Coronary Artery Segmentation

Besides knowing a path through the coronary arteries, it is also important to segment their whole lumen from the surrounding tissue. This information is vital for the assessment of atherosclerosis. First of all, the lumen can be used as a mask to limit the plaque detection within a CT data set to a certain region. Furthermore, the morphology of the lumen itself can be analyzed. In general, plaques cause a change of the vessel wall, either an outward expansion referred to as positive remodeling or a vessel lumen shrinkage, also called negative remodeling.

The challenges for coronary artery segmentation are comparable to those for the tracking of coronary arteries. Stenoses or motion artifacts may lead to a premature termination of the segmentation process whereas crossing of adjacent structures with similar intensity characteristics may lead to a leakage of the segmentation process into these undesired regions. Therefore, constraints about the tubular anatomical shape of the coronary arteries are often included in the segmentation process.

For the segmentation of the coronary arteries, we used a graph-cut approach, because graph-cuts have shown to be robust and only few parameters are needed for their adoption to specific problems. The graph-cut algorithm was combined with a shape-prior and a prunning process. For the segmentation, the tracked coronary artery tree was reformatted along its path.