# Get Your Brain Together HCK 02: Anatomic orientation in OME-Zarr NGFF

2023-05-23

## Attendees

- Josh Moore, OME
- David Feng, AIND
- Sharmi, AIND
- Nick TL, Vanderbilt
- Nick Lusk, AIND
- Lydia Ng, AIBS
- Cai McCann,
- Camilo Laiton
- Alan Watson
- Christian
- Christian
- Dzenan Zukic, Kitware
- Beatriz, Cajal Neuro
- Yael Balbastre
- Alan Watson, U Pitt
- Niles, Cajal Neuro
- Satra G, MIT
- Matt McCormick, Kitware

## Discussion

### *Standard, default orientation*?

- LPS?
- RAI?
- RAS? *
- [...]

### Where to store?

- v0.5 later this summer
- More Transformations
- Add orientation to Transform Spec
- https://github.com/ome/ngff/issues/84
- https://github.com/ome/ngff/pull/138#issuecomment-1467186626
- Let's merge Affine Transforms GitHub Flash Mob
- @type
- Scale
- AnatomicalToRAS
- Radial
- captures more information not captured by current transformations
- Death to ambiguity
- Spatial Transcriptomics data storage
- https://spatialdata.scverse.org/
- Add to data officially before v0.5 released
- Propose incremental change to John's transformations proposal may be a good path forward
- New specification
- Lead authors connection with Josh M
- `axes`, `long_name`
- https://github.com/ome/ngff/issues/142
- `long_name` describes axis identifier, direction
- Registration transformations
- Space names
- Many registration transformations that mapping to the same anatomic space
- Define expectations
- Where and how these are handled
- Transformations that go beyond rough anatomical orientation go into separate Zarr paths / groups from the multiscale image group
- Expected sequence or expected conventions for provenance of transformations
- Do not have to re-write data associated with transformations
- https://earthmover.io/

### How to standardize?

- Zarr spec?
- No current mechanism
- "multiscale" proposed, -> OME spec
- Geo-Zarr now adding multiscales as "Zarr"
- https://github.com/zarr-developers/geozarr-spec
- Spatial concepts
- May be difficult to converge with OME-Zarr
- OME spec?
- Pubish specification
- Schema that can be validated

### How to encode?

#### Terms relative to the body or the embryonic forebrain

- https://www.quora.com/Can-you-tell-me-the-key-differences-between-anterior-versus-posterior-ventral-versus-dorsal-and-rostral-versus-caudal-in-terms-of-which-positions-they-are-referring-to-in-anatomical-terminology

#### Direction matrix

- Affine transformation
- Rigid rotation (direction cosines)
- Pre-concatenated (preferred), vs concatenated
- Axis-aligned, human readable metadata
- Nifti, ITK
- Orientation matrix defines how to rotate into a prescribed orientation
- S-form and Q-form
- Has caused confusion
- Scanner and RAS
- and this is the set of nifti intent codes: https://nifti.nimh.nih.gov/nifti-1/documentation/nifti1fields/nifti1fields_pages/qsform.html/document_view

Progression, named, concatenated transformation
1. Scale
2. Translation
3. Rough axis-aligned axes, (matrix of 1's, -1's and 0's ?)
4. Refined orientation / direction matrix

#### Representation, e.g. "LPS", "left-posterior-superior", "left-to-right"

- Towards
- Nifti to and from discussion
- https://docs.google.com/document/d/11gCzXOPUbYyuQx8fErtMO9tnOKC3kTWiL9axWkkILNE/edit#heading=h.mqkmyp254xh6
- some of it was later codified in nitransforms (a python library now): https://nitransforms.readthedocs.io/en/latest/notebooks/isbi2020.html

- Be explicit
- Bring bioimaging and biomedical communities together
- Try to be inclusive
- Enum
- Controlled vocabulary
- https://bids-specification.readthedocs.io/en/stable/appendices/coordinate-systems.html#image-based-coordinate-systems
- Define translations as needed
- Provenance
- who defined the transform / coordinate system
- can you have multiple?
- are they overwritten? (no yes/no answer; all interpretation)