Microsoft Word - ISMRM2009-002937.DOC

INTRODUCTION: Correction of respiratory motion represents a major practical problem in quantitative myocardial perfusion MRI. Several registration methods exist to correct the motion of the heart in myocardial perfusion MRI [1-4]. However, these methods suffer from a number of important limitations: 1) they are limited to correction of translation (and in some cases rotation), 2) they require user-interaction, e.g., to specify the location of the heart, 3) they can be corrupted by image brightness changes caused by the contrast agent, and 4) they are unsuitable for correcting motion artifacts in accelerated dynamic MRI, since aliasing artifacts prevent reliable detection of the motion. The aim of our study was to device a registration method that overcomes these problems. The proposed method uses predictive modeling, controlled by respiratory navigators. The method automatically determines from a set of training frames the correlation between the navigators and the respiratory motion of each pixel. This information is then used to predict and correct respiratory motion of the perfusion series. The main advantage of this strategy is that it separates the registration task from the image content, thereby reducing the influence of the contrast passage and image artifacts.