Segmentation of coronary angiograms using a vesselness measure and evolutionary thresholding

Abstract

This paper presents a new two stage automatic vessel segmentation method of coronary arteries in X-ray angiograms. In the first stage of the method, a multiscale vesselness measure based on the properties of the eigenvalues of the Hessian matrix is used to detect vessel structures. This vesselness measure was compared with three vessel detection methods using the area under the receiver operating characteristic curve. In the second stage, the magnitude response of the vesselness measure is segmented by a new evolutionary thresholding method using the weighted sum method for multi-objective optimization. This evolutionary method was compared with five classical thresholding methods using five statistical measures and a training set of 20 angiograms. Finally, the proposed method was compared with four state-of-the-art vessel segmentation methods. Experimental results provided an area of 0.8970 with the training set, and an average performance of statistical measures of 0.8302 with a test set of 20 angiograms.