N4 intensity normalization introducing dark/bright spots


Is there any way to disable INU correction in fMRIPrep? I am getting dark or bright spots in some subjects in the preprocessed T1 and this is causing issues with T1w segmentation. I am feeding it two T1 images, one labeled T1w and the other is inplane. Both raw images look fine as far as I am able to tell.

I have found previous posts where others have mentioned a similar issue but I have not found any solutions.

I have tried defacing the T1s, using -force-bbr, deleting the working directory and re-running, etc. but nothing has helped.

Interestingly, we have preprocessed this same data previously using version 1.2.3 and did not run into this issue. I was looking through the version changes and cannot tell if there’s anything new since then that would explain this. I believe it is an issue with N4 INU correction.

This is the command I am using most recently:
singularity run --cleanenv $PI_HOME/fmriprep/fmriprep-20.0.6.simg $PI_HOME/FAB/BIDS_data $PI_HOME/FAB/analysis/fmriprep_update participant --participant-label %s -w $SCRATCH/work --omp-nthreads 8 --nthreads 12 --mem_mb 30000 --force-bbr --use-aroma --fs-license-file $PI_HOME/freesurfer/license.txt

Any thoughts? Thank you!