Detection of Red Lesions in Diabetic Retinopathy using Deep Learning

Diabetic Retinopathy (DR) is a major cause of blindness among people affected with diabetes for over 10 years. The two fundamental stages of DR are – Non-Proliferative Diabetic Retinopathy and Proliferative Diabetic Retinopathy. The progress from the former to the latter takes time and early detection can prevent blindness. The early symptoms of DR include Red Lesions which constitute microaneurysms and hemorrhages. This paper targets to detect red lesions at the preliminary stage of DR and also in the later stage. The work implements two state-of-the-art models to detect microaneurysms and hemorrhages. The results obtained show that RetinaNet used, performs better than Faster RCNN with the publicly available dataset Messidor.