CZI inputs in Hedwig pipelines

[philipmac]

CZI format needs to be supported as an input type in Hedwig.

Glencoe Software's bioformats2raw utility (used elsewhere in these workflows) accepts czi as an input type, and can transform this format to our current pyramid file format, namely OME Zarr.

The czi inputs that Hedwig workflows need to transform typically contain multiple images, of greatly varying sizes and types. That is there are some images that are relatively small used to provide context, labels, or other information. There is also at least one that is typically very large. It's expected that the smaller images may not need to be converted to multi resolution format (pyramid), instead creating a png/jpg file scaled to a size that could be displayed as is in a web browser may be sufficient. It should be noted that "Each zarr structure is a Zarr group hierarchy( or folders), of other groups and arrays.", meaning that a single Zarr datastructure could contain more than 1 image, each of which in turn could be multiscale. (This is different to our mrc inputs, which contain only one image at one resolution.)

Minimally the larger image will have to be converted to Zarr format, for viewing in the multi resolution viewer, e.g. Neuroglancer.

bioformats2raw also generates an OME-XML file containing metadata about the image(s). By default this file is found at
example_file_name.zarr/OME/METADATA.ome.xml. This file contains lots of useful information, such as sizes:

<Pixels BigEndian="true" DimensionOrder="XYZCT" ID="Pixels:0" Interleaved="true" PhysicalSizeX="1.0" PhysicalSizeXUnit="Å" PhysicalSizeY="1.0" PhysicalSizeYUnit="Å" PhysicalSizeZ="1.0" PhysicalSizeZUnit="Å" SignificantBits="8" SizeC="1" SizeT="1" SizeX="1960" SizeY="1692" SizeZ="10" Type="int8">

and information about channels, including the number of components, the name of the IF labeling used for each channel and more:

<Channel ID="Channel:0:0" SamplesPerPixel="1"><LightPath/></Channel>

It might be useful for this file to be provided to Hedwig in its entirety, alternatively information may be parsed from it and provided to Hedwig on a piecemeal basis.

Channels
The images contained in a czi and be grayscale, RGB(A?), or multi-channel from fluorescing protein markers. Neuroglancer can be configure to display these different types of image with different custom shaders. Each type of channel image may have different parameters needed for visualization e.g. min/max per channel and channel name. For grayscale this is already in the visual_min_max tool which generates a json file. The algorithm and the json will need to be extended to support multi-channels. For RGB, typically are 0-255 and no custom min/max is needed as the full range is usually visualized.

[philipmac]

Slightly related to: https://github.com/niaid/hedwig-comm-specs/pull/60

[philipmac]

RGB czi - convert to Zarr, key image from zarr (somehow), scale key image for thumb.
For example - a tool might be created that will look the image that is 1 larger than the desired "key image", and 1 larger than the desired "thumb".

The tool will extract the appropriate level by a requested resolution from the OME-ZARR pyramid and save as a (JPEG/PNG - lossless preferred). This file will then get scaled to exact thumb / key image dimensions by graphicsMagick (or imagemagick).