Congyu Liao1, Berkin Bilgic1, Qiyuan Tian1, Jason Stockmann1, Qiuyun Fan1, Siddharth Srinivasan Iyer1,2, Fuyixue Wang1,3, Chanon Ngamsombat1,4, Xiaozhi Cao1, Mary Kate Manhard1, Susie Y. Huang1, Lawrence L. Wald1, and Kawin Setsompop1
1Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States, 2Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA, United States, 3Harvard-MIT Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States, 4Department of Radiology, Faculty of Medicine, Siriraj Hospital, Mahidol University, Thailand
Synopsis
Diffusion magnetic resonance imaging (dMRI) is
a highly sensitive imaging modality, but is limited in spatial resolution and signal-to-noise
ratio (SNR). In this work, we combine an SNR-efficient acquisition and model-based
reconstruction strategies with newly-available hardware instrumentation to
achieve distortion-free in-vivo dMRI at 600-860 µm isotropic voxel size with
high fidelity and sensitivity on a clinical 3T scanner. At this resolution, it is
possible to accurately probe the microstructure of different cortical layers in
the human brain.
Introduction
Submillimeter-isotropic-resolution dMRI has
shown great potential for characterizing gray-matter1,2. However, the current resolution of in-vivo dMRI
(at 1.2-2.0mm-isotropic-resolution) is too low to probe the microstructure in cerebral
cortex, which causes undesirable partial volume effects. The key challenges of submillimeter
in-vivo dMRI are the decreased SNR, physiological noise-induced phase
variations and image distortions from B0-inhomogeneity and
eddy-current. To address these issues, we first propose Blip Up-Down
Acquisition (BUDA) and joint parallel-imaging reconstruction to achieve distortion-free
EPI. Second, we combine BUDA with a self-navigated RF-encoded multi-slab acquisition,
termed “Generalized Slice Dithered Enhanced Resolution” (gSlider)3,4, to enable high-isotropic-resolution dMRI with
high-fidelity and SNR-efficiency. Finally, we minimize g-factor in gSlider-BUDA,
by incorporating slab-by-slab dynamic shimming with a 32-channel AC/DC coil5 into the acquisition to reduce B0-inhomogeneity.
This enabled high-quality distortion-free whole-brain 600μm dMRI on a 3T scanner.Methods
In
BUDA-EPI (Fig.1), two EPI-shots sample complementary subsets of k-space, one
with a positive ky traversal (blip-up) and the other with a negative
traversal (blip-down), to create opposing distortions (Fig.1(B)). As shown in
Fig.1(A), the B0 and eddy-current can be extracted from this
acquisition pair via FSL ‘TOPUP’6. This information is then incorporated into a joint-reconstruction
with a structured low-rank constraint across the two shots to obtain one
distortion-free image. Compared to hybrid-space SENSE7, BUDA-EPI obviates the need for direct phase
estimation and achieves improved reconstruction at high accelerations.
Additionally, a partial Fourier acquisition where EPI-shots sample
complementary k-space sections are incorporated directly into the low-rank
reconstruction, so a short-TE can be attained to minimize T2-related
signal-losses.
We combine BUDA-EPI with gSlider, where both
blip-up and blip-down shots are acquired in an interleaved fashion (Fig.2(A)). Five
consecutive RF-encoding pulses (Fig.2(B)) are used for a total of 10 EPI-shots per
slab across blip-up and -down encodings. The RF-excitations are also combined
with blipped-CAIPI8 (MB=2) to acquire 10 simultaneous slices per
EPI-shot. BUDA reconstruction is first performed for each
RF-encoding, followed by gSlider reconstruction4 to obtain high slice-resolution data.
Acquisitions:
(i)To validate gSlider-BUDA, whole-brain 860μm-data were acquired: FOV:220×220×120mm3, TR/TE=3500/68ms. Two-shots
were collected at MB=2×Rinplane=4 and partial Fourier (p.f.)6/8. Sixty-four
diffusion-directions with b=1000 s/mm2 were acquired.
(ii)To investigate the microstructure of gray-matter,
whole-brain 780μm-data with two shells were
acquired: FOV:220×220×117mm3, TR/TE=4400/76ms. Thirty-two
diffusion-directions with b=1000s/mm2 and ninety-six
diffusion-directions with b=2500s/mm2 were acquired. To reduce TE
and T2* blurring, we used Rinplane=5 with p.f.=6/8 for each shot (resolution
point-spread-function ~1.12x for Rinplane=5 at
effective-echo-spacing=0.196ms, assuming T2*/T2=30/60ms). 3D-MPRAGE was
acquired for surface-based analysis.
(iii) We further push the spatial-resolution to
600μm isotropic using Rinplane=6 per
shot to reduce T2* blurring (point-spread-function ~1.19x) and p.f.=5/8 to
reduce the TE to 61ms. To reduce g-factor in the joint distortion-corrected BUDA-reconstruction,
AC/DC-coil was utilized for slab-by-slab shimming to reduce B0-inhomogeneity
and hence the level of distortion differences between the blip-up and-down data.
