A few questions about the workflows of fMRIPrep

1

Dear experts,

I am relatively new to fMRIPrep and have run it on my data using the following command:

docker run -ti --rm \
        -v ${sour_dir}:/data:ro \
        -v ${targ_dir}:/out \
        -v ${work_dir}:/work \
        -u $(id -u):$(id -g) \
        -v ${TFDIR}:/opt/templateflow \
        -v ${FSLICENSE}:/opt/freesurfer/license.txt \
        -e TEMPLATEFLOW_HOME=/opt/templateflow \
        nipreps/fmriprep:25.2.5 \
        /data /out/fmriprep participant \
        -w /work --fs-license-file /opt/freesurfer/license.txt --nthreads 10 \
        --output-spaces MNI152NLin6Asym:res-2 fsaverage:den-10k

After reading the documentation and inspecting the outputs, I have a few questions regarding the underlying workflows:

  1. Handling of dummy scans

As far as I understand, dummy scans (non-steady-state volumes) are not removed from the BOLD data but are instead indicated in the confounds.tsv file. Do these dummy scans affect subsequent processing steps? For example, do they influence the PCA decomposition when performing CompCor? In post-processing, is the recommended way to account for dummy scans to include the corresponding columns in the design matrix (or regress them out)?

  1. Slice timing correction (STC)

Is STC performed on the raw (input) BOLD data or on the motion-corrected BOLD data?

  1. Denoising in standard vs. native space

By default, the preprocessed BOLD data are resampled to standard space. Given that these outputs are only minimally preprocessed and additional denoising is typically warranted, is it recommended to perform denoising procedures in standard space while the confounding signals are derived in native space?

  1. Resolution selection for registration targets

The --output-spaces option specifies the resampling target. How does fMRIPrep select the registration target? For example, in the command above, I requested the BOLD data to be resampled to the res-2 resolution of the MNI152NLin6Asym template. Which resolution of that template is used to derive the registration warp field? This question might become more important when using custom templates.

Thank you very much for your time and guidance.

Best regards,

Yang Hu

2

Hi @younghoo,

Please read previous response here: Fmriprep --dummy-scans purpose? - #2 by effigies

STC is applied to the motion-corrected BOLD.

You should denoise in whatever space you want to analyze data in.

I believe the T1w registration step uses the best resolution available for the specified template family. The resampling happens separately.

Best,
Steven

1 Like
3

Hi @Steven , Thank you for your help. I found a reply from another post, and it seems that the STC is applied on the raw BOLD?

4

Head motion is estimated on raw BOLD, but applied to slice-time-corrected BOLD.

2 Likes