Myelin Water Imaging (MWI) is a quantitative T₂ relaxation-based magnetic resonance imaging technique¹, which has been validated as a viable biomarker for myelin in the central nervous system (CNS)². The MR signal from tissue in the CNS originates from three main water pools: water trapped between myelin bilayers (T₂≈20ms), intra/extra-cellular water (T₂≈80ms) and cerebrospinal fluid (T₂≈2s). The signal fraction originating from each water pool is determined by analysing the T₂-decay curve using a non-negative least squares (NNLS) fitting-algorithm¹. From this analysis the myelin water fraction (MWF) is obtained as the ratio of the signal originating from water within the myelin bilayers to the total MR signal.

The acquisition time for MWI has improved significantly over the last 20 years, most recently with the implementation of a 3D Gradient Spin Echo (3D-GRASE) sequence enabling full cerebral coverage in 15min³. However, multi-slice MWI in spinal cord (SC) has thus far only employed a 3D Turbo Spin Echo (3D-TSE) sequence with a significantly longer acquisition time⁴. The purpose of this study was to investigate the reliability of using a high resolution 3D-GRASE sequence for MWI in cervical spinal cord in vivo to speed up data acquisition by a factor of three.

MR Experiments: Seven healthy volunteers underwent imaging on a Philips Achieva 3T scanner using a 6-channel spine coil. Scans were centered at the C2/C3 disc and aligned perpendicular to the cord. The scanning protocol included:

(1) 3D-GRASE (32 echoes, echo spacing 10ms, TE_min=10ms, TE_max=320ms, TR=1500ms, refocusing flip angle α_refocus=180°, acquired resolution=1x1x5mm³, FOV=180x40x150mm³ (AP, FH, RL), reconstructed resolution=0.625x0.625x2.5mm³, SENSE factor 2 Right-Left, scan time=8.6min)

(2) 3D-TSE (all parameters matched to 3D-GRASE except for α_refocus=135°, FOV=180x40x135mm³, no SENSE, and reconstructed resolution=0.7x0.7x5mm³, scan time=23.3min)

(3) T₂ weighted multi-echo fast gradient echo (mFFE) (5 echoes, echo spacing 8.2ms, TE_min=6.6ms, TE_max=39.4ms, TR=814ms, acquired resolution=0.8x0.8x2.75mm³, FOV=150x43.75x150mm³, reconstructed resolution=0.3x0.3x2.75mm³, for anatomical segmentation).

To assess reproducibility, the 3D-GRASE sequence was collected twice, at the beginning and end of the scanning protocol.

Image Analysis: Voxel-wise MWF maps were calculated by analyzing the 3D-GRASE and 3D-TSE T₂-decay data using in-house software, with corrections for stimulated echo artifacts as well as regularizing the data in both spatial and T₂ space⁵. Regions of interest (ROIs) were obtained by registering the mFFE scan to the MNI-POLY-AMU spinal cord template from the Spinal Cord Toolbox⁶ and then registering the 3D-GRASE and 3D-TSE scans to the mFFE. Four ROIs from the template (full cord, gray matter (GM), the dorsal column (DC), and the lateral corticospinal tract (LCT)) were registered to mFFE space and MWF within each ROI was determined (Figure 1).

Statistical Analysis: MWF estimates obtain by 3D-GRASE and 3D-TSE were compared using linear regression and Bland-Altman analysis. Reproducibility of the 3D-GRASE sequence was evaluated by calculating the coefficient of variation (COV) in MWF between the two 3D-GRASE experiments.

One subject was excluded from the results due to severe motion artifacts rendering ROI registration unreliable. 3D-GRASE and 3D-TSE MWF maps were qualitatively similar (Figure 2) and linear regression demonstrated a strong positive correlation between 3D-GRASE and 3D-TSE MWF (Figure 3, R²=0.69, p<0.001). A Bland-Altman plot showed no biases for the various ROIs (Figure 4). Average whole cord segment MWF was 23% (SD=1.5%) for 3D-GRASE and 24% (SD=3%) for 3D-TSE. 3D-GRASE showed good scan-rescan reproducibility with average COV’s ranging from 6.2% in whole cord to 12.8% in GM (Table 1).The MWF results obtained in this study are in agreement with previous studies investigating cervical spinal cord MWF with values between 21.8-26.4%⁷.Bland-Altman analysis (Figure 4) of 3D-GRASE and 3D-TSE showed that the 95% confidence intervals for the mean difference overlaps zero, indicating there is no significant difference between the 3D-GRASE and 3D-TSE results. Regression analysis (Figure 3) also shows that the unity line, i.e. the one-to-one correlation, falls within the 95% confidence interval of the linear fit, further supporting the hypothesis that these techniques produce comparable results. Our study is the first successful implementation of a 3D-GRASE sequence for human cervical spinal cord MWI in vivo. 3D-GRASE MWF is reproducible and in good agreement with MWF values reported in previous studies. Using 3D-GRASE for MWI significantly reduces the acquisition time compared to the matched 3D-TSE sequence. Our findings demonstrate that cervical cord myelin water data can reliably be collected in clinically feasible scan times.