The Detection of Diabetic Retinopathy in Human Eyes Using Convolution Neural Network (CNN)

Abstract

A medical situation in diabetic patients is recognised as diabetic retinopathy, primarily involving human eye. Provenance of diabetic retinopathy is because of high blood glucose levels in past prolonged course of era called as diabetes mellitus. Diabetic retinopathy dataset has 5 levels of images present with levels numbered from 0 to 5 with initial level having mild signs of retinopathy to the last levels having no retinopathy. In high-resolution pictures of retina, system should segregate the pictures whether the patient has no diabetic retinopathy or has diabetic retinopathy. Originally, the pictures should be preprocessed by rotations and also need to be resized to standard image size so that the system can process the images with same efficiency. Then, deep learning approach of convolutional neural network (CNN) is applied to convert image which tells whether the patient is having diabetic retinopathy or not. The conclusions are alleged to conclude a sensitivity of 95% and a competence of 75%. So, this system can easily analyze retinal images whether healthy patients and diabetic patients diminishing the quantity of surveys of specialists.