Computation of coronary perfusion territories from CT angiography

Abstract

The assessment of myocardial viability in the left ventricle (LV) of patients with coronary artery disease is essential for determining prognosis and planning appropriate therapy. Typically, population-based models are used to associate regions containing non-viable (dead) myocardial tissue with a particular coronary artery. We propose a technique to automatically generate patient specific coronary distribution maps based on noninvasive multi-slice computed tomography (MSCT). We associate a region with the closest coronary using geodesic distance measurement over the LV epicardial surface. Projection of the coronary territories onto a flat disk allows our maps to be compared to standard models (i.e., the 17-Segment Model). This technique was applied to seven cases and an experienced cardiac radiologist examined the results for validation. Additionally, we register the patient-specific coronary territory maps with myocardial viability maps derived from noninvasive magnetic resonance imaging (MRI).