Dear Mrtrix users/developers,
I am performing the fixel-based analysis for my dataset, but have a question for calculating the intersection mask…
On the information page the following mrmath command is mentioned with ‘min’ option:
mrmath input_all_warped_masks_multiple_inputs min output_template_mask_intersection
However, if I inspect this intersection image, it rather looks like a sum image (the ‘sum’ option creates exactly the same image by the way)…
Whereas if I use the option ‘product’ instead of ‘min’, this rather looks like the intersecting mask that is appropriate in all subjects…
Is this possible?
A second question I have regarding group comparisons of FC, FD and FDC is:
Should only FC be log-transformed to be normally distributed, or FD also?
And should the multiplication of both (FDC) be log_FC x log_FD <=> rather FC x FD <=> rather log_FC x FD ?
Thank you in advance!
With kind regards,
I just tested this scenario for you, and there’s indeed something wrong. Thanks for reporting; it’s clearly a bug!
Strangely, what does seem to work correctly is:
mrcat input_all_warped_masks_multiple_inputs - -axis 3 | mrmath - min output_template_mask_intersection -axis 3
Until the bug has been looked into and resolved, I’d personally advise to go with this one as a quick fix. The bug seems to at least relate in a way to the multiple input behaviour of
mrmath, so even though the product option seems to work correctly, I wouldn’t fully trust it for now…
For your other question, let’s see if we can grab @Dave’s attention to write a good post about this.
Only FC should be log transformed for analysis. However when computing FDC use the raw FC values (i.e. not log transformed).
I’ll be improving the documentation for fixel-based analysis of FD, FC and FDC in the near future, after we have made a switch to a new file format for storing fixels and associated data. I’ll post about it when I do.
In case you have not already seen it, Fig. 2 of our recent paper may help to explain the flow of the FBA steps in the MRtrix documentation.
Hi Thijs and Dave,
Thank you both for your fast responses!
@Dave , I read the paper indeed and very much appreciate your work!
I’m really enthusiastic about Mrtrix btw !
No worries @Charlotte. By the way, the bug you discovered has been resolved in the mean time; if you update your MRtrix installation, the “min” operator should now work correctly in the scenario you described!