Paper Title (use style: paper title) Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1175 https://internationalpubls.com Development of Deep Learning Framework for Multiclass Lung Infection Detection and Classification from Chest X-Ray Imaging with Real-Time Integration Prof. Dr.M.D.Rokade 1, Shermale Hema Mohan 2, Prof. Dr. Sunil Khatal 3 1Department of Computer Engineering, Sharadchandra Pawar College Of Engineering, Dumbarwadi, Otur monikarokade4@gmail.com 2Department of Computer Engineering , Sharadchandra Pawar College Of Engineering, Dumbarwadi, Otur shermalehema@gmail.com 3Head of Department, Department of Computer Engineering , Sharadchandra Pawar College Of Engineering, Dumbarwadi, Otur Article History: Received: 12-01-2025 Revised: 15-02-2025 Accepted: 01-03-2025 Abstract: This study presents an advanced deep learning-based system for the detection of lung infections using chest X-ray images, designed to support accurate and efficient medical diagnostics. The system operates through three primary phases: training, testing, and classification. During the training phase, pre-processed chest X-ray images are fed into a robust feature extraction module built on convolutional neural network (CNN) architectures like ResNet or VGG. This module leverages convolutional layers, max pooling, residual blocks, and global average pooling to extract both fine-grained and structural features from the images, enabling the model to capture subtle patterns critical for diagnostic accuracy. Advanced pre-processing techniques such as noise reduction and image normalization further enhance input quality, ensuring reliable model performance across diverse datasets. In the testing and classification phases, unseen X-ray images are processed through the same pipeline, ensuring consistency. Optimized feature representations are utilized to classify images into multiple categories, including COVID-19, viral pneumonia, bacterial pneumonia, tuberculosis, and normal cases. The system integrates explainability tools like Grad-CAM to visualize decision-making, fostering trust among clinicians. Designed with real-world applications in mind, this system supports real- time deployment through cloud-based APIs or edge devices, enhancing its practical utility. By combining state-of-the-art CNN architectures, multi-class classification, and explainability, the proposed framework provides a scalable and reliable solution for improving diagnostic workflows and advancing healthcare delivery. Keywords: Deep learning, Lung infection detection, Chest X-ray analysis, Convolutional neural networks, Medical diagnostics. I. INTRODUCTION Lung infections, including pneumonia, tuberculosis, and COVID-19, are major global health concerns with high mortality rates. Timely and accurate detection of these infections is critical for mailto:monikarokade4@gmail.com mailto:shermalehema@gmail.com Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1176 https://internationalpubls.com effective treatment and management. Traditionally, diagnosing lung infections has relied on clinical symptoms, physical examinations, and laboratory tests. However, chest X-rays (CXR) are commonly used for the initial screening and diagnosis due to their accessibility, cost-effectiveness, and ability to provide detailed visual information about the lungs [1]. Despite their widespread use, the manual interpretation of chest X-rays can be time-consuming and prone to human error, especially in a clinical setting with limited resources. This limitation highlights the need for advanced computational methods that can automate and enhance the diagnostic process, improving accuracy and reducing the time required for diagnosis. In recent years, deep learning, particularly Convolutional Neural Networks (CNNs), has demonstrated remarkable success in medical image analysis. CNNs have shown significant promise in tasks such as image classification, object detection, and segmentation. These networks excel in automatically learning hierarchical features from images, which makes them ideal for tasks such as detecting lung infections in X-ray images. By leveraging large-scale datasets and powerful computation resources, CNNs can identify subtle patterns that may be missed by human clinicians [2]. The potential for deep learning models to assist radiologists in diagnosing lung infections in real- time has become a focal point for medical research and innovation. The core strength of deep learning models, such as CNNs, lies in their ability to extract high-level features from raw image data. In lung infection detection, these networks are capable of learning complex patterns that distinguish between different types of infections or between infected and healthy lungs. Several CNN architectures, including VGG16, ResNet, and Inception, have been used successfully for medical image classification tasks [3][4]. These architectures, with their deep layers and specialized design, enable the model to capture intricate