Bold-T1 registration with fmriprep 23.2.0

Software Support
,
1

Hello,
We are trying to re-preprocess data using a newer version of fmriprep, but we have a couple of instances in which the registration between the bold images and the anatomical image looks worse in the newer version (23.2.0) compared to the previous version that we had used (20.2.0). I’m including screenshots for the same subject and same resting-state run, preprocessed using the two versions I mention before. In the image preprocessed with fmriprep 23.2.0 the bold signal extends outside of the brain in occipital regions, and this doesn’t happen in the image preprocessed with fmriprep 20.2.0. This same pattern is appearing in multiple runs, but not all of them. Could anyone offer any insight on why this is happening, why it’s happening with the newer version of fMRIprep and not the older version we used, and what could be done to prevent the signal extending outside of the brain?

Thanks!

Screenshot 2024-03-22 at 2.17.27 PM
Screenshot 2024-03-22 at 2.17.27 PM1503×1160 138 KB

2

Hi @dianaperez25 , and welcome to Neurostars!

Could you share the fmriprep command you use for both version and also how do you use fmriprep (singularity, docker, bare-metal installation)?

3

Thank you! Happy to be here. We use fmriprep via singularity on a HPC.

This is the command for the newer version:

singularity run --cleanenv \
    -B /gpfs/research/grattonlab,$APPTAINERENV_TEMPLATEFLOW_HOME:/home/.cache/templateflow \
    /gpfs/research/grattonlab/singularity_images/fmriprep-23.2.0.simg \
    ${BIDS_DIR} \
    ${BIDS_DIR}/${DERIVS_DIR} \
    participant --participant-label ${subject} \
    --skip-bids-validation \
    -w ${WORK_DIR} \
    --fs-license-file /gpfs/research/grattonlab/singularity_images/freesurfer_license.txt \
    --output-spaces MNI152NLin6Asym:res-2  \
    --ignore slicetiming --fd-spike-threshold 0.2 \
    --notrack \

And for the older version:

singularity run --cleanenv \
    -B /gpfs/research/grattonlab,$SINGULARITYENV_TEMPLATEFLOW_HOME:/home/.cache/templateflow \
    /gpfs/research/grattonlab/singularity_images/fmriprep-20.2.0.simg \
    ${BIDS_DIR} \
    ${BIDS_DIR}/${DERIVS_DIR} \
    participant --participant-label ${subject} \
    --skip-bids-validation \
    -w ${WORK_DIR} \
    --fs-license-file /gpfs/research/grattonlab/singularity_images/freesurfer_license.txt \
    --output-spaces MNI152NLin6Asym:res-2 \
    --ignore slicetiming --fd-spike-threshold 0.2 \
    --notrack \
4

Please use the latest versions in each minor series, so 20.2.7 and 23.2.1. Without that, there’s the potential that you’re wasting time debugging a problem that’s already been fixed.

5

Thank you for those details.

Next couple of questions (after you follow the advice from @effigies :wink: ): what could affect the bold-T1 registration are the SDC correction of bold images and if SBref images were used for coregistration.

  • What kind of fieldmap do you have in your dataset and were they used for susceptibility distorsion correction (SDC)? You can see those details in the HTML output, in the summary of each bold run.
  • Do you have sbref images in your dataset? There was a change in the way they are used between both version that could affect the quality of the registration.
6

Hi, thank you so much for your response. We use a phase difference map and don’t have sbref images.