Detection of Diabetic Retinopathy Using Deep Convolutional Neural Networks

Diabetic retinopathy (DR) is the most prevalent disease among diabetic patients. DR affects retinal blood vessels and causes total loss of vision if it is not treated earlier. Around 50% of the diabetic population is affected by DR and thus necessitates the need for DR detection. Manual processes available consume more time, and hence a convolutional neural network is employed to detect DR at an earlier stage. The proposed method has three convolutional layers and a fully connected layer. This method gives higher accuracy (94.44%) with reduced hardware requirement than conventional approaches to detect and classify DR into five stages, namely no DR, mild, moderate, severe, and proliferative DR.