🫁 Lungs segmentation in mice CT scans

We provide a YoloV8 model for the segmentation of the lungs region in mice CT scans. The model was trained on 2D slices and can be applied slice by slice to produce 3D segmentations.

[Installation] [Model weights] [Usage]

This project is part of a collaboration between the EPFL Center for Imaging and the De Palma Lab.

Installation

We recommend performing the installation in a clean Python environment. Install our package from PyPi:

pip install mouselungseg

or from the repository:

pip install git+https://github.com/EPFL-Center-for-Imaging/mouselungseg.git

or clone the repository and install with:

git clone git+https://github.com/EPFL-Center-for-Imaging/mouselungseg.git
cd mouselungseg
pip install -e .

Model weights

The model weights (~6 Mb) are automatically downloaded from this repository on Zenodo the first time you run inference. The model files are saved in the user home folder in the .mousetumornet directory.

Usage

In Napari

To use our model in Napari, start the viewer with

napari -w mouselungseg

Open an image using File > Open files or drag-and-drop an image into the viewer window.

Sample data: To test the model, you can run it on our provided sample image. In Napari, open the image from File > Open Sample > Mouse lung CT scan.

Next, in the menu bar select Plugins > Lungs segmentation (mouselungseg) to start our plugin.

As a library

You can run a model in just a few lines of code to produce a segmentation mask from an image (represented as a numpy array).

from mouselungseg import LungsPredictor

lungs_predict = LungsPredictor()

segmentation = lungs_predict.predict(your_image)

As a CLI

Run inference on an image from the command-line. For example:

uls_predict_image -i /path/to/folder/image_001.tif

The command will save the segmentation next to the image:

folder/
    ├── image_001.tif
    ├── image_001_mask.tif

To run inference in batch on all images in a folder, use:

uls_predict_folder -i /path/to/folder/

This will produce:

folder/
    ├── image_001.tif
    ├── image_001_mask.tif
    ├── image_002.tif
    ├── image_002_mask.tif

Issues

If you encounter any problems, please file an issue along with a detailed description.

License

This project is licensed under the AGPL-3 license.

This project depends on the ultralytics package which is licensed under AGPL-3.

Related projects

• Mouse Tumor Net | Detect tumor nodules in mice CT scans.
• Mouse Tumor Track | Track tumor nodules in mice CT scans.

Acknowledgements

Special thanks go to Quentin Chappuis for his contribution to the ideation and exploration of the data and for developing the preliminary code that laid the foundations for this project during the course of his Bachelor project in Fall 2023.