Automatic micro-tubule cut detector.
Code associated to the paper "Cut-Detector: A Tool for Automated Temporal Analysis of Late Cytokinetic Events" available from bioRxiv.
demo_v162.mp4
This napari plugin was generated with Cookiecutter using @napari's cookiecutter-napari-plugin template.
It is highly recommended to create a dedicated conda environment, by following these few steps:
-
Install an Anaconda distribution of Python. Note you might need to use an anaconda prompt if you did not add anaconda to the path.
-
Open an Anaconda prompt as admin to create a new environment using conda. We advice to use python 3.10 and conda 23.10.0, to get conda-libmamba-solver as default solver.
conda create --name cut_detector python=3.10 conda=23.10.0
conda activate cut_detectorOnce in the dedicated environment, our package can be installed via pip:
pip install cut_detectorAlternatively, you can clone the github repo to access to playground scripts.
git clone https://github.com/15bonte/cut-detector.git
cd cut-detector
pip install -e .We highly recommend to use GPU to speed up segmentation. To use your NVIDIA GPU, the first step is to download the dedicated driver from NVIDIA.
Next we need to remove the CPU version of torch:
pip uninstall torch torchvisionThe GPU version of torch to be installed can be found here. You may choose the CUDA version supported by your GPU, and install it with conda. This package has been developed with the version 11.6, installed with this command:
conda install pytorch==1.12.1 torchvision==0.13.1 pytorch-cuda=11.6 -c pytorch -c nvidianapari can be launched by just running napari in the terminal.
conda activate cut_detector
napariOnce open, click on Plugins > Cut Detector > Single Video.
Note that Folder is the equivalent of Single Video, but it processes all .tif files contained in a specific folder rather than a single video file. Divisions Matching and Distribution Comparison are more specialized features intended for advanced users and developers. Additional details can be found in the developers folder.
The video test_data.tif can be downloaded from Zenodo for testing. Once downloaded, open it in napari.
Before running the analysis, several settings can be modified:
img layer: The napari layer to be analyzed. By default, this corresponds to the currently opened file.Use default segmentation model?: Enabled by default. If you want to use your own Cellpose segmentation model, specify it using theSelect filebutton on the following line.Save cell division movies?: Option to save individual cell division movies as.tiffiles. These files will be saved in the folder selected using theChoose directorybutton.Directory to save results: Use theChoose directorybutton to select the directory where the results will be saved as a.csvfile.Debug mode: If enabled, temporary files and folders are not deleted but instead moved to the results folder. For regular use, it is recommended to leave this option disabled.Display segmentation and tracking: If enabled, napari layers displaying the segmentation and tracking results will be created. This option can be useful when the results appear unexpected or incorrect.Pixel size (nm): The default value corresponds to thetest_data.tifvideo.
Finally, click on the Run Whole Process (Single Video) button to run Cut Detector.
To update cut-detector to the latest version, open an Anaconda prompt and use the following commands:
conda activate cut_detector
pip install cut-detector --upgradeEach detected cell division is labeled with one of the following categories:
- NORMAL: Division happening as expected, where (at least) 1 micro-tubule cut is detected.
- NO_MID_BODY_DETECTED: Along the cell division, no mid-body was detected on the MKLP1 channel. This category encompasses different cases: the detection may have failed, the mid-body may not express the fluorescence, or this may not actually be a division.
- MORE_THAN_TWO_DAUGHTER_TRACKS: Tripolar division. This category encompasses both actual tripolar divisions and wrong identifications of daughter cells (mainly caused by segmentation issues).
- NEAR_BORDER: Division close to the border of the image, hence ignored as it is likely to be difficult to detect micro-tubule cuts. A division is classified as NEAR_BORDER as soon as the distance between 1 detected mid-body and the border of the image is less than 20px.
- NO_CUT_DETECTED: Division whose mid-body was detected, but with all micro-tubule bridges classified as "No cut". Likely to be at the end of the video, cells dying before the end of division, or cells going out of frame.
- TOO_SHORT_CUT: First micro-tubule cut detected before or at 50 minutes. Ignored as this is very unlikely, so it is probably caused by a wrong division detection.
Division movies start at the maximum between:
- Mother cell start frame
- 10 frames before the end of metaphase
Division movies end at the minimum between:
- Last frame of any of the daughter cells
- Metaphase of any of the daughter cells
Contributions are very welcome. Tests can be run with tox, please ensure the coverage at least stays the same before you submit a pull request.
Scripts required to improve any of Cut Detector tasks can be found in the folder developers.
Distributed under the terms of the BSD-3 license, "cut-detector" is free and open source software
If you encounter any problems, please file an issue along with a detailed description.
If you found our work useful, please consider citing:
@article{bonte2025cut,
title={Cut-Detector: A Tool for Automated Temporal Analysis of Late Cytokinetic Events},
author={Bonte, Thomas and Dubois, Lucas and Gagna, Paul and Dibsy, Rayane and Petrovi{\'c}, An{\dj}ela and Advedissian, Tamara and Serres, Murielle and Cuvelier, Fr{\'e}d{\'e}rique and Crouigneau, Marie and Sassoon, Nathalie and others},
journal={bioRxiv},
pages={2025--06},
year={2025},
publisher={Cold Spring Harbor Laboratory}
}