ROI registration to fmriprep output




I see that people have asked similar questions so I’m sorry to ask again–I would like to extract ROIs from the bold data after fmriprep preprocessing to then run simple GLMs and MVPA. I am interested in cortical and subcortical regions.

I have a couple of questions:

  1. Are there ROIs available in MNI152NLin2009cAsym space for ROI extraction? I understand this isn’t part of the package, but I was looking in fslview and see MNI152Lin but I’m not sure which one matches this one. Or does this exist in a different software package?
  2. What are the general principles regarding which space is best for bold analysis–fsaverage, MNI, etc.–does it depend on just if you’re using cortical vs. subcortical ROIs? Since I am interested in both, I want to keep it consistent.

Thanks so much!



FMRIPREP uses MNI152NLin2009cAsym template as coregistration target because it’s superior sharpness and tissue contrast makes coregistration easier. However, the differences between various MNI templates in terms of localization of individual points should be smaller than the spatial resolution you would expect in a typical fMRI experiment. Thus I would not worry about using a segmentation/atlas defined on one MNI template to extract timeseries in data coregistered to a different MNI template.

fsaverage is a surface only space (cortical only), MNI152NLin2009cAsym is volume (thus cortical and subcortical regions). FMRIPREP also supports CIFTI outputs (which use fsaverage for cortical regions and MNI152NLin2009cAsym for subcortical - see --cifti-output command line flag). There is currently no consensus in the field about whether surface or volume based analysis is better. Some groups argue they are comparable (when comparing ANTs and freesurfer) and some that surface is superior (when comparing FNIRT vs freesurfer).