Design and Develop A System for Detecting Diabetic Retinopathy Using CNN Model

Abstract

Diabetic retinopathy (DR) is one of the most threats to diabetes, causing damage and blindness to the retina. It destroys the retinal tissue blood vessels, resulting in fluid leakage and deformation of vision. DR contains four stages (Mild non-proliferative retinopathy, Moderate Non-Proliferative Retinopathy, Severe non-proliferative retinopathy, and Proliferative Retinopathy). Therefore, appropriate screening and treatment at an early point of DR will profoundly avoid serious vision loss. So, this paper proposes a study on an automated diabetic retinopathy detection system which will estimate in which stage is the diabetic retinopathy from eye retina images for diabetic’s patients by using deep learning techniques, where we demonstrate the use of Inception V3 a convolutional neural network (CNN) architecture in training fundus image and detecting the DR class. This system has a significant role in the health care sector where it will help the ophthalmologist working in the hospitals in diagnosis diabetic retinopathy and this system will be considered as a functional, time and cost-effective method for automated screening of diabetic retinopathy (DR). Also, we have reported some literature reviews for different research studies and conducted a comparison between three CNNs models which are VGG16, Inception V3, and ResNet-50.