Decoupling effects of fiber dispersion and microscopic anisotropy fundamentally change interpretations of DTI in multiple sclerosis
Kasper Winther Andersen1, Samo Lasič1,2, Henrik Lundell1, Markus Nilsson3, Daniel Topgaard4, Filip Szczepankiewicz5,6,7, Hartwig Roman Siebner1,8,9, Morten Blinkenberg10, and Tim B Dyrby1,11

1Copenhagen University Hospital Hvidovre, Danish Research Centre for Magnetic Resonance, Hvidovre, Denmark, 2Random Walk Imaging, AB, Lund, Sweden, 3Clinical Sciences, Lund, Department of Radiology, Lund University, Lund, Sweden, 4Division of Physical Chemistry, Department of Chemistry, Lund University, Lund, Sweden, 5Medical Radiation Physics, Lund University, Lund, Sweden, 6Harvard Medical School, Boston, MA, United States, 7Radiology, Brigham and Women’s Hospital, Boston, MA, United States, 8Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark, 9Department of Neurology, Copenhagen University Hospital Bispebjerg, Copenhagen, Denmark, 10Danish Multiple Sclerosis Center, Copenhagen University Hospital Rigshospitalet, Copenhagen, Denmark, 11DTU Compute, Technical University of Denmark, Lyngby, Denmark


Fractional anisotropy from diffusion tensor imaging (DTI-FA) has frequently been used to probe changes in white matter microstructure, but is also heavily affected by axonal fiber dispersion. μFA removes fiber dispersion effects and thereby estimates the microscopic anisotropy. Here, we found lower μFA in normal appearing white matter in multiple sclerosis patients as compared with healthy controls. In addition, μFA correlated significantly with age, disability and cognitive performance. These relations could not be established with DTI-FA. Our results indicate that μFA could be used as a powerful biomarker for diseases related to micro-structural changes in white matter as well as in studies of the healthy brain.


Changes in brain tissue microstructure are frequently probed using the fractional anisotropy derived from diffusion tensor imaging (DTI-FA). A multitude of studies have characterized how pathology affect DTI-FA in diseases like multiple sclerosis (MS)1, schizophrenia2, Alzheimer’s disease3, as well as describing normal brain development4, ageing5 and learning related plasticity6. DTI-FA is sensitive to microstructural brain alterations in white matter (WM) such as demyelination and axonal degeneration. Unfortunately, DTI-FA it is also modulated by axonal fiber dispersion and is lower in areas of crossing, kissing, and bending fibers7. Since DTI-FA entangle effects of both microstructure and fiber dispersion, it is impossible to infer effects of microstructure alone. New diffusion encoding techniques such as double diffusion encoding8–10 and isotropic diffusion encoding11–13 allow estimation of the underlying microscopic fractional anisotropy (μFA) by removing the fiber dispersion through signal representations14,15. We hypothesize that μFA better than DTI-FA relates to disease progression and cognitive performance in both healthy controls (HC) and MS patients. We compare μFA with DTI-FA in a group of MS patients as well as HC across a large age span and show the importance of decoupling fiber dispersion in microstructural anisotropy analysis.


Here we used data from 40 MS (14 primary progressive (PPMS), 26 relapsing remitting (RRMS)) patients and 27 HC. All subjects gave written consent. Cognitive performance was evaluated with Symbol Digit Modalities Test (SDMT) and disability in MS was measured with Expanded Disability Status Scale (EDSS). MR images were recorded on a Philips Achieva 3T scanner using a 32-channel head-coil. Structural scanning included T1- and T2-weighted and FLAIR, which were used to manually delineate lesions. Diffusion weighted imaging was recorded with 50 different diffusion encoding directions at b=1000s/mm2 and voxelwise DTI-FA maps were derived using FSL. µFA maps were estimated using qMAS acquisition and diffusional variance decomposition (DIVIDE)15,16. The WM was segmented using FreeSurfer, and MS lesions were masked from the WM ROIs. DTI-FA and μFA maps were co-registered to the structural T1w image and parameters were averaged over HC WM and MS normal appearing WM. Group-differences were evaluated using ‘hemisphere’ (left, right) by ‘group’ (HC, RRMS, PPMS) repeated measures ANOVA (rmANOVA) with age as covariate. We further correlated μFA and DTI-FA with age, SDMT and EDSS.


Group differences

We found remarkably different results when comparing groups with μFA and DTI-FA in the NAWM (Fig1). μFA showed clear group differences (rmANOVA, F(2,63)=9.0, p<0.001), with HC having higher μFA compared with both RRMS and PPMS patients. There were no main effect or interactions of hemisphere. With the DTI-FA, however, there were no significant main effect of group (F(2,63)=2.25, p=0.11), hemisphere (F(1,64)=0.001, p=0.98), nor group by hemisphere interaction (F(2,64)=0.06, p=0.98). So, while the μFA is able to differentiate groups, DTI-FA fails to reveal group differences in the NAWM.


In both the HC and MS groups, μFA correlated negatively with age (HC left: r=-0.41, p=0.03, HC right: r=-0.58, p=0.002; MS left: r=-0.35, p=0.03, MS right r=-0.50, p=0.001). For DTI-FA, however, this relationship with age was only significant in HC (left: r=-0.55, p=0.003, right: r=-0.41, p=0.03) but not in the MS group (left: r=-0.09, p=0.59, right: r=-0.12, p=0.47) (Fig2).


In the MS group, the μFA in right (r=-0.43, p=0.005) and left (r=-0.32, p=0.048) hemisphere both correlated negatively with EDSS, meaning that lower μFA values were associated with higher disability. This association, however, was not significant when including age as a covariate. Again, with DTI-FA, this correlation was not significant (left: r=-0.16, p=0.33; right: r=-0.22, p=0.17).


