Hi,
Interesting report. I have the feeling that I read already several reports showing worse results when using the latest versions of FMRIPREP (v21 and v22) compared to earlier version (v20).
For example:
- Buggy SDC correction with Syn-SDC – version 21.0.2
- The SDC of fMRIPrep v21-22 seems not robust!
- Incorrect/dark functional images after SDC with SyN · Issue #2777 · nipreps/fmriprep · GitHub
- SDC extreme corrections in fMRIprep 22.0.1 vs 20.0.3
- fMRIPREP output - distortion correction and BOLD mask - #5 by abudzinska
- Unusual susceptibility distortion correction for fmriprep v21.0.2?
- Preprocessing failing to finish with 22.0.0
- Inaccurate brain mask when enabling --use-syn-sdc
But I must say that it is not what I observed with our datasets where we use pepolar SDC.
Did you acquire fieldmaps for your dataset? If yes, Why do you chose to use SyN?
In your case I can make 2 suggestions:
- removing
--fs-no-reconallmay improve the bold-to-T1w registration - the combination
--ignore fieldmaps --use-syn-sdcis not functional with fmriprep v22.0.0 but was fixed in v22.0.2. See here.
Let us know if you get better results! Thanks!