Quantification of demyelination and remyelination with diffusion MRI: specific in vivo White Matter Tract Integrity metrics agree with electron microscopy-derived features
Ileana O Jelescu1, Magdalena Zurek1, Kerryanne V Winters1, Jelle Veraart1, Anjali Rajaratnam1, Nathanael S Kim1, James S Babb1, Timothy M Shepherd1, Dmitry S Novikov1, Sungheon G Kim1, and Els Fieremans1

1Center for Biomedical Imaging, Radiology, New York University School of Medicine, New York, NY, United States


White Matter Tract Integrity (WMTI) metrics derived from diffusion data provide a compartment-specific characterization of white matter. Here, we evaluated the specificity of the axonal water fraction (AWF) and extra-axonal radial diffusivity (De,┴) by assessing their correlations to metrics derived from electron microscopy (EM), in the splenium of control, cuprizone-treated and recovering mice. As the model predicted, the WMTI-derived AWF correlated very strongly with the EM-derived AWF, but not with the g-ratio, while De,┴ correlated with the g-ratio, but not with the EM-derived AWF. WMTI parameters are therefore promising biomarkers for specific biophysical aspects of white matter pathology in vivo.


Monitoring of myelin damage and repair is of great clinical importance for demyelinating diseases. Conventional MRI measures (e.g. T2, MTR and radial diffusivity (RD)) are sensitive, but non-specific to the WM injury, whereas WM tract integrity (WMTI) metrics1 derived from diffusion kurtosis imaging (DKI) can provide compartment specific estimates of WM properties, such as axonal water fraction (AWF) and extra-axonal radial diffusivity (De,┴) (Fig. 1). Cuprizone fed mice provide an excellent model of demyelination2. In vivo changes in both conventional MRI measures and WMTI parameters during cuprizone-induced demyelination and subsequent recovery were previously reported in the mouse corpus callosum3-5. Here, we examine the correlations between WMTI parameters and electron microscopy (EM) measures.


We conducted an 18-week longitudinal study on 34 female C57 mice (8 weeks old at baseline) separated into 3 groups: "Control" group received standard chow; "Cpz6" and "Cpz12" groups received cuprizone for 6 and 12 weeks, respectively (Fig. 2). MRI: In vivo MRI was performed using a 7-T Bruker BioSpec system. Diffusion-weighted images of one 0.8-mm thick mid-sagittal slice were acquired (b=1000 and 2000 s/mm2, 30 directions each, and 6 b=0 images) at 112-μm in-plane resolution. DKI analysis and derivation of the WMTI parameters were performed1. The splenium was segmented using a semi-automated in-house software tool. EM: Subsets of animals were sacrificed and perfusion-fixed at 6, 12 and 18 weeks (Fig. 2). The brains were extracted and further prepared for EM. Nanometer-thick slices from the splenium were imaged using a Phillips CM12 transmission electron microscope at 2.36 nm spatial resolution. Electron micrographs were analyzed using in-house developed macros in ImageJ and CellProfiler. The EM metrics extracted were myelin volume fraction (MVF), myelinated axon volume fraction (mAVF) and total axonal volume fraction (tAVF). An "aggregate" g-ratio was derived as $$$\sqrt{tAVF/(MVF+tAVF)}$$$. Because there is a conceptual difference between AVF and AWF (the former refers to a volume fraction, the latter is a fraction of "MR visible" water, which does not account for the myelin space), the EM-derived $$$mAWF=mAVF/(1-MVF)$$$ and $$$tAWF=tAVF/(1-MVF)$$$ were further calculated. The EM image analysis was performed for three randomly chosen mice per timepoint and diet group, and eight micrographs per animal, resulting in 423 myelinated and 1623 unmyelinated axons processed on average per animal. Correlations: The mean and standard deviation in the splenium were calculated for all MRI and EM measures in each animal. Partial Spearman correlations (ρ) were calculated between the MRI and the EM measures, using mouse weight at baseline as a covariate. A linear (Pearson) correlation (r) was also computed between WMTI-derived AWF and EM-derived tAWF to evaluate their direct agreement.

