A Transparent Diagnosis Model for Diabetic Retinopathy Using Explainable AI

Diabetic retinopathy is the leading cause of vision problems and partial blindness, and its signs are complicated and variable, making detection difficult. Standard diagnosis is mainly reliant on expert interpretation of retinal images, which can lead to bias and mistake, highlighting the need for improved diagnostic approaches. Despite the fact that conventional AI technologies considerably improve diagnostic skills, their black box aspect limits healthcare practitioners' confidence and optimizes their usage in practice. Grad-CAM++ is a new technique that addresses the black box problem by improving model knowledge, allowing users to grasp the reasoning behind specific decisions. Using XAI approaches, this work creates a diagnostic model for diagnosing diabetic retinopathy, hence improving transparency. The proposed model's convergent plots validate the learning process, which results in 94.4% better diagnostic precision than previous approaches.