DEEP LEARNING-DRIVEN PNEUMONIA DETECTION SYSTEM USING CONVOLUTIONAL NEURAL NETWORKS

Pneumonia remains a critical global health challenge, particularly in regions with limited access to expert radiological and medical services at large. This study therefore presents the development of a Deep Learning-Based Pneumonia Detection System that leverages Convolutional Neural Networks (CNNs) for automatic analysis of chest X-ray images, with primary emphasis on accurate and efficient detection rather than long-term prediction, geared towards improving the efficiency of medical practitioners in detecting pneumonia. The system was developed using Python (Flask framework) for backend integration, HTML, CSS, and JavaScript for the frontend interface, and TensorFlow for model development, with PostgreSQL used for database and data storage. The core detection algorithm is based on a CNN architecture that performs feature extraction through convolutional and pooling layers, followed by classification using fully connected layers with ReLU activation and Softmax output functions. The model was trained using cross-entropy loss and optimized to minimize classification error. Input chest X-ray images undergo preprocessing steps including resizing (e.g., 224×224), normalization, and data augmentation to improve model robustness. The system performance was evaluated using standard metrics such as accuracy, precision, recall, F1-score, and AUC, while classification outcomes are supported by a confidence score (risk score) indicating the probability of pneumonia presence. To enhance interpretability, Grad-CAM was employed to generate heatmaps that highlight regions of interest influencing model decisions. Experimental results demonstrate that the system achieves high detection accuracy of 76% with reliable sensitivity and specificity, effectively distinguishing pneumonia-infected images from normal cases. The system reduces diagnostic time, minimizes human errors, and provides consistent results through automated analysis. In conclusion, the proposed system offers a scalable, efficient, and interpretable solution for pneumonia detection using deep learning. It demonstrates the practical applicability of CNN-based models in medical imaging and provides a valuable decision-support tool for healthcare professionals, particularly in resource-constrained environments