I got what I believe are reasonable results from susceptibility distortion correction with fMRIprep version 21.0.2. Here is a snapshot of the fieldmap from the output report with v21.0.2:
I used singularity containers for both versions of fMRIprep and, other than changing the container/version, I ran the same command to process the same data each time. I was also sure to use a clean working directory with each run.
Does anyone have an idea what might be causing the different results?
In short, SDCflows (the SDC software fMRIPrep uses) has been undergoing some major revisions. You can see the change log here. Without knowing more about your fieldmaps, hard to say much more.
Thanks. We have two magnitude images (one for each echo) and a single phase difference image. It’s odd that the revisions to SDCflows would introduce what look like errors. Hopefully someone else may have encountered a similar issue and can weigh in. I’m also happy to provide any additional information that would be helpful.