I’m re-analysing a dataset with fmriprep that has been preprocessed before using a custom pipeline. While fmriprep seems to do an excellent job for most subjects, for some the results look quite problematic, although there are no errors reported.
For this particular subject, fMRIPrep fails for two out of four runs, outputting heavily distorted
preproc_bolds (see below). Judging by the generated report, things go south after fieldmap registration. I added screenshots from one of the problematic runs below.
I don’t have much experience with fmri-preprocessing and since there are no errors, I’m unsure what went wrong. The data has been preprocessed before with no problems.
Is this a common problem? Where should I start looking? I’m happy to share reports and files, but don’t want to post them publicly for privacy reasons.