RetinoScan: An Explainable Deep Learning Framework for Multi-Disease Retinal Screening

Retinal diseases are a major cause of vision impairment worldwide, making early detection essential to prevent vision loss. Diabetic retinopathy (DR), glaucoma, and age-related macular degeneration (AMD) can progress without noticeable symptoms, highlighting the importance of timely retinal screening. This paper presents RetinoScan, an explainable deep-learning framework for multi-disease retinal screening using retinal images. The framework employs image preprocessing and augmentation techniques, followed by EfficientNetV2-S for feature extraction and classification. It assesses diabetic retinopathy, glaucoma, and age-related macular degeneration. Grad-CAM is used to generate visual explanations highlighting image regions that influence model predictions. The framework also provides prediction confidence scores and a clinical-style report to support the interpretation of results. RetinoScan aims to support accessible retinal screening and assist healthcare professionals in early assessment. It is intended as a decision-support tool and does not replace professional medical diagnosis.