brain_3_1 19 An Optimized Clustering Approach for Automated Detection of White Matter Lesions in MRI Brain Images M. Anitha Department of Computer Science and Engineering, Sri Krishna College of Technology, Coimbatore, India anitha_murugaiyan@yahoo.com Prof. P. Tamije Selvy (SG) Department of Computer Science and Engineering, Sri Krishna College of Technology, Coimbatore, India tamijeselvy@gmail.com Abstract: Settings White Matter lesions (WMLs) are small areas of dead cells found in parts of the brain. In general, it is difficult for medical experts to accurately quantify the WMLs due to decreased contrast between White Matter (WM) and Grey Matter (GM). The aim of this paper is to automatically detect the White Matter Lesions which is present in the brains of elderly people. WML detection process includes the following stages: 1. Image preprocessing, 2. Clustering (Fuzzy c-means clustering, Geostatistical Possibilistic clustering and Geostatistical Fuzzy clustering) and 3.Optimization using Particle Swarm Optimization (PSO). The proposed system is tested on a database of 208 MRI images. GFCM yields high sensitivity of 89%, specificity of 94% and overall accuracy of 93% over FCM and GPC. The clustered brain images are then subjected to Particle Swarm Optimization (PSO). The optimized result obtained from GFCM-PSO provides sensitivity of 90%, specificity of 94% and accuracy of 95%. The detection results reveals that GFCM and GFCM- PSO better localizes the large regions of lesions and gives less false positive rate when compared to GPC and GPC-PSO which captures the largest loads of WMLs only in the upper ventral horns of the brain. Keywords: Geostatistics, fuzzy clustering, particle swarm optimization, possibilistic clustering, white matter changes 1. Introduction Medical imaging is the technique used to create images of the human body for clinical or medical science that produce images of the internal aspect of the body. Magnetic Resonance Imaging (MRI) is one of the medical imaging techniques. MRI of brain is highly sensitive for detecting all forms of White Matter abnormalities. Non-specific changes in the White Matter appear frequently on MRI in elderly patients presenting with either stroke or cognitive impairment. In general, human brain consists of main components namely, White Matter (WM), Grey Matter (GM) as shown in figure 1. Figure 1.WM and GM of brain BRAIN. Broad Research in Artificial Intelligence and Neuroscience Volume 3, Issue 1, February 2012, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print) 20 Neuronal tissue containing mainly long, myelinated axons is known as White Matter. Closely packed neuron cell bodies form the Grey Matter. Grey Matter is in grey color because of the grey nuclei that comprises the cells. Myelin is responsible for the white appearance of White Matter. White Matter Lesions (WMLs) are commonly found in patients with Multiple Sclerosis (MS), Cerebrovascular Disease (CVD), stroke, and other neurological disorders. It is believed that the total volume of the lesions and their progression relate to the aging process as well as disease process. Therefore, quantification of White Matter Lesions is very important in understanding the aging process and diagnosis and assessment of these diseases. 1.1. Automated Quantification of WML It Non-specific changes in the White Matter appear frequently on CT and MRI in elderly patients presenting with either stroke or cognitive impairment, but are also commonly seen in healthy elderly individuals. Evaluation of WMLs in MRI is conventionally performed using skill and knowledge of experts [2]. This manual assessment on WML results in different ratings [1], [3], which make it non reproducible and difficult for a general agreement. Manual assessment of WM lesions is not only time consuming but also shows high inconsistency among human raters. To overcome this drawback of inaccurate prediction, clustering models like Fuzzy-set, Possibilistic and Geostatistic frameworks are proposed for automated detection of White Matter changes. The proposed clustering models are derived by extending the objective functions of FCM and Possibilistic approach with a Geostatistical (spatial) model. The clustered images are further optimized using Particle Swarm Optimization (PSO) in order to obtain the most optimist solution. Particle Swarm Optimization (PSO) is a population based optimization tool which provides the most optimist solution. 