Metadata: User stories for named channels

[tischi]

I have the following user story:

Batch analysis of many multi-channel images, where the channel that contains the nuclei should be segmented. The data is messy in a sense that the nucleus channel sometimes is the first one, and sometimes the second one. To deal with this it seems useful to me to have something like int channelIndex = omeZarrDataset.getChannelIndex( String channelID ), where channelID in my case would be "nuclei". To be able to do this, I think some optional metadata inside the OME-Zarr that would allow one to store and retrieve a mapping from a channel name to its index along the channel axis could be useful.

What do people think?

Does anyone else have a need for named channels?

ping @bogovicj @joshmoore

[d-v-b]

Here's how this situation is handled in the climate sciences. Basically you associate the axis of the data with another piece of data (in this case, an array of strings), which gives meaning to each element along the axis. This would actually be useful for non-discrete axes, too, and it would greatly simplify the specification of coordinate transformations (instead of defining a function that transforms one grid to another, you would just store the output grid directly)

Personally I think we should adopt as much of the climate science stuff as possible! I think their framework could solve a lot of our problems.