Dictionary-free MR Fingerprinting reconstruction of balanced-GRE sequences. Sbrizzi, A., Bruijnen, T., van der Heide, O., Luijten, P., & Berg, C. A. T. v. d. arXiv:1711.08905 [physics], November, 2017. arXiv: 1711.08905
Dictionary-free MR Fingerprinting reconstruction of balanced-GRE sequences [link]Paper  abstract   bibtex   
Magnetic resonance fingerprinting (MRF) can successfully recover quantitative multi-parametric maps of human tissue in a very short acquisition time. Due to their pseudorandom nature, the large spatial undersampling artifacts can be filtered out by an exhaustive search over a pre-computed dictionary of signal fingerprints. This reconstruction approach is robust to large data-model discrepancies and is easy to implement. The curse of dimensionality and the intrinsic rigidity of such a precomputed dictionary approach can however limit its practical applicability. In this work, a method is presented to reconstruct balanced gradient-echo (GRE) acquisitions with established iterative algorithms for nonlinear least-squares, thus bypassing the dictionary computation and the exhaustive search. The global convergence of the iterative approach is investigated by studying the transient dynamic response of balanced GRE sequences and its effect on the minimization landscape. Experimental design criteria are derived which enforce sensitivity to the parameters of interest and successful convergence. The method is validated on simulated and experimentally acquired MRI data.
@article{sbrizzi_dictionary-free_2017,
	title = {Dictionary-free {MR} {Fingerprinting} reconstruction of balanced-{GRE} sequences},
	url = {http://arxiv.org/abs/1711.08905},
	abstract = {Magnetic resonance fingerprinting (MRF) can successfully recover quantitative multi-parametric maps of human tissue in a very short acquisition time. Due to their pseudorandom nature, the large spatial undersampling artifacts can be filtered out by an exhaustive search over a pre-computed dictionary of signal fingerprints. This reconstruction approach is robust to large data-model discrepancies and is easy to implement. The curse of dimensionality and the intrinsic rigidity of such a precomputed dictionary approach can however limit its practical applicability. In this work, a method is presented to reconstruct balanced gradient-echo (GRE) acquisitions with established iterative algorithms for nonlinear least-squares, thus bypassing the dictionary computation and the exhaustive search. The global convergence of the iterative approach is investigated by studying the transient dynamic response of balanced GRE sequences and its effect on the minimization landscape. Experimental design criteria are derived which enforce sensitivity to the parameters of interest and successful convergence. The method is validated on simulated and experimentally acquired MRI data.},
	language = {en},
	urldate = {2021-02-12},
	journal = {arXiv:1711.08905 [physics]},
	author = {Sbrizzi, Alessandro and Bruijnen, Tom and van der Heide, Oscar and Luijten, Peter and Berg, Cornelis A. T. van den},
	month = nov,
	year = {2017},
	note = {arXiv: 1711.08905},
	keywords = {Physics - Medical Physics},
}

Downloads: 0