AI-Based Diabetic Retinopathy Detection using Explainable Deep Learning Models

Diabetic Retinopathy (DR) is a critical complication of diabetes, often leading to vision loss if not diagnosed early. With the increasing global prevalence of diabetes, there is an urgent need for scalable, accurate, and real-time DR diagnostic tools. This study presents a Convolutional Neural Network (CNN)-based solution using the ResNet50 architecture for classifying DR into five stages: No DR, Mild, Moderate, Severe, and Proliferative DR. Our system integrates Grad-CAM, an Explainable AI (XAI) technique, to enhance interpretability by highlighting regions in retinal fundus images that influence the classification. The solution is deployed through a Gradio web interface, ensuring usability in real-time clinical settings. The model was trained and tested using public datasets like APTOS and EyePACS and achieved high accuracy and interpretability, making it suitable for telemedicine and mobile health (mHealth) applications.