Arnold Julian Vinoj Benjamin1,2, Pedro A. Gómez3,4, Mohammad Golbabaee1, Zaid Mahbub2, Tim Sprenger4, Marion I. Menzel4, Michael Davies1, and Ian Marshall2
1School of Engineering, Institute for Digital Communications, University of Edinburgh, Edinburgh, United Kingdom, 2Centre for Clinical Brain Sciences, University of Edinburgh, Edinburgh, United Kingdom, 3Computer Science, Technische Universität München, Munich, Germany, 4GE Global Research, Munich, Germany
Synopsis
This study shows the practical implementation of an
accelerated Cartesian Magnetic Resonance Fingerprinting (MRF) scheme using a
multi-shot Echo Planar Imaging (EPI) readout. Its performance is compared with conventional
spiral MRF and the fast convergence of accelerated iterative reconstructions
for this method is shown.
Purpose
The main
purpose of this study is to show that a highly accelerated Cartesian MRF scheme
using a multi-shot EPI readout (i.e. multi-shot EPI-MRF) can produce good
quality multi-parametric maps such as T1, T2 and proton density (PD) in a
sufficiently short scan duration that is similar to conventional MRF1. This multi-shot
approach allows considerable subsampling while traversing the entire k-space
trajectory, can yield better SNR,
reduced blurring, less distortion and can also be used to collect higher
resolution data compared to existing single-shot EPI-MRF implementations2, 3. The generated
parametric maps are compared to an accelerated spiral MRF implementation with
the same acquisition parameters to evaluate the performance of this method4. Additionally, an
iterative reconstruction algorithm is applied to improve the accuracy of parametric
map estimations and the fast convergence of EPI-MRF is also demonstrated5. Methods
The
scanning was performed on a 3T GE MR750w scanner with a 12 channel receive
only head RF coil (GE Medical Systems, Milwaukee, WI). 16-shot EPI-MRF datasets
and spiral datasets with 89 interleaves and golden angle rotations were
acquired from a tube phantom (Diagnostic Sonar, Livingston, UK) consisting of
tubes with different T1 and T2 values and a healthy volunteer using the inversion
recovery (IR) prepared (Tinv = 18 ms) Quantitative Transient-state
Imaging (QTI) sequence, using a linear ramp flip angle (FA) variation from 1° to 70° for 500
frames4. The Gx and Gy
gradient of the multi-shot EPI trajectory were balanced (Fig. 1) to ensure that
the residual magnetization remained constant for every shot and the crusher gradient (Gz) was applied to introduce gradient spoiling. The TR
was set to 16 ms for all the datasets to enable comparison between multi-shot EPI-MRF
and spiral MRF. The acquisition time for a single slice was 9 s. Both
acquisitions had 22:5 x 22:5 cm Field of View (FOV), 128 x 128 matrix size, 1.3
mm in-plane resolution and 5 mm slice thickness. The reconstruction was
performed by the CS based dictionary matching method with iterative
reconstructions to generate quantitative T1, T2 and PD maps5, 6. The dictionaries were
calculated using the Extended Phase Graph (EPG) model7. Results
Fig.
