Skip preprocessing steps in fmriprep / reuse existing data?


Is it possible to skip the normalisation to a template and instead use each individuals native space? Would that work if one flagged --output-space T1w? And in that case, would the output still be motion corrected (and include the confounds.tsv)?

Also, if one already has fmriprep preprocessed data, is it possible to reuse some of the output in order to speed up the process? When I try to pass the command on existing data it throws an error, maybe I’m missing a flag I would need to include?

Thank you!


Yes, that is how to do that.

Yes. Spatial normalization is the last step that is done, so there are not separate confounds for different spaces.

Yes. If relevant, Freesurfer outputs will automatically be reused if they are in the default location (i.e. where they were made when you first ran fMRIPrep). If you reuse the same working directory as your previous runs, then a lot of steps will not have to be recalculated. You can pass in your working directory with -w argument (if you specified one in your original run, do not change it). In addition, there is the --anat-derivatives flag which could point to the subject’s fmriprep/$SUB/anat folder.

Hard to know without seeing your new command, but in theory your new command should just be the same as the old one, with specifying a different output-space.