details in chest X-rays, which is essential for distinguishing between diseases with similar symptoms but different underlying causes. However, the use of CNNs in healthcare applications, particularly for lung infection detection, is not without its challenges. One of the main concerns is the interpretability of the results. While deep learning models can produce highly accurate predictions, the "black-box" nature of these models makes it difficult to understand how decisions are made. This lack of transparency is particularly problematic in the healthcare domain, where medical professionals must trust and understand the reasoning behind diagnostic suggestions. Recent advancements in explainability methods, such as Grad-CAM and SHAP, have been developed to address this issue by visualizing the decision-making process of deep learning models [5]. These techniques allow clinicians to gain insights into the regions of the X-ray image that influenced the model’s classification, increasing trust and usability. Another challenge in deploying deep learning models for lung infection detection is ensuring the robustness and generalizability of the model across diverse patient populations and imaging conditions. A model trained on a specific dataset may perform well on similar data but struggle when exposed to unseen cases or images from different sources. To overcome this limitation, data augmentation and transfer learning techniques are commonly employed. Data augmentation techniques, such as rotation, flipping, and rescaling, help increase the diversity of training data, improving the model’s ability to generalize [6]. Transfer learning, on the other hand, involves using pre-trained models on large datasets and fine-tuning them on smaller, domain-specific datasets to achieve better accuracy in a particular application [7]. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1177 https://internationalpubls.com In addition to technical challenges, the real-world deployment of deep learning models in healthcare settings requires careful consideration of factors such as integration with existing systems, computational requirements, and regulatory compliance. Real-time detection of lung infections using chest X-rays involves processing large volumes of medical images in a timely manner, which can place significant demands on computational resources. Edge computing, where models are deployed on local devices for faster inference, or cloud-based APIs, which offer scalable solutions, are potential solutions to address these challenges [8]. Moreover, healthcare regulations and standards must be adhered to, ensuring that the system meets the necessary medical certification and data privacy requirements. The overarching goal of this research is to develop an advanced deep learning-based system that can accurately detect a wide range of lung infections, including COVID-19, pneumonia, and tuberculosis, from chest X-ray images. By incorporating state-of-the-art deep learning techniques, explainability features, and real-time deployment capabilities, this system aims to revolutionize healthcare delivery. The proposed system not only aims to improve diagnostic accuracy but also to provide clinicians with the tools they need to make informed decisions quickly and confidently, ultimately improving patient outcomes and contributing to the fight against lung infections worldwide. II. OBJECTIVE 1. To study the effectiveness of deep learning models in detecting lung infections from chest X-ray images. 2. To study the application of Convolutional Neural Networks (CNNs) for automated lung disease classification. 3. To study the impact of explainable AI techniques on improving model trust in medical diagnostics. 4. To study the integration of real-time deep learning systems for clinical deployment in hospitals. 5. To study the comparative performance of multi-class classification models in detecting various lung infections. III.LITERATURE SURVEY Paper Title Authors Publication/Source Year Techniques Used Summary Pneumonia Detection and Classification using CNN and VGG16 Poosa Praveen Kumar, Yashwanth Renukunta, Krishna Tej Chinta, Aditya Jangam, Dr. P. Chandra International Journal for Research in Applied Science & Engineering Technology (IJRASET) 2023 CNN, VGG16, VGG19, RESNET-50, RESNET- 101 This study contrasts deep learning models for pneumonia detection from chest X-rays, selecting the optimal model based Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1178 https://internationalpubls.com Sekhar Reddy on performance for practical application. Identify Type of Lung Infection from Lung Patients X- RAY Image Leveraging Computer Vision Mohamed Mahyoub, Thomas Coombs, Manoj Jayabalan, Jamila Mustafina, Abir Hussain 15th International Conference on Developments in eSystems Engineering (DeSE) 2023 CNN, Pre- trained models, Custom CNN Explores CNN-based approaches to classify COVID-19, pneumonia, and normal lung conditions using X-ray images, achieving