Novel compressive sensing MRI methods with combined sparsifying transforms

Compressive sensing (CS) is an emerging technique for fast MRI, which relies on the sparsity constraint of the underlying image to reduce the data acquisition requirement. Sparsifying transforms, such as total variation (TV), wavelet, curvelet, have been used in CS-MRI as regularization terms. Linear weighted summations of these regularization terms have also been used and tested. However, tuning the weights for individual terms is complicated and time-consuming. In this paper, a novel method that uses combined sparsifying transforms is proposed. This method applies transforms sequentially. It can avoid the artifacts associated with a single transform, as well as save the time of tuning the weights. Simulated results using in-vivo data show that the proposed method is efficient while providing similar or improved reconstruction quality.