fMRIPrep --fs-no-reconall / re-running preprocessing

1

Hi there,

I am running fMRIprep on a dataset, and want to rerun preprocessing on a subject. For example, in my case to have a different output-space (e.g., T1w), or to now ignore slicetiming.
A couple questions:

  1. If my subject has already been through the fMRIprep processing stream and thus the surface reconstruction has already been created, can I use the --fs-no-reconall flag to skip the surface reconstruction? Or, does that flag interfere with other preprocessing measures that take freesurfer, and the surface, into consideration (e.g., if using output-space fsnative)? I’m looking to speed up the process for the 2nd time around.
  2. Alternatively, does fMRIprep just look to see if there is a freesurfer directory for the subject, and if so, skips creating the surface reconstruction because it already exists?
  3. If I want to create a new preprocessing stream for a subject (e.g., now ignore slicetiming) should I delete (or rename) the fmriprep/subject/tasksession directories so as to have a new dataset but keep the analyzed anatomical data (i.e., keep fmriprep/subject/anat & freesurfer/subject)?

Running fMRIprep locally using docker.

Thanks,
Elissa

2
  1. […] I use the --fs-no-reconall flag to skip the surface reconstruction? Or, does that flag interfere with other preprocessing measures that take freesurfer, and the surface, into consideration (e.g., if using output-space fsnative)? I’m looking to speed up the process for the 2nd time around.
  2. Alternatively, does fMRIprep just look to see if there is a freesurfer directory for the subject, and if so, skips creating the surface reconstruction because it already exists?

The latter. If you use the same output directory, which still has the previous run’s freesurfer/ directory in it, then we will simply use that, and not re-run FreeSurfer. If you use --fs-no-reconall, we will not use surfaces at all, such as coregistration with bbregister or surface sampling.

  1. If I want to create a new preprocessing stream for a subject (e.g., now ignore slicetiming) should I delete (or rename) the fmriprep/subject/tasksession directories so as to have a new dataset but keep the analyzed anatomical data (i.e., keep fmriprep/subject/anat & freesurfer/subject)?

Apart from Freesurfer, the output is saved every time, and not reused. It’s the working directory (passed with the -w flag) that allows reuse of portions of the fMRIPrep workflow. That is, if I re-run the exact same version with the exact same options, but delete <output dir>/fmriprep, it will just check the structure of the working directory, and re-populate the output directory.

Hope this helps.

2 Likes