3234

A novel diffusion MRI acquisition scheme for optimal orientation sampling and susceptibility distortion correction at 7T
Antoine Legouhy1, Martina F Callaghan2,3, Hui Zhang1, and David L Thomas2
1CMIC, University College London, London, United Kingdom, 2UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, 3Imaging Neuroscience, University College London, London, United Kingdom

Synopsis

Keywords: Data Acquisition, Diffusion/other diffusion imaging techniques, Distortion correction

Motivation: 7T provides higher SNR compared to clinical field strengths, but susceptibility-induced distortion compromises diffusion imaging at 7T. Current methods to accurately correct distortion require doubling of scan time.

Goal(s): Develop an efficient acquisition scheme to enable distortion correction of multi-direction diffusion images without increased scan time.

Approach: Two ‘half-density’ acquisitions with opposing phase-encoding directions are collected, such that their gradient orientations uniformly sub-sample the sphere in a complementary manner. Spherical harmonics are used to interpolate image pairs for distortion correction.

Results: The proposed method provides distortion correction comparable to the full blip-up/blip-down approach, in half the acquisition time, while retaining the orientation density.

Impact: As the clinical benefits of 7T become established, it is crucial for all sequence types to be available for use. The methods proposed here enable advanced distortion-corrected diffusion imaging to be routinely and efficiently included in 7T imaging protocols.

Introduction

7T MRI provides higher signal, which is particularly beneficial for SNR-limited methods such as diffusion-weighted imaging (DWI). However, imaging at 7T also has associated issues; in particular, by using EPI for image acquisition, DWI suffers from significant spatial distortions in the phase-encoding (PE) direction due to increased B0 inhomogeneity.

To correct this, paired data sets with PE applied in opposite directions are typically acquired (known as ‘blip-up/blip-down’ (BU/BD), indicating the switched polarity of the EPI PE blips). Distortions can then be corrected using post-processing algorithms such as FSL-TOPUP1 or HySCO2, to achieve anatomical fidelity. However, for optimal image restoration, this approach requires the dataset to be acquired twice, doubling the scan time; this is undesirable for high angular resolution DWI since acquisition times are already lengthy, and previous studies have shown that acquiring images with many different gradient directions is preferable acquiring repeats/averages3.

In this work, we investigate an alternative BU/BD acquisition scheme which enables accurate distortion correction without increased scan time. We show that this new approach performs comparably to using the full BU/BD scheme, and is superior to using a BU/BD b=0 image only, as is commonly done at 3T.

Methods

Whole brain imaging of 7 volunteers was performed on a 7T Siemens Terra scanner, using a multi-band diffusion-weighted EPI sequence (CMRR, University of Minnesota)4. Imaging parameters were: 1.5mm isotropic resolution, 96 slices, multiband factor 2, GRAPPA 3. Two acquisition protocols were run with PE in both the A>>P and P<<A directions:
  • DTI5: single shell 64$$$\times$$$b=1000 s/mm2 (plus 1$$$\times$$$b=0) [5min48s]
  • NODDI6: 3 shell 8$$$\times$$$b=300; 32$$$\times$$$b=700; 64$$$\times$$$b=2000 (plus 13$$$\times$$$b=0) [9min36s]

For distortion correction, the following approaches were investigated (see Figure 1):
  • Reference: BU/BD for all directions.
    Using all the acquired data; all b-vectors have ‘blip-up’ and ‘blip-down’ images.
    This represents the ‘gold standard’ but requires double the desired scan time.
  • Baseline: BU for DW for all directions, BD additionally only for b=0.
    Using all the b vectors with ‘blip-up’ PE and the additional b=0 images acquired with ‘blip-down’.
    This is a commonly used approach for diffusion at 3T.
  • Proposed: BU for half the directions, BD for the other half (split in a complementary manner; see Figure 1), and both BU/BD for b=0.
    A novel approach where the orientation sampling is split into two equally-sized complementary subsets, each covering the sphere as uniformly as possible in an interleaved manner. One subset is acquired with ‘blip-up’, the other with ‘blip-down’. To complete BU/BD pairs for distortion correction, missing images are interpolated from the acquired images through their spherical harmonic representation7.