2 shows the highly subsampled aliased images of the tube phantom and healthy
volunteer at different frames/repetitions indexes t along with the CS-based
reconstruction that removed the aliasing and provided a better visualization of
the signal temporal dynamics. Fig. 3 shows the T1 and T2 sensitivity of the
sequence for discriminating dictionary elements using a linear ramp FA
variation from 1° to 70° for 500 frames. Fig. 4
shows the generated T1, T2 and PD maps for the tube phantom and healthy
volunteer for both the 16 shot EPI-MRF and spiral MRF acquisitions. Fig. 5
shows the qualitatively improved parametric estimations of T1, T2 and PD maps
for the tube phantom and healthy volunteer after the application of an iterative
projection algorithm.Discussion and Conclusion
Fig. 3a
shows that the T1 sensitivity is high throughout the acquisition, is enhanced
by the initial inversion pulse and occurs mostly at lower flip angles whereas
Fig. 3b shows that the T2 sensitivity occurs mostly at higher flip angles (>
25°). The T1 and T2 values of EPI-MRF and spiral
MRF in Fig. 4 and Fig. 5 for grey matter
(GM) and white matter (WM) are very similar to each other, in close agreement
to those reported in literature8. However, there is an
underestimation of cerebrospinal fluid (CSF) T2 values in both EPI-MRF and
spiral MRF and slight aliasing artefacts also appear in the T2 and PD maps of
EPI-MRF. This is because the encoding scheme used in the acquisition is comparatively
less sensitive to T2 variations than T1 variations. The use of an optimized FA train instead of a
linear ramp may yield more accurate T2 values and may potentially suppress the
ghosting artefacts in the T2 and PD maps. Fig. 5
shows that the accuracy of the parametric maps was improved by the use of
iterative reconstruction algorithms. The iterative reconstructions of multi-shot
EPI-MRF converges very quickly (35 s) compared to spiral MRF (~ 4 minutes) and could
therefore result in a very fast implementation on the scanner. This could be
further improved by the use of an adaptive iterative algorithm9. The convergence of
spiral acquisition is slow (which means more iterations) because spiral
sampling is ill-posed10. Moreover, each iteration
is more expensive because spiral sampling uses costlier non-uniform fast
Fourier Transform (NUFFT) compared to FFT in EPI. In addition, higher
resolution data can be acquired using a multi-shot EPI acquisition because it
does not suffer from blurring artefacts that become more pronounced in spiral
acquisitions at longer readout durations11. Acknowledgements
The research leading to these results has
received funding from the European Union's H2020 Framework Programme
(H2020-MSCA-ITN-2014) under grant agreement no 642685 MacSeNet and the Engineering and Physical Sciences Research Council (EPSRC) platform grant,
number EP/J015180/1.References
[1] Y. Jiang, D. Ma, N. Seiberlich, V.
Gulani, and M. A. Griswold, "MR fingerprinting using fast imaging with
steady state precession (FISP) with spiral readout," Magnetic resonance in medicine, vol. 74, pp. 1621-1631, 2015.
[2] B. Rieger, F. Zimmer, J. Zapp, S. Weingärtner, and L. R.
Schad, "Magnetic resonance fingerprinting using echo‐planar imaging: Joint quantification of T1 and T2∗ relaxation times," Magnetic
resonance in medicine, vol. 78, pp. 1724–1733,
2017.
[3] O. Cohen, M. Sarracanie, M. S. Rosen, and J. L. Ackerman,
"In Vivo Optimized Fast MR Fingerprinting in the Human Brain," in Proc Intl Soc Mag Reson Med, #430, 2016.
[4] P. A. Gomez, G. Buonincontri, M. Molina-Romero, J. I. Sperl,
M. I. Menzel, and B. H. Menze, "Accelerated parameter mapping with
compressed sensing: an alternative to MR Fingerprinting," in Proc Intl Soc Mag Reson Med, #1167,
2017.
[5] M. Davies, G. Puy, P. Vandergheynst, and
Y. Wiaux, "A compressed sensing framework for magnetic resonance
fingerprinting," SIAM Journal on
Imaging Sciences, vol. 7, pp. 2623-2656, 2014.
[6] M. Davies, G. Puy, P. Vandergheynst, and
Y. Wiaux, "Compressed quantitative MRI: Bloch response recovery through
iterated projection," in Acoustics,
Speech and Signal Processing (ICASSP), 2014 IEEE International Conference on,
2014, pp. 6899-6903.
[7] M. Weigel, "Extended phase graphs:
dephasing, RF pulses, and echoes‐pure and
simple," Journal of Magnetic
Resonance Imaging, vol. 41, pp. 266-295, 2015.
[8] J. P. Wansapura, S. K. Holland, R. S.
Dunn, and W. S. Ball, "NMR relaxation times in the human brain at 3.0
tesla," Journal of magnetic
resonance imaging, vol. 9, pp. 531-538, 1999.
[9] M. Golbabaee, Z. Chen,
Y. Wiaux and M. E. Davies,"Cover tree for fast MR
fingerprint Recovery", Proceedings of the IEEE workshop
on Machine Learning for Signal Processing (MLSP),
2017, arXiv:1706.07834.
[10] C. C. Cline, X. Chen, B. Mailhe, Q. Wang,
J. Pfeuffer, M. Nittka, et al.,
"AIR-MRF: accelerated iterative reconstruction for magnetic resonance
fingerprinting," Magnetic Resonance
Imaging, vol. 41, pp. 29-40, 2017.
[11] K. T. Block and J. Frahm, "Spiral
imaging: a critical appraisal," Journal
of Magnetic Resonance Imaging, vol. 21, pp. 657-668, 2005.