Tour
Back to Table of Contents
Tractography has become a valuable tool for neuroscientists because it is an in vivo way to examine white matter connectivity in the human brain that is individualized to the particular anatomy of each subject. This avoids some of the difficulties of voxelwise approaches, and allows you, for example, to compare the properties of analogous white matter tracts across individuals even in the presence of large baseline differences in brain structure.
In a standard approach (below), the initial set of all the streamline “fibers” in the brain is pruned down through the use of multiple inclusion masks (orange) to result in a virtual dissection of the white matter tract of interest (here, the left corticospinal tract). Color encodes directionality, with the maxim “XYZ = RGB” explaining that fibers running left-right will be colored red, anterior-posterior will be colored green, and inferior-superior will be colored blue.
Once these virtual dissections have been performed on all of the subjects, the standard protocol is to then compare some measure of their quality between groups. For instance, we might investigate whether the mean FA – a metric related to myelination – is lower in a group of patients as compared to typical controls. A presentation of these results would look something like this:
While the above “tract-averaged” approach has led to many interesting results in the literature, it doesn’t capture the large degree of variation in FA that exists along white matter tracts. This variation is easy to see if we overlay FA onto the tract group instead of directionality:
Using this corticospinal tract dissection as an example, the within-tract variability in FA might be something like ±0.14 (SD). This far exceeds both the between-subject variability in tract-averaged FA (something like ±0.03), as well as the typical tract-averaged effect sizes that are reported in the literature (effects of 0.03-0.05 are typical). Further, we can see that the greatest component of the within-tract variance exists along the longitudinal axis of the curving tract spine. This suggests that developing ways to analyze DTI metrics along tracts will allow us to:
- Improve detail in brain mapping studies
- Increase power to detect statistical effects (both in general, since the model residuals will be decreased in tracts showing this type of along-tract detail, but, in particular, the type of focal effects that could average themselves out in a tract-averaged analysis).