Viewing local OME-Zarr data

[ruaridhg]

Find instructions on loading OME-Zarr data into Napari (if doesn't exist/not up-to-date then update instructions)
Local version of OMERO/Neuroglancer?

[ruaridhg]

OMERO Desktop client is deprecated, but more generally you can interact with OMERO through Python scripts/CLI.

[K-Meech]

Some initial points from loading Zarr data into Napari:

• Napari has zarr-python as a dependency, and can open zarr arrays directly by default (no extra plugins needed). You can drag and drop a .zarr dir onto the Napari window to open it, or use File > Open Folder. This seems to use dask_array.from_zarr underneath

• Napari's add_image etc. functions can directly support various types of arrays, including zarr-python zarr arrays. So you can load a local zarr array as follows:

import napari
from pathlib import Path
import zarr

heart = zarr.open_array(Path("path/to/heart.zarr"))

viewer = napari.Viewer()
new_layer = viewer.add_image(heart)
napari.run()

This also works for remote zarrs e.g.:

store = zarr.storage.FsspecStore.from_url(
   'gs://ucl-hip-ct-35a68e99feaae8932b1d44da0358940b/A186/lung-right/4.26um_VOI-3_bm18.ome.zarr/',
   read_only=True
)
remote = zarr.open_group(store=store, mode='r')

# View a specific resolution level 
level = "6"

viewer = napari.Viewer()
new_layer = viewer.add_image(remote[level])
napari.run()

• For multi-scale viewing (interactively switching between different resolution levels in an ome-zarr) - there is the napari-ome-zarr plugin. This can open a full remote image e.g.:

viewer = napari.Viewer()
viewer.open(
    "https://storage.googleapis.com/ucl-hip-ct-35a68e99feaae8932b1d44da0358940b/A186/lung-right/4.26um_VOI-3_bm18.ome.zarr/", 
    plugin="napari-ome-zarr"
)
napari.run()

BUT the viewing is very choppy / hangs a lot with a decently large image (such as the one used above). It is pretty much unusable at this speed, so I think we would have to demo either a small ome-zarr image, or pulling out smaller pieces / lower resolution levels to display in the viewer.

There is experimental support for async viewing which can be enabled by setting the NAPARI_ASYNC=1 environment variable. This keeps the viewer more responsive, but the data still takes a very long time to load as you zoom in / move to different z planes. Performant multi-scale viewing is very much still a work in progress in napari, and some relevant issues are linked here.

[ruaridhg]

Add example to appendix

[K-Meech]

Pending PR for this in the heftie textbook repo

[K-Meech]

I'll go ahead and close this issue, as the small napari example has been merged into the HEFTIE textbook appendix now.