carltonlab-napari-tools

About the plugin

Work in progress, please expect bugs V0.1.0

This napari plugin is a tool designed to computationally count the number of foci in microscopy of C. elegans gonads. It has two operational methods, manual and automatic counting. It uses ndevio to extract the required metadata and multiview-stitcher to stitch the gonad tiles into a single image.

The tool was designed for RAD-51 foci counting in 4D images (CZYX) since it is the most common assay for quantification of DSBs in C. elegans but can be used to count other foci in any image.

It features multi-gonad project creation where all genotypes and gonads are mixed after individual nuclei are selected (manual or automatic) and are scored in a blind manner to minimize human bias.

We recommend using the automatic process, which uses a trained cellpose model to segment the nuclei. By default, it uses our trained model: https://bioimage.io/#/artifacts/sneaky-panda.

For the automatic counting, it uses spotiflow with a default model and then point filtering based on user provided parameters such as expected foci volume, channel (usually DAPI) co-localization and more.

It'll export the data as plots that can be organized by genotype similar to the conventional RAD-51 scoring figures commonly used in papers.

It also writes all the intermediate files for troubleshooting, data archive and manual inspection. Formats are OME-Zarr for tiles and stitched images, TIFF files for cropped nuclei, CSV files for foci coordinates and feature tables. All which are easy to inspect even without the tool.

You can also do the entire counting and refinement of the data using the manual process.

Installation

Install from PyPI or napari Hub

Install the plugin in napari:

• In napari, open Plugins → Install/Uninstall Plugins…, search for carltonlab-napari-tools, select it, and click Install. Restart napari if prompted.

• If Install/Uninstall Plugins… isn't available, install the napari plugin manager in the same environment as napari.

• Or, in the Python environment used by napari, run:

  python -m pip install carltonlab-napari-tools
  

The commands above install the plugin's base dependencies. For automatic segmentation, install the matching extra in the same Python environment:

For CPU:

python -m pip install "carltonlab-napari-tools[full-cpu]"

For CUDA 12:

python -m pip install "carltonlab-napari-tools[full-cuda12]"

Depending on your system, you may need a different PyTorch build for CUDA.

Install from source

Clone the repository and enter its directory:

git clone git@github.com:carltonlab/carltonlab-napari-tools.git
cd carltonlab-napari-tools

Select what segmentation you'll be using (Automatic workflow only).

For the manual workflow, install napari and Qt without the segmentation models:

uv sync --extra all

For automatic segmentation on CPU, use:

uv sync --extra full-cpu

For automatic segmentation with CUDA 12, use:

uv sync --extra full-cuda12

We highly recommend using a GPU for the segmentation. Depending on the setup, you might need to install a different PyTorch version.

For our workflow, with approximately 2 × 50 × 1024 × 1024 (CZYX) images, segmentation and foci counting use about 6 GB of VRAM.

Running

Launch napari from the environment where you installed the plugin:

• If you installed it through napari's plugin manager, pip, or conda, launch napari as you normally do.
• If you installed from source using uv, run this from the repository directory:

uv run napari

Then, use the Plugins menu to launch the tool.

Contributing

Contributions are very welcome.

Acknowledgements

This plugin uses ndevio for image metadata and multiview-stitcher for image stitching.

If you use multiview-stitcher in published research, cite its Zenodo record.

License

Distributed under the terms of the BSD-3 license, "carltonlab-napari-tools" is free and open source software