Space-time variant weighted regularization improves motion reconstruction in compressed sensing accelerated cardiac cine MRI

Abstract

In compressed sensing (CS) dynamic MRI, temporal sparsity is commonly exploited introducing a temporal regularization that affects the dynamic behavior of the images in moving regions. While previous works proposed to combine different sparsity terms accounting for dynamic and static regions in the image, in this work we propose a methodology to dynamically adapt the temporal regularization according to the presence of motion. The proposed method is based on a robust registration technique for non-rigid motion estimation. A variable Density k-space sampling it is applied to highly accelerated breath-hold cine data.