MR Fingerprinting: Reconstruction Considerations
Debra McGivney1

1Radiology, Case Western Reserve University, United States

Synopsis

This lecture will outline various considerations associated with the quantification of tissue properties in the framework of MR fingerprinting

Objectives

This lecture will provide an overview of reconstruction methods for magnetic resonance fingerprinting (MRF), including some of the challenges that arise and their proposed solutions.

Methods

Traditional quantitative methods for MR generally involve measuring one tissue property at a time, estimated through exponential curve fitting. The novel framework of MR fingerprinting (MRF) allows for the simultaneous quantification of multiple tissue properties using one data acquisition.

In the original implementation, the problem of computing tissue properties values from the MRF signal evolutions was solved using pattern matching with a precomputed dictionary of simulated signal evolutions, using a wide range of tissue property values as inputs. This pattern matching, which uses the inner product as a measure of similarity between the acquired pixel signal evolution and the dictionary elements, is an exhaustive search, which guarantees that the best match from the dictionary will be found. One challenge which is associated with this method is that of dictionary size. Dictionary size is dependent upon several factors, including how many tissue properties the MRF sequence is sensitive to as well as the number of time points acquired. Additionally, since the dictionary is a discrete representation of signal evolutions associated with these tissue properties, the step sizes used for each property will impact the size of the dictionary and the accuracy of the matching. Various methods have been proposed to ensure accurate tissue property mapping in spite of the dictionary size.

Another challenge that arises in MRF is related to the sampling patterns typically used, which can cause severe aliasing artifacts seen in the reconstructed signal evolutions. Though the pattern matching method described previously is capable of producing accurate tissue property maps, this requires longer acquisitions to mitigate the effects of the aliasing artifacts. To this end, several iterative methods have been proposed to reduce the aliasing artifacts seen in the signal evolutions, which can result in shorter acquisitions required for accurate mapping.

This lecture will focus on the aforementioned issues in MRF reconstruction and the various methods that have been proposed to solve these challenges. These include a variety of approaches, including dictionary compression, fast dictionary matching, iterative reconstructions, and low rank methods.

Discussion/Conclusion

MRF offers a novel framework for computing tissue property values within a single acquisition. Estimating these tissue properties is a mathematical and computational challenge due to several factors, however, it is an active area of research with many promising results.

Acknowledgements

The author would like to acknowledge research support from Siemens Healthcare as well as NIH grants 1R01EB016728-01A1 and 5R01EB017219-02.

References

1. Assländer, J. et al. Low rank alternating direction method of multipliers reconstruction for MR fingerprinting. Magn. Reson. Med. 79, 83–96 (2018).

2. Cauley, S. F. et al. Fast group matching for MR fingerprinting reconstruction. Magn. Reson. Med. 74, 523–528 (2014).

3. Cline, C. C. et al. AIR-MRF: Accelerated iterative reconstruction for magnetic resonance fingerprinting. Magn. Reson. Imaging 41, 29–40 (2017).

4. Davies, M., Puy, G., Vandergheynst, P. & Wiaux, Y. A Compressed Sensing Framework for Magnetic Resonance Fingerprinting. SIAM J. Imaging Sci. 7, 2623–2656 (2014).

5. Deshmane, A., McGivney, D., Jiang, Y., Ma, D. & Griswold, M. Enforcing a physical tissue model for partial volume MR fingerprinting. in Proceedings of the 24th Annual Meeting, ISMRM 2998 (2016).

6. Doneva, M., Amthor, T., Koken, P., Sommer, K. & Börnert, P. Matrix completion-based reconstruction for undersampled magnetic resonance fingerprinting data. Magn. Reson. Imaging 41, 41–52 (2017).

7. Hamilton, J., Deshmane, A., Griswold, M. & Seiberlich, N. MR fingerprinting with chemical exchange (MRF-X) for in vivo multi-component relaxation and exchange rate mapping. in Proceedings of the 24th Annual Meeting, ISMRM 431 (2016).

8. Jiang, Y., Ma, D., Seiberlich, N., Gulani, V. & Griswold, M. MR fingerprinting using fast imaging with steady state precession (FISP) with spiral readout. Magn. Reson. Med. 74, 1621–1631 (2015).

9. Ma, D. et al. Magnetic resonance fingerprinting. Nature 495, 187–192 (2013).

10. McGivney, D. et al. SVD compression for magnetic resonance fingerprinting in the time domain. IEEE Trans. Med. Imaging 33, 2311–2322 (2014).

11. McGivney, D. et al. Bayesian estimation of multicomponent relaxation parameters in magnetic resonance fingerprinting. Magn. Reson. Med. doi:10.1002/mrm.27017 (2017).

12. Pierre, E. Y., Ma, D., Chen, Y., Badve, C. & Griswold, M. A. Multiscale reconstruction for MR fingerprinting. Magn. Reson. Med. 75, 2481–2492 (2016).

13. Pierre, E., Griswold, M. A. & Connelly, A. Fast analytical solution for extreme unaliasing of MR fingerprinting image series. in Proceedings of the 25th Annual Meeting, ISMRM 1353 (2017).

14. Yang, M. et al. Low rank approximation methods for MR fingerprinting with large scale dictionaries. Magn. Reson. Med. 79, 2392–2400 (2018).

15. Zhao, B., Setsompop, K., Ye, H., Cauley, S. F. & Wald, L. L. Maximum likelihood reconstruction for magnetic resonance fingeprinting. IEEE Trans. Med. Imaging 35, 1812–1823 (2016).

16. Zhao, B. et al. Improved magnetic resonance fingerprinting reconstruction with low-rank and subspace modeling. Magn. Reson. Med. 79, 933–942 (2018).

Proc. Intl. Soc. Mag. Reson. Med. 26 (2018)