5th World Conference on Smart Trends in Systems, Security and Sustainability

Authors - A.Aruna Kumari, Santosh Kumar Henge
Abstract - The Diabetic Retinopathy is one of the eye condition that may cause vision loss, blindness to the eyes quickly if not treated at the earliest in the patients having Diabetes Milletus type-1 and Type-2. This can be treated when diagnosed at the early stages of the disease with screen monitoring and treatment like laser etc. If not diagnosed in early stages can cause the impaired vision and may also lead to blindness. To avoid blindness, vision loss, the "Early Treatment Diabetic Retinopathy Study (ETDRS)" helps to detect eye disease at the earliest and prevent the consequences of the disease . As a progressive disease, The Diabetic retinopathy divided into two stages one is "non proliferative diabetic retinopathy" (NPDR) and "proliferative diabetic retinopathy" (PDR). Non-proliferative diabetic retinopathy has four stages, those are normal, mild, moderate, severe which has the threshold as 0,1,2,3 respectively for No Diabetic Retinopathy, mild, moderate, severe, but if the threshold value is 4 it is a proliferative diabetic retinopathy is last stage of the DR. Unlike the available methods, the proposed work aims at diagnosis of all the stages of Diabetic Retinopathy where the retinal image preprocessing followed by the analysis using Convolutional Neural Network Classifier which classifies into different stages of the disease based on the threshold. RESTNET is used for the image tuning and the Hyper parameter tuning is done for detailed analysis. The Kaggle public dataset is used to build the model and would give the higher performance when compared to the existing ones.