A Comprehensive Analysis of Diabetic Retinopathy Using Deep Learning Techniques

Diabetic Retinopathy (DR) is a diabetes-related retinal eye disease that affects retina and tissues at the back of the eyes, leading to compromised vision. DR occurs due to high blood sugar levels impairing blood vessels in retina. Diabetes is caused due to insufficient amounts of insulin in blood, which increases glucose levels, making it one of the most common disorders worldwide. People with diabetes experience damages to the nerves and blood vessels which also affect the heart, eyes, brains and kidneys. The stages of DR are: Non-Proliferative Diabetic Retinopathy (NPDR) that includes mild to moderate and severe NPDR, while PDR involves the evaluation of new and abnormal vessels in retinal fundus images. Deep Learning (DL) techniques like Long Short-Term Memory (LSTM), Deep Neural Network (DNN), transformer models, and Bi-directional LSTM(Bi-LSTM) are used to classify and detect DR in images. The evaluation metrics: specificity, sensitivity, accuracy, and Fl-score are used to assess DR classification and detection using aDL method.