What to do when fmriprep goes wrong?

So I’m new to fmriprep, and it seems like an incredibly useful tool, but something appears to have gone wrong in the ANTS brain extraction. If I were running each of the composite tools by hand, I could tweak the thresholds for the brain extraction until it looked right, but how do you do something like this with fmriprep?

Report view below:

fmriprep commandline:
fmriprep --participant_label pixar001 --output-space template --no-freesurfer -w work BIDS test_fmriprep participant

using singularity 2.3 with
singularity shell docker://poldracklab/fmriprep:latest


1 Like

Wow - that is some impressive wrap around! Not sure which ants brain extraction option could fix this, but if you could share this nifti file we can see what can be done. Have you tried the admin skull stripping option? --no-skull-strip-ants

PS Always use a version tag when specifying Docker container to make sure you run the same version across all participants

1 Like

Thanks, Chris. It is impressive wrap-around!

I haven’t tried the AFNI skull stripping yet, because I was hoping to use ANTS for everyone and to keep a consistent protocol, but I can certainly try. FSL’s BET does a good job, if you tweak the threshold from the default.

The nifti file is here, if you’d like to take a look at it!

Hi @toddt, what do you mean by “FSL’s BET does a good job”? Does it wrap the mask of the frontal lobe through the border of the Y axis?

Even if we managed to fix the skull-stripping (whatever “fix” means here as regards including the cropped out of the frontal lobe in the posterior of the image), I’m afraid that all downstream steps in fmriprep will fail whenever the T1w is involved.

For this case, I would try to fix the wraparound itself. And more importantly, I’d try to identify all the participants with that wraparound just in case we are acquiring many of our images with it.