Best practice for using freesurfer parcellations as subcortical ROIs


I’m planning to use subject-specific freesurfer parcellations, run as part of fmriprep, as ROIs for a task-based analysis, but I think I’m missing something in creating these ROI masks.

I’ve been extracting the masks from the freesurfer output using mri_binarize, e.g. for left amygdala:

mri_binarize --i ${freesurfer_dir}/${subj}/mri/aparc.a2009s+aseg.mgz \
          --match 18 \
          --o ${masks_dir}/${subj}_left-amygdala_raw.nii.gz

Then, I had then been resampling to BOLD (MNI) space as follows:

3dresample -master ${preprocessed_dir}/${subj}/func/${subj}_${ses}_${task}_bold.nii.gz \
          -input ${masks_dir}/${subj}_left-amygdala_raw.nii.gz\
          -prefix ${masks_dir}/${subj}_left-amygdala_resampled.nii.gz

Before using 3dmaskave to extract the beta-weights.

However, I think I may be missing something as when I try to view these masks they are not in the correct location in MNI space.

So, my questions are:

  1. As far as I can tell, the freesurfer parcellations are in T1 space, is that correct?
  2. If so, then is there a way to re-run fMRIPrep or freesurfer so that they’re output in MNI space, or do I need to extract as above but transform the into MNI space using the transformation calculated by fmriprep (instead of the 3dresample command above)?
  3. Finally, what is the best way to do this? I’ve seen various methods, but mri_label2vol seems most common. Is that what would be recommended?

Hopefully that all makes sense, please let me know if any clarification needed!