Advice on how to compare fiber density

Hi MRtrix experts,
I’m trying to compare wm tracts connecting 2 ROI’s before and after an ablative neurosurgical procedure. I’ve been spending a lot of time with the documentation on the website, but I’m still a bit murky on the best way to do this. I’ve followed the instructions on the website up until :
cd …/template

tckgen -angle 22.5 -maxlen 250 -minlen 10 -power 1.0 wmfod_template.mif -seed_image
template_mask.mif -mask template_mask.mif -select 20000000 -cutoff 0.10 tracks_20_million.tck

tcksift tracks_20_million.tck wmfod_template.mif tracks_2_million_sift.tck -term_number 2000000

fixelcfestats fd files.txt design_matrix.txt contrast_matrix.txt tracks_2_million_sift.tck stats_fd
fixelcfestats log_fc files.txt design_matrix.txt contrast_matrix.txt tracks_2_million_sift.tck stats_log_fc
fixelcfestats fdc files.txt design_matrix.txt contrast_matrix.txt tracks_2_million_sift.tck stats_fdc

I guess I’ve gotten myself confused, because the .tck files I’ve produced are at the template level, but I want to compare for each subject how the amount of structural connectivity between 2 regions has changed from before to after. Any advice would be greatly appreciated! (I have Patient1_Pre Patient 1_Post Patient 2_Pre Patient 2_Post etc)

Hi @BennyD,

This way you pose the question really gets at the heart of the bifurcation of how one goes about “quantifying structural connectivity”.

The documented Fixel-Based Analysis pipeline provides all of the requisite steps for performing statistical inference of quantitative parameters at the granularity of individual fixels. However you instead express the desire to quantify:

… the amount of structural connectivity between 2 regions …

This is actually a considerably different definition. This latter definition I have a manuscript I’ve been threatening to publish for the last five years; I promise we’ll get there eventually (pinging @alan-connelly…)

Once you are operating on the premise of quantifying connectivity of a specific pathway of interest, there are then two different ways of proceeding:

  1. Obtain quantitative estimates of connection density within the space of each individual using quantitative tractography techniques e.g. ACT / SIFT.

  2. Define a fixel mask in template space corresponding to the pathway of interest; within that mask, for any particular quantitative measure of interest (e.g. FD, FC, FDC), calculate the mean value within that mask for each subject.

In either case, you obtain a single scalar measure for that pathway per subject, to which any statistical analysis can then be performed. Option 2 is probably the easier of the two, and would require considerably less computation, especially given you already have spatial & fixel correspondence across all of your data.

Hope that irons out the confusion!


Dear @rsmith,

Thank you very much for the explanation, but how can I Define a fixel mask space? I tried to do it but I didn’t get it ? Could you clarify a little bit more,

Your hel is well appreciated.


Hi Abir,

A “fixel mask” is simply a fixel data file (see format here) for which the values contained within are binary, i.e. 1 or 0. There are multiple ways in which such data could be derived; I answered a relevant question today here.