Review on Detection and Classification of Diabetic Retinopathy using Fundus Images

Diabetic Retinopathy (DR) is a prevalent outcome of diabetic mellitus. It leads to lesions forming on the retina, impairing eyesight. Most likely, blindness can be avoided if the DR condition is discovered at an initial stage. Although DR is a non-reversible condition, early identification and intervention can greatly lower the risk of vision loss. Fundus images are used to manually detect DR, which is a laborious and error-prone procedure. Today in assessing and categorizing medical images, deep learning has emerged as the most efficient method, even surpassing human performance, common image processing methods, and other computer-aided detection systems. For this study, the most recent approaches of deep learning using fundus images to classify and detect DR have been reviewed. The freely accessible DR Datasets consisting of fundus images and performance metrics have also been discussed. We concluded with some important current and future trends of DR. Further, several open problems and limitations of DR that need further study are discussed.