MRtrix: Diffusion tractography in crossing fiber regions

Abstract

In recent years, diffusion-weighted magnetic resonance imaging has attracted considerable attention due to its unique potential to delineate the white matter pathways of the brain. However, methodologies currently available and in common use among neuroscientists and clinicians are typically based on the diffusion tensor model, which has comprehensively been shown to be inadequate to characterize diffusion in brain white matter. This is due to the fact that it is only capable of resolving a single fiber orientation per voxel, causing incorrect fiber orientations, and hence pathways, to be estimated through these voxels. Given that the proportion of affected voxels has been recently estimated at 90%, this is a serious limitation. Furthermore, most implementations use simple ''deterministic'' streamlines tracking algorithms, which have now been superseded by ''probabilistic'' approaches. In this study, we present a robust set of tools to perform tractography, using fiber orientations estimated using the validated constrained spherical deconvolution method, coupled with a probabilistic streamlines tracking algorithm. This methodology is shown to provide superior delineations of a number of known white matter tracts, in a manner robust to crossing fiber effects. These tools have been compiled into a software package, called MRtrix, which has been made freely available for use by the scientific community.

Beyond fractional anisotropy: Extraction of bundle-specific structural metrics from crossing fiber models

“…Finally, we computed the diffusion tensor, the three eigenvectors, and the FA value for each voxel. On this dataset we performed an eighth-order SH approximation of the fODF using MRtrix (Tournier et al, 2012; http://www.brain.org.au/software/mrtrix/). After applying CSD to the signal attenuation we normalized the result to the number of fibers in the deconvolution kernel.…”

Individual differences in structural and functional connectivity predict speed of emotion discrimination

“…Unidirectional probabilistic streamlines between amygdala seed masks and fusiform gyrus inclusion masks were computed for each hemisphere using tckgen, the FODs derived from CSD, an exclusion mask derived from each individual's CSF, and the iFOD2 algorithm with a step size of 1.15 and a cut-off of 0.15 (Tournier et al, 2010(Tournier et al, , 2012. After visual inspection, streamlines were converted to track-density images with tckmap, thresholded using mrthreshold, binarized with fslmaths, and used as masks to extract the mean values for each individual from the FA images using fslstats (Calamante et al, 2010).…”

Mapping somatosensory connectivity in adult mice using diffusion MRI tractography and super-resolution track density imaging

“…org.au/software/) (Tournier et al, 2012), as well as in-house modifications of this software package. A probabilistic streamlines method (SD_PROB) was employed in combination with the constrained spherical deconvolution (CSD) technique to model multiple fibre orientations (Tournier et al, 2007).…”