Early Diabetic Retinopathy Detection Using Convolution Neural Network

Abstract

One type of diabetes is diabetic retinopathy, which results in vascular abnormalities that can result in blindness. Because the effects of this disease are irreversible, early detection is essential because unchecked eye disease can lead to blindness. A crucial first step in automated screening for diabetic retinopathy is the detection of microaneurysms in digital color fundus images. Normal, mild, moderate, severe, and PDR are the five DR stages or grades (proliferative diabetic retinopathy), colored fundus images are often examined by highly qualified specialists to identify this catastrophic condition. Manual diagnosis of this illness (by clinicians) takes time and is error-prone. Numerous computer vision-based methods for automatically identifying DR and its various stages from retina images have so been developed. This study uses fundus-colored images to demonstrate an automated strategy for detecting this condition early and classifying its severity using a convolution neural network.