Toward interrogating relationships between grey and white matter measures using Fixel Track-
Weighted Imaging and Fixel-Based Analysis
Robert E Smith , David Raelt , David N Vaughan , Fernando Calamante , and Alan Connelly
The Florey Institute of Neuroscience and Mental Health, Heidelberg, Australia, Department of Neurology, Austin Health, Melbourne, Australia, Department of
Medicine (AH/NH), The University of Melbourne, Australia
Neuroimaging studies assessing white and grey matter are most typically performed as independent analyses. The relationships between white and grey
matter abnormalities are therefore poorly understood. We present a novel framework for interrogating relationships between quantitative measures
derived from grey matter analysis, and diusion MRI-based, bre-specic white matter measures.
Quantitative measures estimated from neuroimaging typically focus on either grey or white matter. As such, these measures are analysed using
techniques specically tailored for the biological tissue of interest. This however limits the extent to which grey and white matter measures may be
analysed in combination, despite their intrinsic relationship via neuronal connectivity. Here we propose a framework for combining such measures into a
common space, by using updates to the Track-Weighted Imaging (TWI) framework to project grey matter measures into the white matter based on
streamlines tractography, thereby working toward statistical analysis of these data in conjunction with quantitative and bre-specic measures /
methods designed specically for white matter analysis in the presence of crossing bres .
The proposed framework operates as follows (Figure 1):
1. White matter analysis:
1.1. Generate whole-brain tractogram in native subject space (rather than using warped images in template space ).
1.2. Perform image registration to a common template space, ideally based on a higher-order diusion model .
1.3. Dene the target xels (specic bre population elements within voxels) for this analysis in template space.
2. Grey matter analysis:
2.1. For each streamline in the tractogram, sample the grey matter parameter(s) of interest at the two streamline endpoints; either using a native surface
mesh representation, simply sampling at the streamline endpoints e.g. , or 'extending' the streamline endpoints to sample from grey matter e.g. .
2.2. Typically take some statistic of the two endpoint values (e.g. mean) as the 'TWI factor' for that streamline.
2.3. Transform subject tractogram to template space based on the non-linear displacement eld derived during symmetric registration. Unlike FOD-
based spatial normalisation, where modulation of bre density based on bre orientation must be performed explicitly , this approach implicitly
modulates streamlines density appropriately depending on their orientation relative to eld deformations (Figure 2) .
2.4. Perform Fixel Track-Weighted Imaging (TWI), with per-streamline factors dened in step 2.2, and xels dened in group average template space in
step 1.3 acting as the targets for streamlines mapping (common across all subjects).
Data for 28 healthy controls were acquired on a Siemens 3T Trio system. Image data used included T1-weighted image at 0.9mm isotropic resolution,
DWI (b=3000 s.mm , 60 directions, 2.5mm isotropic), and a pair of b=0 images with reversed phase encoding for eld inhomogeneity estimation .
Subject-specic reconstruction included: Cortical thickness estimation using FreeSurfer ; Fibre Orientation Distributions estimated using Multi-Tissue
Constrained Spherical Deconvolution ; tractogram reconstruction using the iFOD2 probabilistic streamlines algorithm and the Anatomically-
Constrained Tractography framework ; quantitative streamline weights derived using the SIFT2 algorithm .
A population-specic Fibre Orientation Distribution (FOD) template image was constructed using previously described methods .
For demonstration purposes, for each streamline, the mean cortical thickness at the streamline endpoints (using only one endpoint if the other did not
terminate at the cortex) was used as the TWI factor for that streamline. During mapping, the mean TWI factor for streamlines traversing a particular xel in
template space was taken as the nal value for that xel.
The white matter quantitative value of interest for this demonstration was the combined measure of Fibre Density and
Cross-sectional area (FDC) throughout the template xel mask.
Figure 3 demonstrates how the proposed framework provides both white and grey matter quantitative measures in a common space: that of xels in a
group average WM template. An example zoomed view of this information is presented in Figure 4.
By drawing grey matter quantitative information directly into the space in which white matter quantitative measures are analysed, the measures become
directly comparable in this higher-dimensional space, without the loss of complex white matter bre crossing information. Although this process may
result in grey matter information that is 'smooth' along the relevant white matter pathway, such smoothness is a preferable trait, and in fact smoothing is
typically applied explicitly .
1 1 1,2 1,3 1,3
1 2 3
Future developments to the Fixel-Based Analysis (FBA) framework will enable direct multivariate statistical analysis of data such as that shown.
Connectivity-based Fixel Enhancement (CFE) will be used for hypothesis testing in template space, incorporating both grey and white matter
quantitative measures. Combined with the exibility of the Track-Weighted Imaging (TWI) framework (particularly mapping to target xels as presented
here), and the General Linear Model (GLM) with permutation testing , this framework will enable hypothesis formation and testing where (any)
quantitative measures from both white and grey matter are of interest.
Examining relationships between such measures has the potential for improved sensitivity to
neurological diseases, compared to assessing the grey or white matter alone.
