Using Convolutional Neural Network to Detect Diabetic Retinopathy in Human Eye

Abstract

A medical condition in diabetic patients is recognized as diabetic retinopathy; it is a primarily involved disease in human eye. This disease needs to be detected by ophthalmologist which can cause incorrect results and also in a tedious process. Hence, a need for automated detection of these diseases without human intervention is necessary such that drawbacks from misconception from the humans are prevented. The data set used for the same is being downloaded from Kaggle website [11]. The systems detect this disease by scanning the retinal images of human to check whether the patient is having diabetic retinopathy or no. Also, a fewer pre-processing of the data set’s images needs to be done in order to get the images to a standardized form. Later, the convolutional neural network (CNN) is being used because of the fact that neural networks acquire the approach of biological brain which understands and learns the various patterns and nerves in human eye to conclude whether the patient has diabetic retinopathy or not. This system uses 1600 and 400 images as training and testing data, respectively. The accuracy of this system comes out to be 80%.