Abstract
Hypertensive Retinopathy is a retinal disease that results in vision loss and this disease is closely related to high blood pressure. With the advancement of computer technology in medical science, the development of automatic systems for detecting ocular diseases such as hypertensive retinopathy and diabetic retinopathy is becoming possible. Most of the retinal diseases results in change of blood vessels caliber. Arteriovenous Ratio, which is the ratio of arteriolar diameter to venular diameter, is an important parameter for the diagnosis of hypertensive retinopathy. Blood vessels are classified into arteries and veins which is a crucial step for the calculation of Arteriovenous Ratio. Current paper gives a novel method for differentiation of vessels. Firstly a region of interest is defined by localization of optic disk. Then the blood vessels lying within region of interest is segmented using bottom hat filtering. In the second phase 40 statistical and texture features are extracted from the segmented vessels. Scaled conjugate gradient back propagation neural network is applied on the best 18 features for classifying vessels into arteries and veins. Locally available fundus images dataset is used for performance evaluation of the proposed methodology Accuracy of 84.5% is obtained. Comparison of the proposed method with other work is done.