Automatic classification of retinal vessels into arteries and veins

Abstract

Many studies have shown that irregularities on the retinal vascular structure may be used to diagnose several pathologies such as diabetes, hypertension, stroke or heart diseases. Since these alterations affect differently arteries and veins, a precise characterization of both types of vessels is a key issue in the development of automatic diagnosis systems. This paper proposes a simple methodology to classify the vessels automatically based on a pre-segmentation and a local clustering strategy. The vessels are classified locally in overlapping areas centered on the optic disc and the results of each area are combined to make the final decision. The accuracy of the proposed methodology was proven using 20 images manually labeled by experts. The results of the tests showed high sensitivity and specificity. Copyright © 2011 ACTA Press.