Fod2fixel still not working

Hi. I’m following the ‘Fibre density and cross section -Multi tissue CSD’ tutorial. I’ve finished running through the entire pipeline just to discover that almost all the files are mostly empty (they do have 1-10 random voxels with values). The only exceptions are the index.mif files that at all stages of the pipeline show the brain. When I try to find where the error is I found out that the first file that is empty is the fd.mif file that is generated in step 14 for each subject (in the directory fixel_in_template_space_NOT_REORIENTED). I also did noticed that the files subject2template_warp.mif and template2subject_warp.mif seem funky in mrview (see attached image - maybe this could be the problem).

Any ideas on how to proceed?

1 Like

Hi @polivaoren ,
Would you please load wmfod.mif in mrview & under tools, you will have fixel plot, maybe you can drag your fd.mif there. I hope that’s not empty.

You may also check mrstats fd.mif. I hope you get some values as outcome.

Also if you like to see fd.mif in mrview, you can use fixel2voxel fd.mif mean fd_mean.mif. I hope it’s not empty.


Great. I was using overlay. When I used fixel plot instead, I can see the data. New question: which of the many files am I expected to look at for a group comparison (zstat, fwe, tstat etc)? I can’t find any location in the documentation that elaborate on that.

Also, how do I limit the analysis to a ROI. Do I need to modify the wmfod_template.mif (and run the analysis almost from the beginning)?

These are basic statistical values for the GLM contrast you perform, at each fixel:
Zstat = Z value (population statistic)
Tstat = T value (sample statistic)
fwe = 1 - family-wise error corrected p-value

You can provide a mask to limit the analysis to the ROI, using the -mask flag

Is the zstat corrected for family wise error? Can we use it for thresholding (I noticed it automatically appears in the thresholding section of the fixel tool)? or do I use the cfe file for thresholding (and if yes, how do I do the thresholding given the high values)?

Hello again,

Have answered in the other post - but just to be clear, the Z value is just an effect size, it does not say anything about the likelihood of finding that effect size if the null hypothesis is true.