Conventions for spatial tranformations#47
Conversation
|
That's clear (at least for me). One important thing for the warps would be to mention that yes it represents the transformation from A->B, but it should be documented somewhere what is the representation (is it XYZ displacement in RASMM or LPSMM or VOX). Also the ITK convention is to write the displacement in an image interpolation friendly way (it is actually B->A internally). Maybe these notion should be writen in the doc so future module can explain them (case by case) to avoid confusion. |
|
There is 2 options to inform about the transformation space : either we (A) add entities in the filename or (B) add a json sidecar to the mix. I'd tend to try to prevent name bloating with (A), but (B) adds some undesirable management hurdles as well. There is a mechanism in BIDS to define a common sidecar for multiple files using the common entities present in the filenames that solves parts of it, but adds an extra layer of complexity. I don't have a perfect solution for this, maybe we should brainstorm and PoC. |
|
About the ITK convention, I already gave a link to the ANTs wiki, which has a beautiful description of the datastructures and conventions. I can add a sentence to push people towards it to learn more. |
gagnonanthony
left a comment
There was a problem hiding this comment.
Looks clear to me as well! I noticed a small typo, otherwise perfect job 💯
New documentation section on spatial transformations conventions :