Classification of Diabetic Retinopathy using Convolutional Neural Network

Abstract

Diabetic Retinopathy is a scenario in medical field which leads to the rise of damage of blood vessels in the retina which is due to diabetes mellitus. The suitable detection for this kind of problems and care to be done immediately in order to prohibit loss of sight in a person. Presently, diagnosing Diabetic Retinopathy manually is a time- consuming process where they require experienced clinicians to examine the digital-colored fundus images. Here, we have proposed a machine learning technology using Convolutional Neural Network (CNN) approach which has emerged as an operative productive tool in medical image examination for the classification and detection of Diabetic Retinopathy (DR) in real-world. The different layers which are used to detect the brain tumor are conv2D, Activation, MaxPooling2D, Dense and Flatten. The set used here considers 750 retinal images, with 600 training images and the test set considers 150 images with the accuracy of 82.75% which ran for 80 epochs.