SDC (PE blips) leaves ventral pfc unwarped

Summary of what happened:

SDC correction via varying phase-encoding blips.
Ventral pfc is not corrected, remains stretched. “Corrected” image ends up looking worse.
Only a subset of participants affected; the artifact is consistent across them. Curious if others have encountered similar. There is no immediate pattern present in the fieldmaps.

Command used (and if a helper script was used, a link to the helper script or the command generated):

singularity run --cleanenv fmriprep_21.0.1.sif fmriprep /home/hamilton/bids/tmp/ /home/hamilton/bids/derivatives/fmriprep participant --participant-label obs01 --fs-license-file /home/hamilton/bids/code/license.txt -w /home/hamilton/bids/derivatives/fmriprep_work -vvv --mem 16000 --n_cpus 12 --nthreads 24 --output-spaces anat MNI152NLin2009cAsym --notrack --resource-monitor


21.0.1 (data is few years old. similar issue crops up with a dataset processed with 23.0.2)

Environment (Docker, Singularity / Apptainer, custom installation):


Screenshots / relevant information:

Difficult to understand what is going on here. Fmriprep is using the PEPOLAR method to correct those images, right? Do the images used for SDC correction look correct, with similar readout times as the BOLD EPI images, similar shim setting and well aligned with bold images? The images for PEPOLAR correction are Spin Echo EPI?

What are your acquisition settings? Each spacing, total readout time, parallel acceleration?

On the Fieldmap, the two red blob visible in sagittal slices: x=24, x=10, x=-3 and x=-17, look strange rather unusual to me?

For the above subject, I do not have shim setting info. Readout time for the spin-echo EPI pair was 0.046 while BOLD was 0.014. Images are well-aligned.

For another subject, one whose BOLD improved w/ SDC, readout time for the spin-echo EPI pair was 0.046 while BOLD was 0.024. Images also well-aligned.

Thank you for this information. Normally for SDC with topup (as it is the case with fmriprep >=21) , different readout times between SE EPI pairs and BOLD images should not be a problem.
But to have different Total Readout Times for SE EPI pairs and the bold images is not standard and that may be the origin of what you are seeing?

I would check with the newest fmriprep version (24.0.0) as SDC was improved along the way.

1 Like