High-resolution quantitative magnetic resonance imaging of the human cervical spinal cord at 7T
Aurélien Massire1,2,3, Manuel Taso1,2,3,4, Maxime Guye1,2, Jean-Philippe Ranjeva1,2,3, and Virginie Callot1,2,3

1Centre de Résonance Magnétique Biologique et Médicale (CRMBM), UMR 7339, CNRS, Aix-Marseille Université, Marseille, France, 2Centre d'Exploration Métabolique par Résonance Magnétique (CEMEREM), Hôpital de la Timone, Pôle d’imagerie médicale, AP-HM, Marseille, France, 3iLab-Spine - Laboratoire international - Imagerie et Biomécanique du rachis, Marseille, France, 4LBA, UMR T24, Aix-Marseille Université, IFSTTAR, Marseille, France

Synopsis

A high-resolution multi-parametric MRI protocol dedicated to 7T cervical spinal cord (SC) investigation using a commercial prototype transceiver radiofrequency coil array is proposed. This work pushes forward SC quantitative MRI by reporting T1/T2/T2* relaxation times mapping as well as diffusion tensor imaging metrics at the C3 cervical level on a cohort of ten healthy volunteers. Automatic segmentation and registration of these multi-parametric acquisitions to SC templates enable group studies with quantitative evaluation within regional WM tracts and GM horns never reported so far at 7T. This study lays the groundwork for improved characterization of degenerative SC pathologies at ultra-high field.

Introduction

Although quantitative MRI (qMRI) techniques have proven huge potentials to provide information about spinal cord (SC) pathologies from both microstructural and functional perspectives [1], there is still a strong need for pathological biomarkers and predictive factors of accurate prognosis. By enabling very high image resolution and enhanced tissue contrast acquisitions, ultra-high field (UHF) imaging offers new opportunities to investigate neurological diseases, as recently demonstrated in the brain [2]. In the SC, only few studies have been conducted so far, mainly reporting advantages for anatomical description [3,4]. The present work pushes forward SC qMRI at 7T by evaluating T1, T2 and T2* relaxation times as well as fractional anisotropy (FA) and mean diffusivities (λ//, λ^, MD), on a cohort of ten healthy volunteers. The SNR available at 7T was traded to achieve very high in-plane and 3D spatial resolution compared to conventional clinical imaging, thereby enabling fine SC substructure findings. Furthermore, automatic segmentation and registration of these multi-parametric acquisitions to SC templates enabled group studies with quantitative evaluation of regional WM tracts and GM horns. Such high-resolution morphometric and structural data pave the way for future clinical SC qMRI at UHF.

Methods

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

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

- Anatomical imaging: sagittal 2D T2-weighted (T2-w) TSE sequence (0.6x0.6x2 mm3) used for accurate axial positioning.

- Relaxometry mapping: coronal 3D T1-w MP2RAGE sequence (0.7x0.7x0.7 mm3, C1-C7 coverage), axial 2D T2*-w GRE sequence with multiple TEs (0.18x0.18x3 mm3, 12 slices) and TE-stepped axial 2D T2-w segmented spin echo EPI sequences (4 TE, 0.8x0.8x3 mm3, 12 slices, pulse triggered) at the C3 level.

- Diffusion Tensor Imaging (DTI): two opposite phase-encoding acquisitions using a single-shot EPI sequence (0.8x0.8x3 mm3, b-values: (0,800) s/mm2, 12 slices, 12 directions, pulse triggered) at the C3 level.

- Total acquisition time, including system adjustments (B0 shimming, B1+ calibration): 50 min/subject.

- Image post-processing (see Figure 1): semi-automated, using the Spinal Cord Toolbox (SCT) [5], Matlab (The Mathworks, Natick, MA, USA) & FSL (FMRIB, Oxford, UK).

- Quantification within specific WM [6] and GM regions of interest (ROIs) at the C3 cervical level: after registrations of all parameter maps within a common reference space (MNI-Poly-AMU template [7]), non-linear co-registrations with the AMU40 template [8], and arithmetic sum on all subjects.