2. Related Works Anbeek et al. [4] proposed k-nearest neighbors algorithm (k-NN) for automatic segmentation of WMLs. This is a supervised learning method and used the information from T1-weighted, inversion recovery (IR), proton density-weighted (PD), T2-weighted, and fluid attenuation IR (FLAIR) scans in order to estimate the probability of voxels. T1-weighted scans are a standard basic scan, in particular differentiating fat from water - with water darker and fat brighter. This is one of the basic types of MR contrast and is a commonly run clinical scan. T2 - weighted scans are another basic type. Like the T1-weighted scan, fat is differentiated from water - but in this case fat shows darker, and water lighter. For example, in the case of cerebral and spinal study, the CSF (cerebrospinal fluid) will be lighter in T2-weighted images. By combining the results of these techniques, binary segmentation results are obtained from the selected threshold values, and therefore the relation between an optimal threshold and lesion volume was separately chosen for each patient. A probability mixture model and the Bayesian classifier was used by Khayati et al. [5] inorder to extract normal tissue, abnormal tissue and cerebrospinal fluid (CSF) which serves primary purpose like buoyancy, protection and chemical stability. Normal tissue refers to White Matter and Grey Matter of brain whereas abnormal tissue refers to lesions of brain in FLAIR-MR images. This method does not focus on the lesions of small size or irregular shape. Lao et al. [6] proposed an approach for segmenting WML in which support-vector machine (SVM) classifier was used in order to classify new scans, and post processing analysis was carried out to eliminate false positives. The strength of SVM-based classifiers is the ability to separate overlapping features, but selecting effective features for classifying a particular difficult problem is one of the key issues in pattern classification that should be first identified, where the results are based on only expert-defined information. This method is less accurate. Therefore WML image intensities cannot be visually distinguished. Lesions are irregular voxels that do not belong to GM, WM and CSF and can be classified as outliers in grey and White Matter regions. This method was proposed by Seghier et al. [7]. After M. Anitha, P. Tamije Selvy - An Optimized Clustering Approach for Automated Detection of White Matter Lesions in MRI Brain Images 21 combining the segmentation and normalization of images, fuzzy clustering was applied to identify outlier voxels as lesions in normalized grey and White Matter segments. Spatial smoothing was done using Gaussian kernel, which affect the sensitivity and specificity of the method. Hernandez et al. [8] presented a multispectral MRI approach for segmenting normal and abnormal brain tissue. The procedure was carried out by combining pairs of different MRI sequences and modulated them in the red-green color space to enhance the tissue discrimination. Fully automated method for CSF, GM and WM segmentation based on multimodal MRI data is optimized and extended with WML segmentation was proposed by R. de Boer, H. A. Vrooman, F. van der Lijn [9]. V.K.Panchal1, Harish Kundra, Ms. Jagdeep Kaur [10] presented a comparative study of Particle Swarm Optimization (PSO) and Unsupervised Clustering Techniques. In order to overcome the shortcomings of traditional clustering algorithms such as local optima and sensitivity to initialization, a new Optimization technique, Particle Swarm Optimization is used in association with Unsupervised Clustering techniques. Hesam Izakian, Ajith [11] Abraham proposed hybrid fuzzy c-means and fuzzy-PSO. FCM algorithm is integrated with FPSO algorithm to form a hybrid clustering algorithm called FCM- FPSO which maintains the merits of both FCM and PSO algorithms. FCM-FPSO algorithm applies FCM to the particles in the swarm every number of iterations/generations such that the fitness value of each particle is improved. Dian Palupi Rini, Siti Mariyam Shamsuddin and Siti Sophiyait Yuhaniz [12] proposed a basic variant of Particle Swarm Optimization, in order to improve speed of convergence and quality of solution found by PSO. Saeed Vaneshani and Hooshang Jazayeri-Rad [13] proposed a method for finding the optimum membership functions of a fuzzy system using particle swarm optimization (PSO) algorithm. A synthetic algorithm combined from fuzzy logic control and PSO algorithm is used to design a controller for a continuous stirred tank reactor (CSTR) with the aim of achieving the accurate and acceptable desired results. Renske de Boer, Michiel Schaap, Fedde van der Lijn, Henri A. VroomanMarius de Groot, Aad van der Lugt, Arfan Ikra Meike, W. VernooijMonique, M.B. Breteler Wiro and J. Niessen [14] presented a framework for the construction of weighted structural brain networks, containing information about connectivity, which can be effectively analyzed using statistical methods. 3. Proposed Scheme This paper mainly focuses on automated detection of White Matter Lesions of brain using fast and efficient clustering algorithms. The goal of clustering a medical image is to simplify the representation of an image into a meaningful image and makes it easier to analyze. As a first step, MRI brain image is pre-processed using Contrast Stretching technique which is one of the efficient image enhancement techniques. The pre-processed image is subjected to clustering. The clustering algorithms include Fuzzy c-means Clustering (FCM), Geostatistical Possibilistic Clustering (GPC) and Geostatistical Fuzzy Clustering Model (GFCM). However clustering techniques are sensitive to initialization and are easily trapped in local optima. In order to obtain an optimized result, the clustered images are undergone optimization. Particle swarm optimization (PSO) is a stochastic global optimization tool which is used in many optimization problems. Figure 2 represents overall process of automatic detection of WMLs of brain. Since MS lesions present different characteristics from lesions in elderly individuals there are many clustering models to determine the accuracy but those methods are not directly applicable to predict the accurate lesions because of the decreased contrast between White Matter and Grey Matter in elderly people. The proposed clustering models are derived by extending the objective functions of FCM and Possibilistic clustering with a Geostatistical (spatial) model. These algorithms are applied to real magnetic resonance images and is shown to be more robust to noise and other artifacts than competing approaches. BRAIN. Broad Research in Artificial Intelligence and Neuroscience Volume 3, Issue 1, February 2012, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print) 22 Figure 2. Overall process of the system 3.1. Image Enhancement (Pre-processing) The procedure done before processing by correcting image from different errors is preprocessing. Image enhancement is one of the image preprocessing techniques. The aim of image enhancement is to improve the interpretability or perception of information in images for human viewers, or to provide `better' input for other automated image processing techniques. It consists of collection of techniques that seek to improve the visual appearance of an image or to convert the image to a form better suited for analysis by a human or machine [15]. Contrast stretching is the image enhancement technique that is commonly used for medical images. Contrast stretching process plays an important role in enhancing the quality and contrast of medical images. Different types of contrast stretching techniques include local contrast stretching, global contrast stretching, partial contrast stretching, bright and dark contrast stretching. Among these techniques, bright contrast stretching is imposed on the brain image. 