Registration of Cardiac Magnetic Resonance Perfusion Data as a Basis for Quantification of Myocardial Perfusion

Quantification of cardiac magnetic resonance (CMR) myocardial perfusion remains time consuming since it requires manual intervention to compensate for motion. Thus, the aim of this study was to test an automated registration method. We studied 10 patients who had rest and stress CMR perfusion exams. For both exams, three short-axis slices were selected. Then, a rigid edge based registration algorithm was performed. Its quality was assessed 1) qualitatively by comparing the k-means clustering maps obtained before and after registration. 2) quantitatively by estimating noise amplitude within the myocardium. Registration substantially improved myocardial symmetry and heart structures identification on the k-means maps in 12/16 slices at rest and 22/27 slices at stress. It reduced noise amplitude from 48 +/- 26 to 28 +/- 10 at rest (p<0.05) and from 53 +/- 13 to 31 +/- 10 during stress (p<0.05). Our method performed successfully on both rest and stress CMR perfusion data.