Raw data were corrected for eddy current distortions using FSL-EDDY8, and susceptibility distortion fields were estimated from b=0 images using FSL-TOPUP1. For the proposed method, uniform orientation sub-sampling was performed using Camino9,10, and spherical harmonic interpolations were performed on EDDY-corrected data using MRtrix7,11. Finally, susceptibility correction was applied to each BU/BD image pair (BU only for the baseline).
The quality of correction was evaluated for the baseline and proposed methods by comparing distortion-corrected signal intensities and diffusion metric maps with the reference full BU/BD method.

Results

Compared to the baseline acquisition scheme, the proposed method shows a closer signal intensity match to the reference data set (Figure 2/3). This has direct repercussions on the accuracy of the estimation of diffusion metrics for both DTI and NODDI (Figure 4/5). The differences are especially apparent in areas of strong distortion, such as the inferior frontal lobe, brainstem and posterior cerebellum.

Discussion

Our results demonstrate that spherical harmonic interpolation can be used to generate synthetic images suitable for BU/BD distortion correction. Spherical harmonics do not require multi-shell, and are well established, fast to estimate and widely available. Other interpolation methods could also be used for this purpose, potentially embedding all the b-values in a single representation and improving performance e.g. Gaussian process12 or SHORE13. However, the requirement for multi-shell data sets would prevent application to the most common situation of single-shell DTI. Previous work has used interpolation to convert between different orientation sampling schemes, to investigate their equivalence for estimating diffusion metrics14, as well as for outlier detection15. The latter could be incorporated into the approach developed here to recover images lost to motion-induced dropout, enabling restoration of the BU/BD pair and subsequent distortion correction. Lastly, the proposed scheme is also likely to be useful for diffusion imaging data at 3T; we are currently investigating this application.

Conclusion

We have introduced a new acquisition scheme for multi-direction diffusion imaging at 7T which enables both high angular resolution b-shell sampling and accurate susceptibility distortion correction.

Acknowledgements

The Wellcome Centre for Human Neuroimaging is supported by core funding from Wellcome [203147/Z/16/Z]. This work was funded by the UCL Wellcome Institutional Strategic Support Fund 3 (204841/Z/16/Z).

AL and HZ are additionally supported by the Fonds de la Recherche Scientifique (F.R.S.-F.N.R.S) and the Fonds Wetenschappelijk Onderzoek – Vlaanderen (F.W.O.) under the Excellence of Science (EOS) Project (MEMODYN, No. 30446199).

References

1. Andersson JL, Skare S, Ashburner J. How to correct susceptibility distortions in spin-echo echo-planar images: application to diffusion tensor imaging. NeuroImage 2003; 20(2):870-88;

2. Clark IA, Callaghan MF, Weiskopf N, Maguire EA, Mohammadi S. Reducing Susceptibility Distortion Related Image Blurring in Diffusion MRI EPI Data. Front Neurosci. 2021 Aug 5; 15:706473

3. Jones DK. The effect of gradient sampling schemes on measures derived from diffusion tensor MRI: a Monte Carlo study. Magn Reson Med 2004; 51: 807–815.

4. https://www.cmrr.umn.edu/multiband/

5. Le Bihan, D, Mangin JF, Poupon C, Clark CA, Pappata S, Molko N, Chabriat H. Diffusion tensor imaging: concepts and applications. Journal of Magnetic Resonance Imaging 2001; 13: 534-546

6. Zhang H, Schneider T, Wheeler-Kingshott CA, Alexander DC. NODDI: practical in vivo neurite orientation dispersion and density imaging of the human brain. NeuroImage 2012; 61(4): 1000-1016

7. Tournier JD, Calamante F, Connelly A. Robust determination of the fibre orientation distribution in diffusion MRI: non-negativity constrained super-resolved spherical deconvolution. NeuroImage 2007; 35, 1459–1472

8. Andersson, JLR, & Sotiropoulos, SN. An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging. NeuroImage 2016; 125, 1063–1078.

