Detection of Diabetic Retinopathy Using Deep Learning

The human eye is a complex organ responsible for vision, enabling us to perceive the world in intricate detail. Diabetes, a metabolic disorder characterized by chronic high blood sugar levels, can lead to severe complications, including heart disease, kidney failure, and vision impairment. One such vision-threatening condition is Diabetic Retinopathy (DR), a progressive disorder that can cause blindness if left untreated. Early detection is crucial for preventing further retinal damage and preserving vision. The proposed methodology exploits the VGG-16 deep learning model, known for its robust feature extraction capabilities, to accurately classify DR stages. To address the class imbalance in the dataset, Synthetic Minority Over-sampling TEchnique and Tomek Links are employed. The model is trained on a dataset of 88702 retinal images, categorized into five DR stages: No DR, Mild DR, Moderate DR, Severe DR, and Proliferative DR. Performance evaluation metrics, including accuracy, precision, recall, F1 score, and support, are analyzed and discussed to validate the effectiveness of the proposed approach.