I have one further question regarding lesion masks in BIDS format. This is related but distinct to issue 27100 and 2962. Consider that all individuals are only observed once (hence no session folder), but for each person we define three lesion maps: one maps chronic injury, one acute injury and one both. As a further wrinkle, we use the TRACE and ADC sequences to draw the injury, though these raw DWI scans direct from the scanner are not defined by BIDS (so would have to be included in the bids ignore). How would we define this? Does this make sense:
I think TRACE and ADC maps can be BIDS valid and defined with the rec-TRACE/ADC label, as these are off-the-scanner reconstructions of the image (so it would still have the dwi suffix). You would still want to filter them out of DWI pipelines like QSIPrep because you do not want them to be treated like a raw DWI images.
BIDS will not like the something like desc-acute-lesion because that is two hyphenated words in a row (when it should be alternating between hyphen and underscore). You could make it something like desc-AcuteLesion instead. However, the validator doesn’t operate on derivatives yet and the lesion mask file type is not fully codified into BIDS yet as far as I can tell, so at this point, you can do what is convenient for you.
@Steven can you provide any sample BIDS datasets that show this usage (e.g. I could not find any on OpenNeuro or the examples). Since these are non-directional reconstructions, I assume they do not include the bval or bvec files (or do these use the maximal b-value used in their creation).
As a concrete example, are you suggesting this is the valid structure:
I will look for example datasets, but yes that is what I was proposing. Now that you mention it, I believe the validator will throw errors if any _dwi file is not accompanied by bval/bvecs, which do not really apply to a TRACE/ADC image. So I guess you could include a bval/bvec with a single b0 volume to keep it BIDS valid, but that feels a little silly.
I guess this may be one of the only times the Inheritance Principle will not have unintended consequences for .bvec files. Since all of my DWI data is non-directional single volume data, I can have a single .bval and .bvec file in the bids root that will act as a surrogate for all the dwi images in all the folders, right? In general, I think the inheritance principle can have unintended consequences for bvec files, as it reports vectors in image space, so an identical sequence will create different bvecs for different image angulation.
@Steven After further reflection, I think the masks belong in the dwi not anat derivatives folder as they are drawn on diffusion images, so if you agree with this, I will label this as solved: