Single-band reference images for fmriprep and dcm2bids - intendedFor fields?

,
1

We are using a multi-band acquisition on a Philips scanner for task-based and resting-state fMRI. We had been acquiring SBREF images for all of our scans, though I’ve never really been sure how many to get. (One per BOLD run? One per session? One per scan type?)

We were troubleshooting something else, and I realized that although I listed all of the BOLD runs in the “IntendedFor” field for dcm2bids, we found that the sbref was only applied for the scan matching its name. For instance, if we had the following BOLD scans:

BOLD Scans
sub-2304_ses-1_task-encoding_run-01_bold.nii.gz
sub-2304_ses-1_task-encoding_run-02_bold.nii.gz
SBREF Scan
sub-2304_ses-1_task-encoding_run-01_sbref.nii.gz

Only run-01 would have the sbref associated with it. Even though the sidecar for the SBREF says:

    "IntendedFor": [
        "func/sub-2304_ses-1_task-encoding_run-01_bold.nii.gz",
        "func/sub-2304_ses-1_task-encoding_run-02_bold.nii.gz",
        "func/sub-2304_ses-1_task-encoding_run-03_bold.nii.gz",
        "func/sub-2304_ses-1_task-encoding_run-04_bold.nii.gz"
    ],

I then saw a response on neurostars post that fmriprep uses the matching file name (Fmriprep warning: No single-band-reference found for - #9 by jsein). I could envision a few ways around this - e.g., renaming our sbref to match each scan separately…but maybe that is NOT the proper use of SBREFs and we should’ve been collecting one-per-BOLD scan?

2

IntendedFor is not defined for sbrefs. The only way to associate them is if the filenames are the same up to the bold or sbref (or dwi and sbref, in the case of diffusion images). The strong (but apparently implicit) assumption was that the sbref would be taken immediately prior to the multiband EPI series, so that you could with reasonable safety use the sbref or first volume of the EPI interchangeably, where the only difference is contrast.

If you would like to use these with fMRIPrep, probably the best thing to do is to register the initial BOLD volumes to the first volume of run-01_bold, and then use those registrations to resample the sbref into alignment with each BOLD series.

1 Like
3

Thanks, I was worried about something like this when we set the scans up. The critical data point would be how closely the sbref was acquired viz. the BOLD scan. So, if they were acquired far apart, they’re likely not as useful for registration. Sounds like we should forego sbref use for scans that weren’t acquired with them directly prior? It’s not as though the registrations were poor without the sbref for those scans that evidently weren’t registered with them by fmriprep.

4

If the registration and SDC are fine without the sbref, then you should be fine to ignore them. You can use --ignore sbref.

1 Like