Differential Diagnosis of Retinal Edema Based on OCT Image Analysis

Abstract

The aim of the work is differential diagnosis of retinal edema, study of deep learning methods and their application to image analysis. The application of convolutional neural networks to the task of semantic segmentation of retinal layers was investigated and its efficiency in selecting two selected layers (pigment epithelium and retina) was proved. A disease classifier based on intelligent analysis of the layers extracted by the neural network was implemented. A proof of its applicability for differential diagnosis of retinal edema was presented. The accuracy of disease prediction amounted to 90%.