Cardiovascular segmentation based on Hough transform and heuristic knowledge

Nowadays cardiovascular diseases are one of the most major causes of mortality. Computed Tomography Angiography (CTA) is a very useful imaging tool for cardiovascular disease diagnosis. So it is important to analyze CTA images well. This paper proposed a new method for fully automatic cardiovascular segmentation based on combination of Hough transform and region growing algorithm. It is a robust method which segments ascending aorta, descending aorta, and left ventricle concurrently. Comparing to the manual method which is done by cardiologist and previous automatic and semi- automatic works, our method is faster, more accurate, and fully automatic. This procedure also can be applied to coronary segmentation. The validation of the acquired cardiovascular images is evaluated by a cardiologist. By evaluating 10 datasets, which contain about 5000 images, the accuracy of the method is 97.3% comparing to the gold standard. Our gold standard is the images segmented by cardiologist. In addition, average elapsed time is 0.18s per image.