Detection of Diabetic Retinopathy with Retinal Images using CNN

Abstract

Diabetic Retinopathy (DR) is an eye condition that develops in diabetics, causing retinal damage and, in the long term, visual impairment. It has been predicted that 40 million people in the World could be blind due to Diabetic Retinopathy by 2025. DR is currently being tested manually by ophthalmologists, which is a time-consuming operation. Therefore, in this paper, a Deep Learning based Algorithm is proposed for the Automatic Detection of Diabetic Retinopathy by sorting high-resolution fundus images. Specifically convolutional neural network (CNN) technique is used to train the dataset that classifies the retinal images into infected and unaffected images. The dataset used to train the model is comprised of 757 colored retinal images and the proposed model is tested upon 151 images. The simulation results are presented to validate the proposed scheme. It has been demonstrated that the proposed CNN-based algorithm can achieve 99.5% accuracy, 97.6% sensitivity, and 91.24% specificity as compared to the existing algorithms.