Evaluate napari plugins for basic image processing

[K-Meech]

For day 2, we'll need Napari plugins that can do basic operations like: image histograms, gaussian blur, thresholding, connected components, opening/closing

Some options to try out:

• napari-segment-blobs-and-things-with-membranes (scikit-image based)
• napari-simpleitk-image-processing (itk based)
• napari-zelda
• napari-matplotlib (for image histograms)
• napari-brightness-contrast (for image histograms)

Definition of done

• [ ] plugins cover the required basic operations
• [ ] Easy to install and use (e.g., on M1/2 chips)
• [ ] Histograms ideally work well for 2D/3D + RGB images

[p-j-smith]

I've had a look at some of the plugins and have some thoughts:

napari-zelda

• Is not yet tested on Python 3.10+ (as one of its dependencies has only just added support for it)
• Has some pre-defined workflows that could be useful, but they are not the most intuitive or straightforward to use

napari-allencell-segmenter

• developed by the same team that make napari-aicsimageio
• easy to install, but only supports single 3D images (no timeseries)

napari-partseg

• easy to install
• lots of tools for segmentation and calculating statistics; some tools have lots of parameters that can be set (which is probably useful but perhaps overwhelming for students on the course)
• no support for timeseries

napari-segment-blobs-and-things-with-membranes

• Has a large suite of tools for doing the sorts of operations we're interested in (Gaussian blur, thresholding, etc.)

• Easy to install using pip (although I don't have an M1 Mac so can't test that out):

  python -m pip install napari-segment-blobs-and-things-with-membranes

• Actually provides a long list of plugins (available via Plugins > napari-segment-blobs-and-things-with-membranes > <plugin list>)

• Using some of napari's builtin data (File > Open sample > napari builtins > Human mitosis), can label cells using the following workflow (each step is the name a separate plugin from the above list):

gaussian blur
threshold otsu
split touching objects
connected component labelling
manually split labels / manually merge labels

napari-simpleitk-image-processing

• Similar to napari-segment-blobs-and-things-with-membranes - has a large suite of tools for doing the sorts of operations we're interested in

• Easy to install using pip:

  python -m pip install napari-simpleitk-image-processing 

• this also provides a long list of plugins (available via Plugins > napari-simpleitk-image-processing > <plugin list>)

• Using the same builtin data (Human mitosis), can label cells and calculate statistics using the following workflow (again each step is the name a separate plugin from the above list):

gaussian blur
theshold otsu
connected component labelling
label statistics

Summary

napari-segment-blobs-and-things-with-membranes and napari-simpleitk-image-processing are the most promising in terms of being easy to use and having the functionality we need.

In differences between the two:

• napari-segment-blobs-and-things-with-membranes has a tool to split touching objects before the segmentation step, and another to manually relabel data
• napari-simpleitk-image-processing can calculate some simple cell statistics. Plus, the connected component labelling seems to perform slightly better compared to napari-segment-blobs-and-things-with-membranes (with fewer neighbouring cells being incorrectly connected)

So we might want to use both napari-segment-blobs-and-things-with-membranes and napari-simpleitk-image-processing?

[K-Meech]

Thanks @p-j-smith - this is a great overview! Seems like napari-segment-blobs-and-things-with-membranes and napari-simpleitk-image-processing should cover everything we need for day 2.

[K-Meech]

I'll close this issue now, as all episodes that use these napari plugins are merged