Mapping the myelin g-ratio: promises and pitfalls
Jennifer SW Campbell1, Ilana R Leppert1, Mathieu Boudreau1, Sridar Narayanan1, Julien Cohen-Adad2,3, G B Pike1,4, and Nikola Stikov2,5

1Montreal Neurological Institute, Montreal, QC, Canada, 2Ecole Polytechnique, University of Montreal, Montreal, QC, Canada, 3Functional Neuroimaging Unit, University of Montreal, Montreal, QC, Canada, 4Hotchkiss Brain Institute, University of Calgary, Calgary, AB, Canada, 5Montreal Heart Institute, Montreal, QC, Canada

Synopsis

The aggregate myelin g-ratio is a function of the myelin volume fraction (MVF) and the fiber volume fraction (FVF). While this relationship holds in theory, obtaining precise and accurate MRI measures of the MVF and FVF remains a challenge. Most MVF mapping techniques have been linearly correlated with histology, but the literature suggests that the slope and intercept are acquisition dependent. In this work, we focus on three magnetization transfer (MT) derived MVF metrics (MTR, MT_sat and qMT) and explore how improper calibration of the MVF estimates propagates to the aggregate g-ratio. The result of an incorrect MVF calibration is not simply loss in sensitivity to g-ratio changes, but rather g-ratio trends that are statistically significant, incorrect, and highly dependent on the fiber volume fraction changes.

Target audience

Researchers and clinicians interested in non-invasive characterization of tissue microstructure

Purpose

The aggregate myelin g-ratio (see Fig. 1) is a function of the myelin volume fraction (MVF) and the fiber volume fraction (FVF) ($$$g = \sqrt{1 - \frac{MVF}{FVF}}$$$)1, 2. While this relationship holds in theory3, obtaining precise and accurate MRI measures of the MVF and FVF remains a challenge. In this work, we focus on three magnetization transfer (MT) derived MVF metrics and explore how improper calibration of the MVF estimates propagates to the aggregate g-ratio.

Methods

We performed magnetization transfer and diffusion imaging (2mm isotropic resolution) on one healthy subject and one multiple sclerosis (MS) patient. For the MS patient, T2W and PDW images were acquired for lesion segmentation. For each subject we calculated three MRI myelin metrics: magnetization transfer ratio (MTR)4, MT_sat5 (similar to MTR, but accounting for T1 effects), and the fractional pool size (F) obtained from quantitative MT6 (10-point sampling of the z-spectrum). To relate each metric to the absolute MVF, we used a combined MT/histology dataset in cynomolgous macaques as the ground truth7, with the scaling constants for F, MTR and MT_sat chosen to match the MVF from histology, assuming a zero intercept. First we computed correlations in brain parenchyma between the three myelin metrics. Then we calculated FVF using the NODDI model of diffusion8 ($$$FVF = MVF + AVF$$$), where AVF is the axon volume fraction, $$$AVF=(1-MVF)(1-v_{iso})v_{ic}$$$. Next, we computed the aggregate g-ratio with each of the three metrics, and looked at g-ratio differences between healthy white matter, normal-appearing white matter (NAWM) and MS lesions. Finally, we performed a theoretical computation varying the slopes (c) and intercepts (b) in the MVF model (MVF = c * MRI_metric + b) to see how improper MVF calibration would affect the aggregate g-ratio measurement.

Results

Figure 2 shows the MVF derived from MTR, MT_sat, and F in a healthy brain. The trends are similar, but the percent difference between healthy white and grey matter was 15.02% for MTR, 40.08% for MT_sat, and 45.86% for F. Figure 3 shows that F is better correlated with MT_sat than with MTR, partly due to the larger dynamic range of MT_sat, but also because the relationship between F and MTR in parenchyma is non-linear (MTR ~ T1 * F). Figure 4 shows aggregate g-ratio maps in the MS patient computed using MTR, MT_sat and F. When using F or MT_sat, the aggregate g-ratio was 0.76 in NAWM and 0.8 in lesions (p < .001). When using MTR, the g-ratio trend was reversed (0.76 in NAWM and 0.65 in lesions, p < .001). This reversed trend (when using MTR) indicates thicker myelin in lesions, which is highly unlikely. Figure 5 shows how improper MVF calibration can bias the g-ratio measures via the FVF metric.

