Image Processing Techniques for the Quanti cationof Atherosclerotic

This paper describes the design and implementation of an oo-line, non-invasive, automated method for the examination and follow up of the arteriosclerotic changes due to hypertension, with the help of digital image processing of fundus images. This method would help in evaluating the eecacy of various treatments on the regression and reversion of arteriosclerotic lesions. This method, in interaction with appropriate knowledge bases, can be used at the clinical practice for monitoring hypertensive patients on a frequent basis, hence it aims at minimum discomfort of the patient, bypassing even the regular uorescein injection for fundus image enhancement. Our method is based on segmenting the vasculature by identifying the centerline of each vessel utilizing the idea that vessels present a ridge in cross-sectional intensity prooles. Therefore, such a ridge can be detected along the vessels, as if there was three-dimensional information. Once the vasculature is segmented we present image-based measuring techniques for length to next bifurcation, vessel calibre, wall thickness and we introduce a novel measure of tortuosity. All our measurements are automatic, with minimal assumptions, and they are calibrated by means of the papilla which is considered of standard size. To achieve this, we implement a locating technique for nding and measuring the papilla on fundus images.