Thanks, that clarifies that fMRIPrep is misreading the BIDS query. I have opened https://github.com/poldracklab/fmriprep/issues/1251. Please follow up there.
Thank you! Appreciate all your help.
Would it be a problem to use fmriprep 1.0.4 in the meantime while the bug is fixed? Or should I wait and use the most latest version?
(Just in case, below are the additional options that I’m using in fmriprep)
Hi, sorry about the delay. I looked into this. The problem is that your dataset path has an special name in it (
sourcedata) that is excluded from indexing. We can look into the best way to improve the exclusion rules, but renaming your directory would be the quickest fix.
Thanks for clarifying this. It had me hung up too. The BIDS spec led me to believe that I should have the input nifti in a directory named
sourcedata (and that file structure wasn’t an issue in earlier versions of fmriprep). Throwing a warning might help other people.
sourcedata is for files you used to generate your BIDS directory, for example, the raw DICOM files off of your scanner, or subject recordings that you turn into reaction time or response categorical variables for
events.tsv, or similar. The BIDS directory itself has no particular requirements to its path. So strictly speaking it’s not wrong to have
sourcedata in your path, but right now fMRIPrep can’t handle it.
Incidentally, you can make wording suggestions to the BIDS specification to help us make this clearer:
This change was in response to other issues, where people put their derivatives directly into
<bids_dir>/derivatives, which then led to derivatives being indexed as normal; because
<bids_dir>/sourcedata/. While these are valid directories, they’re not where files to be preprocessed should be placed, so we ignored them. Unfortunately, that uses a regular expression, which is hard to make more specific. The next release of pybids should have this feature built in, and I’ve added tests to ensure that it won’t be affected by your path.