[BUG] Inconsistent description of space entity specification

[tkkuehn]

Describe your problem in detail.

Opening this issue as suggested by @Remi-Gau in this Neurostars topic. In writing a BIDS app that needs to handle masks, and that I want to be as permissive as possible while still following the spec, I ran into the following (apparent) inconsistency:

• The specification section about masks says the space entity SHOULD have the value orig if there hasn’t been a transformation.
• The definition of the space entitity says that the value of the space entity MUST come from one of the modality specific lists in this appendix.
• That appendix includes, for images not registered to a standard template coordinate space, the following options for space: individual, study, or scanner (which is the implicit default, and space should just be omitted in this case). Notably, orig is not included.

It's not clear to me from the spec that orig is even an allowed value of the space entity, and it seems possible that it's redundant with scanner (although this is where I start to get out of my depth).

Describe what you expected.

The following would be helpful in determining how to properly use the space entity in a mask:

• Either that the definition of the space entity is relaxed to explicitly allow any string, or that orig is listed as an allowed value up in the coordinate system appendix.
• A clearer recommendation between use of orig and scanner in the coordinate systems appendix and masks section.

BIDS specification section

• Definition of the space entity
• Masks
• Coordinate system appendix

[tsalo]

The BIDS maintainers discussed this in a meeting, and we're thinking that the space entity definition should be relaxed, as you said, and scanner should be replaced with orig across the board.

For the entity definition, I think we should say that standard spaces must be limited to the allowed values, while non-standard spaces should use the defined values when possible.

[tkkuehn]

That all seems reasonable to me, with the additional note that currently:

• the masks section says that space-orig SHOULD be used if there hasn't been a transformation
• and the coordinate system appendix says that space-scanner SHOULD NOT be used at all, in favour of not specifying the space entity.

These recommendations seem to conflict if both of them are talking about space-orig, so I think as part of the recommended change to use orig everywhere, one of them should be reworded or removed. My personal leaning would be to prefer to use space-orig over omitting it in masks, but users with more experience than me may have a different preference.

[tsalo]

I'm not sure that any tools actually use space-orig. For the most part, I think they just follow the recommendation in place for space-scanner, so I think maybe the space-orig recommendation for masks is largely ignored. I don't have a strong opinion about it though.

[CPLambert]

Also just raised this as an issue with confusion over "individual" vs "orig" spaces without realising there was another term in play with "scanner". I think there are a lot of datasets already using the space-orig term (e.g. https://www.medrxiv.org/content/10.1101/2021.12.09.21267554v1.full), plus it pervades throughout the BIDS recommendations.

It seems the term "orig" is used more on the structural imaging/segmentation side vs. where-as fMRI-orientated crowd use "individual" instead - That probably says something deep about the different ways we view our data =)

Anyhow, agree "orig" should be accepted/preferred term.