Screening for Diabetic Retinopathy Using Computer Vision and Physiological Markers

Abstract

A direct computer vision approach enabled high-quality identification of exudates and DHs in an independent data set of fundus images. The methods are readily generalizable to other clinical manifestations of DR. The results justify a blinded clinical trial of the system to prove its capability to detect, diagnose, and, over the long term, monitor the state of DR in individuals with diabetes.