An Empirical Study of Deep Learning Techniques for Diabetic Retinopathy Detection

Presently, a competent professional must perform a thorough manual examination of digital color fundus photographs of the retina in order to diagnose diabetic retinopathy (DR). The delayed results result in missed evaluation, inaccurate communication, and postponed therapy by the time technicians complete their evaluations, which is often a day or two later. By looking for lesions connected to the vascular anomalies brought on by the disease, doctors can spot DR. This strategy works well, but it uses a lot of resources. Where DR detection is most important and there is a high prevalence of diabetes in the area, there is typically a lack of the requisite information and resources. As the number of persons with diabetes increases, the infrastructure required to prevent DR-related blindness will become increasingly more insufficient. In order to diagnose and classify diabetic retinopathy, this paper aims to develop and deploy image processing and ML algorithms to analyze fundus images.