). the application of dwi to the post mortem setting is appealing as it does not require the administration of an exogenous contrast agent. a recent pilot study of 15 paediatric cases suggested that lung adc values at pm mri may be a useful marker of post mortem interval (time since death; pmi) which has forensic relevance, but other body organs have not been comprehensively evaluated. the aim of this study was therefore to evaluate the relationship between pmi and body organ adc values in a larger cohort of subjects across a wider gestational range in the setting of perinatal death. whole body perinatal postmortem mri with dwi sequences were performed at 1.5t, with b values of 0, 500, 1000 mm 2 /s. mean adc values were calculated from regions of interest (rois) placed in the lungs, myocardium, spleen, renal cortex, liver and psoas muscle. the values were measured by two independent readers, correlated against gestational age and post mortem interval (pmi) using the pearson product-moment correlation coefficient. bland-altman plots were created, and the limits of agreement used to assess the inter-observer agreement of mean adc values. results: eighty fetal deaths and stillbirths were imaged with mean gestational age 31.5 weeks (range: 20 -41 weeks). the mean pmi was 8.9 days (range 2-20 days). there was a weakly positive correlation between pmi and mean lung adc (r 2 =0.03) and spleen adc (r 2 =0.08). no correlation was found with between adc and pmi for the other body organs. there was reasonable inter-observer agreement between the two readers, with mean adc difference 11.8 mm 2 /s (+/-135.1 mm 2 /s). perinatal lung and splenic adc values show a mild increase with increasing pmi. together with other imaging parameters, this may be useful to evaluate organ-specific changes which occur in the post mortem period, particularly in a forensic setting. further research is needed to understand the organ-specific changes which occur in the post-mortem period. usefulness of combined grey-scale and color doppler ultrasonography(us) findings in the evaluation of acute pyelonephritis in children k. lee, j.h. lee; anyang/kr objective: us diagnosis of apn in children can give a valuable information to the clinicians for the early treatment. but the problem of us in the diagnosis of apn is wide range of sensitivity, which is 11-69%. the purpose of this presentation is to evaluate the usefulness of grey-scale us and color doppler us in the diagnosis of acute pyelonephritis in children. from march 2007 to february 2014, 154 children(308 kidneys), 108 boys and 46 girls, aged 2 weeks to 9 years (mean age, 7.7 months) underwent kidney us as an initial diagnostic tool for acute pyelonephritis and follow up dmsa scintigraphy within a week. criteria for acute pyelonephritis on grey-scale image were focal/diffusely increased/decreased echogenicity or loss of corticomedullary differentiation. on color doppler sonography, the criterion was decreased color flow. we classified the us diagnosis of apn into 4 categories. definite, suggestive, possible and normal. when above two grey-scale us criteria and color doppler us criterion are seen, we classified it as 'definite'. when one of greyscale us and color doppler us finding are seen, it was classified as 'suggestive' of apn. 'possible' apn was abnormal finding either on grey-scale or color doppler us. 'normal' was no abnormal findings on grey-scale and color doppler us. we compared above findings with dmsa scan, which is considered as gold standard for diagnosing apn. statistical analysis was performed on all 308 kidneys. the overall sensitivity of our study was 68%(88/129) and specificity was 72%(123/177). the positive predictive value for each definite, suggestive, possible groups were 86%, 60%, and 38% respectively. the negative predictive value for normal group was 75%, which means the false ppv was 25%. the p-value of the definite and suggestive was statistically significant, but the possible was statistically insignificant. in the diagnosis of apn in children, abnormal us finding either on greyscale or color doppler us is not optimal. abnormal us findings both grey-scale us and color doppler us showed good association with dmsa scan and statistically significant. combined grey-scale and color doppler us findings can give a more reliable information in the diagnosis of apn in children. the greater degree of gastric and/or duodenal wall thickening and increased echogenicity are helpful sonographic features in differentiating congenital duodenal anomalies from malrotation. our findings confirm the superiority of us vs ugi for evaluation of duodenal obstruction in neonates and evaluation of gastric and duodenal wall must be added to the constellation of other features to be assessed on us examinations. a measure of renal morphology as an indicator for potential renal failure a.c. eichenberger, p. grehten, c. kellenberger; zurich/ch this study introduces a measure of renal morphology, herein labelled split renal volume (srv), that should be applied as an