competitive accuracy with state-of- the-art. Deep- learning CNNs with Transfer Learning for COVID-19 Lung Infection on pCXR Shreeja Kikkisetti, Jocelyn Zhu, Beiyi Shen, Haifang Li, Tim Q. Duong Montefiore Medical Center and Albert Einstein College of Medicine 2023 CNN, Transfer Learning, Portable Chest X-ray (pCXR) Utilizes transfer learning in CNN to classify COVID-19 and other lung infections from portable chest X-rays, enhancing diagnostic accuracy through deep learning. Pneumonia Detection Using Deep Learning Methods Hrushikesh S. Rajankar, Pavan S. Patil, Karan V. Nakum, Gopal S. Paraskar Sinhgad Academy of Engineering, Pune 2022 Deep Learning, CNN Demonstrates CNN's application for pneumonia detection to reduce misdiagnosis, focusing on efficient model design and Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1179 https://internationalpubls.com predicting patient health from chest X-rays. Pneumonia Detection and Classification using Deep Learning Models Dr. Sunil L. Bangare International Journal of Advanced Research in Science, Communication, and Technology (IJARSCT) 2022 CNN, Transfer Learning, Multi-class Classification Presents a multi-class classification model using deep learning for pneumonia detection on X-rays, emphasizing transfer learning for robust predictions. IV. PROPOSED SYSTEM The proposed system leverages advanced deep learning techniques to automate the detection and classification of lung infections from chest X-ray (CXR) images. The system’s primary goal is to enhance the diagnostic process by providing accurate and timely predictions of various lung conditions, including viral pneumonia, bacterial pneumonia, tuberculosis (TB), and COVID-19. The proposed architecture integrates sophisticated image pre-processing, feature extraction using Convolutional Neural Networks (CNNs), and a multi-class classification approach that categorizes the X-ray images into specific diagnostic labels. Below, the components of the system are described in detail: Fig.1 System Architecture Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1180 https://internationalpubls.com 1. Data Acquisition and Preprocessing The system begins with the acquisition of chest X-ray images, which are sourced from publicly available medical datasets such as the COVID-19 Radiography Database, the NIH Chest X-ray dataset, and other specialized repositories. These datasets typically contain labeled X-ray images of healthy and infected lungs, which serve as the foundational data for training the model. The images undergo several preprocessing steps to ensure uniformity and quality before feeding them into the deep learning model. The preprocessing steps include: • Resizing: All images are resized to a consistent shape (e.g., 224x224 pixels) to standardize input dimensions. • Normalization: Pixel values are normalized to a range between 0 and 1 to improve the stability of the model during training. • Noise Reduction: Techniques such as Gaussian filtering or median filtering are applied to reduce noise and enhance the quality of the X-ray images. • Data Augmentation: To mitigate overfitting and improve the generalization of the model, various augmentation techniques are applied. These include random rotations, flipping, zooming, and brightness adjustments. Such transformations help simulate diverse real-world conditions in the training data and increase model robustness. 2. CNN Feature Extraction Module The core of the proposed system is the CNN-based feature extraction module. Convolutional Neural Networks (CNNs) are widely used in image classification tasks due to their ability to automatically learn spatial hierarchies of features from raw images. This module is designed to extract relevant features from X-ray images that are indicative of different lung infections. The CNN feature extraction pipeline in the proposed system includes the following steps: • Convolutional Layers: The network begins with convolutional layers that apply various filters to the input images. These filters are designed to capture low-level features such as edges, textures, and basic patterns in the X-ray images. Convolutional operations are repeated across multiple layers, progressively capturing more complex features. • Activation Functions: Each convolutional layer is followed by an activation function (typically ReLU), which introduces non-linearity and helps the network model more complex patterns in the data. • Pooling Layers: Max-pooling layers are introduced after convolutional layers to reduce the spatial dimensions of the image, thereby decreasing computational complexity while preserving the most important features. This helps in reducing overfitting by preventing the model from becoming overly sensitive to minute changes in the input images. • Residual Blocks: To avoid the vanishing gradient problem and ensure that deeper networks can learn effectively, residual blocks are employed. These blocks allow the model to learn both the Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1181 https://internationalpubls.com residual (or difference) and the main features, making it easier to train deeper models and extract finer-grained features from X-ray images. • Global Average Pooling: Instead of using fully connected layers at the end of the CNN, the system uses global average pooling to produce a compact representation of the extracted features. This reduces the dimensionality of the feature maps and prevents overfitting while retaining critical information. 