Efficacy of deep learning techniques in detecting Diabetic Retinopathy

Diabetic Retinopathy (DR), a primary factor in blindness globally, affects over 93 million people. Early detection is crucial, but it often remains asymptomatic until advanced stages, posing a challenge for timely treatment. Diabetic individuals commonly experience a health condition called Diabetic Retinopathy in the eye retina. Currently, DR diagnosis is manual and intensive, requiring expert evaluation of retina photos. Long-term increased blood sugar levels in the retina, a disease linked to diabetes mellitus, the primary reason behind diabetes retinal necrosis (DR). The main objective is to inevitably identify patients with and without DR. the following study introduces a computerized design using DL models of Convolutional Neural Network(CNN) to categorize fundus photos, enhancing the speed and accuracy of DR identification. Subsequently, a Convolutional Neural Network is taught with the previously processed picture using a DL technique to determine the individual's diabetes status. In this study, we employed the methodology on a set of data sourced from Kaggle, as well as the images collected by the EyePACs. In this paper, the achieved accuracy is 99%. By classifying retinal images into those with diabetes and those who are healthy, the system which is an automated like this can reduce the amount of reviews left for medical professionals.