MRI of the spinal cord (SC) at 7T has recently demonstrated its ability to provide high-quality anatomical images [1], with consecutive better characterization of pathologies [2]. Although routinely used to diagnose pathologies and probe SC microstructure at lower field [3], diffusion tensor imaging (DTI) has however, to the authors’ knowledge, not been reported so far on the SC at 7T. Two main reasons may explain this absence of investigation. Firstly, single-shot echo-planar imaging (ss-EPI), which is well established as the method of choice for DTI, is seriously limited at high field by the increased level of artifacts associated with magnetic susceptibility variations at tissue interfaces. Secondly, SC imaging also presents its own challenges such as physiological pulsatility of the cerebrospinal fluid, partial volume effects due to the small physical dimensions of the cord and last, severe magnetic susceptibility-induced image distortions coming from adjacent bone structures. Several original approaches, largely applied in the brain, were proposed to reduce image distortions [4,5]. One of them, which includes post-processing of two ss-EPI acquisitions with opposed phase-encoding directions [6,7], has been used in this study. The work demonstrates that a quite simple yet thoughtful implementation of ss-EPI at 7T combined with distortion correction post-processing can generate high-resolution axial DTI images of the cervical SC with added value compared to lower field standard protocols for an identical acquisition time.

- Whole-body actively-shielded 7T system with an eight-channel transceiver cervical SC coil array.

- Seven healthy volunteers (22±2 years) scanned with approval of the local Ethic Committee.

- Anatomical imaging: sagittal 2D T₂-w TSE sequence (0.6x0.6x2 mm³) for positioning (Figure 1.a), and axial 2D T₂*-w GRE sequence with multiple TEs (4 TE, 0.18x0.18x3 mm³, 12 slices). The sum of squares of all echoes (Figure 1.e) was used for anatomical reference after resampling to DTI resolution (c3D, ITK-SNAP) (Figure 1.f).

- DTI with two opposite phase-encoding acquisitions: R>L (Figure 1.b) and L>R (Figure 1.c) using ss-EPI sequence, at the C3 level (0.8x0.8x3 mm³, b-values: (0,800) s/mm², 12 slices, 12 directions, TE 53ms, PAT 3, 5 averages, pulse trigger, 3-4 concatenations depending on heart rate, and ~7 min acquisition).

- Total acquisition time, including system adjustments (2^nd order B₀ shimming: FWHM~150 Hz, B₁⁺ calibration): 30 min/subject.

- DTI post-processing using FSL Topup (FMRIB) [7] and FSL DTIFIT.

- Quantification within specific WM and GM regions of interest (Figure 1.j) after manual segmentation using FSL.

- Last subject also enrolled for 3T acquisitions: optimized “3T DTI” protocol (~7 min acquisition, 0.9x0.9x10 mm³, same b-values, 6 slices, 30 directions, pulse trigger, 4 averages, 3 concatenations; Figure 1.k), and “7T-like” protocol (Figure 1.l).

Figures 1.b to 1.d illustrate how well Topup corrects image distortions (subject #7). Outlines of the cord parenchyma were manually drawn in R>L (Figure 1.b), L>R (Figure 1.c) and in corrected (Figure 1.d) images, and subsequently compared to the reference cord outline (Figure 1.f). Related zoom (Figure 1.g) depicts all outlines, with significant overlap between the reference (red) and the corrected (green) outlines, showing accurate distortion corrections. The sum of squares of all GRE echoes provided clear delineation of the GM butterfly (Figure 1.e). DTI metrics, here MD (Figure 1.h, mean diffusivity) and FA (Figure 1.i, fractional anisotropy) maps enable cord and GM butterfly visualization respectively, with exquisite delineation of the posterior horns (Figure 1.I). Figure k. and l. exhibit FA maps of subject #7 acquired at 3T for “3T DTI” and “7T-like” protocols, respectively. The “3T DTI” protocol was acquired within the same acquisition time but with a lower spatial resolution. It shows decent SNR and GM butterfly delineation yet slice thickness was 3.33 times larger, precluding visualization of small tissue alteration. The “7T like” protocol was too noisy to interpret data. Quantitative results for all subjects at 7T (Figure 2) were found comparable to values measured at 3T in this work and coherent with [8]. Furthermore, higher spatial resolution and slice thickness helped minimizing partial volume effect while obtaining DTI quantification within the posterior horns.A quite simple implementation of ss-EPI at 7T combined with distortion correction post-processing can generate useful high-resolution axial DTI images of the cervical SC. Obtained results confirmed the added value of the 7T in terms of image resolution, metrics evaluation in precise ROIs and total sequence coverage, compared to lower field standard protocols for an equivalent acquisition time. This study lays the groundwork for high-resolution DTI of the SC at 7T. The proposed methodology can direcly be used to explore degenerative SC pathologies and quantify tissue alterations. Future studies will focus on higher spatial resolution and refined sequences.