Classification of retinal vessels into arteries and veins for detection of hypertensive retinopathy

Abstract

The damage to the retina of the eye caused by high blood pressure is known as hypertensive retinopathy (HR). The advancement in automatic retinal image analysis has made it possible to detect & diagnose this disease at an early stage. To indicate the presence and severity of Hypertensive Retinopathy, Arterio-Venous ratio (AVR) is an important measurement. For the calculation of AVR (the ratio of retinal arterioles to venuoles), classification of vessels into arteries and veins is an essential step. This paper proposes a novel technique for classifying the vessels into arteries and veins. The proposed system uses binary vessel mask and optic disc (OD) center localization to extract Region of Interest (ROI) around OD. Feature extraction modules extract a number of intensity based features followed by classification which classifies vessels as arteries and veins. The features are extracted by focusing on the visual difference between the arteries and veins in different color spaces. The proposed system is tested on a locally gather fundus image dataset. An accuracy of 81.3% is achieved.