Motion compensated dynamic MRI reconstruction exploiting sparsity and low rank structure

In this paper, we propose a motion compensated dynamic magnetic resonance imaging (MRI) reconstruction method based on compressed sensing. First, a motion compensation method is used to improve the sparsity in temporal finite difference domain and the nuclear norm of the low rank property. Furthermore, the effective regularization terms are designed to enforce the low rank structure of dynamic scenes and sparsity in finite difference domain along spatial and temporal dimension simultaneously. To efficiently solve the proposed corresponding optimization problem, we decouple this problem into four sub-problems. Demons algorithm and Fast Composite Splitting Algorithm (FCSA), iterative shrinkage thresholding algorithm (ISTA) and conjugate gradient (CG) algorithm are employed to efficiently solve these sub-problems. The performance of the proposed method was evaluated on dynamic cardiac MRI dataset and experimental results demonstrate its effectiveness and robustness comparing with the current methods in CS dynamic MRI reconstruction.