Abstract:
This paper presents an automatic method to find and segment the boarders of the left and right ventricles (LV and RV) from the middle short axis (SA) slice of a functional cardiac MRI study. The segmentation is implemented using a new atlas-based registration framework which is able to integrate boundary, intensity and anatomical information. The method is shown to be accurate, robust and reliable in a clinical trial of 330 patients with a range of cardiac and vascular disease. The method failed in only two cases, and showed high agreement with expert observers in the remaining 328.
