Two basic questions, but better be safe than sorry:
After fmriprep, the BOLD needs brain extraction, right?
This is before BET:
(New users can only put one image, so… I can add it later, if it’s of interest)
My data were normalized to MNI 2009c (which is not what FSL uses, I understand). I wonder if you have an opinion about whether anybody is going to be shocked if I use the Harvard-Oxford atlas in FSL? I want to create masks of the hippocampus and other MTL, frontal, lateral temporal, parietal regions.
On a side note, in case it’s useful to others:
After using BET on the fmriprep BOLD data, I could view that data in FSLeyes (but not before). It’s an easy fix for me anyway (I had raised that issue on github #1067).
The only times FSL complained about the orientation of images or something of that order (at 3rd level), I suspect it was because I didn’t do #2 properly (or at all, actually).
Now, all looks really good! I’ll report later on success (hopefully), but in the meantime, I processed the data of 28 participants with a multiband BOLD sequence. The outcome looks great to me so far! Thanks a lot to everyone that made fmriprep possible!
The _preproc files are not skullstripped, but FMRIPREP also provides a brain masks (see _brainmask files). Instead of running BET you can just apply these masks using fslmaths.
As far as I know FSL uses MNI152NLin6Sym template, but it’s very comparable to the MNI152NLin2009cAsym that FMRIPREP uses. Any discrepancies between the two will most likely be smaller than the original resolution of your functional data so there is nothing to worry about.
I saved the mask and resampled to the BOLD data next (to use the mask in the analysis). I’m not sure if the order matters (e.g., I can see that it would be possible to resample the MNI152 instead), but I haven’t tested that.
I have a similar issue segmenting the output of an aggregate-level analysis into function regions. Specifically I want to multiply masks obtained from Neurosynth with my output file (from FSL’s randomise) to get discrete regions, and in order to do this want to resample the Neurosynth masks to the same voxel size as my output file. I was planning to use 3dresample (but am open to other techniques) - is there any downside to this method?