3. Classification Module Once the features have been extracted, they are passed to the classification module, which is responsible for determining the class label of the input image. The classification module is a fully connected layer that processes the high-level features extracted by the CNN. This phase involves: • Feature Optimization: The extracted features undergo optimization techniques, such as feature scaling or dimensionality reduction, to enhance the model’s discriminative power. This ensures that the features are adequately represented for the classification task. • Multi-Class Classification: The system classifies the X-ray images into one of several categories: Viral Pneumonia, Bacterial Pneumonia, COVID-19, Tuberculosis, or Normal. A softmax activation function is typically used at the output layer to provide probability scores for each class. The model outputs the class with the highest probability, indicating the predicted lung condition. The classification phase utilizes a multi-class loss function, such as categorical cross-entropy, which allows the model to optimize its weights for accurate class predictions. Additionally, the system may be trained using a combination of labeled datasets to improve accuracy and reduce class imbalance. 4. Explainability and Visualization Tools One of the key features of the proposed system is its integration of explainability methods, which help clinicians interpret the model's decisions. Given that medical professionals rely heavily on the reasoning behind diagnosis, the system incorporates tools such as: • Grad-CAM (Gradient-weighted Class Activation Mapping): Grad-CAM is used to visualize the areas of the chest X-ray image that contributed most significantly to the model's decision. By highlighting the regions of interest, Grad-CAM helps clinicians understand which parts of the image (e.g., lungs, heart, or surrounding structures) were most important in detecting specific infections. • SHAP (Shapley Additive Explanations): SHAP values are calculated to explain the contribution of each feature to the final decision, allowing for more granular understanding of the model’s predictions. By interpreting the model's output in terms of individual pixel or feature importance, SHAP provides transparency to the decision-making process. These explainability features are crucial for gaining the trust of medical professionals and ensuring that the system’s predictions are both understandable and actionable. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1182 https://internationalpubls.com 5. Real-Time Deployment and Integration The system is designed for real-time deployment in clinical settings. The proposed system can be integrated with hospital information systems (HIS) or Picture Archiving and Communication Systems (PACS), enabling seamless data flow between the system and other healthcare management tools. Real-time detection allows medical professionals to receive instant diagnostic results, facilitating quicker treatment decisions. For deployment, two primary options are considered: • Edge Computing: In resource-constrained environments, the system can be deployed on local edge devices, such as low-cost GPU-powered workstations or medical imaging devices. This approach minimizes latency by allowing the model to perform inference directly on the device, reducing the dependency on cloud infrastructure. • Cloud-Based Deployment: For hospitals with robust IT infrastructure, the model can be deployed as a cloud service, offering scalable solutions to process large volumes of medical images. This approach allows for easy updates to the model and enables remote access for healthcare providers in different locations. 6. Evaluation and Performance Metrics The performance of the proposed system is evaluated using several key metrics: • Accuracy: The overall accuracy of the system in classifying lung infections is computed by comparing the model’s predictions to the ground truth labels in the test dataset. • Precision, Recall, and F1-Score: Precision and recall are calculated for each class to measure how well the model distinguishes between different types of lung infections and minimizes false positives or false negatives. The F1-score, which combines both precision and recall, provides a balanced measure of model performance. • Confusion Matrix: A confusion matrix is used to visualize how well the model performs across all classes and identify potential areas for