Myelin biomarkers in the healthy adult brain: Correlation, reproducibility, and the effect of fiber orientation

Abstract

We investigated the correlation, reproducibility, and effect of white matter fiber orientation for three myelin-sensitive MRI techniques: magnetization transfer ratio (MTR), inhomogeneous magnetization transfer ratio (ihMTR), and gradient and spin echo-derived myelin water fraction (MWF). Methods: We measured the three metrics in 17 white and three deep grey matter regions in 17 healthy adults at 3 T. Results:We found a strong correlation between ihMTR and MTR (r = 0.70, p < 0.001) and ihMTR and MWF (r = 0.79, p < 0.001), and a weaker correlation between MTR and MWF (r = 0.54, p < 0.001). The dynamic range in white matter was greatest for MWF (2.0%-27.5%), followed by MTR (14.4%-23.2%) and then ihMTR (1.2%-5.4%). The average scan-rescan coefficient of variation for white matter regions was 0.6% MTR, 0.3% ihMTR, and 0.7% MWF in metric units; however, when adjusted by the dynamic range, these became 6.3%, 6.1% and 2.8%, respectively. All three metrics varied with fiber direction: MWF and ihMTR were lower in white matter fibers perpendicular to B 0 by 6% and 1%, respectively, compared with those parallel, whereas MTR was lower by 0.5% at about 40 • , with the highest values at 90 • . However, separating the apparent orientation dependence by white matter region revealed large dissimilarities in the trends, suggesting that real differences in myelination between regions are confounding the apparent orientation dependence measured using this method. Conclusion:The strong correlation between ihMTR and MWF suggests that these techniques are measuring the same myelination; however, the larger dynamic range of MWF may provide more power to detect small differences in myelin.

Data-driven characterization and correction of the orientation dependence of magnetization transfer measures using diffusion MRI

“…Note that each following plot uses the same pattern of circles for original measures and squares for corrected measures, as well as a dashed line for the polynomial fit. Furthermore, the variability of the characterization is assessed in Supporting Information and the reproduced results from Morris et al 19 are presented. With these insights, all the following single-subject results come from subject 2, session 4, with a FA threshold of 0.5, a bin width of 1 degree for single-fiber voxels, a bin width of 10 degrees in the case of voxels of crossing fibers for increased voxel count, and only MTR and ihMTR.…”

“…However, this ignores a large fraction of WM voxels, as crossing fibers are estimated to compose 60 to 90% of the total number of voxels in a typical brain 22,23 . For this reason, we extended the typical single-fiber characterization methods 16,19 to be sensitive to multiple orientations in a single voxel, using fODFs instead of the average diffusion tensor from DTI.…”

“…In the same vein, the correction of ihMTR in crossings is not as good as MTR, because we were not able to use the f δ m max ( p 1 ) factor taking into account the varying maximum amplitude. Another concern of the method is the difference in orientation dependence curves between WM structures, as highlighted throughout the manuscript and by Morris et al 19 . Indeed, the whole WM characterization approach used in this work might not be fully adequate for every single WM bundle around the brain.…”

“…This characterization method assumes an homogeneous distribution of myelin around WM, which is not necessarily true, as pointed out by Morris et al 19 . Indeed, an extension of COMMIT allowing myelin streamline decomposition 34 revealed variations of myelination between WM bundles.…”

“…Using their results, they also simulated the variation of ihMTR with respect to the angle ( θ a ) between a fiber bundle and the main magnetic field, in the case of a cylindrical myelin sheath. However, an analysis on in vivo data from a multi-subject dataset showed a different behavior of ihMTR in this regard 19 . Nevertheless, this study also exhibits the dependency of MTR, which seems to follow a trend similar to the simulated ihMTR from Morris et al 18 and the local dipolar field calculated by Girard et al 17 .…”