Microaneurysms and Exudates Detection in Retinal Images using Deep Neural Network

Abstract

One frequent diabetes consequence that affects the eyes is diabetic retinopathy(DR). The most frequent reasonfor blindness in working-age adults in the world is diabetic retinopathy. One in three diabetics has diabetic retinopathy to some extent. DR affects nearly all persons having type 1 diabetes and more than 60% of people having type 2 diabetes to some extent after 20 years of diabetes. In the US, approximately 4.2 million persons, 40 and older have diabetic retinopathy. In the United States, 12% of all new occurrences of blindness are caused by diabetic retinopathy.A 95% reduction in the risk of blindness is possible with a diabetic retinopathy early detection and treatment. The retina's appearance, the presence or absence of Microaneurysm and Exudates, and the degree of participation are all taken into account in the grading process. It has been demonstrated that deep neural networks (DNNs) work well for automatically grading diabetic retinopathy. The features like Microaneurysm and Exudates that are diagnostic of various stages of diabetic retinopathy are taught to these DNNs utilizing vast datasets of retinal pictures and accompanying grading information.It has been demonstrated that deep neural networks (DNNs) are efficient in automatically detecting diabetic retinopathy from retinal pictures. Sensitivity, specificity, precision, accuracy, and Kappa value are used to measure how well DNNs work in detecting diabetic retinopathy; these values are 95.74%, 92.31%, 96.77%, 94.74%, and 0.87, respectively.