indicator for potential renal failure and eventual surgical treatment of obstructive uropathy in children. current practice applies dynamic contrast enhanced functional renal imaging (fri) with complex post-processing methods. fri generates a measure of split renal function (srf). reduced values of srf under 45% are currently considered to be an indicator for surgical treatment. this retrospective study compares the accuracy of srv with the accuracy of srf as methods for assessing potential renal failure. materials: srv is a quotient of volumetric measurements. total renal volume is described by the sum of parenchymal volume and intra-renal collecting system volume. srv is designated in this study as the quotient of two ratios: first, the ratio of total renal volume to parenchymal volume of the left kidney; and second, the ratio of total renal volume to parenchymal volume of the right kidney. twenty-two children were studied: 16 (age 3.1±4.5y) with unilateral asymptomatic intrinsic uretero-pelvic-junction obstruction (upjo), and 6 normal controls (age 6.6±4.0y). all subjects underwent mr urography at 1.5t, which provided estimates of srf and srv for each of the 44 examined kidneys. the sensitivity and specificity of both srf and srv for predicting surgical management were determined by comparing the indicators with an expert review panel's decision to operate. the panel was blinded to values of srv. results: when a cut-off value of 45% for srf was used, the resultant sensitivity and specificity of srf for the detection of kidneys at risk were found to be 44% and 86%. the values of srv ranged between 0.3 and 3.5. it was found that a value greater than 1.1 indicated kidneys at risk. when the cut-off value of 1.1 for srv was used, the resultant sensitivity and specificity of srv for the detection of kidneys at risk were both 100%. in this small population, srv proved to be 100% accurate and is superior to srf for detecting kidneys at risk of failure due to obstruction. routine application of srv promises to simplify mr urography by obviating dynamic contrast enhanced imaging studies. further prospective studies are necessary in order to select an optimal cut-off value of srv. factors that can distort the dj flexure mimicking malrotation v. bhalla 1 , s. mohan 2 , k.a. bradshaw 2 , m. thyagarajan 2 ; 1 stoke-on-trent/uk, 2 birmingham/uk to highlight the varied radiological appearances and position of the duodenal-jejunal flexure in children and to discuss its importance in assessing for malrotation materials: retrospective analysis of the multiple fluoroscopic examinations performed in the assessment for malrotation over the past 5 years in a busy tertiary centre results: the classic position of the dj flexure is to the left of left pedicle of l1 and at the level of the duodenal bulb on frontal views and posterior (retroperitoneal) on lateral views. however variations of the normal location can appear, particularly on frontal views, in the upper gi series that can mimic malrotation which has shown to be more common in neonates. we present cases with examples to illustrate the variability in position due to various causes and its implications in the diagnosis of malrotation and volvulus. our case mix includes patients with excessively distended stomachs, large bowel obstruction, renal pelvic dilatation, repeated naso-jejunal and gastro-jejunal tube insertion and in patients post liver transplantation. malrotation and its assessment have serious management and prognostic implications. this presentation demonstrates that the imaging features can be varied, and knowledge about factors distorting the position of the dj flexure is vital in the accurate management of neonates presenting with bilious vomits. retrospective study of prospectively collected data performed at a single tertiary paediatric institution over a 16.5 year period. a total of 1000 consecutive patients, aged <18 years, were reviewed who underwent native renal biopsy. all biopsies were performed within the interventional radiology department. all patients had renal disease requiring a renal biopsy for diagnosis. outcome measures include technical success, early and late complications and the adequacy of histological samples. in addition, age, body weight, glomeruli number, histological data, number of cores, size of the biopsy needle, use of co-axial needle and the rate of tract embolisation/plugging were recorded. results: from september 1999 to april 2016, 1000 patients (mean age: 10.07 years +/-4.65; range 0.14 -18.0 years) underwent native renal biopsy. one hundred ninety-one patients were <5 years of age. nine hundred forty-six patients (94.6%) had a biopsy of the right kidney, 53 patients (5.3%) had a biopsy of the left kidney and 1 patient (0.1%) had a biopsy of a horseshoe kidney. five hundred fifteen patients were female (51.5%). seven hundred sixtynine patients (76.9%) had the procedure performed under general anaes-thetic and 227 of patients (22.7%) had the procedure performed under local anaesthetic (+/-sedation/entonox). mean number of passes of the core biopsy needle through the renal capsule was 2.7. a 16 gauge core biopsy needle was used in 100% of the patients. 