Results and Discussion

All EM and MRI measures were consistent with expected trends of cuprizone intoxication and recovery at all timepoints (Figs. 3 and 4). Acute cuprizone exposure (6 weeks) was characterized by "patchy" demyelination (Figs. 1C and 3A) and significant decrease in AWF, and chronic exposure (12 weeks) by widespread demyelination and significant increase in De,┴.

Correlations are summarized in Figure 5. RD correlated significantly with EM-derived MVF, mAWF, tAWF and g-ratio. Among WMTI metrics, the AWF correlated most strongly with tAWF, followed by mAWF and MVF; AWF did not correlate significantly with the g-ratio (although a trend was present). Conversely, De,┴ correlated most strongly with MVF and g-ratio; De,┴ did not correlate significantly with mAWF and tAWF.

The correlations are consistent with results from previous simulations6, which suggest AWF is particularly sensitive to random axonal loss and/or patchy demyelination, while De,┴ is particularly sensitive to homogeneous demyelination, and hence to the g-ratio .

The WMTI-derived AWF is by design a measure of the same quantity as the EM-derived tAWF. Remarkably, the two measures exhibited a very strong linear correlation, and their relationship was well fit by a linear model with a quasi-zero intercept (0.014 ± 0.120). The slope however was not unity (0.58 ± 0.17), indicative of a scaling effect between the two modalities. One explanation for this scaling could be non-linear shrinkage caused by chemical fixation in the EM images: the extra-axonal space shrank more than the intra-axonal space. Another explanation, this time affecting AWFWMTI, could be a higher transverse relaxation rate in the intra- vs. the extra-axonal space7 which would introduce a non-negligible $$$e^{-TE\cdot\Delta R_{2}}$$$ weighting.


The correlations between EM and WMTI metrics support the specificity associated with AWF and De,┴ as expected from the model. The WMTI model parameters can be derived from data acquired in under ten minutes on clinical scanners, and are therefore promising candidates as sensitive and specific biomarkers for routine monitoring of microstructural changes in demyelinating pathologies.


This work was supported by the NIH grant R21 NS081230.


1. Fieremans et al., Neuroimage 2011, 58(1):177-88.

2. Matsushima and Morell, Brain Pathol 2001, 11(1):107-16.

3. Falangola et al., NMR in Biomed 2014, 27(8):948-57.

4. Jelescu et al., Proc of the ISMRM 2015, p. 3000.

5. Guglielmetti et al., Neuroimage 2015, 125:363-377.

6. Novikov and Fieremans, Proc of the ISMRM 2012, p. 1829.

7. Beaulieu et al., Magn Reson Imaging 1998, 16(10):1201-10.


The WMTI model for compartment specific water diffusivities. The compartments are color-coded, and axial (A) and transverse (B) views of the white matter bundle illustrate axial intra-axonal (Da), axial extra-axonal (De,||) and radial extra-axonal (De,┴) diffusivities. Panel C illustrates a case of patchy demyelination, typical of cuprizone intoxication.

Experimental timeline for number of in vivo MR datasets (black) and ex vivo EM (red) at each timepoint and in each group. CPZ = cuprizone.

Differences in diet arms from EM images. A: Representative EM images from each diet arm at the 12-week timepoint. B: Mean (± standard deviation) of EM measures, per group and timepoint. The measures follow the expected trend with intoxication and recovery.

Mean (± standard deviation) MR measures in each group, at each timepoint. Barplots highlight the differences between diet arms at a given timepoint - when significant (based on ANCOVA), the differences are marked with asterisks. *: p < 0.05; **: p < 0.01; ***: p < 0.001.

A: Spearman partial correlations (coefficient ρ and p-value) between EM and MR outcomes, covarying for mouse weight at baseline. Significant correlations are highlighted in green (p < 0.05). B: Representative scatter plots. A bold frame indicates a significant correlation. The correlations are essentially driven by the differences between diet groups.

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