Remove the first four scans of fMRI sequence BEFORE/DURING fMRI prep

Hi there,

This is a pretty simple question (newbie here).

I want to exclude the first four dicom acquisitions of my fMRI sequences from my NIFTI conversions that I put into fMRIprep. Is there a way to do this in Dcm2niix or Dcm2Bids? or does fMRIprep somehow automatically do this?

Thanks,

Jen

Hi Jen,

I believe you can just set the number of —dummy-scans to be 4 in your fmriprep command. If not, I think FSL’s fslroi can do something similar to your nifti files. Alternatively you can process as normal and use one-hot encoded vectors in your design matrices to exclude the first four volumes from further analysis (similar to how motion outliers are treated).

Hope this helps,
Steven

Hi Steven

I have the same question. and I did this by setting the number of --dummy-scans to be 4 in my fmriprep commnand. However the output of preproc_bold file still have the full time points/volumes (240) rather than the expected 236 times points. Maybe I misunderstood the ‘–dummy-scans’ option?
and is it necessary to delete the first few dummy time points? Some articles do this and some don’t. and if I deleted them after preprocess, does it make sense? because dummy scans have impacts on slice timing correction.

Thank you!
Jun

He @wenjun and welcome to Neurostars!

Have you looked in the resulting confounds.tsv files? --dummy-scans should add column regressors that you can use to regress out the dummy scans in your first level models.

It depends on what your analysis and task (or no-task) design is, there is not one golden rule.

I would not delete alter your preprocessed nifti files, because then there will be a mismatch between your scan length and confounds file length.

Do they? I though slice timing correction is a within-volume operation, so dummy scans would not have an impact on the scans afterwards.

Best,
Steven

Hi Steven,

Thank you for the time reply :smiley:
Now I get it.
And if I did not set the ‘–dummy-scans’ option, fmriprep will automatically estimate the number of dummy scans for each subject, and the numbers may vary from individual to individual. Will the difference affect the next regression analysis?(I’m not there yet)

Sorry for the unclear description. I saw the words in fmriprep docs " These nonsteady states (also called dummy scans) typically show greater T1 contrast and higher average intensity, and therefore potentially are detrimental if used in the interpolation of slice timing corrections. Hence, nonsteady states are discarded by the slice timing correction tool (in this case, AFNI’s 3dTShift)"

https://fmriprep.org/en/latest/faq.html#error-in-slice-timing-correction-insufficient-length-of-bold-data-after-discarding-nonsteady-states

Many thanks,
Jun

Hi,

Yes, fMRIPrep will automatically return non steady-state volume regressors which you can similarly include in your analyses. As long as no one has too many volumes to regress out, this variability shouldn’t affect your analyses more.

You can also make your own similar one-hot encoded regressors for volume regression if you did want to explicitly remove the first X volumes but forgot to add the --dummy-scans argument to fmriprep.

Best,
Steven

I got it! Thanks Steven!