3.2. Fuzzy C-Means Clustering (FCM) Fuzzy c-means has been a very important tool for image processing in clustering objects in an image [16]. Fuzzy c-means (FCM) clustering is an unsupervised method derived from fuzzy logic that is suitable for solving multiclass and ambiguous clustering problems. Fuzzy c-means (FCM) clustering is an unsupervised technique that has been successfully applied to feature analysis, clustering, and classifier designs in fields such as astronomy, geology, medical imaging, target recognition, and image segmentation. An image can be represented in various feature spaces, and the FCM algorithm classifies the image by grouping similar data points in the feature space into clusters. This clustering is achieved by iteratively minimizing a cost function that is dependent on the distance of the pixels to the cluster centers in the feature domain. There are many acceleration techniques for FCM; there are very large data versions of FCM that utilize both progressive M. Anitha, P. Tamije Selvy - An Optimized Clustering Approach for Automated Detection of White Matter Lesions in MRI Brain Images 23 sampling and distributed clustering; there are many techniques that use FCM clustering to build fuzzy rule bases for fuzzy systems design; and there are numerous applications of FCM in virtually every major application area of clustering. FCM clustering algorithm is used to calculate the minimization of the fuzzy objective function [11]. It works by assigning membership to each data point corresponding to each cluster center on the basis of distance between the cluster center and the data point. The fuzzy clustering of objects is described by a fuzzy matrix µ , with n rows and c columns in which n is the number of data objects and c is the number of clusters. µ ij, the element in the ith row and jth column in µ , indicates the degree of association or membership function of the ith object with the jth cluster. The characters of µ are as follows: µ ij [0, 1], i=1, 2…., n, j=1, 2…., c (1) c where, ∑µ ij =1, i=1, 2…., n (2) j=1 0<∑ µ ij 1) is a scalar termed the weighting exponent and controls the fuzziness of the resulting clusters and '||xi – vj||' is the Euclidean distance between i th data and j th cluster center. The zj, centroid of the jth cluster, is obtained using (5). n Zj = ∑ (µ ij) m xi n (5) i=1 ∑ (µ ij) m i=1 3.2.1. FCM Algorithm S1 Randomly select ‘c’ cluster centers. S2 Compute the Euclidean distance, ||xi-vj||. S3 Calculate the fuzzy membership according to the constraints of Eq. (1), (2) and (3). S4 Calculate the fuzzy center according such that (5). S5 Repeat steps 2) and 3) until the minimum ‘J’ value is incorporated, such that (4). 3.3. Geostatistical Possibilistic Clustering (GPC) Although FCM is a very useful clustering method, its memberships do not always correspond well to the degree of belonging of the data [17], and may be inaccurate in a noisy environment. To improve this weakness of FCM and to produce memberships that have a good explanation for the degree of belonging for the data, Possibilistic approach was proposed. It is a variation over fuzzy clustering where the membership to clusters can be seen as a degree of typicality membership matrix U, uih ∈ [0, 1]. Possibilistic clustering algorithms prove the fact that it can be applied for one cluster at a time. N c N c BRAIN. Broad Research in Artificial Intelligence and Neuroscience Volume 3, Issue 1, February 2012, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print) 24 JGP (U, v) = ∑ ∑ (uij) m [d(xi, vj)] 2 + ∑ 1/(ej) 2 ∑ (1 – uij) m (6) i=1 j=1 j=1 i=1 where, (ej ) 2 is the kriging (geostatistical) variance of estimating vj using xi, i = 1…N-1. 