Diabetic Retinopathy Detection Using ResNet-50 and ResNet-152 Deep Learning Approach

Diabetic Retinopathy (DR) is a leading cause of preventable blindness, stemming from damage to retinal blood vessels due to prolonged diabetes. Early detection is critical to mitigate vision loss, yet conventional diagnostic methods are labor-intensive, costly, and inaccessible in under-resourced regions. This research presents a deep learning-based framework for automated detection and classification of DR stages using retinal fundus images. Leveraging advanced preprocessing techniques such as Contrast Limited Adaptive Histogram Equalization (CLAHE), green channel extraction, and vessel segmentation, coupled with fine-tuned models like ResNet-50 (87.5% accuracy), ResNet-152 (89.3% accuracy the system ensures robust classification accuracy. Evaluated on expert-annotated datasets, the system demonstrated high sensitivity, specificity, and reliability in identifying Diabetic Retinopathy stages, from mild abnormalities to proliferative cases. Designed for real-time analysis through a user-friendly interface, this solution addresses key challenges such as dataset imbalances, model interpretability, and compliance with medical privacy regulations.