We are grateful to the National Health and Medical Research Council (NHMRC) (400121) of Australia and the Victorian Government’s Operational
Infrastructure Support Program for their support.
1. Calamante, F.; Tournier, J.-D.; Smith, R. E. & Connelly, A. A generalised framework for super-resolution track-
weighted imaging. NeuroImage, 2012, 59, 2494-2503
2. Raelt, D. A.; Tournier, J.-D.; Smith, R. E.; Vaughan, D. N.; Jackson, G.; Ridgway, G. R. & Connelly, A. Investigating white matter bre density and
morphology using xel-based analysis. NeuroImage, 2016, j.neuroimage.2016.09.029
3. Christiaens, D.; Dhollander, T.; Maes, F.; Sunaert, S. & Suetens, P. The Eect of Reorientation of the Fibre Orientation Distribution on Fibre Tracking.
Computational Diusion MRI, 2012
4. Raelt, D.; Tournier, J.-D.; Fripp, J.; Crozier, S.; Connelly, A. & Salvado, O. Symmetric dieomorphic registration of bre orientation distributions.
NeuroImage, 2011, 56, 1171-1180
5. Raelt, D. A.; Smith, R. E.; Ridgway, G. R.; Tournier, J.-D.; Vaughan, D. N.; Rose, S.; Henderson, R. & Connelly, A. Connectivity-based xel enhancement:
Whole-brain statistical analysis of diusion MRI measures in the presence of crossing bres. NeuroImage, 2015, 117, 40-55
6. Calamante, F.; Smith, R.E.; Liang, X.; Zalesky, A.; Connelly, A. Track-weighted dynamic functional connectivity (TWdFC): a new method to study dynamic
connectivity. In Proc ISMRM 2016:0308
7. Tozer, D.; Chard, D.; Bodini, B.; Ciccarelli, O.; Miller, D.; Thompson, A. & Wheeler-Kingshott, C. Linking white matter tracts to associated cortical grey
matter: A tract extension methodology. NeuroImage, 2012, 59, 3094-3102
8. Andersson, J. L.; Skare, S. & Ashburner, J. How to correct susceptibility distortions in spin-echo echo-planar images: application to diusion tensor
imaging. NeuroImage, 2003, 20, 870-888
9. Dale, A. M.; Fischl, B. & Sereno, M. I. Cortical Surface-Based Analysis: I. Segmentation and Surface Reconstruction. NeuroImage, 1999, 9, 179-194
10. Dhollander, T.; Connelly, A. A Novel Iterative Approach to Reap the Benets of Multi-Tissue CSD from Just Single-Shell (+b=0) Diusion MRI Data. In
Proc ISMRM, 2016, 3010
11. Tournier, J.-D.; Calamante, F. & Connelly, A. Improved probabilistic streamlines tractography by 2nd order integration over bre orientation
distributions. In proc ISMRM, 2010, 1670
12. Smith, R. E.; Tournier, J.-D.; Calamante, F. & Connelly, A. Anatomically-constrained tractography: Improved diusion MRI streamlines tractography
through eective use of anatomical information. NeuroImage, 2012, 62, 1924-1938
13. Smith, R. E.; Tournier, J.-D.; Calamante, F. & Connelly, A. SIFT2: Enabling dense quantitative assessment of brain white matter connectivity using
streamlines tractography. NeuroImage, 2015, 119, 338-351
14. Winkler, A. M.; Ridgway, G. R.; Webster, M. A.; Smith, S. M. & Nichols, T. E. Permutation inference for the general linear model. NeuroImage, 2014, 92,
Schematic of Fixel TWI operation, including: Registration of subject Fibre Orientation Distribution (FOD) image to template; generation of streamlines
tractogram in subject space; extraction of streamline TWI factors based on grey matter information in subject space; transformation of subject
tractogram to template space; direct mapping of transformed streamlines to xels in template space. The result is an image containing a quantitative
value per xel in template space, which is therefore amenable to comparison with other images dened in the same space.
Demonstration of the benets of applying spatial non-linear transformations directly to streamlines data. In each tracking example (left column), if the
eld compression / expansion is orthogonal to the streamlines direction, then the voxel-wise streamlines density is intrinsically modulated appropriately;
however, if this compression / expansion is along the same direction as the streamlines, then the streamlines lengths are altered, but the voxel-wise
streamlines density is unaected.
Fixel data from the rst four subjects in the group, displayed in template space; axial slice. Fibre Density and Cross-section (FDC; top row) is a white
matter-based measure, whereas the xel Track-Weighted Imaging (TWI) parameter derived here (bottom row) is a grey matter-based measure; but
because the grey matter information is propagated into white matter based on streamlines tractography, both sources of information are available
within the same template space.
Zoomed region of template image showing the data contributed by a single subject. For each individual xel in template space, the subject contributes
one value corresponding to the Fibre Density and Cross-section (FDC) measure, and one value derived from xel Track-Weighted Imaging (TWI), based on
cortical thickness at streamlines endpoints in this particular case. Note that within any particular voxel in template space,not only may there be multiple
xels, but the values of the quantitative measures of interest may vary between those xels.
Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)