Predicting Eye Blindness by Detecting Exudates in the Retina of Human Eye

Abstract

Diabetic retinopathy is a condition where a person suffering from diabetes starts to loosen his vision slowly as the severity of the disease increases gradually. We can diagnose this condition by the fundus image of the retina of the human eye; although it is very complicated for doctors to predict the conditions just by seeing the fundus images. By detecting diabetic retinopathy at the earliest, we can protect patients from losing their vision. Hence, this study proposes a computer-assisted diagnosis based on digital retinal image processing to assist people who detect diabetic retinopathy beforehand. The main objective is the automatic classification of the degree of non-proliferative diabetic retinopathy in any retinal image. Nowadays, there are many methods with which we can simplify this problem. One of the best ways is to use machine learning to predict diabetic retinopathy. In machine learning, we can also use many techniques, so we have used CNN and SVM, which are very popular image classification techniques.So the proposed system predicts diabetic retinopathy disease using a deep convolutional neural network and support vector machine with the help of fundus images of the human eye's retina. These models can deliver superior performance for early-stage diagnosis, thus helping the doctors todiagnose the diabetic retinopathy disease accurately.