Discussion

Most MT-based MVF mapping techniques have been linearly correlated with histology9-12, particularly in MS brain, where the dynamic range is larger due to pathology. However, the literature suggests that the slope and intercept are acquisition dependent, and that neither linearity nor monotonicity is guaranteed13,14. Figure 5 shows how important it is to get an accurate MVF estimate, even if the relationship is linear. Being off by a small amount will result in interpreting FVF changes as g-ratio changes, and vice versa. When using MTR as a surrogate for myelination, nonlinearity and the conflicting effects of magnetization transfer and T1 changes severely confound quantitative g-ratio computation. In healthy and normal-appearing white matter the g-ratio is relatively constant, so improper MVF calibration gives reasonable values. However, in lesions MTR drops much less than F (15% vs. 50%), whereas AVF drops by more. This makes it seem like little myelin was lost (despite large losses), so the result is an artifactually lowered g-ratio. On the other hand, MT_sat produces similar percent differences to qMT in white matter and MS lesions, making it a clinically feasible candidate for g-ratio mapping (one quarter of the qMT scan time and only 25% longer than conventional MTR).

Conclusion

The result of an incorrect MVF calibration is not simply loss in sensitivity to g-ratio changes, but rather g-ratio trends that are statistically significant, incorrect, and highly dependent on the fiber volume fraction changes. MT_sat and F seem to be better suited for g-ratio mapping (compared to MTR), but great care must be taken in interpreting aggregate g-ratio values in light of the sensitivities demonstrated in this abstract.

Acknowledgements

Study supported by the Quebec Bio-Imaging Network (QBIN 8436-0501), Fonds de Recherche du Québec - Nature et Technologies (FRQNT 2015-PR-182754), the Natural Sciences and Engineering Research Council of Canada (NSERC), and the Canadian Institute of Health Research (CIHR).

References

1. Stikov, N., et al., Bound pool fractions complement diffusion measures to describe white matter micro and macrostructure. Neuroimage, 2011. 54(2): p. 1112-21.

2. Stikov, N., et al., In vivo histology of the myelin g-ratio with magnetic resonance imaging. Neuroimage, 2015. 118: p. 397-405.

3. West, K.L., et al., A revised model for estimating g-ratio from MRI. Neuroimage, 2015.

4. Dousset, V., et al., Experimental allergic encephalomyelitis and multiple sclerosis: lesion characterization with magnetization transfer imaging. Radiology, 1992. 182(2): p. 483-91.

5. Helms, G., et al., High-resolution maps of magnetization transfer with inherent correction for RF inhomogeneity and T1 relaxation obtained from 3D FLASH MRI. Magn Reson Med, 2008. 60(6): p. 1396-407.

6. Levesque, I.R. and G.B. Pike, Characterizing healthy and diseased white matter using quantitative magnetization transfer and multicomponent T(2) relaxometry: A unified view via a four-pool model. Magn Reson Med, 2009. 62(6): p. 1487-96.

7. Stikov, N., et al., Quantitative analysis of the myelin g-ratio from electron microscopy images of the macaque corpus callosum. Data Brief, 2015. 4: p. 368-73.

8. Zhang, H., et al., NODDI: practical in vivo neurite orientation dispersion and density imaging of the human brain. Neuroimage, 2012. 61(4): p. 1000-16.

9. Schmierer, K., et al., Magnetization transfer ratio and myelin in postmortem multiple sclerosis brain. Ann Neurol, 2004. 56(3): p. 407-415.

10. Schmierer, K., et al., Quantitative magnetization transfer imaging in postmortem multiple sclerosis brain. J Magn Reson Imaging, 2007. 26(1): p. 41-51.

11. Dula, A.N., et al., Multiexponential T2, magnetization transfer, and quantitative histology in white matter tracts of rat spinal cord. Magn Reson Med, 2010. 63(4): p. 902-909.

12. Thiessen, J.D., et al., Quantitative MRI and ultrastructural examination of the cuprizone mouse model of demyelination. NMR Biomed, 2013. 26(11): p. 1562-81.

13. Vavasour, I.M., et al., Is the magnetization transfer ratio a marker for myelin in multiple sclerosis? Journal of Magnetic Resonance Imaging, 2011. 33(3): p. 710-718.

14. Gareau, P.J., et al., Magnetization transfer and multicomponent T2 relaxation measurements with histopathologic correlation in an experimental model of MS. Journal of Magnetic Resonance Imaging, 2000. 11(6): p. 586-595.

Figures

Figure 1: White matter model defining the g-ratio, axon volume fraction (AVF), myelin volume fraction (MVF), and fiber volume fraction (FVF).

Figure 2: Maps of myelin volume fraction (MVF) derived from MRI metrics: magnetization transfer ratio (MTR), MT_sat, and quantitative magnetization transfer (qMT). MVFMTR has a narrower dynamic range compared to the other two.

Figure 3: Correlation between MTR and F (r = .59, p<.001), and MT_sat and F (r = .77, p < .001).

Figure 4: Myelin g-ratio in an MS patient computed using MTR, MT_sat and qMT. The arrow shows a lesion with a low g-ratioMTR (compared to NAWM), but the same lesion has an elevated g-ratio when computed using MT_sat and qMT.

Figure 5: Simulations showing that improper MVF calibration (MVF = c * MRI_metric + b) results in artifactual g-ratio variations, driven by changes in the FVF.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
1501