Converting just a field of a plate converted using bf2raw

[psobolewskiPhD]

Over in the bf2raw repo I asked:

Is there a way to take the whole plate omezarr generated by bf2raw and use raw2ometiff to convert select, individual fields? I can hack up the dir structure and METADATA.ome.xml to accomplish this, but was hoping there was a cleaner way. Passing: cellos_data.zarr/C/7/24 as the input doesn't work.

@melissalinkert tipped me off to:

Closest would be a recently-added feature in raw2ometiff 0.8.0-rc1:

#135

so something like raw2ometiff -f cellos_data.zarr/OME/METADATA.ome.xml cellos_data.zarr/C/7/24 C7-24.ome.tiff in theory should work. That feature was designed for label images, though, not this specific use case, so may not be quite what you have in mind.

If you eventually want to convert all wells and fields, but have each one end up in a separate file, then raw2ometif --split ... would also be an option.

I tested this and it worked-ish.
I get an OME-TIFF, but when I try to open in Fiji I get a ton of spam in the console for missing planes from all the fields that were not converted during the "Reading file header". After several minutes I gave up.
I'm using a test dataset with just a few wells, so I assume for a full plate this would be worse
I also noticed the file is significantly larger than if I do the initial convert to zarr using --series to just pass the single field.
I guess all of the METADATA.ome.xml is in the header and this is the issue?

Edit: If I open it in Python using tifffile the data looks to be identical to the individually exported ome-tiff. So it does appear that it's just the header is loaded with too much metadata.

[melissalinkert]

At the moment, that's entirely expected behavior.

The -f option here indicates what metadata to use - so if OME/METADATA.ome.xml has a plate and/or more Images than are being converted, that additional metadata will still show up in the output OME-TIFF.

Before we consider further changes, it would be useful to know what your target workflow or use case is. There are already a couple of ways to get one field per OME-TIFF that are better supported (bioformats2raw --series or raw2ometiff --split), so it would be good to know why those won't work and why passing a single multiscales to raw2ometiff is needed.

[psobolewskiPhD]

I'll double check with the end users, but my understanding is that they are using zarr for their data processing pipelines, but then for select fields they want OME-TIFF to share out with others or use in Fiji.
So the workflow is something like convert plate from phenix -> zarr -> select fields -> convert to OME-TIFF
Right now our workaround is to convert those select fields to zarr using --series and then convert those.

Can you point me to some documentation or an example using --split? The issue is they don't want to convert the whole plate zarr to OME-TIFF, just selected fields.

Thanks!