3.3.1. GPC Algorithm S1 Randomly select ‘c’ cluster centers according to (5). S2 Calculate the possibilistic membership. S3 Calculate the spatial variability. S4 Incorporate spatial variability into objective functions as in (6). S5 Minimize the objective functions (a small value of difference) then stop. 3.4. Geostatistical Fuzzy c-means Clustering (GFCM) The fuzzy C-means objective function is generalized to include a spatial penalty on the membership functions. The fuzzy C-means algorithm (FCM) has been utilized in a wide variety of image processing applications such as medical imaging and remote sensing. Its advantages include a straightforward implementation, fairly robust behavior, applicability to multichannel data, and the ability to model uncertainty within the data. A major disadvantage of its use in imaging applications is that FCM does not incorporate information about spatial context, causing it to be sensitive to noise and other imaging artifacts. Therefore geostatistical fuzzy clustering is proposed. The advantages of the new method are the following: (1) it yields regions more homogeneous than those of other methods, (2) it removes noisy spots, and (3) it is less sensitive to noise than other techniques. It is derived by extending the into the FCM objective function. Clustering is a two-pass process at each iteration [18]. The first pass is the same as that in standard FCM to calculate the membership function in the spectral domain. In the second pass, the membership information of each pixel is mapped to the spatial domain, and the spatial function is computed from that. The FCM iteration proceeds with the new membership that is incorporated with the spatial function. The iteration is stopped when the maximum difference between two cluster centers at two successive iterations is less than a threshold. A distinctive observation of the incorporation of the geostatistical modeling into the fuzzy clustering is that it is able to accurately detect the WMLs as the regions of interest of an elderly population and provides the bidirectional association between depression and vascular disease. The main aim of is to minimize the objective function, where kriging variance is incorporated It is a derived function. N c N c n c JGF(U, v)=∑∑(uij) m [d(xi,vj)] 2 +∑(uij) m ∑(ej) 2 -∑λi (∑uij-1) (7) i=1j=1 i=1 j=1 i=1 j=1 3.4.1. GFCM Algorithm S1 Randomly select ‘c’ cluster centers. S2 Calculate the fuzzy membership. S3 Calculate the spatial variability. S4 Incorporate spatial variability into objective function such that (7). S5 Minimize the objective functions by setting a Lagrangian function. 3.5. Particle Swarm Optimization (PSO) Particle swarm optimization is a heuristic global optimization method. It is developed from swarm intelligence and is based on the research of bird and fish flock movement behavior [12]. While searching for food, the birds are either scattered or go together before they locate the place where they can find the food. While the birds are searching for food from one place to another, there is always a bird that can smell the food very well, that is, the bird is perceptible of the place where the food can be found, having the better food resource information. Because they are M. Anitha, P. Tamije Selvy - An Optimized Clustering Approach for Automated Detection of White Matter Lesions in MRI Brain Images 25 transmitting the information, especially the good information at any time while searching the food from one place to another, conduced by the good information, the birds will eventually flock to the place where food can be found. As far as particle swam optimization algorithm is concerned, solution swam is compared to the bird swarm, the birds’ moving from one place to another is equal to the development of the solution swarm, good information is equal to the most optimist solution, and the food resource is equal to the most optimist solution during the whole course. The most optimist solution can be worked out in particle swarm optimization algorithm by the cooperation of each individual. The particle without quality and volume serves as each individual, and the simple behavioural pattern is regulated for each particle to show the complexity of the whole particle swarm. This algorithm can be used to work out the complex optimist problems. Due to its many advantages including its simplicity and easy implementation, the algorithm can be used widely