Make mask from FBA results

Hi all,

I have recently done FBA analysis, and after running fixelcfestats on my sample I got a region with significant differences between groups (pvalue_corrected < 0.05) that I visualize with the corresponding track scalar file (tsf). I would like to keep studying this area in each subject (for example to calculate the mean FA) but I can’t perform a mask from the thresholded tsf to run mrstats FA.mif -mask.

These are the commands I ran:

fixelcfestats fd file_fd.txt design_matrix.txt contrast_vec.txt tracks_SIFT.tck stats_fd
fixel2tsf stats_fd/fwe_pvalue.mif tracks_SIFT.tck stats_fd/fwe_pvalue.tsf
tsfthreshold stats_fd/fwe_pvalue.tsf 0.95 stats_fd/fwe_pvalue_mask.tsf

I understand I can’t run mrstats because the fwe_pvalue_mask.tsf file ins’t an image.
Do you have any idea to get the mask of this region?
Or another idea to calculate some metrics (FA, MD, etc) within this area for each subject?

Thanks,
Cheers.-
Stella

Good question… I don’t think we have a way to do this in MRtrix3, as far as I can tell – maybe @rsmith or @thijsdhollander may correct me here. The natural place for such a functionality would be in tckmap, but this hasn’t been implemented.

But your question would probably be better addressed by mapping the significant fixels to voxels directly. The TSF approach is simply a way of generating a more visually appealing representation of the results. In your case, I’d say you could just use fixel2voxel from your stats_fd/fwe_pvalue.mif directly to get the max FWE p-value per voxel, then mrthreshold to get a mask of significant voxels. Would that give you what you’re after…?

1 Like

Hi Donald,

Thanks for your response. Good idea, I think it could work!

Yep, fully agreed; this is the most direct and easiest way to go at it, as far as I know. We’re starting to use this more often in our and collaborators’ studies as well; but we prefer to stick to “fixel ROIs”, rather than mapping to voxels. Using voxel2fixel combined with sub-regions (either drawn or from another source), and intersecting the result with the significant fixels (e.g. mapping both to binary fixel ROIs and multiplying), it’s also possible to select fixel “sub-ROIs” from an FBA result; or to parcellate them, etc… Or using tractography to define bundles, map them to fixels via tck2fixel, threshold, and again potentially intersect with FBA results.

The “general” trick is always the same: try to get your ROIs (from FBA, tractography, atlases, drawing, etc…) first all to your (template) fixels, and then benefit of the fixel format to perform logical combinations of the fixel ROIs.