Accelerated dynamic MRI via inter-frame motion estimation

Chuqing Cao, Ying Sun

Abstract: The sparsity of MR images has been utilized to significantly undersample k-space measurements for accelerated MRI. In dynamic MRI, besides the spatiotemporal structures of images, the motion information should be considered to improve the reconstruction performance. Motivated by this, we propose a new method to recover dynamic MR images using partial k-space data based on the estimation of inter-frame motion. Our method consists of three main steps: single frame reconstruction, inter-frame motion estimation, and image sequence recovery. In contrast to algorithms which use a single reference frame for motion estimation, the motion information of each image in a dynamic MRI sequence is estimated according to adjacent frames. Since motion is estimated from the reconstructed images, the recovery process is robust against both noise and artifacts. The proposed method was evaluated on two dynamic MRI datasets, and compared with several state-of-the-art reconstruction methods. Experimental results demonstrate the effectiveness and robustness of the proposed method.