HIgh B-value and high Resolution Integrated Diffusion (HIBRID) Imaging

Qiuyun Fan¹, Aapo Nummenmaa¹, Jonathan R. Polimeni¹, Thomas Witzel¹, Susie Y. Huang¹, Van J. Wedeen¹, Bruce R. Rosen^1,2, and Lawrence L. Wald^1,2

¹Massachusetts General Hospital, Boston, MA, United States, ²Harvard-MIT Division of Health Sciences and Technology, MIT, Cambridge, MA, United States

Synopsis

The cerebral cortex is rich in gyral folding. Axonal fibers take sharp turns when bending into the cortex. High resolution diffusion MRI is needed to characterize cortical structures in finer scale, while high b-value is desired to resolve complex white matter structures. We examined the impact of imaging resolution on characterizing the radial diffusion pattern in cortex, and proposed to improve the HIgh B-value and high Resolution Integrated Diffusion (HIBRID) imaging by incorporating information about each voxel’s proximity to the cortex. The combined data demonstrated the desired features from both high resolution and high b-value diffusion imaging.

PURPOSE

The cerebral cortex is rich in gyral folding. Axonal fibers take sharp turns when branching off from major White Matter (WM) bundles and bending into the cortex. Thus, high spatial resolution diffusion MRI is desired for resolving fiber paths near detailed cortical structures (such as the interior of gyri). On the other hand, high b-value is beneficial in improving the angular resolution of crossing structures in WM¹. While spatial resolution and b-value are necessarily traded-off due to finite SNR, efforts have been made to combine diffusion data acquired with multiple spatial resolutions and/or b-values to fuse the advantages from both^2-4. In this work, we examined the relationship between spatial resolution and the radiality measure⁵ of diffusion throughout cortical depths. We propose to perform the HIgh B-value and high Resolution Integrated Diffusion (HIBRID) imaging by incorporating the spatial location of each voxel relative to the cortex.

METHODS

Data were acquired from a healthy adult subject on the Siemens 3T Connectom Scanner with a 64-channel head coil⁶. A spin echo-Echo Planar Imaging (SE-EPI) sequence was used to acquire diffusion data with 1.0mm, 1.5mm and 2.0mm isotropic resolution at b=1500 s/mm² (60 DW directions, TR/TE=6700/69ms) and 2.0mm isotropic resolution at b=8000 s/mm² (120 DW directions, TR/TE=3000/60ms). b=0 images were interspersed every 12 DW images. Data were collected with R>>L and L>>R phase encoding directions. Other parameters include Partial Fourier: 7/8, FLEET-ACS GRAPPA^7,8 iPAT=3, MB/SMS^9,10 factor=2. An MEMPRAGE¹¹ image volume was collected for cortical surface reconstruction. Total acquisition time was approximately 80 minutes.

TOPUP¹² was used to correct for susceptibility distortions. The b8k dataset was upsampled and concatenated with the 1.0mm-b1.5k dataset for eddy current correction using EDDY¹³. For the spatial resolution comparison at b=1500 s/mm², eddy current correction was performed separately in their native spatial resolutions to avoid further smoothing due to interpolations. The cortical pial and white-gray boundary (WGB) surfaces were calculated using FreeSurfer^14,15. Surfaces at various cortical depths were also calculated¹⁶. The b=0 image was then registered to the T1w image using boundary-based registration¹⁷, and the transformation obtained was used to register the surface meshes with diffusion images. The curvature of the WGB-surface was used to group the vertices into gyral crown (curvature ≤ -1/3), wall (-1/3<curvature<1/3) and fundi (curvature≥1/3) (see Figure 1).

To combine the 1.0mm-b1.5k data with 2.0mm-b8k data, the weighting on the 1.0mm-b1.5k data was set to 1 in cortical gray matter (GM). For the rest of the brain, the distance from the WGB-surface, $$$d$$$, was used to determine the weighting on the 1.0mm-b1.5k data, $$$w$$$, where $$$w=max\left\{ \frac{d_{0}-d}{d_{0}}, e^{-\frac{d}{d_{1}}}\right\}$$$, so that the weighting on the 1.0mm-b1.5k data smoothly decays as it is further away from the WGB (Figure 2). The weighting was then applied on q-ball ODFs^18,19 calculated separately for 1.0mm-b1.5k data (L_max=4, λ=0.006) and 2.0mm-b8k data (L_max=10, λ=0.001), to form the combined dataset.

Jackknife resampling was performed to examine the variance of ODF by leaving 20% DW directions out each iteration for 100 iterations. The mean ODF across Jackknife samples was caculated. For each DW direction, the min-max amplitude of the mean ODF (i.e. contrast) over standard deviation across iterations (i.e., noise) was then averaged across all DW directions and used as contrast-to-noise ratio (CNR) estimates.

RESULTS

As expected, the primary diffusion directions inside cortical GM are largely radial to the WGB-surface (Figure 1,3). Increased spatial resolution demonstrated a faster transition between non-radial diffusion in the WM to highly-radial diffusion in the GM (Figure 1). The proposed distance-based weighting (Distance-W, Figure 2) was compared with universally equal weighting (Half-Half), and showed higher CNR in WM in the Jackknife resampling analysis (Figure 4). The mean ODF (opaque, Figure 5) + 3 times standard deviation (transparent) across Jackknife samples demonstrated reasonable reliability in the combined data using Distance-W.

DISCUSSION

The radiality measure improved as spatial resolution increased from 2.0mm to 1.0mm, suggesting sub-mm resolution may continue to benefit characterization of detailed cortical microanatomy. GM signal content at b=8000 s/mm² is lower compared to b=1500 s/mm², which renders it more susceptible to partial voluming with WM signal at the WGB, and ODFs close to the pial surface are more variable across Jackknife samples. The Distance-W combined data demonstrated reasonable reliability in both cortical GM and subjacent WM. The CNR in the center of the brain was relatively low due to low SNR further from the coil detectors; to address this, the potential of ultrahigh-field imaging may be investigated in future work.