81.9% of the patients had three or less passes of the biopsy needle though the renal capsule. the overall complication rate was 2.6% (n= 26). 1.2% (n= 12) of patients had a non-diagnostic biopsy. fifty-five patients underwent a post biopsy ultrasound due to clinical concerns. twenty patients developed perinephric haematoma (19 were treated conservatively; one underwent embolisation and subsequent nephrectomy). four patients developed arteriovenous fistulas. two patients developed post procedure infections (one at the skin site and one a perinephric collection). histology results were reviewed in all patients. the mean number of glomeruli obtained was 26.97 (range 2-87). glomerulonephritis was the most common histological diagnosis (n=464; 46.4%) conclusion: renal biopsy is an extremely useful diagnostic tool for renal disease. there is no published data of this size assessing the outcome of native renal biopsies in the paediatric population. jr usa 1979 a. lassrich germany 1979 j. sauvegrain france 1982 c. fauré france 1982 a. giedion switzerland 1983 e. willich germany 1984 r. astley united kingdom ringertz sweden 1994 d.g. shaw united kingdom 1996 r. lebowitz usa 1996 b. lombay hungary pena spain gold medallists london/united kingdom the dutch group of paediatric radiologists, the hague/the netherlands 1981 g. stake ringertz (espr) & d. kirks (spr) chicago/united states future espr meeting italy european courses of paediatric radiology (ecpr) genoa/italy (neuroradiology) 2000 r.fotter, graz/austria (abdomen) brussels/belgium (thorax) 2006 j-n. dacher paediatric musculoskeletal imaging) references: 1. 2014-stellungnahme-lnt-modell.pdf [internet]. [zitiert 2 an evaluation of paediatric projection radiography in ireland contrast imaging -> application -dectris background ionizing radiation and the risk of childhood cancer: a census-based nationwide cohort study best practices in digital radiography communicating radiation risks in paediatric imaging kinderradiologie-besonderheiten des strahlenschutzes diagnostic imaging and ionizing radiation -canadian nuclear safety commission epidemiology without biology: false paradigms, unfounded assumptions, and specious statistics in radiation science (with commentaries by inge schmitz-feuerhake and christopher busby and a reply by the authors) european guidelines for ap/pa chest x-rays: routinely satisfiable in a paediatric radiology division? eurosafe imaging together -for patient safety image gently campaign back to basics initiative: ten steps to help manage radiation dose in pediatric digital radiography hostens j, u. a. in-vivo dark-field and phase-contrast x-ray imaging safety commission cn. linear-non-threshold model optimisation of paediatric chest radiography optimizing digital radiography of children radiation exposure in diagnostic imaging: wisdom and prudence, but still a lot to understand radiation shielding for diagnostic radiology strahlenhygienische aspekte bei der röntgenuntersuchung des thorax the image gently pediatric digital radiography safety checklist: tools for improving pediatric radiography the standardized exposure index for digital radiography: an opportunity for optimization of radiation dose to the pediatric population gastroenterology and radiology records were searched to identify ibd patients with colonic strictures. all patients underwent an mre within 3 months of colonoscopy. the following colonic parameters were evaluated: bowel wall thickening with luminal narrowing, pre-stenotic bowel dilatation, bowel wall enhancement, and diffusion restriction (if performed). colonoscopy and operative notes were correlated. results: fourteen patients met the inclusion criteria, one with 2 colonic strictures. bowel wall thickening with luminal narrowing at the site of the reported stricture was present in all cases. pre-stenotic bowel dilatation (>3.0 cm) proximal to the reported stricture was present in 11/15 cases. using luminal narrowing and prestenotic dilatation as criteria for diagnosis of a colonic stricture, 11/15 cases were therefore positive on mre. when comparing to colonoscopy, mre diagnosed colonic strictures in 8/12 cases (67%). in the six patients who had surgery, mre accurately diagnosed colonic strictures in 5/6 cases (83%). conclusion: mre is not the primary modality for colonic evaluation, yet diagnosing colonic pathology on mre, particularly strictures, may be beneficial for the referring gastroenterologist in the assessment of these patients. potential strictures on colonoscopy did not agree with mre in all cases, but when correlating with surgery 83% of colonic strictures were accurately diagnosed in a small subset. although mre is not optimized for the evaluation of the colon, colonic strictures can be recongnized in children with crohn's disease.disorders of sexual differentiations in neonates: standardized sonographic evaluation and proposal of a