This paper presents a novel methodology for precise detection of both sides of the arterial walls (lumen border and vessel border) in multi-slice computed tomography (MSCT) angiography for coronary artery disease analysis. First, a 3D region growing strategy with a volume limitation quickly determines the center of the coronary artery for the region of interest. The vessel centers are then aggregated to form the central curve of the vessel. For each of the points along the central curve, the volumetric data is processed to generate a target image slice (TISL) which is perpendicular to the curve at that point. Because of the optimal orientation, the TISL image yields good dimensions for the vessel at that point. Organizing the TISL images linearly as a target image stack (TIST) forms a straightened longitudinal version of the vessel, and the TIST longitudinal cross-section produces a good anatomic structure relation image. Secondly, a proposed dual snake model is then used to simultaneously detect the inner and outer vessel wall borders for each TISL. The automatic initialization of the deformable model is achieved by integrating detected vessel boundaries on four TIST longitudinal images based on a minimum cost path method. For validation, vessel diameter and area determined automatically by the proposed method were compared with those defined manually by experts and by corresponding conventional angiography (r = 0.9160 ∼ 0.9740, p < 0.0001). The proposed method achieved sub-pixel accuracy. The entire process is computationally fast and simple. The proposed system (methodology) could provide a fast and reliable clinical tool for physicians in the diagnosis and treatment planning of coronary diseases.
|Number of pages||7|
|Journal||Journal of Medical and Biological Engineering|
|Publication status||Published - 2007 Dec 1|
All Science Journal Classification (ASJC) codes
- Biomedical Engineering