Open Zarr File on M1

[jburel]

• Create a conda environment with the required dependencies
• Run the example below

from zarr.storage import FSStore
from itkwidgets import view
import zarr
fsstore = FSStore('https://dandiarchive.s3.amazonaws.com/zarr/7723d02f-1f71-4553-a7b0-47bda1ae8b42')
brainstem = zarr.open_group(fsstore, mode='r')

view(brainstem)

• The following message pops up.
  The git command requires the command line developer tools
  The tools is not installed.

[thewtex]

Hi @jburel, thanks for the report!

Currently, itkwidgets 1.X pre-releases provide M1 support. The 1.X refactor is architected on top of NGFF-Zarr.

The example notebook needed a slight update to install fsspec[http] with zsh in #553 .

The 1.X pre-releases are currently only available on PyPI. However, please try this environment.yml, which worked on my M1:

name: itkwidgets-zarr-demo
channels:
  - conda-forge
dependencies:
  - python
  - jupyterlab
  - pip
  - fsspec[http]
  - pip:
    - itkwidgets[lab]>=1.0a15

[will-moore]

Hi @thewtex - thanks for that.
I tried with that environment.yml and the image at https://uk1s3.embassy.ebi.ac.uk/idr/zarr/v0.4/idr0062A/6001240.zarr
but didn't get anything to show up, using your demo at ome/ngff#139 (comment) as an example:

Maybe OME-NGFF images aren't yet supported directly by itkwidgets?

[thewtex]

Hi @will-moore , thanks for testing!

There is a tweak required in your case -- itkwidgets[lab] -> itkwidgets[notebook] for Jupyter Lab vs Jupyter Notebook.

Here is what results (note -- the data does take a while to load from the USA at least -- and we are working on better feedback in the viewer that data is loading).

Untitled_.Oct.13.2022.10_08.AM.webm

Maybe OME-NGFF images aren't yet supported directly by itkwidgets?

itkwidgets is built around OME-NGFF and will load it directly! 🎇 If the data is not OME-NGFF, it is dynamically generated on the fly. Details:

itkwidgets/itkwidgets/integrations/__init__.py

Lines 41 to 129 in 339b0ee

	# NGFF Zarr
	if isinstance(image, zarr.Group) and 'multiscales' in image.attrs:
	return image.store
	min_length = 64
	if label:
	method = Methods.DASK_IMAGE_NEAREST
	else:
	method = Methods.DASK_IMAGE_GAUSSIAN
	store, chunk_store = _make_multiscale_store()
	if HAVE_MULTISCALE_SPATIAL_IMAGE:
	from multiscale_spatial_image import MultiscaleSpatialImage
	if isinstance(image, MultiscaleSpatialImage):
	image.to_zarr(store, compute=True)
	return store
	if isinstance(image, itkwasm.Image):
	ngff_image = itk_image_to_ngff_image(image)
	multiscales = to_multiscales(ngff_image, method=method)
	to_ngff_zarr(store, multiscales, chunk_store=chunk_store)
	return store
	if HAVE_ITK:
	import itk
	if isinstance(image, itk.Image):
	ngff_image = itk_image_to_ngff_image(image)
	multiscales = to_multiscales(ngff_image, method=method)
	to_ngff_zarr(store, multiscales, chunk_store=chunk_store)
	return store
	if HAVE_VTK:
	import vtk
	if isinstance(image, vtk.vtkImageData):
	ngff_image = vtk_image_to_ngff_image(image)
	multiscales = to_multiscales(ngff_image, method=method)
	to_ngff_zarr(store, multiscales, chunk_store=chunk_store)
	return store
	if isinstance(image, dask.array.core.Array):
	ngff_image = to_ngff_image(image)
	multiscales = to_multiscales(ngff_image, method=method)
	to_ngff_zarr(store, multiscales, chunk_store=chunk_store)
	return store
	if isinstance(image, zarr.Array):
	ngff_image = to_ngff_image(image)
	multiscales = to_multiscales(ngff_image, method=method)
	to_ngff_zarr(store, multiscales, chunk_store=chunk_store)
	return store
	if HAVE_TORCH:
	import torch
	if isinstance(image, torch.Tensor):
	ngff_image = to_ngff_image(image.numpy())
	multiscales = to_multiscales(ngff_image, method=method)
	to_ngff_zarr(store, multiscales, chunk_store=chunk_store)
	return store
	# Todo: preserve dask Array, if present, check if dims are NGFF -> use dims, coords
	# Check if coords are uniform, if not, resample
	if HAVE_XARRAY:
	import xarray as xr
	if isinstance(image, xr.DataArray):
	# if HAVE_MULTISCALE_SPATIAL_IMAGE:
	# from spatial_image import is_spatial_image
	# if is_spatial_image(image):
	# from multiscale_spatial_image import to_multiscale
	# scale_factors = _spatial_image_scale_factors(image, min_length)
	# multiscale = to_multiscale(image, scale_factors, method=method)
	# return _make_multiscale_store(multiscale)
	return xarray_data_array_to_numpy(image)
	if isinstance(image, xr.Dataset):
	# da = image[next(iter(image.variables.keys()))]
	# if is_spatial_image(da):
	# scale_factors = _spatial_image_scale_factors(da, min_length)
	# multiscale = to_multiscale(da, scale_factors, method=method)
	# return _make_multiscale_store(multiscale)
	return xarray_data_set_to_numpy(image)
	if isinstance(image, np.ndarray):
	ngff_image = to_ngff_image(image)
	multiscales = to_multiscales(ngff_image, method=method)
	to_ngff_zarr(store, multiscales, chunk_store=chunk_store)
	return store
	raise RuntimeError("Could not process the viewer image")

