Deep Diabetic Retinopathy Detection System (DDRDS) using Convolutional Neural Network: A Comparative Study

Abstract

Diabetic Retinopathy (DR) is a medical condition in the retina of human eye, triggered due to diabetes mellitus which causes formation of lesions in the retina and leads to blurred vision and even blindness. The statistical data estimations show 80% of diabetic patients, suffering from protracted diabetes, also suffers from DR. Hence, early DR evaluation and assessment can reduce susceptibility to severe blindness, especially amongst the working generation. The process of physical diagnosis is laborious, inefficient and liable to cause error, and the lack of resources and expert opinions, makes early detection and treatment infeasible. Thus, advanced intelligent systems using innovative Machine Learning (ML) techniques such as Deep Learning (DL) are proposed by researchers. This paper proposes an intelligent system named Deep Diabetic Retinopathy Detection System (DDRDS) which employs four Deep Convolutional Neural Networks (DCNNs), for fundus image classification, for early detection of DR.