International Journal of Advanced Research in Computer and Communication Engineering

Abstract: Diabetic Retinopathy (DR) is a progressive ocular disorder resulting from prolonged diabetes, in which damage to the retinal blood vessels can eventually cause permanent vision impairment. Early diagnosis is crucial, yet access to routine screening remains inadequate, especially in rural and resource-constrained communities. RetinoAI is an automated deep-learning framework designed to detect DR and classify its severity using retinal fundus photographs. The system uses Convolutional Neural Networks (CNNs) enhanced through transfer learning, accompanied by an optimized preprocessing pipeline incorporating image resizing, normalization, and augmentation to improve visual quality and model robustness. The DR classification module identifies disease grades ranging from No DR to Proliferative DR and generates corresponding confidence levels. To improve interpretability, RetinoAI employs Grad- CAM-based visual explanations that highlight important retinal regions contributing to the model’s decision, supporting clinicians with clearer insight into AI-assisted diagnoses.

Keywords: Deep Learning, Diabetic Retinopathy, Convolutional Neural Networks, Fundus Image Analysis, Grad- CAM, Medical Imaging.

Downloads: | DOI: 10.17148/IJARCCE.2025.1411123