Misalignment of brain tissue segmentation of the T1w

[angelw55]

Hi fmriprep experts,

I am using fmriprep 1.4.0 to preprocess our anatomical and functional data. I found something wrong with the brain tissue segmentation and the spatial normalization of the T1w. But I found that it is okay for the surface reconstruction done by FreeSurfer, and the alignment of functional and anatomical data. Please refer to the attached figures. How to solve this problem?

Thanks for your help in advance.

Best,
Angel

[oesteban]

Please re-run fMRIPrep without reusing the working directory.

BTW, the latest release 1.5.0 contains some improvements that might help with that very strong bias that the standard normalization plot is showing.

[angelw55]

Thanks for your advice. I will try to rerun it or use the latest version of fmriprep.

BTW, can I install several versions of fmriprep on the same computer? I have run fmriprep 1.4.0 for 60 more subjects already. I don't want to switch the version very frequently.

[effigies]

You can have multiple docker images (or multiple Python environments, if you're installing the Python packages natively), but we highly recommend using the same version of fMRIPrep for all subjects in an analysis. It's annoying to have to re-run when you find a bug that exhibits in only one subject, but part of the point of fMRIPrep is to help make preprocessing consistent. Using multiple versions runs counter to that.

[oesteban]

Yes, if your ..._desc-preproc_T1w.nii.gz images look good with a different viewer (i.e., they don't show a very strong bias), then I would keep 1.4.0 fixed.

[angelw55]

FYI, I tried the fmriprep 1.5.0, but the errors still persist. There is something wrong with the spatial normalization of the T1w. The ..._desc-preproc_T1w.nii.gz images don't look good. So can we still use the preprocessed fMRI data?

[effigies]

We're not here to tell you what you can and can't do, but if I found a report like that, I would try to figure out how it happened, fix the issue if possible, and if fixing the issue for those subjects involved upgrading fMRIPrep, I would rerun preprocessing on all subjects.

[angelw55]

Thanks again for your advice and support. I will probably give up this subject.