Chapter 12 - Diabetic retinopathy and maculopathy lesions

Abstract

Diabetic retinopathy (DR) and maculopathy lesions are biomarkers of disorder that is the main cause of sight loss in a working-age population. They are the consequence of a continuous exposure to excessively high level of sugar in blood and systemic high blood pressure. DR is an asymptotic disease that manifests itself through the appearance of lesions, which localization is a major indicator of eye’s health. This chapter reviews the latest fundus photography-based studies in computational health care, which focused mainly in the analysis of the spatial distribution of DR appearing in the retinal surface, and developed algorithms to automatically detect DR lesions. The analysis highlights the need to introduce automated systems to the clinical practice, to support ophthalmologists’ daily activities. In fact, a prompt detection of lesions can reduce the risk of developing sight impairment, especially when lesions’ first appearance is below retinal horizontal meridian. Furthermore, lesion detection and segmentation are currently manually performed by the figure of the ophthalmologist, and the severity is assessed by retinopathy graders and screeners, upon visual inspection of the ocular fundus. This is a highly error-prone and subjective process, which highlights the necessity of creating automated systems that can be deployed in clinical scenarios. In this context, deep learning approach has been shown to perform at a level comparable to humans.