Extraction of Coronary Arterial Tree Using Cine X-ray Angiograms

Coronary angiography is still the most common modality for physicians to assess the severity of vessel narrowing or stenosis during percutaneous coronary intervention procedure. Accurate quantitative analysis of coronary arteries in digital angiographic images is valuable and important to clinical needs. Computerassisted extraction of a set of major arteries or the entire coronary arterial tree from two-dimensional (2D) angiograms is regarded as a crucial process. Once the vessels are identified, additional techniques may be applied to obtain quantitative information including severity of stenosis, three-dimensional representation of the vascular tree, motion analysis of the coronary arteries, or blood-flow analysis [1-8]. The major difficulty in automatic extraction of coronary arterial structures in angiogram lies in (1) low signal-to-noise ratio due to poor X-ray penetration, (2) vessel overlaps, and (3) superimposition of other tissues such as ribs, spine, or cardiac chambers. Traditional signal-based edge detection algorithms [915] were unable to effectively or accurately detect the desired structures. The existing methods specific to vessel extraction can be categorized into (i) modelbased [16-18] (ii) tracking-based [19-21], (iii) classifier-based [22], and (iv) filter-based [23-25] techniques. In model-based methods, the coronary arterial tree is produced based on a pre-defined coronary artery model in the form of a graph structure. In tracking-based methods, the process proceeds with an initial start-of-search location followed by an automatic tracking process by ABSTRACT