With dynamic-shimming, 36 diffusion-directions with b=1000s/mm2 and
3 averages were acquired, with FOV:220×220×120mm3, for a total
acquisition time of 90 minutes. All in-vivo measurements were performed on a 3T
Siemens Prisma scanner.
Post-processing:
Colored-FA maps were generated using FSL6. For cortical depth analysis, the intermediate
surfaces between white and pial surfaces were generated by FreeSurfer9 from the MPRAGE data. Then the diffusion principal
eigenvectors were aligned to the MPRAGE data10 and projected to intermediate surfaces to
calculate the radiality2. For the multi-shell data, ODF was estimated utilizing
multi-tissue-multi-shell CSD method11.Results
Figure3
shows the results of 860μm-data. As white arrows indicate in Fig.3(A), the
individual blip-up and -down reconstructions show significant distortions.
Compared to the hybrid-space SENSE, the residual artifacts are eliminated in
BUDA. Figure3(B) shows the high-fidelity whole-brain diffusion-weighted images
(DWI), colored-FA maps and averaged DWIs in three orthogonal views, demonstrating
closely matched volumes compared to the reference.
Figure4(A) shows the radiality map at the different
surfaces on the inflated brain surface. Radiality is low at the white-gray
boundary, higher at the middle cortical depths and lower at the pial surface
with the low radiality of somatosensory cortex (S1) (indicated by blue arrow) at
all cortical depths which reveals the tangential fibers in the S1 area. Figure4(B)
shows the ODF in S1 and primary motor cortex (M1). Consistent with previous
studies1,2, we find a primarily radial orientation in M1
and tangential fibers to the local cortical surface orientation in S1.
Figure5(A) shows the 1/g-factor maps of
single-shot EPI, gSlider-BUDA without and with dynamic shim, and with perfect
shim (no B0-inhomogeneity) at Rinplane=6. Compared to the
single-shot EPI, BUDA significantly reduces the g-factor penalty at the cost of
large g-factor in severe B0-inhomogeneity areas. With dynamic
shimming, B0-variation is reduced by>50%, which results in 28%
g-factor improvement (Fig.5(B)), approaching performance of the idealized
perfect-shimming case (effectively Rinplane=3). Figure 5(C) shows
the zoom-in view of a slice of averaged DWI, displaying the high-resolution
capability of the 600µm-data.Discussion and conclusion
In this work, we developed gSlider-BUDA with dynamic
shimming to achieve rapid distortion-free submillimeter-isotropic-resolution dMRI. In comparison to standard
dMRI, the proposed method has provided improved g-factor and image-fidelity,
which is beneficial for co-registration of MPRAGE and research on cortical
analysis. This will push in-vivo dMRI toward the submillimeter-scale of
cortical layers to transfer new insights from post-mortem imaging to in-vivo
imaging.Acknowledgements
This work was supported in part by NIH research
grants: NIH R01EB019437, R01EB020613, R01 MH116173, U01EB026996 and U01EB025162References
1. Leuze CWU, Anwander A, Bazin PL, et
al. Layer-specific intracortical connectivity revealed with diffusion MRI.
Cereb. Cortex 2014;24:328–339.
2. McNab JA, Polimeni JR, Wang R, et al. Surface-based
analysis of diffusion orientation for identifying architectonic domains in the
in vivo human cortex. Neuroimage 2013;69:87–100.
3. Setsompop K, Fan Q, Stockmann J, et al. High-resolution in
vivo diffusion imaging of the human brain with generalized slice dithered
enhanced resolution: Simultaneous multislice (gSlider-SMS). Magn. Reson. Med.
2018;79:141–151.
4. Liao C, Stockmann J, Tian Q, et al. High‐fidelity,
high‐isotropic‐resolution diffusion imaging through gSlider acquisition with
and T 1 corrections and integrated ΔB0/Rx shim array. Magn. Reson. Med.
2019;83:56–67.
5. Stockmann JP, Witzel T, Keil B, et al. A 32-channel
combined RF and B0shim array for 3T brain imaging. Magn. Reson. Med.
2016;75:441–451.
6. Jenkinson M, Beckmann CF, Behrens TEJ, Woolrich MW, Smith
SM. Fsl. Neuroimage 2012;62:782–790.
7. Zahneisen B, Aksoy M, Maclaren J, Wuerslin C, Bammer R.
Extended hybrid-space SENSE for EPI: Off-resonance and eddy current corrected
joint interleaved blip-up/down reconstruction. Neuroimage 2017;153:97–108.
8. Setsompop K, Gagoski BA, Polimeni JR, Witzel T, Wedeen VJ,
Wald LL. Blipped-controlled aliasing in parallel imaging for simultaneous
multislice echo planar imaging with reduced g-factor penalty. Magn. Reson. Med.
2012;67:1210–1224.
9. Fischl B. FreeSurfer. Neuroimage 2012;62:774–781.
10. Greve DN, Fischl B. Accurate and robust brain image
alignment using boundary-based registration. Neuroimage 2009;48:63–72.
11. Jeurissen B, Tournier J-D, Dhollander T, Connelly A,
Sijbers J. Multi-tissue constrained spherical deconvolution for improved
analysis of multi-shell diffusion MRI data. Neuroimage 2014;103:411–426.