Developing a Transparent Diagnosis Model for Diabetic Retinopathy Using Explainable AI

Diabetic retinopathy (DR) is a major global health issue, accurate and timely mass screening to prevent vision loss. Although deep learning (DL) models have shown expert-level performance in automatically grading diabetic retinopathy (DR) from fundus images, their inherent "black box" nature presents a significant barrier to their use in clinical settings. Clinicians need diagnostic explanations to ensure patient safety, confirm the model's reliability, and maintain accountability. This is a requirement that systems lacking transparency cannot fulfill. This study presents the creation and testing of a Transparent Diagnosis Model (TDM) for Diabetic Retinopathy, using Explainable Artificial Intelligence (XAI) methods. Our approach combines a customized Convolutional Neural Network (CNN) for classification with the Grad-CAM method to create visual explanations at the same time.