Matteo Cencini1,2, Pedro A Gómez3, Mohammad Golbabaee 4, Rolf F Schulte5, Giada Fallo1,2, Luca Peretti1,2, Michela Tosetti2,6, Bjoern H Menze3, and Guido Buonincontri2,6
1University of Pisa, Pisa, Italy, 2Imago7 Foundation, Pisa, Italy, 3Technical University of Munich, Munich, Germany, 4University of Bath, Bath, United Kingdom, 5GE Healthcare, Munich, Germany, 6IRCCS Stella Maris, Pisa, Italy
Synopsis
Transient-state
imaging techniques such as MR Fingerprinting allow for simultaneous
quantification of tissue properties by using variable acquisition parameters in
conjunction with undersampled non-Cartesian trajectories. Several implementations
exist in literature, relying on two- or three-dimensional sampling readouts.
Here, we studied the effect of the spatial encoding on quantification. In
addition, an evaluation of the impact on parametric maps of anti-aliasing
techniques (k-space weighted image
constrast) was performed, both in vitro and
in vivo.
Introduction
Transient-state
MRI techniques such as MR Fingerprinting are an emerging tool in the field of
quantitative MR imaging, due to their multiparametric quantification
capability, their robustness to motion and system imperfections1. One
distinct feature of transient-state techniques is the great flexibility in the
choice of acquisition parameters for contrast encoding (such as variable Flip
Angle and variable TE/TR schedules) and k-space
trajectories for spatial encoding (both two- and three dimensional2–5).
Here,
we studied the most common 2D/3D readout strategies (i.e. 2D Radial/Spiral, 3D
Radial/Spiral). In addition, we assessed the impact on quantification of an
anti-aliasing processing step, specifically the k-space weighting technique6.Methods
Trajectories
We compared 2D/3D Spirals and 2D/3D Radial
trajectories. All the readouts were based on rotations of an individual trajectory
interleave from one frame to the next. For both 2D acquisitions, the interleave was
rotated by a golden angle around the z-axis.
For 3D Spiral acquisitions, trajectories were generated by encoding 2D spirals
inside one (xy) plane and rotating this
plane around the z-axis (see Cao et al.7).
Finally, in 3D Radial k-space acquisitions, random permutations of spoke indexes
were used to achieve full spatial coverage. Specific details for the waveforms
are shown in Figure 1.
Acquisition and reconstruction
All
the data were acquired on a GE HDxt 1.5T scanner using an 8-ch receiver coil, using FISP-based magnetic resonance fingerprinting2. Both
phantom (Eurospin TO5 phantom8) and
data from a healthy volunteer were acquired. The flip angle schedule is shown
in Figure 2a, while fixed TE/TR = 0.46ms/12ms and 2.08ms/12ms were used
respectively for spiral and radial acquisitions.
The
reconstruction pipeline is represented in Figure 2. Dictionary matching with
EPG simulations9 was
performed, obtaining quantitative T1, T2 and PD maps. To assess the accuracy of
the measurements, phantom nominal values were used as well as 2D FISP-MRF
acquisition with the original schedule from Jiang et al2.
Antialiasing
We
evaluated an anti-aliasing
pre-processing step before SVD compression, named here k-space weighted
view-sharing (KW), based on k-space
weighted image contrast6. This
method allows, given an appropriate weighting function, to share data
among adjacent frames to increase the sampling of individual frames (Figure 1
and 2), thus reducing aliasing, without compromising the contrast.
We set the weights as a function of the sampling density (i.e. proportional to
the trajectory DCF), enabling us to compare readouts with distinct density
patterns (i.e. spirals and radials).
We
also validated the entire pipeline in
silico, simulating the acquisition of time-domain data of a single
representative white matter voxel (T1 = 600 ms, T2 = 70 ms) using all trajectories.
