Hello,
I've been working on taking the outputs of the fmriprep BOLD preprocessing tool (https://fmriprep.readthedocs.io/en/latest/index.html) and coercing them into the fsfast structure so that first levels can be run with mkanalysis-sess and selxavg3b-sess. It's been a bit trickier than anticipated and I was hoping to get some help verifying my current approach (attached to this email).
Question 1:
fmriprep outputs the preprocessed BOLD data in the T1 space (mni305 is not an option). However, the typical spatial normalization workflow uses a registration matrix from the bold to T1 to transform to talairach. Is there a way to do this transformation without having to specify a registration matrix given the functional data is already in the T1 space? My current approach is to use mri_convert with the --apply_transform flag pointing to the talairach.xfm registration matrix in the subject's recon folder, but I am not sure if this is a reasonable workaround.
Question 2:
fsfast normally calculates the global mean value of the raw data to be used for intensity normalization. However, with fmriprep I only have access to the preprocessed data in T1 space. Would extracting the global mean value for intensity normalization from the preprocessed BOLD in T1 space be valid?
Question 3:
Is there a way to do first levels in the native subcortical space? If so, what is the filename structure expected (maybe mni305.2mm gets replaced with the subject name as with the surface data)? I've only seen examples transforming to mni305 space, but ideally an option would be to just transfer the preprocessed BOLD in T1 space and run first levels staying in the native T1 space.
Thanks for your help!
Matt