Manually converting ABCD task fMRI to BIDS

Similar to a previous question I asked, my lab has a set of ABCD task-fmri and associated fieldmaps which contain multiple reverse-PE images that need to be manually split after dcm2niix conversion.

The fieldmaps consist of interleaved “blip-up” and “blip-down” spin-echo images which are used for SDC correction (i.e. through topup).

The first volume of each task-fMRI bold sequence is a gradient-echo sequence. It was explained to me that this is not used for B0/SDC correction (due to dropout in the image), however it is used to “check” the accuracy of the distortion field estimated from the spin-echo images.

Labeling the spin-echo images seems simple enough (i.e. sub-01_dir-[AP/PA]_epi.nii.gz), however I am somewhat confused what I should do with the gradient echo images. Should I just discard these? Should I instead label the spin-echo maps as “magnitude1,2” and the gradient echo images as “phase1,2”?

You say that:

the first volume of the task-fMRI bold sequence is gradient echo-sequence.

Just to be sure to be on the same page: the whole run of the task-fMRI bold sequence consists also of volumes acquired with a gradient-echo sequence.

You speak specifically of the first volume of each task-fMRI sequence: is it saved individually? May be it is a SBref image, that is a calibration image use to reconstruct multi-band images, which has more contrast than gradient-echo images within the fMRI run and can be used as target for motion correction and also be used to coregister the bold space with the anatomical space.
In that case, such an image can be saved with _sbref suffix with the same name as the bold sequence it is related to.

A side note: gradient-echo EPI sequence may also be used for SDC correction and was shown to be useful. See this paper for instance:
Schallmo, M.-P.; Weldon, K. B.; Sponheim, S. R.; Olman, C. A. Assessing Methods for Geometric Distortion Compensation in 7T Gradient Echo FMRI Data; preprint; Neuroscience, 2020. https://doi.org/10.1101/2020.07.02.184515.

Labeling the spin-echo images seems simple enough (i.e. sub-01_dir-[AP/PA]_epi.nii.gz ), however I am somewhat confused what I should do with the gradient echo images. Should I just discard these? Should I instead label the spin-echo maps as “magnitude1,2” and the gradient echo images as “phase1,2”?

If those first volume are really SBref images, I would advise to keep them and name them with the _sbref suffix as I said above.
See here for details about the SBref images in BIDS:
https://bids-specification.readthedocs.io/en/stable/04-modality-specific-files/01-magnetic-resonance-imaging-data.html#task-including-resting-state-imaging-data

The bold sequences are acquired with an 8 channel head coil, so I don’t think they can be classified as single-band reference…

It may be possible to use a multiband sequence with a 8ch coil. If you implemented the ABCD protocol, they do use multi band sequence. If it is not single band reference image, you have to understand what is this extra image coming with your fmri sequence.