in the fields such as function optimization, the model classification, machine study, neutral network training, the signal procession, vague system control, automatic adaptation control etc. 3.5.1. Advantages of Particle Swarm Optimization (1)PSO is based on the intelligence. It can be applied into both scientific research and engineering use. (2)PSO have no overlapping and mutation calculation. The search can be carried out by the speed of the particle. During the development of several generations, only the most optimist particle can transmit information onto the other particles, and the speed of the researching is very fast. (3)The calculation in PSO is very simple. Compared with the other developing calculations, it occupies the bigger optimization ability and it can be completed easily. (4) PSO adopts the real number code, and it is decided directly by the solution. The number of the dimension is equal to the constant of the solution. 3.5.2. PSO Steps S1 Initialize PSO parameters ( c1,c2,w), and N particles (x,pbest,gbest,v) S2 Initialize X, V, pbest for each particle and gbest for the swarm. Vid=w*vid + c1*rand()*(Pid-Xid) + c2*rand()*(Pgd-xid) (8) Xid=Xid+Vid, (9) where, X and V - position and velocity of particles, w - inertia weight, c1 and c2 - positive constants, called acceleration coefficients, P - number of particles in the swarm, r1 and r2 - random values in range [0, 1] . S3 Determine the cluster centre. S4 Determine the fitness of each particle: f(x) =K/Jm therein K is a constant and Jm is the objective function S5 Update pbest and gbest. S6 Update velocities and positions (particles) S7 Terminate on convergence or at end of iteration S8 Go to step 2 4. Experimental Results The performance of WML quantification is evaluated using clustering algorithms. When the image is pre-processed, contrast of the image is enhanced. The resulting enhanced image is clustered using the effective clustering algorithms. Figure 3 represents the input image for WML detection. In order to increase robustness, the noisy medical image is pre-processed. Figure 4 depicts the pre-processed image. Bright contrast stretching, which is one of the image enhancement (pre-processing) techniques is applied. After pre-processing the enhanced image is subjected to clustering. Three clustering models are proposed to provide accurate results. The algorithms include BRAIN. Broad Research in Artificial Intelligence and Neuroscience Volume 3, Issue 1, February 2012, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print) 26 (1) FCM, (2) GPC and (3) GFCM. As a next step, the clustered brain images are optimized using Particle Swarm Optimization (PSO). Figure 3. MRI of FLAIR image Figure 4. Enhanced image Figure 5(a). WML detection using FCM Figure 5(b). WML detection using FCM-PSO Figure 6(a). WML detection using GPC Figure 6(b). WML detection using GPC-PSO The first method proposed is fuzzy-c means clustering. When FCM is applied, numerous small regions around the lobes (ventral, parietal, occipital and temporal) are falsely detected as shown in figure 5(a). Figure 6(a) shows the extracted WMLs which are detected using M. Anitha, P. Tamije Selvy - An Optimized Clustering Approach for Automated Detection of White Matter Lesions in MRI Brain Images 27 Geostatistical Possibilistic Clustering. GPC could better capture the largest loads of WMLs only in the upper ventral horns and it failed to detect other smaller regions of the White Matter changes. When GFCM is applied, the regions containing White Matter Lesions are accurately quantified as shown in figure 7(a). This method provides the best results when compared to FCM and GPC, and provides accuracy of 93% as it incorporates the spatial information. The clustered images are further optimized using Particle Swarm Optimization (PSO). Figure 5(b), 6(b) and 7(b) shows the optimized results obtained using the hybrid methods FCM- PSO, GPC-PSO and GFCM-PSO respectively. Figure 7(a). WML detection using GFCM Figure 7(b). WML detection using GFCM-PSO All scans obtained from different image clustering models are manually