Ricardo Otazo1, Cagdas Bilen2, Yao Wang2, Leon Axel1, Daniel K. Sodickson1
1Bernard and Irene Schwartz Center for Biomedical Imaging, NYU School of Medicine, New York, NY, United States; 2Department of Electrical Engineering, Polytechnic Institute of NYU, Brooklyn, NY, United States
Low-rank matrix completion is proposed as a new generalized approach to combining compressed sensing and parallel imaging by jointly exploiting implicit temporal and coil correlations without an explicit sparsifying transform or coil calibration procedure. A low-rank k-t matrix can be obtained by concatenating overlapping k-space blocks from consecutive time points and multiple coils to form the different columns. Reconstruction of k-t undersampled data is performed using an iterative singular value thresholding algorithm. We demonstrate the feasibility of reconstructing undersampled cardiac cine data.