The data were then reconstructed to obtain a point spread function (PSF) in
image space for each temporal coefficient, with and without KW view-sharing,
and compared to a fully-sampled case. Results
In silico validation
Figure
3 shows the results of the PSF analysis, showing that all undersampled readouts
approximate the fully-sampled PSF for the first SVD coefficient, whereas
view-sharing significantly improves the PSF with respect to zero-filling in
lower energy coefficients, where the sidelobes of the PSF were reduced by an
order of magnitude.
In vitro validation
T1/T2
quantification results are summarized in Figure 4. All the acquisitions,
regardless of the readout trajectories, give comparable results to the
reference. In addition, while KW view-sharing does not negatively affect
quantification accuracy it reduces of standard deviations with respect to
zero-fill reconstruction.
In vivo validation
Figure
5 shows
the results for in vivo acquisitions.
All trajectories benefit from this anti-aliasing pre-processing step. However,
for extreme undersampling factors such as in the 3D Radial case, aliasing
artifacts cannot be completely recovered even with the use of view-sharing,
therefore leading to blurred T2 maps. Overall, the best results were obtained
with 3D spiral projections in combination with KW view-sharing.Discussion and conclusion
In
this work, we studied different 2D and 3D non-Cartesian trajectories for transient-state
acquisitions at the basis of MRF, and evaluated the impact of non-iterative
anti-aliasing technique on quantification.
Overall,
quantification was agnostic with respect to the specific sampling strategy and
anti-aliasing pre-processing step used. Phantom results were in good agreement
both with gold-standard based nominal values and FISP-MRF based measurements. This
represents an important result, since it enables the comparison of images
obtained from these different spatial encoding strategies.
While
all the trajectories achieved a similar accuracy, spiral trajectories showed
substantially reduced undersampling artifacts, resulting in a diminished
standard deviation of the measurements. Conversely, quantification errors due
aliasing errors for highly undersampled trajectories (such as 3D Radial) were
still visible. However, all trajectories (including 3D Radial) benefitted from
the KW view-sharing processing, resulting in a substantial reduction of
aliasing artifacts.
In
conclusion, transient-state MRI acquisitions provided accurate estimations
irrespectively of the specific k-space
trajectory used, while the use of KW view-sharing could mitigate undersampling
artifacts in all the cases here tested.Acknowledgements
Support from the Italian Ministry of Health and the Tuscany Region under the project “Ricerca Finalizzata”, Grant n. GR-2016-02361693.
References
1. Ma, D. et
al. Magnetic
resonance fingerprinting. Nature 495, 187–92 (2013).
2. Jiang, Y., Ma, D.,
Seiberlich, N., Gulani, V. & Griswold, M. A. MR fingerprinting using fast
imaging with steady state precession (FISP) with spiral readout. Magn.
Reson. Med. 74, 1621–1631 (2015).
3. Cloos, M. A. et al.
Multiparametric imaging with heterogeneous radiofrequency fields. Nat.
Commun. 7, 12445 (2016).
4. Liao, C. et al. 3D
MR fingerprinting with accelerated stack-of-spirals and hybrid sliding-window
and GRAPPA reconstruction. Neuroimage 162, 13–22 (2017).
5. Cao, X. et al. Fast
3D brain MR fingerprinting based on multi-axis spiral projection trajectory. Magn.
Reson. Med. 82, 289–301 (2019).
6. Song, H. K. &
Dougherty, L. k-Space weighted image contrast (KWIC) for contrast manipulation
in projection reconstruction MRI. Magn. Reson. Med. 44, 825–832
(2000).
7. Xiaozhi Cao, Congyu Liao,
Qing Li, Huihui Ye, Hongjian He, J. Z. Fast 3D MR fingerprinting with spiral
projection acquisition for whole brain quantification imaging. in Proceedings
of the 26th Annual Meeting of ISMRM 1017 (2018).
8. Lerski, R. A. & de
Certaines, J. D. Performance assessment and quality control in MRI by Eurospin
test objects and protocols. Magn. Reson. Imaging 11, 817–833
(1993).
9. Weigel, M. Extended phase graphs: Dephasing, RF pulses, and
echoes - Pure and simple. J. Magn.