A study of deep learning methods to diagnose Diabetic Retinopathy

Abstract: Damage to the retina that impairs vision is a Diabetic Retinopathy (DR), a common side effect of diabetes mellitus and it may cause blindness if not caught in time. Regrettably, DR is irreversible, and therapy can only preserve eyesight. If DR is identified and handled right away, visual loss may be prevented. In contrast to computer-aided diagnostic techniques, it takes a lot of time, effort, money, and misdiagnosis for ophthalmologists to manually diagnose DR retina fundus photos. When it comes to medical picture analysis and classification, deep learning has quickly become among the most well-liked methods due to its improved performance. In medical image processing, convolutional neural networks (CNN) are most well-liked and successful deep learning (DL) technology. In this paper, we analyse and examine the advanced methods for applying deep learning to identify and categorise DR colour fundus images. Furthermore, an overview of the fundus retina DR colour datasets that are currently accessible has been provided. We also talk about some of the other difficult subjects that need additional research.

Keywords: Diabetic Retinopathy, Diabetes, Deep Learning, Convolutional Neural Networks, Retina.