ranked based on values in table 1. Table 2 represents WML detection rates of optimized images. FCM, GPC and GFCM clustering methods and hybrid optimized methods (FCM-PSO, GPC-PSO and GFCM-PSO) are applied on a dataset of 208 images and ranking is done in terms of under detected, over detected, properly detected as shown in figure 8 and figure 9. The number of images detected properly in GFCM is comparatively high than FCM and GPC. The optimized result of GFMC provides accurate detection of WMLs and it properly detects 195 images. Table 1. WML Detection Rates (%) of FCM, GPC and GFCM S.No Model Over detected Under detected Properly detected 1 FCM 121 120 144 2 GPC 120 112 165 3 GFCM 74 92 193 Table 2. WML Detection Rates (%) of FCM-PSO, GPC-PSO and GFCM-PSO S.No Model Over detected Under detected Properly detected 1 FCM-PSO 118 115 152 2 GPC-PSO 116 106 171 3 GFCM-PSO 62 81 195 BRAIN. Broad Research in Artificial Intelligence and Neuroscience Volume 3, Issue 1, February 2012, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print) 28 Figure 8. Performance analysis of FCM, GPC and GFCM Figure 9. Performance analysis of FCM-PSO, GPC-PSO and GFCM-PSO In this paper, clustering algorithms are evaluated in terms of sensitivity (Se), specificity (Sp) and accuracy (Acc). Taking Table III into account the metrics are defined as Se = TP - (4) TP+FN Sp = TP - (5) TN+FP Acc= TP+TN - (6) TP+FN+TN+FP where, TP-True Positive, FN-False Negative, TN-True Negative, FP-False Positive. The performance results show that GFCM provides sensitivity (Se) of 89%, specificity (Sp) of 93% and overall accuracy (Acc) of 93% when compared to FCM and GPC. In turn, the optimized results of GFCM provide overall accuracy of 95%. Figures 8(a), 8(b) and 8(c) shows the comparison results of clustering models and optimization technique in terms of Se, Sp and Acc. 0 50 100 150 200 250 OVER DETECTED UNDER DETECTED PROPERLY DETECTED FCM GPC GFC 0 50 100 150 200 250 OVER DETECTED UNDER DETECTED PROPERLY DETECTED FCM-PSO GPPC-PSO GFCM-PSO M. Anitha, P. Tamije Selvy - An Optimized Clustering Approach for Automated Detection of White Matter Lesions in MRI Brain Images 29 Table 3. Performance results of FCM, GPC, GFCM and hybrid methods on FLAIR image database S.No Model Sensitivity(%) Specificity(%) Accuracy(%) 1 FCM 121 120 144 2 GPC 120 112 165 3 GFCM 74 92 193 4 FCM-PSO 47 54 53 5 GPC-PSO 72 78 81 6 GFCM-PSO 90 94 95 Figure 8(a). Comparing Se, Sp and Acc of Figure 8(b). Comparing Se, Sp and Acc of FCM and FCM-PSO GPC and GPC-PSO Figure 8(c). Comparing Se, Sp and Acc of GFCM and GFCM-PSO 5. Conclusion The proposed Fuzzy c-means clustering, Geostatistical Possibilistic clustering and Geostatistical Fuzzy c-means clustering methods are used for automatic detection of WMLs in brains of elderly people. The incorporation of the geostatistical estimate variance into the objective functions of fuzzy clustering and possibilistic clustering algorithms is relatively a simple and effective procedure for implementation and can be further explored using various advanced kriging systems in multivariate geostatistics. Experimental results using the MRI data of elderly individuals shows the advantages that Geostatistical Fuzzy c-means clustering is the best and effective approach for extracting White Matter Lesions. More accurate results are obtained by GFCM whereas GPC and FCM provide more false positives in brain image and they are less sensitive to noise. In this paper, in order to overcome the shortcomings clustering techniques, particle swarm algorithm is integrated with clustering algorithms. Experimental results over datasets show that the 0 10 20 30 40 50 60 Se(%) Sp(%) Acc(%) FCM FCM-PSO 64 66 68 70 72 74 76 78 80 82 Se(%) Sp(%) Acc(%) GPC GPC-PSO 86 87 88 89 90 91 92 93 94 95 96 Se (%) Sp (%) Acc (%) GFCM GFCM- PSO BRAIN. Broad Research in Artificial Intelligence and Neuroscience Volume 3, Issue 1, February 2012, ISSN 2067-3957 (online), ISSN 2068 - 0473 (print) 30 proposed hybrid method, i.e., GFCM-PSO is efficient and can reveal very encouraging results in terms of quality of solution