Coronary Artery Tracking
For the tracking of coronary arteries in CT coronary angiography data sets, we address several prerequisites to our algorithm. First of all, it must be able to deal with atherosclerosis which is reflected by intensity changes within the coronary arteries caused by stenoses or different types of plaques. Further, it should be able to handle motion artifacts. Although the latest CT system generation provides good temporal resolution, motion artifacts still cannot be excluded which result in spatial offsets of the coronary arteries between slices in a CT data set such that an automatic tracking stops in general, and has to be manually re-initialized. Finally, the coronary artery tracking algorithm should be fast and robust.
Besides the tracking of the coronary arteries, we further wanted to extract the coronary artery tree, i.e. the hierarchical branching information between the tracked vessels. This information allows - in a next step - a labeling of the extracted branches according to the American Heart Association scheme and therefore simplifies the automatic annotation of detected lesions within the tracked coronary tree.
To deal with all the aforementioned requirements, we developed a semi-automatic coronary artery tracking based on the live-wire approach. This approach allows user-interaction to overcome motion artifacts or to bridge stenoses and it is further flexible in its design such that it can be easily extended with additional cost functions. As current cost function, we apply an intensity and gradient feature.
