Xingfeng Shao1, Stefan M Spann2, Kai Wang1, Lirong Yan1,3, Stollberger Rudolf2, and Danny JJ Wang1,3
1Mark & Mary Stevens Neuroimaging and Informatics Institute, Keck School of Medicine, University of Southern California, Los Angeles, CA, United States, 2Institute of Medical Engineering, Graz University of Technology, Graz, Austria, 3Department of Neurology, University of Southern California, Los Angeles, CA, United States
Synopsis
Ultra-high field allows ASL to achieve higher
spatial resolution due to increased SNR and prolonged T1 relaxation. We present a single-shot 3D GRASE pCASL technique
with 12-fold acceleration using time-dependent 2D CAIPI sampling strategy, and
reconstruction of the label/control time series with joint spatial and temporal
total-generalized-variation (TGV) regularization. 2D CAIPI under-sampling pattern increases
temporal incoherence between measurements which allows joint reconstruction of
the highly accelerated ASL time series. Combining the advantages of ultra-high field strength,
pTx coils, accelerated acquisition and advanced reconstruction, whole-brain CBF
map with 2 mm isotropic resolution was obtained within 5 mins.
Introduction
Pseudo-continuous
arterial spin labeling (pCASL) with segmented 3D acquisitions has been
recommended for clinical implementations with moderate spatial resolution
(>3 mm) at 3T due to the low SNR.1 Ultra-high field allows ASL to achieve
higher spatial resolution due to increased SNR and prolonged T1 relaxation.2,3 However, standard segmented 3D acquisition is not ideal for higher resolution
due to increased motion sensitivity and image blurring effect. In this study,
we present a motion-robust single-shot 3D GRASE pCASL technique at 7T with a 12-fold
acceleration using time-dependent 2D CAIPI sampling strategy to increase the
temporal incoherence.4 To exploit this, the whole label/control time series were
jointly reconstructed using spatio-temporal total-generalized-variation (TGV)
regularization.4 The combination of ultra-high field strength, parallel RF
transmit (pTx) coils, accelerated acquisition and advanced reconstruction
yields high resolution (2 mm isotropic) whole brain cerebral blood flow (CBF)
maps within an acquisition time of 5 mins.Methods
Two
healthy subjects (2M, age=31.0±5.0 yrs) underwent MRI scan on
a 7T Siemens Terra scanner with a NOVA 8-channel parallel transmit (8-pTX) and
32-channel receiver (Rx) head coil. A standard and an accelerated ASL acquisition
were performed with the following common imaging parameters: FOV=200 mm, matrix
size=96×96, 40 slices with 20% slice
oversampling, resolution = 2.1×2.1×2 mm3. pCASL
parameters were: FA=250, Gmean = 0.6 mT/m, Gratio = 10, labeling
duration = 1500 ms and post-labeling delay = 1800 ms. Position of imaging
volume is shown in figure 1 (a). pCASL labeling plane was placed to be
approximately perpendicular to both carotid and vertebral artery based on both
structural (figure 1 (a)) and TOF MRA images (figure 1 (b)). Labeling efficiency
was derived by Bloch equation simulations assuming flow velocity = 40 cm/s at
carotid artery. Background suppression (BS) was applied to suppress gray and
white matter signals. SAR was monitored by FDA approved vendor software and was
within the first level (3.2W/kg on head).
For
the standard acquisition with segmented readouts: turbo factor = 12, EPI-factor
= 31, 12 segments, TE = 22 ms and TR = 6000 ms. One M0 and one pair of
label/control images were acquired in 3 mins 36 secs. Single-shot 3D-GRASE
acquisition was achieved with 12-fold acceleration using a time-dependent 2D-CAIPI
under-sampling pattern to increase the temporal incoherence between
measurements, as illustrated in Figure 2. Coil sensitivity maps were estimated
using ESPIRiT5 from combined k-space of 12 measurements. The whole 4D under-sampled
label/control time series were jointly reconstructed using spatial-temporal TGV
regularized reconstruction.4 For the accelerated acquisition, two M0-images
and 24 pairs of label/control images were acquired in 5 mins. Motion correction
was performed using SPM126 and ASL-Toolbox.7,8 CBF maps were calculated
according to ASL white paper.1Results and discussion
Average
B0 = -85 Hz and average B1 = 51.9% with True-Form B1-shimming at the pCASL
labeling plane, as shown in Figure 3 (a) and (b). Figure 3 (c) shows simulated
magnetization of spins flowing through the labeling plane. Labeling efficiency
= 67%, which is lower than typical pCASL at 3T mainly due to B1 insufficiency.
