Dynamic variations of coronary arterial curvilinearity have been very difficult to study in the current angiograms. However, there is both experimental and clinical evidence showing that vessel extraction is very useful for the purposes of surgery treatment and clinical study. We propose in this paper an automated algorithm to locate the outline of the coronary artery tree blood vessels in angiograms. The proposed approach is a useful tool for physicians. The algorithm automatically segments the coronary arteries from the cineangiogram followed by an accurate extraction of the vessel features. Such a pre-processing in conjunction with the matched (Gaussian) filter can greatly improve the results. The proposed segmentation algorithm consists of two major processes: (1) Gaussian filter for blood vessels and (2) thresholding. We evaluated our algorithm by testing it on a raw data set of 100 angiogram images and the results were validated by two ways. First, hand-labelled annotations of ground truth segmentation of 20 images; and the results showed that our algorithm is better in detecting features of arteries even with poor contrast that bare eye could not recognize it in the hand-labelling. Secondly, by making a questionnaire to validate the efficient illustration output. The hand-labelling matched our results by 98% of the output while the validation rate of questionnaire was 90.84%. We conclude that our improved algorithm is efficient for extracting the coronary artery tree vessels including the tiny ones. In addition, the algorithm was fast to extract vessels that it takes between 14 to 15 s per image.
Key words: Angiocardiography, coronary artery segmentation, matched filter, adaptive thresholding, vessel extraction.
Copyright © 2023 Author(s) retain the copyright of this article.
This article is published under the terms of the Creative Commons Attribution License 4.0