9. P. A. Cook, Y. Bai, S. Nedjati-Gilani, K. K. Seunarine, M. G. Hall, G. J. Parker, D. C. Alexander, "Camino: Open-Source Diffusion-MRI Reconstruction and Processing", International Society for Magnetic Resonance in Medicine, Seattle, WA, USA, p. 2759, May 2006

10. Cook, P. A., Symms, M. R., Boulby, P. A., & Alexander, D. C. (2006). Ordering diffusion-weighted MRI measurements improves results from partially completed scans. In Proceedings of the Scientific Meeting of the International Society for Magnetic Resonance in Medicine (p. 1035).

11. Tournier, J-D, Smith RE, Raffelt D, Tabbara R, Dhollander,T, Pietsch M, Christiaens D, Jeurissen B, Yeh C-H & Connelly A. MRtrix3: A fast, flexible and open software framework for medical image processing and visualisation. NeuroImage 2019; 202, 116137

12. Andersson JL, & Sotiropoulos SN. Non-parametric representation and prediction of single- and multi-shell diffusion-weighted MRI data using Gaussian processes. NeuroImage 2015 122, 166–176.

13. Ozarslan E, Koay C, Shepherd T, Blackband S, Basser P. Simple harmonic oscillator based reconstruction and estimation for three-dimensional q-space MRI. 2009 ISMRM 17th Annual Meeting 1396

14. Yeh F-C and Verstynen TD. Converting Multi-Shell and Diffusion Spectrum Imaging to High Angular Resolution Diffusion Imaging. 2016 Front. Neurosci. 10:418

15. Nath V, Schilling KG, Hainline AE, Parvathaneni P, Blaber JA, Lyu I, Anderson AW, Kang H, Newton AT, Rogers BP, Landman BA. SHARD: Spherical Harmonic-based Robust Outlier Detection for HARDI Methods. Proc SPIE Int Soc Opt Eng. 2018 Mar; 10574:105740X.

Figures

Figure 1. Diagram of the reference, baseline and proposed acquisition schemes for DWI as well at subsequent image processing pipelines.

Figure 2. Error in the DW signal after distortion correction: For each of the 7 subjects, absolute difference between the scaled signal of the undistorted images from the baseline (blue) or proposed (green) acquisition scheme, and the scaled signals of the reference acquisition scheme. Results are averaged across all volumes of a given b-value. The DW signals are scaled by their respective average b=0 signal.

Figure 3. Error in the DW signal after distortion correction: For subject 1, absolute difference between the scaled signal of the undistorted images from the baseline or proposed acquisition scheme, and the scaled signals of the reference acquisition scheme. Results are averaged across all volumes of a given b-value. The DW signals are scaled by their respective average b=0 signal.

Figure 4. Error in the diffusion metrics: For each of the 7 subjects, absolute difference between the diffusion from the baseline (blue) or proposed (green) acquisition scheme, and the diffusion metrics of the reference acquisition scheme. The DTI acquisition has been used for the DTI metrics (MD, FA, V1), the NODDI acquisition has been used for the NODDI metrics (NDI, ODI). V1 stands for the principal direction, the error is an angle in degrees.

Figure 5. Error in diffusion metrics: For subject 1, absolute difference between the diffusion metrics from the baseline or proposed acquisition scheme, and the diffusion metrics of the reference acquisition scheme. The DTI acquisition has been used for the DTI metric (MD), the NODDI acquisition has been used for the NODDI metrics (NDI, ODI). MD errors are multiplied by a factor 1000.

Proc. Intl. Soc. Mag. Reson. Med. 32 (2024)
3234
DOI: https://doi.org/10.58530/2024/3234