- Statistical analyses: paired t-test and HSD Tukey-Kramer test with JMP9 (SAS, Cary, USA) (p-values <0.05 significant).

Results

Figure 2.a exhibits MP2RAGE-derived T1-w image (sagittal view), where spatial coverage of the coil from cerebellum to the T1 level can be appreciated. Corresponding T1 map (Figure 2.b) is directly computed with Bloch equation. Figure 2.c and 2.d show sagittal TSE and axial GRE acquisitions. By combining high SNR and CNR, the sum of squares of all GRE echoes enables excellent delineation of the GM butterfly and visualization of fine anatomical details (see zoom). Resulting average maps (Figure 3), where GM butterfly (T1, T2*-w images: 3.d and 3.e / T1, T2*, and FA maps: 3.a, 3.f and 3.h) and cord parenchyma (T2 and MD maps: 3.g and 3.i) can be clearly visualized, enable group study within specific ROIs (3.b and 3.c). Quantitative results (T1, T2, T2*, λ^, λ//, FA and MD) at C3 cervical level in all ROIs are summarized in Figure 4. Statistical differences were observed between WM and GM structures (see table). FA, λ^ and λ// measurements also showed statistical differences between motor pyramidal tracts and sensitive gracile/cuneate tracts.

Conclusion

A high-resolution multi-parametric MRI protocol dedicated to 7T cervical SC investigation using a commercial prototype transceiver radiofrequency coil array is proposed. For the first time, high-resolution qMRI data including T1/T2/T2* relaxation times mapping and DTI metrics are reported (C3 level): clear delineation of the SC substructures (WM/GM) and regional statistical differences could be observed (sensory/motor tracts, anterior GM). Such ultra-high-field multi-parametric MR protocol opens great perspectives for further clinical investigations of SC degenerative and traumatic diseases. Further developments will focus on protocols enabling multiple level investigations, construction of high-resolution 7T SC templates and parallel transmission pulse designs to release full potential of the coil.

Acknowledgements

Fundings: ANR-11-EQPX-0001, A*MIDEX-EI-13-07-130115-08.38-7T-AMISTART & A*MIDEX ANR-11-IDEX-0001-02.

References

[1] Wheeler-Kingshott et al., Neuroimage 84, 2014. [2] van der Zwaag et al., NMR Biomed, 2015. [3] Sigmund et al., NMR Biomed 25, 2012. [4] Dula et al., Mult Scler, 2015. [5] Cohen-Adad et al., OHBM, 2014. [6] Lévy et al., Neuroimage 119, 2015. [7] Fonov et al., Neuroimage 102, 2014. [8] Taso et al., Neuroimage 117, 2015.

Figures

Figure 1: Image processing on multi-parametric data acquired at 7T. A: T1 data processing (blue). B: T2* data processing (green). C: DTI data processing (red). D: T2 data processing (orange). E: Registrations to templates (yellow) to work in a common reference space and corresponding multi-parametric outputs in ROIs.

Figure 2: MP2RAGE acquisition (a: T1-w image; b: corresponding quantitative T1 map). c: Sagittal TSE image with GRE axial positioning. d: T2*-w image (GRE, sum of squares) with zoom on anatomical details (nerve roots: blue, ligaments: purple, blood vessels: red, dura mater: green and pia mater: yellow).

Figure 3: Qualitative (T1-weighted and T2*-weighted) and quantitative (T1, T2, T2*, FA and MD) axial images in the template space at the C3 level. Specific ROIs (blue: whole WM, orange: whole GM; red: anterior GM, yellow: sensitive WM and green: lateral motor WM).

Figure 4: Quantitative results (T1, T2, T2*, λ^, λ//, FA and MD) at C3 cervical level in all ROIs. Statistical differences: * between whole WM and GM, † between WM tracts and anterior GM horns and ‡ between “sensitive” and “motor” WM tracts.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1130