improvement. • AUC-ROC Curve: The area under the Receiver Operating Characteristic curve (AUC-ROC) is evaluated to understand the model’s ability to distinguish between classes at different thresholds. By integrating these metrics, the proposed system ensures comprehensive evaluation and continuous improvement in performance. The proposed system aims to advance the detection of lung infections through a state-of-the-art deep learning architecture. By combining efficient preprocessing, feature extraction using CNNs, multi- class classification, explainability tools, and real-time deployment, the system not only enhances diagnostic accuracy but also fosters trust among healthcare professionals. This holistic approach has the potential to revolutionize the medical imaging field, particularly in resource-limited settings, and improve patient outcomes globally. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1183 https://internationalpubls.com Datasets: • The dataset curated by Paul Mooney consists of 5,856 frontal chest X-ray images. It includes 1,583 images showing healthy lungs and 4,273 images with indications of pneumonia. • The Shenzhen dataset, assembled by Scott Mader, comprises 662 frontal X-ray images. Of these, 326 images represent healthy lungs, while 336 images show signs of tuberculosis infection. • Mohamed Hany's dataset includes 907 lung CT-scan images. This dataset is composed of 215 images with no signs of cancer and 692 images of patients diagnosed with various types of lung cancer, such as adenocarcinoma, squamous cell carcinoma, and large cell carcinoma. V. RESULT & ANALYSIS The results of the proposed system demonstrate its effectiveness in accurately detecting and classifying lung infections, including COVID-19, pneumonia, tuberculosis, and cancer, from chest X- ray and CT images. By leveraging advanced deep learning architectures such as pre-trained CNN models, the system achieves high precision, recall, and F1-scores across all categories, ensuring reliable diagnostic outcomes. The incorporation of extensive preprocessing techniques, such as noise reduction and normalization, enhances the quality of input images, leading to robust feature extraction. The system's ability to analyze diverse datasets and generalize effectively across various demographics highlights its scalability and adaptability to real-world scenarios. Additionally, the inclusion of explainability tools like Grad-CAM enables medical professionals to interpret the decision-making process, fostering trust and facilitating informed clinical decisions. With real-time integration capabilities and high classification accuracy, the system sets a benchmark for leveraging AI in medical imaging diagnostics, significantly contributing to improving patient care and early disease detection. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1184 https://internationalpubls.com Fig.2 Disease Prediction Fig.3 Chart showing Accuracy VI. ADVANTAGES Improved Diagnostic Accuracy: The system enhances diagnostic precision by accurately identifying various lung infections, such as pneumonia, tuberculosis, and lung cancer. Time Efficiency: It reduces diagnostic time, enabling quicker decision-making for healthcare professionals, especially in busy clinical environments. Scalability: The model can be deployed in diverse healthcare settings, from small clinics to large hospitals, facilitating widespread use. Explainability: Incorporates explainable AI techniques, allowing healthcare professionals to interpret and trust the model's decisions. Real-Time Integration: The system can be integrated into hospital management systems for real- time diagnostic support, improving patient outcomes. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1185 https://internationalpubls.com VII. DISADVANTAGES Data Quality Dependency: The system's performance heavily relies on the quality and variety of the datasets, making it susceptible to inaccuracies if the training data is limited or noisy. Computational Requirements: Deep learning models, particularly CNNs, require significant computational resources, which may limit accessibility in resource-constrained environments. Generalization Challenges: The model might struggle to generalize across all patient demographics and imaging variations, potentially leading to incorrect classifications in some cases. Limited Dataset Coverage: The current datasets may not cover all lung diseases comprehensively, limiting the model's ability to detect less common conditions. VIII. FUTURE SCOPE In the future, we aim to expand the system's capabilities by incorporating multi-modal data, including CT scans, MRI images, and patient clinical records, to enhance diagnostic accuracy and provide a more holistic view of patient health. Additionally, integrating advanced techniques such as Federated Learning could allow the model to learn from diverse datasets across multiple institutions while maintaining patient privacy. We also plan to integrate