[will-moore]

Thanks - that works great!

I also tried a v0.3 OME-NGFF as part of filling out the growing table at https://github.com/will-moore/ngff/blob/ngff_tools_table/ngff-tools.md

https://hms-dbmi.github.io/vizarr/?source=https://uk1s3.embassy.ebi.ac.uk/idr/zarr/v0.3/idr0079A/9836998.zarr

I got some strange behaviour (don't know if this is v0.3-specific or not)..

To start with I got a low resolution view of the whole image, but as I was viewing it (just rotating and panning, no other settings changed) the higher resolution levels loaded but only showed a cropped part of the image...

Just another thought... I wonder if it would be possible to configure a page like with vizarr above so that I can open an image directly in itkwidgets without needing to run a notebook etc? It would make testing easier and we could then link from all our samples at https://idr.github.io/ome-ngff-samples/ directly to view in itkwidgets!

[PaulHax]

Hi @will-moore

I'm helping with this project. Your right, not a great experience with datasets that the viewer does not find scale metadata for. Higher resolution images grow in size as each voxel is 1 unit and the camera does not reset to encompass the new resolution. Then it takes a while to build the portion of the image in view of the camera. Appreciate any advice you have.

I like your link idea. Satisfaction:
https://kitware.github.io/itk-vtk-viewer/app/?rotate=false&fileToLoad=https://uk1s3.embassy.ebi.ac.uk/idr/zarr/v0.4/idr0062A/6001240.zarr

[will-moore]

Thanks @PaulHax that's great. I opened a PR to add itk-vtk-viewer links to all the v0.4 Images in https://idr.github.io/ome-ngff-samples/

[jburel]

Sorry for not responding earlier, I was on annual leave.
I will adjust some of my notebooks cc @will-moore

[jburel]

@PaulHax @thewtex we are having our community meeting over the next few weeks [1]
We have sessions on the 10 of November going over the viewers currently opening zarr files. We will encourage people developing viewers to present their tools (~10mins) and discuss what's needed from the specification point of views.
Will you interested to join us? The meeting is free

[1] https://www.openmicroscopy.org/events/ome-community-meeting-2022/

[thewtex]

@jburel that's fantastic! We are excited to participate!

[jburel]

Great. Registration closes on Friday 04/11. We will send zoom link and info to the participants