Segmentation based registration of myocardium in cardiac perfusion images
DiBella E, Adluru G
University of Utah, University of Utah
Registration and segmentation of myocardial perfusion images are important steps to extract qualitative and quantitative information. The two steps can be used to improve each other. We propose a method to achieve accurate registration of myocardium which uses a level set segmentation method to improve registration. A coarse registration is first performed to aid the automatic segmentation of the left ventricular blood pool and the segmentation is then used to improve the registration of the left ventricular myocardium. The method is tested on cardiac data and compares well with manual registration.