EM-DR: Efficient-Mamba Vision Model for Early Detection and Classification of Diabetic Retinopathy

Diabetic Retinopathy (DR) is one of the leading causes of blindness in the world and requires prompt and accurate diagnosis to treat it properly. Recent advancements in Deep Learning (DL) have made the automation of DR detection possible, which has increased the efficiency of diagnosis. However, the existing models suffer from misclassification, sensitivity to image quality, and high computational demands. This project proposesaCustom CNN model that uses EfficientNet with MedMamba architecturesto improve the classification of DR in five stages: No DR, Mild, Moderate, Severe, and Proliferative DR. This model combines sophisticated feature extraction and optimized network design to overcome the drawbacks of existing models. Leveraging the publicly available diabetic retinopathy datasets, the model achieved 91.44% accuracy with a better validation score of 84.12%. Such results further reinforce the success of the proposed model in terms of DR stage categorization with potential application towards early detection and clinical usage.