This PR is intended to supersede #1831 with additional content, as well as resolve #1862.
Rather than incrementing the spec with different aspects of scanner-generated DWI derivatives, as happened previously with #1725 and now #1831, it makes more sense to me to establish what that set is, and resolve the specification against the entirety of that set.
I have additionally created an exemplar dataset that shows all of the derivatives that can be generated by the product diffusion sequence on the Siemens XA platform:
https://github.com/bids-standard/bids-examples/pull/452
If anybody knows of other derivatives that can be generated by other sequences / platforms, it would make sense to add those as extra subjects within that dataset.
While the sequence exports can include data encoding the outcomes of the tensor fit itself, those are exported in a proprietary format, cannot be trivially converted to NIfTI, and are therefore excluded from this list.
Some points requiring discussion (hence draft PR):
[ ] The "exponential ADC" map is explained here. I've confirmed that if you compute the ratio of the trace-weighted image to the mean b=0 image you get the same result. This is the name that parameter is given in the protocol editor on the scanner console, but it's not a unique name; it can also be called "attenuation coefficient" or "attenuation factor". Most likely "expADC" is the most sensible suffix, but requires agreement.
[ ] I've de-capitalised the existing TRACE suffix, and set these new suffices, based on my thinking registered over on bids-2-devel where capitalisation should apply to acronyms.
[ ] I've not confirmed compatibility between the proposed exemplar dataset and the proposed changes to the spec. I presume this can be done, just not something I've done before.
[ ] One of the scanner-generated derivatives is labelled as "TENSOR_B0". This is not precisely equivalent to the mean b=0 image; rather, it is the estimated b=0 signal intensity based on the tensor model fit ("S0" is more commonly used for this term in diffusion model space). So it's just a T2-weighted image, but it's arisen from a tensor model fit. Therefore in the exemplar dataset, I've named these data "*_desc-tensor_T2w.*". I presume this is not currently permissible; I would personally prefer this kind of encoding as opposed to giving it its own suffix, but if others dislike it we'll need to figure out some different encoding.
This PR is intended to supersede #1831 with additional content, as well as resolve #1862.
Rather than incrementing the spec with different aspects of scanner-generated DWI derivatives, as happened previously with #1725 and now #1831, it makes more sense to me to establish what that set is, and resolve the specification against the entirety of that set.
I have additionally created an exemplar dataset that shows all of the derivatives that can be generated by the product diffusion sequence on the Siemens XA platform: https://github.com/bids-standard/bids-examples/pull/452 If anybody knows of other derivatives that can be generated by other sequences / platforms, it would make sense to add those as extra subjects within that dataset.
While the sequence exports can include data encoding the outcomes of the tensor fit itself, those are exported in a proprietary format, cannot be trivially converted to NIfTI, and are therefore excluded from this list.
Some points requiring discussion (hence draft PR):
expADC
" is the most sensible suffix, but requires agreement.TRACE
suffix, and set these new suffices, based on my thinking registered over on bids-2-devel where capitalisation should apply to acronyms.*_desc-tensor_T2w.*
". I presume this is not currently permissible; I would personally prefer this kind of encoding as opposed to giving it its own suffix, but if others dislike it we'll need to figure out some different encoding.