How to re-run a subset of participants without using cached data


I preprocessed a dataset and three of the participants did not complete correctly. Digging into the logs and reports, each of the three detail a gzip error regarding one of the intermediate files. I have a gut feeling that the corrupted gz files are due to unceremoniously terminating the fmriprep process during a maintenance window on our cluster. If that’s the case, they simply need to be re-run. How might I re-run fmriprep on those three participants without fmriprep detecting the cached work? Is there a flag I can add to the run? Or should I instead remove those subject folders from the fmriprep output folder then re-run fmriprep?

T. Bryan Jackson

You can use a new work directory (-w argument in your command). If you want to take extra precautions you can also rename the subject folders in the output directory you want to redo (such as BADsub-XX), then fmriprep will create a new subject folder for the subject.