reading grid h. lerisson, e. amzallag -bellenger, f.e. avni, m. cartigny; lille/fr to propose a systematic and structured sonographic approach in neonates with disorders of sexual differentiation (dsd) materials: review of the us pelvic, external genital and adrenal findings in 20 consecutive patients with clinical suspicion of dsd evaluated in the neonatal period. the us survey included: the uterus (absent or visible -with or without hormonal impregnation), the vagina (absent or present (complete or partial)), the gonads (ovaries, testis or unsetermineddysgenetic ) as well as the adrenals (normal, too small or enlarged). the us conclusions were correlated with the endocrinological and genetical work-up of each patient results: twenty cases of dsd have been included us had correctly identified the presence of a uterus in 11 patients. there was one false positive case; 6 among the 11 patients did not show the physiological hormonal impregnation. the 5 vaginal anomalies were correctly evaluated. the gonads were defined correctly as normal testis in 6 patients, normal ovaries in 4 and dysgenetic gonads in 4. they could not be visualized in 6 patients. adrenals were considered normal in 17 patients (one false negative), hypertrophied in 2 and small in one patient. to compare hepatic 2d shear wave elastography (2d swe) in children between free-breathing and breath-hold conditions, in terms of measurement agreement and time expenditure. a cohort of 57 children (12.7±4.3 years) who underwent standardized 2d swe between may and october 2015 were retrospectively evaluated. liver elastograms were obtained under free-breathing and breath-hold conditions and time expenditure was measured. median stiffness, interquartile range (iqr), and iqr/median ratio were calculated based on 12, six, and three elastograms. results were compared using pearson correlation coefficient, intraclass correlation coefficient (icc), bland-altman analysis, and student's t. median liver stiffness under free-breathing and breath-hold conditions correlated strongly (7.22±4.5kpa vs. 7.21±4.11kpa; r=0.97, p<0.001). time to acquire 12 elastograms with free-breathing was lower than that with breath-holding (79.3±32.5sec vs. 143.7±51.8sec, p<0.001). results for median liver stiffness based of 12, six, and three elastograms demonstrated very high agreement for free-breathing (icc 0.993) and for breath-hold conditions (icc 0.994). hepatic 2d swe performed with free-breathing yields results similar to the breath-hold condition. with a substantially lower time requirement, which can be further reduced by lowering the number of elastograms, the free-breathing technique may be suitable for infants and less cooperative children not capable of breath-holding. abstract: pelvi-ureteric junction obstruction (pujo), classified into intrinsic and estrinsic is one of the most frequent urological diseases affecting the pediatric population. extrinsic causes include the presence of crossing vessels, kinks or adhesions. in cases with extrinsic obstruction of puj, colour doppler ultrasound (cd-us) can detect the presence of crossing vessels. in presence of crossing vessels pyeloplasty or vascular hitch can be performed. the aim of the study is to analyze the sensitivity of cd-us and magnetic resonance urography (mru) in visualizing crossing vessels in extrinsic pediatric hydronephrosis in order to decide the correct diagnostic pathway and evaluate in the pre-operative phase which surgical technique and approach (open, laparoscopic or robotic) is the ideal to be performed. a retrospective review of medical records for patients who underwent surgical treatment for hydronephrosis from august 2006 to february 2016 was performed. a descriptive statistical analysis was performed. the presence of crossing vessels at surgery was considered the gold standard. the sensitivity was calculated for both the imaging techniques as a measure of accuracy, evaluating the ratio between the positive cases divided by the those with aberrant vessels identified at surgery. results 220 clinical charts were reviewed. crossing vessels identified at surgery were 73 (33,2% of pujo). the median age was higher in the group with crossing vessels compared to the group without crossing vessels (p< 0,0001). the sensitivity of cd-us was higher compared to mru (93,3% vs 71,7%). before the surgical time knowing which technique and approach have to be managed in hydronephrotic patients with crossing vessels could be very important. according to our preliminary datacollection cd-us has got a higher sensitivity and could be the gold standard technique. study limitations include the absence of specificity, positive and negative predictive values. in the future it could be useful to perform a double blind trial in which children with moderate-severe hydronephrosis will be subjected to both imaging techniques to evaluate not only the sensitivity, but also the specificity, the positive predictive value and the negative predictive value conclusion: conclusions in the pre-operative phase, cd-us