pepolar scheme using SBRef considered as sophisticated.

[bpinsard]

What happened?

Running fMRIPrep 21.9.1 on a BIDS dataset with func/*sbref.nii.gz and fmap/*acq-sbref*_epi.nii.gz paired with B0FieldIdentifier raises NotImplementedError: Sophisticated PEPOLAR schemes are unsupported.

https://github.com/nipreps/fmriprep/blob/master/fmriprep/workflows/base.py#L511

could be replaced with

            if set(suffices) == {"epi"} or sorted(suffices) == ["bold", "epi"] or sorted(suffices) == ["epi", "sbref"] :

(tested with that patch and it seems to run)
or even maybe

            if set(suffices) == {"epi"} or set(suffices) == {"sbref"} or set(suffices) == {"bold"}  or sorted(suffices) == ["bold", "epi"] or sorted(suffices) == ["epi", "sbref"] :

If we want to allow task ran in both phase-encoding directions to be used for distortion estimation (without any `fmap/*_epi.nii.gz) if SDCFLows supports it.

Let me know if that makes sense and I can submit a small PR.

What command did you use?

apptainer images from repronim-containers, exact command no relevant.

What version of fMRIPrep are you running?

21.0.1

How are you running fMRIPrep?

Singularity

Is your data BIDS valid?

Yes

Are you reusing any previously computed results?

Anatomical derivatives

Please copy and paste any relevant log output.

No response

Additional information / screenshots

No response

[effigies]

This all makes sense to me. I think instead of your second proposal, we want to check that we have two images and that they're some combination of bold, sbref and epi, right?

if len(suffices) == 2 and set(suffices) - {"epi", "bold", "sbref"} == set():

Or possibly less cryptic:

if len(suffices) == 2 and all(suf in ("epi", "bold", "sbref") for suf in suffices):

While these might be less straightforward compared to dir-{AP,PA}_epi.nii.gz (how many volumes should be taken from a BOLD series?) I agree these are not sophisticated. It's one AP (or RL) series and one PA (LR), and that's manageable.

[bpinsard]

This all makes sense to me. I think instead of your second proposal, we want to check that we have two images and that they're some combination of bold, sbref and epi, right?

Exactly, however I think len(suffices) == 2 is not always true, one might combine more than 2 AP/PA files to estimate distortions. Also SDCFLows wrangler auto detection tend to have more than 2 estimator.sources which would explains the original code if I am not mistaken.

[effigies]

I guess it would be good to get @oesteban's input here. What were you considering a sophisticated scheme when you put in this limitation?

[oesteban]

• I believe the reason sbref is not considered is just my ignorance at the time. I think I assumed SBRefs would only be encoded with acq-sbref_bold. However, I just checked the specs and that's definitely wrong.
• I would leave having both a "bold" and an "sbref" reference of the same PE polarity as, at the very least, discouraged, and make the sbref have precedence.

[oesteban]

Exactly, however I think len(suffices) == 2 is not always true, one might combine more than 2 AP/PA files to estimate distortions. Also SDCFLows wrangler auto detection tend to have more than 2 estimator.sources which would explains the original code if I am not mistaken.

This seems correct. However the wrangler will need to drop bold PEs when one or more sbref are available for a given PE direction.