Multi-Modal Deep Learning for Retinal Analysis

We present an AI-driven diagnostic assistant for retinal disease detection that leverages advanced deep learning and explainability techniques. Our framework combines a Vision Transformer for robust classification with a U-Net for precise segmentation, while Grad-CAM provides interpretable heatmaps and Langchain automates report generation for seamless integration with Electronic Health Records. Experimental results demonstrate that our system reliably detects retinal diseases with high accuracy, reducing diagnostic time and enhancing clinical decision-making. This work offers a transparent, scalable solution for early intervention in retinal care.