All at Once - Finger- & Footprints

Mariya Doneva¹

¹Philips Research, Hamburg, Germany

Synopsis

In recent years, the field of quantitative MRI has been expanded by the introduction of MR Fingerprinting as well as several quantitative MRI methods applying extensive signal modeling. This lecture will give an overview of these recently introduced methods.

Objectives

Provide an overview of the modern acquisition and reconstruction techniques for multi-parameter mapping.

Purpose

MRI in current clinical practice is almost entirely based on multiple images with different contrast weighting. The contrast in these images is sensitive to the acquisition parameter and system settings and may vary between sites, vendors, scanners, and sometimes even between images acquired on the same scanner at different times. Absolute quantification of the underlying physical properties that generate the image contrast, such as T1, T2 relaxation times, proton density, diffusion, etc. has the potential of more reproducible MR scans and provides a quantitative value that may be directly used for diagnosis [1]. Another potential application of multi-parameter mapping is synthetic MRI, where MR images at arbitrary contrast can be retrospectively generated from the T1, T2, and proto density maps [2]. In recent years, the field of quantitative MRI has been expanded by the introduction of MR Fingerprinting as well as several QMRI methods applying extensive signal modeling. This lecture will give an overview of these recently introduced methods.

Methods

The recently introduced technique MR Fingerprinting [3] is a generalized approach for MR parameter mapping, aiming at generating a unique MR signal response for each tissue, therefore encoding all relevant tissue parameters in a single measurement. In MR Fingerprinting, Bloch simulations are used to compute a dictionary of the expected signal evolutions for a large set of tissue parameters. The measured signal in each voxel is then matched to the closest entry in the dictionary and the corresponding MR tissue and system parameters are derived. An assumption in MR Fingerprinting is that each tissue or combination of MR tissue parameters generate a unique MR response which serves as a fingerprint for the tissue. MRF is still very new and needs to be clinically validated. However, it has already triggered a lot of research investigating multi-parameter mapping, its advantages and limitations, which also leads to re-thinking more classical parameter mapping approaches.

Parallel to the development of MR Fingerprinting, improvements in the basic footprint of QMRI have been achieved by applying techniques with more accurate signal modeling and/or dictionary-based approaches. In [4] the so called CMR-Footprinting is presented, which uses a dictionary of Bloch simulations for T1 estimation of the T1-MOLLI sequence. In [5] the authors use a similar dictionary-based approach for T2 mapping. Applying a dictionary-based approach for parameter mapping has some desirable properties such as intrinsic denoising, which was also previously utilized in a compressed sensing reconstruction for MR parameter mapping [6]. It also replaces the non-linear parameter estimation problem by a linear search in the dictionary. However, it does introduce discretization of the MR parameters and may result is very large dictionaries. Alongside with the development of these new dictionary based methods, there is also a trend in improving the parameter estimation using classical parameter fits with improved data models and corrections [7]. This lecture aims to give an overview of these new approaches for quantitative MRI and their potential clinical impact.

Discussion and Conclusion

Although quantitative MRI, including multi-parameter mapping is not new, the existing approaches to quantitative MRI are often not sufficiently robust and/or efficient to replace current acquisition of multiple contrast-weighted images [8]. The recent developments of MR Fingerprinting and improved signal modeling in classical parameter mapping are important steps towards achieving this goal. However, further research is needed to prove the efficiency and accuracy of these techniques.

Acknowledgements

No acknowledgement found.

References

1. Deoni, M. "Gleaning multicomponent T1 and T2 information from steady-state imaging data." Magn Reson Med 60 (2008): 1372-1387.

2. Warntjes JBM et al. “Rapid Magnetic Resonance Quantification on the Brain: Optimization for Clinical Usage” Magn Reson Med 60 (2008) 320-329

3. Ma, D. et al. "Magnetic Resonance Fingerprinting." Nature 495 (2013): 187-193.

4. Xanthis C. et al. “CMR-Footprinting: Quantifying Tissue Parameters with Clinical Pulse Sequence Simulations Improves Accuracy-An Example with MOLLI T1 Mapping” ISMRM 2015, # 182

5. Ben-Eliezer, N. et al. „Rapid and accurate T2 mapping from multi-spin-echo data using Bloch-simulation-based reconstruction“ Magn Reson Med 73 (2015): 809-817

6. Doneva, M. et al. „Compressed Sensing Reconstruction for Magnetic Resonance Parameter Mapping.“ Magn Reson Med 64 (2010): 1114-1120.

7. Petrovic, A. et al. “Closed-Form Solution for T2 mapping with Nonideal Refocusing of Slice Selective CPMG Sequences” Magn Reson Med 73 (2015) 818-827

8. Stikov, N. et al. “On the Accuracy of T1 mapping:searching for common ground” Magn Reson Med 73(2015) 514-22

Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)