found. References [1] Barr, A., and Feigenbaum, E. A. (1981), The Handbook of Artificial Intelligence, Volumes 1-3, William Kaufmann Inc. [2] L. O. Wahlund, F. Barkhof, F. Fazekas, L. Bronge, M. Augustin,M. Sj¨ogren, A. Wallin, H. Ader, D. Leys, L. Pantoni, F. Pasquier, T. Erkinjuntti, and P. Scheltens, “A new rating scale for age- related White Matter changes applicable to MRI and CT,” Stroke, vol. 32, pp. 1318–1322, 2001. [3] E. Matsusue, S. Sugihara, S. Fujii, E. Ohama, T. Kinoshita, and T. Ogawa, “White Matter changes in elderly people: MR-pathologic correlations,” Magn. Reson. Med. Sci., vol. 5, pp. 99– 104, 2006. [4] F. Fazekas, “Pathological correlates of incidental MRI White Matter signal hyper intensities,” Neurology, vol. 43, pp. 1683–1689, 1993. [5] P. Anbeek, K. L. Vincken, M. J. P. van Osch, R. H. C. Bisschops, and J. van der Grond, “Automatic segmentation of different-sized White Matter Lesions by voxel probability estimation,” Med. Image Anal., vol. 8, pp. 205–215, 2004. [6] R. Khayati, M. Vafadust, F. Towhidkhah, and S. M. Nabavi, “Fully automatic segmentation of multiple sclerosis lesions in brain MR FLAIR images using adaptive mixtures method and Markov random field model,” Comput. Biol. Med., vol. 38, pp. 379–390, 2008. [7] Z. Lao, D. Shen, D. Liu, A. F. Jawad, E. R. Melhem, L. J. Launer, R. N. Bryan, and C. Davatzikos, “Computer assisted segmentation of White Matter Lesions in 3D MR images using support vector machine,” Acad. Radiol., vol. 15, pp. 300–313, 2008. [8] M. L. Seghier, A. Ramlackhansingh, J. Crinion, A. P Leff, and C. J. Price, “Lesion identification using unified segmentation-normalisation models and fuzzy clustering,” NeuroImage, vol. 41, pp. 1253–1266, 2008. [9] M. delC.V.Hernendez, K. J. Ferguson, F. M. Chappell, and J.M.Wardlaw, “Newmultispectral MRI data fusion technique for white matter lesion segmentation: method and comparison with thresholding in FLAIR images,” Eur. Radiol., Feb. 16, 2010. [10] R. Beare, V. Srikanth, J. Chen, T. G. Phan, J. Stapleton, R. Lipshut, and D. Reutens, “Development and validation of morphological segmentation of age-related cerebral White Matter hyper intensities,” NeuroImage, vol. 47, pp.199–203, 2009. [11] V.K.Panchal, Mr. Harish Kundra, Ms. Jagdeep Kaur, “Comparative study of Particle Swarm Optimization based unsupervised clustering techniques,” IJCSNS International Journal of Computer Science and Network Security, vol.9, No.10, October 2009. [12] Hesam Izakian, Ajith Abraham, “Fuzzy C-means and fuzzy swarm for fuzzy clustering problem,” Expert Systems with Applications , 38 (2011) , 1835–1838.\ [13] Dian Palupi Rini, Siti Mariyam Shamsuddin and Siti Sophiyait Yuhaniz, “Particle Swarm Optimization: Technique, System and Challenges,” International Journal of Computer Applications (0975 – 8887), vol. 14– No.1, January 2011. [14] Saeed Vaneshani and Hooshang Jazayeri-Rad, “Optimized Fuzzy Control by Particle Swarm Optimization Technique for Control of CSTR,” World Academy of Science, Engineering and Technology, vol. 59, 2011 [15] R. de Boer, H. A. Vrooman, F. van der Lijn, M.W. Vernooij, M. A. Ikram, A. van der Lugt, M. M. B. Breteler, and W. J. Niessen, “White Matter lesion extension to automatic brain tissue segmentation on MRI,” NeuroImage, vol. 45, pp. 1151–1161, 2009. [16] Kehong Yuan, Fei Peng and Wufan Chen, “Brain Image Preprocessing and Application,” International Journal of Innovative Computing, Information and Control, vol. 5, pp. 2333—2340 Number 8, August 2009. [17] J. C. Bezdek, “Pattern Recognition with Fuzzy Objective Function Algorithms,” New York: Plenum Press, 1981. M. Anitha, P. Tamije Selvy - An Optimized Clustering Approach for Automated Detection of White Matter Lesions in MRI Brain Images 31 [18] R. Krishnapuram and J.M. Keller, “A possibilistic approach to clustering,” IEEE Trans. Fuzzy Syst., vol. 1, no. 2, pp. 98–110, May 1993. [19] Keh-Shih Chuang a, Hong-Long Tzeng, Sharon Chen a and Jay Wu, Tzong-Jer Chen,” Fuzzy c-means clustering with spatial information for image segmentation,” Computerized Medical Imaging and Graphic,vol. 9–15, 30 (2006).