AUTOMATED RETINAL DISEASE DIAGNOSIS USING CONVOLUTIONAL NEURAL NETWORKS

This project introduces an AI-driven diagnostic system designed to detect retinal diseases from fundus images with high accuracy and operational efficiency. Leveraging advanced deep learning techniques, the system integrates a user-friendly graphical interface to assist clinicians and researchers in early disease detection and classification. It begins with a robust dataset management module that allows users to upload retinal images, associate them with specific classes—including Diabetic Retinopathy (DR), Macular Hole (MH), Normal, and Other Diseases/Conditions (ODC)—and apply essential preprocessing techniques such as image resizing, normalization, and data augmentation. These steps enhance model robustness and dataset diversity. At the system’s core are two distinct model architectures: the first is an existing Deep Neural Network (DNN) utilizing the Stochastic Gradient Descent (SGD) optimizer and composed of multiple convolutional, batch normalization, pooling, and dense layers; the second is a proposed Convolutional Neural Network (CNN) employing the Adam optimizer and incorporating valid padding (AVP), dropout, and additional batch normalization layers to reduce overfitting and improve generalization. Both models are trained and evaluated using standard metrics such as accuracy, precision, recall, and F1-score, with confusion matrices and classification reports highlighting diagnostic performance. Experimental results demonstrate that while both models perform effectively, the proposed CNN with AVP achieves superior accuracy and better differentiation across disease classes, particularly for subtle pathological features. Designed with modularity and scalability in mind, the system supports application across diverse clinical datasets and real-world healthcare environments. Overall, this research underscores the utility of deep learning in medical imaging and offers a practical, scalable solution for automated retinal disease diagnosis, contributing to the advancement of computer-aided diagnosis in ophthalmology.