Accelerated MR Fingerprinting Reconstruction Using Dictionary and Local Low-Rank Regularizations

Magnetic resonance fingerprinting (MRF) is a novel quantitative magnetic resonance imaging technique that simultaneously estimates multiple tissue parameters (e.g. spinlattice relaxation time T
1
, spin-spin relaxation time T
2
, proton density). However, the high undersampling of k-space data, which is employed to reduce scanning time, results in severe artifacts in the recovered MRF data, and subsequently leads to inaccurate estimation of tissue parameters. In this paper, we propose a new MRF reconstruction method in which dictionary and local low-rank priors are combined to help recovering MRF data from its highly undersampled and noisy k-space data. In addition, an acceleration scheme, which utilizes a singular value decomposition (SVD) to project the signal evolutions onto the principal temporal components of the predefined dictionary, is applied to reduce the computational burden of the proposed method. The numerical experiments demonstrate that the proposed method outperforms the state-of-the-art methods for estimation accuracy of tissue parameters with a slight cost of computation time.