Diabetic Retinopathy Detection Using Deep Learning: An Explainable AI Approach

Abstract

Diabetic Retinopathy (DR) is a major cause of blindness in diabetics, and it occurs in more than one-third of the 537 million diabetics patients worldwide (IDF). This paper proposes an automated detection method of DR in retinal fundus images using Convolutional neural networks (CNN): The method is time-saving, does not need experts, and accessible in remote regions. The model was trained on datasets of EyePACS (Kaggle) and IDRiD, including preprocessing (image resizing, normalization, image augmentation) and optimizing hyperparameters.To achieve interpretability, Grad-CAM is used to visualize decision areas. Results: This model is a real-time prediction deployed as a Streamlit web application with 94.8% accuracy, precision/ recall/F1 above 0.93. Grad-CAM outlines the most important features (hemorrhases, microaneurysms). The system has scalable screening under resource- constrained conditions.