Multi-Level Classification of Retinal Disease using SFCN

Abstract

Premature detection of diabetes using retinal images is still challenging. The severity of diabetic retinopathy cannot always be detected or assessed with the current techniques. Our research uses a convolutional neural network to extract the retinal image's properties and pinpoint the stages at which a normal condition becomes an abnormal one in order to identify the presence of diabetes. The retinal images are categorized using a Supervised Fuzzy Clustering Network (SFCN) based on the severity of diabetic retinopathy. Feature learning, reconstruction, and self-supervised fuzzy clustering modules make up the SFC network. The network's ability is secured by the feature learning and reconstruction module, while the fuzzy self-supervision module gives the network guidance. We use retinal dataset into our model to analyse the performance and project the outcome in order to assess the network's effectiveness.