real-time monitoring and decision support systems in clinical environments, enabling proactive alerts for healthcare providers. To address the challenge of dataset limitations, future work will focus on gathering more diverse datasets, including rare diseases, and incorporating synthetic data generation to overcome class imbalance. Furthermore, the implementation of explainable AI methods will be enhanced, providing clearer insights into the decision-making process and increasing trust and acceptance among healthcare professionals. IX. CONCLUSION In conclusion, the proposed advanced deep learning system for lung infection detection using chest X-ray images offers a significant improvement in diagnostic accuracy and efficiency. By leveraging state-of-the-art deep learning techniques such as Convolutional Neural Networks (CNNs), the system demonstrates the ability to classify various lung conditions, including pneumonia, tuberculosis, and COVID-19, with high precision. The integration of explainability features and real-time capabilities further enhances its potential for real-world applications in medical settings. While challenges remain, particularly in dataset diversity and model generalization, the system lays the groundwork for future advancements in automated healthcare diagnostics, contributing to better patient outcomes and streamlining healthcare workflows. X. REFERENCES [1] Zhou, Z. H., Xu, L., & Yang, M. (2020). Deep learning in medical image analysis. Springer. [2] Shin, H. C., Roth, H. R., Gao, M., Lu, L., Xu, Z., Nogues, I., & et al. (2016). Deep convolutional neural networks for computer-aided detection: CNN architectures, dataset characteristics and transfer learning. IEEE Transactions on Medical Imaging, 35(5), 1285- 1298. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1186 https://internationalpubls.com [3] Simonyan, K., & Zisserman, A. (2015). Very deep convolutional networks for large-scale image recognition. Proceedings of the International Conference on Learning Representations (ICLR). [4] He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 770-778. [5] Selvaraju, R. R., Cogswell, M., Das, A., Parikh, D., & Batra, D. (2017). Grad-CAM: Visual explanations from deep networks via gradient-based localization. Proceedings of the IEEE International Conference on Computer Vision (ICCV), 618-626. [6] Shorten, C., & Khoshgoftaar, T. M. (2019). A survey on image data augmentation for deep learning. Journal of Big Data, 6(1), 1-48. [7] Yosinski, J., Clune, J., Nguyen, A., Fuchs, T., & Lipson, H. (2014). How transferable are features in deep neural networks? Advances in Neural Information Processing Systems (NeurIPS), 3320-3328. [8] Rajaraman, A., Sethi, P., & Sharma, S. (2020). Cloud computing in healthcare: A survey and future directions. International Journal of Healthcare Information Systems and Informatics, 15(4), 1-18. [9] Nijhawan, R., Rishi, M., Tiwari, A., & Dua, R. (2019). "A Novel Deep Learning Framework Approach for Natural Calamities Detection." [10] Althomali, O., Al-Zoube, M. A., & Alotaibi, F. S. (2019). "Classification of Pneumonia in Chest Radiographs Using Deep Learning." [11] Esteva, A., et al. (2019). "A Guide to Deep Learning in Healthcare." [12] Al Mamlook, R. E., Chen, S., & Bzizi, H. F. (2020). "Investigation of the Performance of Machine Learning Classifiers for Pneumonia Detection in Chest X-ray Images." [13] Ibrahim, A., Ozsoz, M., Serte, S., Al-Turjman, F., & Yakoi, P. (2021). "Pneumonia Classification Using Deep Learning from Chest X-Ray Images During COVID-19." Cognitive Computation, 1–13. [14] Liang, G., & Zheng, L. (2020). "A Transfer Learning Method with Deep Residual Network for Pediatric Pneumonia Diagnosis." [15] Sirish Kaushik, V., et al. (2019). "Pneumonia Detection Using Convolutional Neural Networks (CNNs)." Proceedings of First International Conference on Computing Communications and Cyber-Security (IC4S 2019). [16] Mahmood, S., Afaq, M., & Sharif, M. (2019). "Pneumonia Detection Using Deep Learning from Chest X-Ray Images." [17] Hosseini-Asl, E., Saeb-Parsy, K., & Graham, S. (2020). "Deep Learning-Based Pneumonia Detection on Chest X-Ray Images." Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 10s (2025) 1187 https://internationalpubls.com [18] Chandra, T., & Verma, K. (2020). "Pneumonia Detection on Chest X-Ray Using Machine Learning Paradigm." [19] Janizek, J., Erion, G., DeGrave, A., & Lee, S. (2020). "An Adversarial Approach for the Robust Classification of Pneumonia from Chest Radiographs." [20] Gabruseva, T., Poplavskiy, D., & Kalinin, A. (2020). "Deep Learning for Automatic Pneumonia Detection." [21] Kulkarni, P., Swaminathan, R., & Thiagarajan, J. J. (2020). "Deep Learning for Pneumonia Detection in Chest X-rays: A Comparative Study." [22] Kaggle. (Online). "Dataset available: Chest X-Ray Pneumonia Dataset" Link: https://www.kaggle.com/datasets/paultimothymooney/chest-xray-pneumonia. https://www.kaggle.com/datasets/paultimothymooney/chest-xray-pneumonia