Something wrong with susceptibility distortion correction and alignment for resting-state data

[angelw55]

Hi fmriprep experts,

I got different errors for another subject using fmriprep 1.4.0. The results of the susceptibility distortion correction and the alignment of functional and anatomical data are weird for the resting-state fMRI data. But it is okay for the task fMRI data for the same subject. Please refer to attached figures. What's wrong?

Thanks,
Angel

[effigies]

This looks like the bug in nipreps/niworkflows#388, which should be fixed in fMRIPrep 1.5.0.

[angelw55]

Thanks for your reply. Since this error just happened for the rest fMRI data and I found the results for the task fMRI data are reasonable, can I still use the preprocessed data for the task fMRI data for subsequent analysis? Can I just only rerun the fmriprep for the rest fMRI data for my case?

[effigies]

It really depends on what you're trying to compare. If your analyses are entirely separate, then re-running on just the rest scans is probably doable. You'll need to be careful to segregate the outputs, though, to make sure you're not partially overwriting your earlier derivatives when running again. And be sure to report the specific versions you've run when detailing any analysis.

[angelw55]

Thanks for your advice. Because I have already run the fmriprep 1.4.0 for 69 subjects and only found some errors for 2 subjects, I really want to keep the consistency of using the same version of fmriprep. Do you think I can fix the errors by rerunning the fmriprep 1.4.0? If no, I will try fmriprep 1.5.0 and report different versions used in my manuscript.

[effigies]

I doubt rerunning will resolve the problem. Another option is just to exclude those two subjects for the runs that show errors. The reduction in power between 69 and 67 subjects for that analysis may be acceptable.

[angelw55]

Just FYI, I also tried the fmriprep 1.5.0 on this subject, and the same errors still persist. I will also give up this subject.

[oesteban]

Does your dataset have SBrefs?