CONCLUSION

The combined data based on Distance-W gained high spatial resolution in cortex while preserving high angular resolution in WM.

Acknowledgements

The work is supported by funding from the National Institutes of Health Blueprint Initiative for Neuroscience Research Grant U01MH093765, NIH NIBIB Grant R00EB015445.

References

1. Fan, Q., et al. Investigating the capability to resolve complex white matter structures with high b-value diffusion magnetic resonance imaging on the MGH-USC Connectom scanner. Brain Connect 4, 718-726 (2014).

2. Fan, Q., et al. Improved diffusion tractography at the cortical boundary using HARDI acquisitions with high-b/low-k in white matter and low-b/high-k within and near the cortex in Proc. ISMRM 0564 (Toronto, Canada, 2015).

3. Sotiropoulos, S.N., et al. RubiX: combining spatial resolutions for Bayesian inference of crossing fibers in diffusion MRI. IEEE Trans Med Imaging 32, 969-982 (2013).

4. Sotiropoulos, S.N., et al. Fusing 3 and 7 Tesla HCP Datasets for Improved Brain Connectivity Analysis. in Proc. ISMRM 0562 (Toronto, Canada, 2015).

5. McNab, J.A., et al. Surface based analysis of diffusion orientation for identifying architectonic domains in the in vivo human cortex. Neuroimage 69, 87-100 (2013).

6. Keil, B., et al. A 64-channel 3T array coil for accelerated brain MRI. Magn Reson Med 70, 248-258 (2013).

7. Bhat, H., et al. Motion Insensitive ACS Acquisition Method for in-plane and Simultaneous Multi-Slice Accelerated EPI. in Proc Intl Soc Mag Reson Med 644 (2014).

8. Polimeni, J.R., et al. Reducing sensitivity losses due to respiration and motion in accelerated echo planar imaging by reordering the autocalibration data acquisition. Magn Reson Med (2015).

9. Feinberg, D.A. & Setsompop, K. Ultra-fast MRI of the human brain with simultaneous multi-slice imaging. J Magn Reson (2013).

10. Setsompop, K., et al. SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (SLIDER-SMS) for high resolution (700 um) diffusion imaging of the human brain. in Proc. ISMRM 0339 (Toronto, Canada, 2015).

11. van der Kouwe, A.J., Benner, T., Salat, D.H. & Fischl, B. Brain morphometry with multiecho MPRAGE. Neuroimage 40, 559-569 (2008).

12. Andersson, J.L., Skare, S. & Ashburner, J. How to correct susceptibility distortions in spin-echo echo-planar images: application to diffusion tensor imaging. Neuroimage 20, 870-888 (2003).

13. Andersson, J.L., et al. A comprehensive Gaussian Process framework for correcting distortions and movements in diffusion images. in Proc Intl Soc Mag Reson Med 2426 (2012).

14. Dale, A.M., Fischl, B. & Sereno, M.I. Cortical surface-based analysis. I. Segmentation and surface reconstruction. Neuroimage 9, 179-194 (1999).

15. Fischl, B. & Dale, A.M. Measuring the thickness of the human cerebral cortex from magnetic resonance images. Proc Natl Acad Sci U S A 97, 11050-11055 (2000).

16. Polimeni, J.R., Fischl, B., Greve, D.N. & Wald, L.L. Laminar analysis of 7T BOLD using an imposed spatial activation pattern in human V1. Neuroimage 52, 1334-1346 (2010).

17. Greve, D.N. & Fischl, B. Accurate and robust brain image alignment using boundary-based registration. Neuroimage 48, 63-72 (2009).

18. Tuch, D.S., et al. High angular resolution diffusion imaging reveals intravoxel white matter fiber heterogeneity. Magn Reson Med 48, 577-582 (2002).

19. Wedeen, V.J., Hagmann, P., Tseng, W.Y., Reese, T.G. & Weisskoff, R.M. Mapping complex tissue architecture with diffusion spectrum magnetic resonance imaging. Magn Reson Med 54, 1377-1386 (2005).

Figures

Figure 1. The radiality histograms at 20% of cortical thickness into WM (-20%), WGB, 20%, 40%, 60%, 80% of cortical thickness into GM, and pial surface. For sulcal wall and fundi, increased spatial resolution revealed a faster transition between non-radial diffusion in the WM to highly-radial diffusion in the GM.

Figure 2. The spatial weighting for the high resolution (1.0mm-b1.5k) data. Closer to the cortex, the high resolution data is weighted more, according to w=max{ (d₀-d)/d₀, exp(-d/d₁) } , while further away from the WGB, the high-b data is weighted more. In this case, d₀= 4mm, d₁= 2mm.

Figure 3. The q-ball ODFs of the 1.0mm-b1.5k data (left) and 2.0mm-b8k data (right) in cortical region, overlaid on the same FA map (1.0mm-b1.5k). The partial voluming with WM is more prominent in the 2.0mm-b8k data, yet crossing structures were not clearly revealed at/near the WGB.

Figure 4. The Jackknife resampling CNR map. The 2.0mm-b8k data showed the best CNR due to larger voxel size and higher diffusion contrast. The 1.0mm-b1.5k data showed relatively poor CNR in the center of the brain. The combined data using Distance-W preserved the high CNR in WM.

Figure 5. The mean ODF (opaque) + 3 times standard deviation (transparent) across 100 Jackknife samples. The mean ODFs are min-max normalized for visualization.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)

1041