Calculating average FBA metrics of specific tracts



I came across this reply by Rob, where he implies that it’s possible to “collapse” the measures (FD, FC, FDC) of a specified pathway and end up with a single measurement for the specified tract for each subject. I’m interested in doing exactly that, but I couldn’t find any more detailed information about that specific idea in the rest of the thread.

So anyway, I produced a fixel mask for a specific region of my FOD template, and ran fixelcfestats successfully using the mask. My question is, is it somehow possible to derive a single (average) measurement of FD/FC/FDC for the masked region for each subject separately? If so, how would I go around doing that?

I hope my question makes sense (I know it’s a bit convoluted ), and thanks for the help!!



Hi @Joe,

If you already have a fixel mask of the “fixel region” you’re after, you’re very close to what you want to get. The key thing to realise is that many MRtrix commands designed to work on “voxel images” can work on fixel data files just as well… as long as those data files come with (a copy of) the same index.mif file. That would be the case already for your fixel mask as well as the data files (FD, FC, FDC), as they also share that requirement of a common index.mif for the fixel stats (and those you’ve been able to run successfully).
Your fixel mask, for instance, would be a binary “fixel data file” sitting in one of your fixel folders. Lets say it’s called “fixelmask.mif”. You’d also have fixel folder with all your FD fixel data files, if you’ve been running FBA analysis. Let’s say a subject’s data file in your FD fixel folder is named “subject001.mif” or something. You can then simply run mrstats (name doesn’t suggest being able to act on fixels, but it works nonetheless!) on subject001.mif and add the -mask option, while providing the latter with fixelmask.mif. The output would give you your mean FD within that fixel mask, for that subject (and a range of other stats as well, like the median, or min/max, etc… even the fixel count in your mask).

To already answer another potential question that might be coming up in this context: if you’re going to be averaging stuff within a (fixel) region, and wondering whether to use FC or log(FC), the answer is log(FC). :wink:



For a further example of this in an actual study: see e.g. , where we used this for further comparison of MCI vs controls, with regions (“fixel masks”) actually being obtained from an FBA of AD vs controls. See for instance figures 6 and 7: even though the regions are shown as streamlines there, these are actually derived from the FBA (fixel) results, and were used also as fixel masks to extract FDC values specifically.


It worked!! Thank you so much for the help :grinning::grinning:


No worries! :blush: