MODELS AND ALGORITHMS UTILIZING DEEP LEARNING FOR THE DETECTIONAND ANALYSIS OF EYE DISEASES BASED ON MEDICAL IMAGING

Abstract

This article describes one of the serious consequences of diabetes, the negative impact on the visual system.The purpose of the work is to review the resources devoted to the problem of diagnosing diabetic retinopathy from eye images using neural networks. Materials and methods.The use of modern methods, approaches and algorithms is considered at stages such as data set collection and preparation, data pre-processing, image recognition task, transfer learning, comparison of methods, model ensembles, system development. Possible promising steps in future research are outlined.Results.During the analysis of publications on methods of diagnosing diabetic retinopathy using neural network-based eye images, the following directions for improving the existing results were identified: increasing the image data set, image pre-processing methods, interpretation of the neural network model, computational power implementation algorithms on mobile devices, classification problems and eye lesion segmentation, false negative and false positive diagnoses, model ensembles, using recurrent and capsule neural networks.