Consistently getting wrong registration in the same dataset

[lalalavi]

What happened?

Hello!

I am writing a test for HALFpipe that runs the same kind of preprocessing for participants coming from different datasets, and the registration is consistently wrong (although the process succeeds without any errors) for one of the datasets. The dataset in question is the ADHD200 dataset (available at https://datasets.datalad.org/adhd200/RawDataBIDS/NeuroIMAGE/ ). This dataset was collected with a 1.5T scanner, but I don't think that is the problem, since I managed to preprocess data from another a 1.5T dataset without issues.

I tried to run different subjects from the ADHD200 dataset, and I also tried doing fslreorient2std before starting preprocessing in fmriprep, but this did nothing to the results. All of the subjects show a weird EPI to structural registration. As an example, these is how the results for sub-3190461 look like:

We re-ran the registrations with FSL BBR on the same subject and it worked fine.

Do you have a hunch on what might be going wrong or what would be useful to try? The script works fine for the other datasets.

What command did you use?

singularity exec --cleanenv --contain --fakeroot \
    --overlay /home/${USER}/overlay.img \
    --bind "${src_directory}":/halfpipe \
    --bind "${output_directory}":/tmp/pytest-of-root \
    halfpipe-latest.sif bash -c ' \
        pytest -tests/workflows/test_consistency.py --full-trace'


I also tried to run it through a bind to a HALFpipe docker image with the same pytest command.

https://github.com/HALFpipe/HALFpipe/blob/cd877a5141410d6696d040979a172fb2c5a0e5a8/tests/workflows/test_consistency.py

What version of fMRIPrep are you running?

20.2.7

How are you running fMRIPrep?

Singularity

Is your data BIDS valid?

Yes

Are you reusing any previously computed results?

No

Please copy and paste any relevant log output.

No response

Additional information / screenshots

No response