Fixel-based Analysis - Feasibilitiy


I would like to conduct a fixel-based analysis on neonatal images acquired with 69 directions and b=700, voxel size 1.0 x 1.0 x 2.0 mm.

Do you think it would be possible with these parameters or is the b value too low?

If it is possible, am I correct in thinking the single shell pipeline would be the best to follow?


Welcome Eleanor!

The b-value is indeed quite low. I don’t have the breadth of experience with different data to know how it will go, but certainly there’s been plenty of FBAs performed at b=1000. I’m also not familiar with the extent to which crossing fibres do or do not manifest at these kinds of ages.

So the two ways in which the limitations of such data arise are:

  1. The estimated FODs, both at a subject-specific level and in the template, may not properly resolve any crossing fibre structures present. There’s a few potential ways that this could influence the analysis. E.g. You may find that by the time you average the FODs in template space, you don’t really have crossing fibre estimates that are faithful to the underlying structure, and so you end up with essentially voxel-aggregate measures, ill-defined fibre orientations, and fixel-fixel connectivity that is not terribly specific because the streamlines in template space kind of go everywhere.

  2. Over and above the issue of being sensitive to any particular effect of interest, even if you were to discover statistical significance of some effect of interest, there is a further issue of specificity in interpreting that result. While you would be able to say in such a circumstance that the diffusion-weighted signal has some variance in concert with your effect of interest, it would be very difficult to label such as genuinely reflecting “fibre density” as is purported to be measured, as the conditions under which that interpretation is valid are not met. But you would still have spatial specificity as to where such differences arise. Note that this applies specifically to the FD metric (and therefore FDC downstream); FC is not sensitive to b-value, as long as the respective fibre orientations can be resolved.

Not having the experience with such data that other members have, I can really only suggest making the attempt, being mindful that given the difference between your data and the kinds of cohorts used to inform that documentation, the outcomes of the constituent pipeline steps will need to be assessed critically.

As you only have one non-zero b-value, you are correct in that the multi-tissue pipeline is not applicable to your data and you should follow the single-shell one.


1 Like