Segmentation based registration of myocardium in cardiac perfusion images

G. Adluru, E. V. DiBella Electrical and Computer Engineering, University of Utah, Salt Lake City, Utah, United States, UCAIR, Department of Radiology, University of Utah, Salt Lake City, Utah, United States, Department of Bioengineering, University of Utah, Salt Lake City, Utah, United States Introduction: Dynamic cardiac perfusion images often suffer from motion artifacts due to respiration. Registration and segmentation of the left ventricular myocardium in the images are important steps to extract useful qualitative or quantitative information about myocardial perfusion. The two steps can be used to improve each other. We propose a robust registration algorithm using a pixel wise minimization of mean squared difference registration method and a level set based segmentation to improve the accuracy of the registration.