could be sufficient for the surgeon to discern between pujo with the presence or the absence of crossing vessels, as it has a higher sensitivity and lower costs compared to mru.urosonography -nonradiant alternative for voiding cystourethrography o.m. fufezan, c.a. asavoaie, s. tatar; cluj-napoca/ro voiding cystourethrography (vcug) was considered the gold standard in the diagnosis and monitoring of vesicoureteric reflux (vur). this method is invasive due to the radiation exposure. in the present the diagnosis of vur can also be established by contrast ultrasound examination, also known as voiding urosnography (vus). the authors will present the role of vus in the diagnosis and grading of the vur and the role of patient position in the detection of vur. the infants and children with congenital anomalies of the urinary tract and/or urinary tract infection have been evaluated with vus. iatrogenic vur, neurogenic bladder and urogenital sinus anomalies were excluded. the presence and the degree of the vur were evaluated. vus has been performed using a protocol similar to the one used for vcug. in conditions of sterile urine, 0.5 ml sonovue and saline solution have been introduced into the bladder until voiding started. the patients were examined both in a supine and an upright position and the following structures have been scanned: urinary bladder, distal part of the ureters and both pelvicaliceal systems during bladder filling, during and after voiding. the visualisation of the ultrasound contrast agent in the upper urinary tract represented a positive vur diagnosis. the grading of the vur has been established based on the same criteria as in vcug. sixty five patients (130 renal units), ages between 2 weeks and 17 years were evaluated (median age ± sd: 3 years ± 4 years and 2 months) through vus. vcug was performed in 6 patients in a maximum of 48 hours after vus. vur has been identified in 35 patients (40.7% renal units). a high vur grade (iv-v) was identified in 19.2% of renal units. for the patients investigated with both methods, the results were concordant in 4 patients. in two patients vur has not been identified by vcug, but was detected during vus. the upright position (in addition to decubitus) revealed vur in 3 renal units in which the reflux was not detected in decubitus. conclusion: vus is extremely useful and reliable in diagnosing and grading vur in pediatrics. the changing of the patient position during examination can improve vur detection.new sonographic features useful in differentiating congenital duodenal anomalies from malrotation: gastric and duodenal wall thickening and hyperechogenicity p. caro dominguez 1 , s. hameed 2 , a. zani 3 , r. moineddin 3 , o.m. navarro kunstmann 3 , a. daneman 3 ; 1 cordoba/es, 2 london/uk, 3 toronto/ca the clinical and plain radiographic differentiation of congenital duodenal anomalies (atresia, web, stenosis) and intestinal malrotation is not always clear. although sonography has been documented as an important diagnostic tool to differentiate these two entities, its role is still not widely appreciated. the purpose of this study was to assess the sonographic features of the gastric and duodenal wall in a large series of neonates with congenital duodenal obstruction as these have not been reported previously. neonates who had surgically proven congenital duodenal anomalies or malrotation were identified from the surgical database in a tertiary pediatric hospital in a period of 15 years (2000) (2001) (2002) (2003) (2004) (2005) (2006) (2007) (2008) (2009) (2010) (2011) (2012) (2013) (2014) (2015) . those with an ultrasound performed within 48 hours of surgery were included in the study. imaging was retrospectively and independently reviewed by two readers in chronological order blinded to final diagnosis. a wall thickness of ≥ 2 mm of a distended loop was considered abnormal. hyperechogenicity was recorded when the wall of the stomach or duodenum was brighter than liver or splenic parenchyma. imaging findings in the group with congenital duodenal anomalies was compared to the group with malrotation using fisher's exact test. one hundred eight neonates were included in the study, 40 with a congenital duodenal anomaly, 49 with malrotation (36 with volvulus) and 19 with both. ugi was performed in 61 neonates who had us. the correct diagnosis was provided only by us in 24 of these 61 newborns (39%), only by ugi in 5 (8%), by both in 26 (43%) and by neither in 6 (10%). ugi was performed in 21 children with malrotation and volvulus, eight were diagnosed only by us, four only by ugi and nine by both. the gastric and/or duodenal wall was significantly thicker and more hyperechoic in neonates with congenital duodenal anomalies than those with malrotation (p<0.0001) [fig 1, table 1 ]. conversely an abnormal relationship between the superior mesenteric artery and vein, abnormal position of the third part of the duodenum and the whirlpool sign were found more commonly in neonates with malrotation than those with congenital anomalies (p<0.0001).