A Survey on Detection of Diabetic Retinopathy Lesions Using Deep Learning

Diabetic Retinopathy is a increasing reason for blindness among working-aged adults in the world. DR is a progressive disease and early detection is very much needed for saving diabetic patients. The blood supply to retina is through tiny vessels which is uncontrolled sugar levels in the blood. When there is a large glucose level in the blood, large lag of oxygen occurs in the cells, as a result blockage occurs in the blood vessels. Any blockage in vessels lead to eye injury and leads to thickening or any abnormal blood vessel which causes DR. DR has many indications like presence of Microaneurysms (MAs), Exudates (EXs), Haemorrhages (HMs), Vascular abnormalities and abnormal growth of retinal blood vessels. This survey paper analyses various reasons for DR, starting from the view of DR, its data sets, Pre-processing techniques followed, extraction of features and detailed analysis of efforts to develop Deep learning models for DR detection.