search

spinalcordtoolbox / PAM50

https://github.com/neuropoly/spinalcordtoolbox
2 stars 1 forks source link

aligning with BIDS #11

Open CPernet opened 1 year ago

CPernet commented 1 year ago

Hey guys,

I'm just putting the link here about the BIDS atlas BEP. This would be fantastic if you could contribute.

From the purely atlas labeling side (https://github.com/spinalcordtoolbox/PAM50/tree/master/atlas) this should be easy, tsv and json for labels and provenance

From the more quantitative size (https://github.com/spinalcordtoolbox/PAM50/tree/master/histology) it would be interesting to see if this works ; we have proposed something for PET and that's somehow similar

thx

jcohenadad commented 9 months ago

Thank you Cyril-- I know we've discussed it internally and we have identified some roadblocks that I can't remember. I will put it on our agenda for our next meeting and we will document here.

joshuacwnewton commented 9 months ago

I know we've discussed it internally and we have identified some roadblocks that I can't remember.

I'm just going to copy over the previous discussion from the email chain, so that the information is all in one place.

@mguaypaq:

I had a look at the BIDS-Atlas BEP, and now I'm trying to see if I can fit the PAM50 template into it, and already I have several questions. So far, it looks like it would require much more re-organization and transformations than just renaming some files. (This would also require changes in the SpinalCordToolbox code which uses the PAM50 template, but that's maybe a separate matter.)

Maybe you'd rather answer these questions (and my underlying confusion) more efficiently in a meeting, but here are some questions to begin with. I know you're more interested in the Quantitative aspects of PAM50, but so far I'm still looking at the "traditional" aspects:

  1. PAM50 comes with some template images (T1-, T2-, and T2*-weighted) so that subject images can be registered with it. But, BIDS-Atlas seems to only allow probseg/dseg/mask as image types. Is there some compliant way to keep the templates together with the atlas? I'll note that the templates are not really "subjects" in a meaningful way, so I don't think that's the solution.

  2. PAM50 seems to use all of probseg/dseg/mask types, with different types for some different structures (spinal cord, white matter, etc.). Is each type a different atlas? But then, we couldn't use consistent numbering across the different types; in other words, the probseg node 0 would not refer to the same structure as dseg node 0, or mask node 0.

  3. PAM50 comes with two ways of representing spinal cord levels: a discrete version (which is probably a dseg?), and a continuous version (from 1.0 to 21.0, with floating point values). What's the data type for the continuous version? It doesn't seem to be a propseg, but maybe it's a mask? And can the two notions of level co-exist in the same atlas?

  4. For the continuous levels, it seems like some of the fields for "relmat.json" files would apply to describe the range and interpretation of values, but the levels are not a relation matrix: each value corresponds to a single voxel, not a pair of nodes. Where would this fit in BIDS-Atlas?

  5. Currentlty, PAM50 is structured as a separate 3D .nii.gz file for each node, with the node number in the filename. For BIDS-Atlas, would they have to be combined as separate volumes inside a single 4D .nii.gz? (Or even 5D, if the 4th dimension is reserved for time?) EDIT by @joshuacwnewton: This may be inaccurate following all spinal levels being merged into a single file, see: https://github.com/spinalcordtoolbox/PAM50/issues/16

Peer:

thank you very much Mathieu for all the comments and feedback, that’s highly appreciated.

I really like how the quantitative data y’all are working with is stressing the atlas proposal concerning its usability and, in turn, required adaptations. I was wondering if we should organize a meeting these days to discuss and address certain points, ie a one hour mini sprint? @ Mathieu, Julien, Agâh et al.: I might be in MTL late May/early June and could stop by at Poly and we could zoom in Cyril. Otherwise, we can of course also schedule a fully virtual meeting. WDYT?

Did we ever get a chance to meet with Peer? I recall that there were some scheduling conflicts with ISMRM at the time...

mguaypaq commented 9 months ago

No, the meeting never happened, and then I think we all just forgot to follow up.

jcohenadad commented 8 months ago

Related to https://github.com/bids-standard/bids-specification/issues/1281