Reconstruction of free-breathing myocardial perfusion MRI using simultanous modeling of perfusion and motion (SMPM) and arbitrary k-space sampling

INTRODUCTION: Respiratory motion of the heart represents a major practical problem in myocardial perfusion MRI. Firstly, the tracer kinetic models used for quantifying perfusion require that the myocardium remains stationary throughout the image time series, thus mandating a prior registration of the images. Secondly, modern k-space undersampling techniques, such as k-t BLAST [1] and UNFOLD [2], suffer from residual aliasing artifacts in the presence of respiration [3]. Therefore, perfusion quantification as well as imaging speed-up techniques would benefit from a computational framework that allows simultaneous modeling of perfusion and motion (SMPM). Ideally, the SMPM framework should work for arbitrary k-space sampling strategies (i.e., it should also work for undersampled k-space data). Awate et al. [4] has presented a framework that models myocardial perfusion for arbitrary k-space sampling based on a one-compartmental perfusion model. This work shows how to also include respiratory motion in that framework, thereby improving its clinical potential. We demonstrate the SMPM concept for a representative freebreathing myocardial perfusion data set using 1) fully sampled k-space data, 2) undersampled Cartesian k-space data, and 3) undersampled radial k-space data.