B1-shimming for pCASL labeling could be improved using advanced pTx configuration
for the next step.
Figure
4 (a) shows CBF maps calculated from standard fully sampled acquisition in
axial, sagittal and coronal views. The high number of segments (tacq = 1 min 12
secs) of the standard acquisition makes the ASL signal very sensitive to motion
and physiological fluctuations especially when spatial resolution is high. Figure
4 (b) shows CBF maps calculated from accelerated acquisition with
spatial-temporal TGV reconstruction. With high spatial resolution and
sufficient SNR, perfusion along cortical gyri can be observed. 2D CAIPI
under-sampling pattern increases temporal incoherence between measurements
which allows joint reconstruction of the highly accelerated label/control time
series. Figure 5 shows forty axial slices of CBF maps acquired with 2D CAIPI readout.
The proposed technique allows whole brain coverage with 2 mm isotropic
resolution in 5 mins at 7T. Conclusion
Standard
segmented 3D ASL suffers from motion and physiological fluctuations during the
long acquisition window especially for high spatial resolution. To overcome
this problem we proposed a single-shot 3D GRASE pCASL sequence with 12-fold acceleration
and time-dependent 2D CAIPI pattern. The temporal incoherence of the sampling
is directly exploited in the reconstruction approach and additional spatial and
temporal constraints on the individual images improved the SNR. By combining
the advantages of ultra-high field strength, pTx coils, accelerated acquisition and
advanced reconstruction, whole-brain CBF map with 2 mm isotropic resolution
obtained within 5 mins.Acknowledgements
This work was supported by National Institute of Health
(NIH) grant UH3-NS100614 and R01-EB028297. NVIDIA Corporation Hardware grant
support.References
[1] Alsop
D C, Detre J A, Golay X, et al. Recommended implementation of arterial
spin‐labeled perfusion MRI for clinical applications: A consensus of the ISMRM
perfusion study group and the European consortium for ASL in dementia[J].
Magnetic resonance in medicine, 2015, 73(1): 102-116.
[2] Zuo Z, Wang R, Zhuo Y, et
al. Turbo-FLASH
based arterial spin labeled perfusion MRI at 7 T [J].
PloS one, 2013, 8(6): e66612.
[3]
Pohmann R, Speck O, Scheffler K. Signal‐to‐noise ratio and MR tissue parameters
in human brain imaging at 3, 7, and 9.4 tesla using current receive coil
arrays[J]. Magnetic resonance in medicine, 2016, 75(2): 801-809.
[4] Spann
S M, Shao X, Wang DJJ, et al. Robust single-shot acquisition of high resolution
whole brain ASL images by combining time-dependent 2D CAPIRINHA sampling with
spatio-temporal TGV reconstruction [J]. Neuroimage, 2019 (accepted).
[5] Uecker
M, Lai P, Murphy M J, et al. ESPIRiT—an eigenvalue approach to autocalibrating
parallel MRI: where SENSE meets GRAPPA[J]. Magnetic resonance in medicine,
2014, 71(3): 990-1001.
[6].
Friston KJ, Ashburner J, Kiebel S, et al. Statistical parametric mapping: the
analysis of funtional brain images. Elsevier/Academic Press 2007.
[7] Wang
Z, Aguirre GK, Rao H, et al. Empirical optimization of ASL data analysis using
an ASL data processing toolbox: ASLtbx. Magn. Reson. Imaging 2008;26:261–269.
[8] Wang
Z. Improving cerebral blood flow quantification for arterial spin labeled
perfusion MRI by removing residual motion artifacts and global signal
fluctuations. Magn. Reson. Imaging 2012;30:1409–1415.