Explainable AI for Retinal Disease Diagnosis: Enhancing Transparency in OCT Image Classification

This research addresses the growing need for transparency and trust in deep learning-based medical diagnosis, particularly in retinal disease classification using OCT (Optical Coherence Tomography) images. While deep learning models offer high accuracy in classifying retinal conditions such as CNV (Choroidal Neovascularization), DME (Diabetic Macular Edema), and Drusen, their "black box" nature makes it difficult for clinicians to rely solely on their predictions without understanding the reasoning behind them. To bridge this gap, the study incorporates Explainable AI (XAI) techniques such as LIME and SHAP, which help interpret the internal decision-making process of CNN-based models. By highlighting key features that led to a diagnosis, these techniques aim to increase the transparency, interpretability, and trustworthiness of AI systems in clinical settings. The paper also emphasizes the importance of early detection of retinal diseases and how XAI-driven models can aid ophthalmologists by providing both high accuracy and visual explanations. This approach promotes clinician confidence in using AI for diagnosis and paves the way for real-world adoption of AI-powered tools in healthcare.