Similarly, μFA correlated with SDMT in both MS (right: r=0.60, p<0.001; left: r=0.52, p=0.001) and HC (right: r=0.47, p=0.02; left (trend) r=0.37, p=0.06). In MS, these associations were unrelated to age as the correlations were also significant when controlling for age. With DTI-FA these correlations were not significant (HC right: r=0.22, p=0.28; HC left r=0.31, p=0.12; MS right: r=0.16, p=0.31; MS left: r=0.09, p=0.68).


Interestingly, whole brain lesion load correlated with μFA in the NAWM (left r=-0.41, p=0.008, right r=-0.56, p=0.002) but not with DTI-FA (left r=0.21, p=0.19; right r=0.18, p=0.27).


Here we show that group studies of DTI-FA are heavily modulated by effects of fiber dispersion, which reduce effect sizes in both group comparisons and correlations. On the other hand, μFA, which is a measure of microscopic diffusion, showed significant group differences and correlations to clinical and cognitive scores as well as aging. Furthermore, we show that µFA in NAWM correlates with lesion load suggesting that µFA can be used as a surrogate measure of the Wallerian degeneration occurring in MS.


  • Danish Multiple Sclerosis Society (A31910 and A27996).
  • VINNMER Marie Curie Industry Outgoing grant (nr. 013-04350).
  • Random Walk Imaging hold patents related to the presented technology.


1. Werring, D. J., Clark, C. A., Barker, G. J., Thompson, A. J. & Miller, D. H. Diffusion tensor imaging of lesions and normal-appearing white matter in multiple sclerosis. Neurology 52, 1626–1626 (1999).

2. Lim, K. O. et al. Compromised white matter tract integrity in schizophrenia inferred from diffusion tensor imaging. Arch. Gen. Psychiatry 56, 367–374 (1999).

3. Yoshiura, T. et al. Diffusion tensor in posterior cingulate gyrus: Correlation with cognitive decline in Alzheimer’s disease. Neuroreport 13, 2299–2302 (2002).

4. Klingberg, T., Vaidya, C., Gabrieli, J., Moseley, M. & Hedehus, M. Myelination and organization of the frontal white matter in children - a diffusion tensor MRI study. Neuroreport 10, 2817–2821 (1999).

5. Charlton, R. A. et al. White matter damage on diffusion tensor imaging correlates with age-related cognitive decline. Neurology 66, 217–222 (2006).

6. Scholz, J., Klein, M. C., Behrens, T. E. J. & Johansen-Berg, H. Training induces changes in white-matter architecture. Nat. Neurosci. 12, 1370–1 (2009).

7. Beaulieu, C. The basis of anisotropic water diffusion in the nervous system - A technical review. NMR Biomed. 15, 435–455 (2002).

8. Jespersen, S. N., Lundell, H., Sønderby, C. K. & Dyrby, T. B. Orientationally invariant metrics of apparent compartment eccentricity from double pulsed field gradient diffusion experiments. NMR Biomed. 26, 1647–1662 (2013).

9. Yang, G., Tian, Q., Leuze, C., Wintermark, M. & McNab, J. A. Double diffusion encoding MRI for the clinic. Magn. Reson. Med. 00, (2017).

10. Shemesh, N. et al. Conventions and nomenclature for double diffusion encoding NMR and MRI. Magn. Reson. Med. 75, 82–87 (2016).

11. Lasič, S., Szczepankiewicz, F., Eriksson, S., Nilsson, M. & Topgaard, D. Microanisotropy imaging: quantification of microscopic diffusion anisotropy and orientational order parameter by diffusion MRI with magic-angle spinning of the q-vector. Front. Phys. 2, 1–14 (2014).

12. Wong, E. C., Cox, R. W. & Song, A. W. Optimized isotropic diffusion weighting. Magn. Reson. Med. 34, 139–143 (1995).

13. Mori, S. & Van Zijl, P. C. M. Diffusion Weighting by the Trace of the Diffusion Tensor within a Single Scan. Magn. Reson. Med. 33, 41–52 (1995).

14. Szczepankiewicz, F. et al. The link between diffusion MRI and tumor heterogeneity: Mapping cell eccentricity and density by diffusional variance decomposition (DIVIDE). Neuroimage 142, 522–532 (2016).

15. Szczepankiewicz, F. et al. Quantification of microscopic diffusion anisotropy disentangles effects of orientation dispersion from microstructure: Applications in healthy volunteers and in brain tumors. Neuroimage 104, 241–252 (2015).

16. Lasič, S., Szczepankiewicz, F., Eriksson, S., Nilsson, M. & Topgaard, D. Microanisotropy imaging: quantification of microscopic diffusion anisotropy and orientational order parameter by diffusion MRI with magic-angle spinning of the q-vector. Front. Phys. 2, 1–14 (2014).


Figure 1. Group means (standard error bars) in (A) FA and (B) μFA (not adjusted for age) in normal appearing white matter. No significant group differences were found with FA. However, μFA is significantly higher in healthy controls compared with both MS groups. μFA are not significantly different between the two MS groups when correcting for age (the two MS groups were not age-matched).

Figure 2. Correlation between (A) FA and (B) μFA and age, expanded disability status scale (EDSS), and symbol digit modalities test (SDMT). With FA, the only significant correlation found was between FA in HC and age. However, with μFA all correlations were significant (except between μFA left hemisphere in HC and SDMT, which was only a trend). The results show that when disentangling the fiber dispersion effects from microscopic anisotropy, we increase the power to detect relations between microstructure and clinical and cognitive variables.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)