Evaluate napari-aicsimageio

[dpshelio]

https://github.com/AllenCellModeling/napari-aicsimageio

Definition of done

Fit for purpose on teaching

• [x] metadata visualisation
• [ ] Open file formats required
• [x] Easy to install and use (e.g., on M1/2 chips)

[p-j-smith]

I had a look at napari-aicimageio this morning and it seems it could be useful for visualising lots of different data formats and their metadata.

Installation

mamba create -n napari -c conda-forge python=3.10 napari bioformats_jar "PySide6<6.5"
mamba activate napari
python -m pip install napari-aicsimageio

I haven't got an M1/2 Mac, but the above worked on my Intel Mac and should work on M1 Macs too.

File formats required

I'm not sure what file formats are needed for the workshop, but the plugin supports reading a widge range of formats.

However, it does not support writing any format. So the plugin is currently useful for visualising images and their metadata, not doing analysis within napai and exporting the results. The underlying library has got writing capabilities, and there's an old issue to add writing support to the plugin. We could consider opening a pr to add writing support to the plugin, at least for file formats we need for the workshop.

Metadata visualisation

It's straightforward to view the image metadata using the plugin:

1. Open the plugin: Plugins > OME Tree Widget (ome-types)
2. Load an image using the plugin: File > Open with plugin > File(s)

The metadata is then displayed in the plugin widget.

Other thoughts

The plugin is developed by the same team that develops the aicimageio, the underlying library. However, the plugin hasn't been updated in over a year even though there have been several updates to aicimageio. The plugin still works for now, but we should encourage the workshop participants to consider alternatives in the future if this one goes unmaintained.

[davecash75]

HI @p-j-smith this was the comment from one of the microscopy experts on this point "Vital to include various data formats and limitations e.g. Tiff (BigTiff) , jpeg, compression types (lossless etc). Converting between types. Proprietary formats and Bio-formats" I have written to them asking if there are any examples he could provide. Are there also any example data in addition to the ones @K-Meech already found in the napari examples that we can start to use here?

[p-j-smith]

aicimageio have some data they use for testing, but it's not available via the napari interface - you need to download it from an s3 bucket:

aws s3 cp s3://aics-modeling-packages-test-resources/aicsimageio/test_resources/resources/ . --recursive

There's about 14 GB of data, so we probably don't want the participants to download it all! There's no bigTiff from what I can see, mostly:

• ome.tiff
• zarr
• nd2
• lif
• czi

Do you know if the microscopy experts have any data they can make available for the workshop?

[MonikaSvata]

• Check the licences
• Consider using histology data

[davecash75]

As mentioned in #5, Omero has some sample data at https://downloads.openmicroscopy.org/images/

[MonikaSvata]

• EICS testing data will not be used due to licencing requirements.

• We won't consider histology data for opening in Napari.

• Tasks left: Testing with data as per the comment above and data from Nick (still waiting for the data).

[p-j-smith]

For the resources mentioned in https://github.com/AllenCellModeling/aicsimageio/discussions/544#discussioncomment-7568092:

• Cell Feature Explorer and the Allen Cell Quilt Package both require a Python package (quilt) to be able to download the data, e.g.

import quilt3
pkg = quilt3.Package.browse("aics/hipsc_single_cell_image_dataset", "s3://allencell")
pkg.fetch("</path/to/image/>")

This might be too much to ask of the participants, especially if we're not requiring Python experience.

• BioImage Archive - there are some lif and czi datasets that could be of use. However, the napari-aicsimageio plugin and shows metadata for OME-TIFF files.

I think something like the MitoCheck OME-TIFF dataset would be good - it's relatively small (126 MB), contains metadata that we can view in napari, and is a 2D timeseries so could potentially be used for both cell counting and/or cell tracking

[MonikaSvata]

Kimberly to review

[K-Meech]

So I took a look at the MitoCheck OME-TIFF dataset - looks really nice + easy to download! It's a nice example of a time series, so I'm using it in the multi-dimensional images episode https://github.com/HealthBioscienceIDEAS/microscopy-novice/pull/31 . Unfortunately it doesn't have a pixel size set (e.g. in microns), so it's not the best example for the 'filetypes and metadata' episode where I want to highlight this.

As an alternative - I was thinking of this yeast dataset from the OME example data (the top one 'Plate1-Blue-A-12-Scene-3-P3-F2-03.czi'). This is small, easy to download and has a lot of metadata that can be browsed in aicsimagio, including pixel size. Also, it's released under a CC 4.0 license as stated in the readme one level up.