A Systematic Review on Diabetic Retinopathy and Common Eye Diseases Detection through Deep Learning Techniques

Abstract

India has approximately 15 million blind people, and the unfortunate reality is that 75 percent of these cases are curable. In India, the doctor-patient ratio is 1:10,000. According to studies, the leading causes of blindness in India are diabetic retinopathy (DR) and glaucoma. Diabetic retinopathy is caused primarily by a person's diabetes and is the leading cause of blindness among working-age people in both developed and developing countries. Glaucoma damages the optic nerve, resulting in blindness. Both diseases are asymptomatic in their early stages, making detection difficult, and if left untreated, they can cause irreversible vision damage. Early detection of diabetic eye disease using an automated system has significant advantages over manual detection as a result of advances in machine learning techniques. A number of advanced studies on diabetic eye disease detection have recently been published. This paper presents a systematic survey of automated approaches to diabetic eye disease detection from a variety of perspectives, including i) available datasets, ii) image pre-processing techniques, iii) deep learning models, and iv) performance evaluation metrics. The survey provides a comprehensive overview of diabetic eye disease detection approaches, including cutting-edge field approaches, with the goal of providing valuable insight into research communities, healthcare professionals, and diabetic patients.