Introduction

Myelin water imaging (MWI) using multi-echo gradient echo (mGRE) is a promising tool for myelin quantification, but it is a highly ill-conditioned and SNR-demanding problem^1-3. Multi-compartment relaxometry MWI (MCR-MWI)⁴, utilising variable flip angle (VFA) acquisition, benefits both from the different apparent transverse and longitudinal relaxations of the free (intra- and extra-axonal) and myelin water pools. MCR-MWI corrects the T₁-saturation dependency on the measured myelin water fraction (MWF) and alleviates the MWI ill-condition problem, resulting in higher SNR MWF maps. Moreover, MCR-MWI can incorporate with diffusion-weighted imaging (DWI) data (MCR-DIMWI) to improve data fitting robustness.

For MCR-(DI)MWI to be adopted in longitudinal studies or multi-centre trials, it has to be reproducible both within-subject and across protocols. In this study, we evaluate the reproducibility of MCR-(DI)MWI in a test-retest experiment (Fig.1a), using 3 sets of protocols with different echo spacings (ΔTE) and repetition times (TR) to examine the cross-protocol repeatability (Fig.1b). We also evaluate the reproducibility of the MWI-related tissue properties across fibre bundles of different subjects.

Methods

Data acquisition
Data acquisition was performed at 3T (Siemens, Erlangen) on 2 heathy volunteers. 3 different protocols were used for two identical imaging sessions (Fig.1a) to evaluate the reproducibility of MCR-(DI)MWI:
1) TR/TE1/ΔTE/nTE=38/2.2/3.07ms/12, TA=2.8min/flip angle (α): short scan time condition;
2) TR/TE1/ΔTE/nTE=50/2.2/3.07ms/15, TA=3.5min/α: data-rich condition;
3) TR/TE1/ΔTE/nTE=55/2.68/3.95ms/13, TA=4min/α: less-performing gradient setup;
with sagittal slice orientation, FOV=168x222x240mm³, res=1.5mm iso., α=[5,10,20,50,70]°, R_CAIPI=5.

B₁ map was acquired using a turbo-flash protocol to correct B₁ field inhomogeneities.

DWI data was acquired in one of the sessions for MCR-DIMWI and bundle-specific analysis with parameters: 2D-MB EPI-DWI, MB=3, res=1.6mm iso., TR/TE=3350/71.20ms, 2-shell (b=0/1250/2500s/mm², 17/120/120 measurements), TA=15mins.

Data processing
DWI data was pre-processed (see Fig.1c). SMT⁵ and a ball-and-stick model⁶ were used to derive white matter (WM) properties for MCR-DIMWI. VFA-mGRE data of each protocol, the DWI-derived maps and B₁ maps were coregistered. Voxel-wise fitting was performed with the MCR-(DI)MWI model to obtain MWF and other tissue properties⁴.

To study the MWF reproducibility in bundle-specific analysis, the MCR-derived maps were registered to the undistorted DWI space using rigid body transformation. Fifty subject-specific WM bundles were obtained from TractSeg⁷, using multi-shell multi-tissue constrained spherical deconvolution⁸. Tractometry was performed to compute the bundle-specific MWF (and other parameters) profile at 100 locations along the tracts. Only the middle 90 sections were used (excluding the first and last 5 sections which have increased partial volume effects), resulting in 4500 WM ROIs/subject.

Data analysis
Test-retest reliability was assessed via Pearson’s correlation coefficient (R) and coefficient of variation (CV, standard deviation of the differences divided by the mean) for both within GRE protocols (across sessions) and across protocols (both within and across sessions) with all 9000 WM ROIs.

Bundle-specific MWF profiles were visually compared on 8 major fibre bundles between MCR-MWI and MCR-DIMWI, and between subjects using the protocol 1 results. These bundles are: splenium/genu of the corpus callosum (CC), left/right optic radiation (OR), left/right inferior longitudinal fascicles (ILF) and left/right corticospinal tracts (CST).

MCR-MWI, beyond MWF, also delivers free water R₁ (R_1,IEW), exchange rate (k_IEWM) and intra-axonal water R₂* maps amongst others. To study the potential of these parameters and their information independence, we examine the subject-averaged parameter profiles along 4 major bundles. Cross-correlation amongst these 4 parameters is also reported using protocol 1 results.

Results and Discussion

Fig.2 shows visually the MCR-(DI)MWI fitting quality across the whole brain for various protocols. The only noticeable difference is in deep grey matter nuclei where all MCR-DIMWI created unrealistically high MWF.

All test-retest analyses show acceptable reliability (Fig.3a: R>=0.7 for MCR-MWI, Fig.3b: R>0.8 for MCR-DIMWI). Within protocol reliability (diagonal) is higher than between protocols: highest reliability for protocol 1 (CV<10% and R=0.85, despite the shortest acquisition time) and lowest for protocol 3 (where longer ΔTE introduced noise in the estimation). The higher MCR-DIMWI test-retest reliability and lower CV are also seen as increased SNR in the MWF maps in Fig.2.

Generally, MCR-MWI and MCR-DIMWI have very similar tract profiles on the same (subject) dataset between the left and right hemispheres (ILF and CST – Fig.4). Good agreements can also be seen between the two subjects in, e.g., OR and the genu of the CC.

Fig.5 shows the subject-averaged MWF tract profile along with other tissue parameters, indicating that the relaxation parameters carry independent information of MWF. Amongst the 4 parameters shown, MWF is the most robust measurement across all protocols, while T_1,IEW and k_IEWM show increased dependency with the use of various TRs. This can be attributed to both the flat cost function associated with these parameters⁹ and the different MT weighting associated with the different TRs¹⁰.

Conclusions

MCR-MWI is a reproducible measure for myelin quantification. Our results suggest that decreasing the echo time range and echo spacing results in more robust MWF maps, which is attributed to the increased physiological noise at longer echo times and the need to fit the short myelin water signal. Incorporating DWI data can improve the repeatability of MCR-MWI in WM. The repeatability measures (R and CV) will be useful for power analysis in future group/longitudinal studies. Future work will focus on studying tissue properties other than MWF which potentially contain other microstructure structure information.