mAIcrobe
mAIcrobe: a napari plugin for microbial image analysis.
mAIcrobe is a comprehensive napari plugin that facilitates image analysis workflows of bacterial cells. Combining state-of-the-art segmentation approaches, morphological analysis and adaptable classification models into a napari-plugin, mAIcrobe aims to deliver a user-friendly interface that helps inexperienced users perform image analysis tasks regardless of the bacterial species and microscopy modality. Built using Python 3.10 and 3.11 and tested on Windows, and macOS.
You can read more about mAIcrobe in our preprint.
Video Showcase

โจ Why mAIcrobe?
๐ฌ For Microbiologists
- Automated Cell Segmentation: StarDist2D, Cellpose, and custom U-Net models. Several pre-trained models also included.
- Deep learning classification: 6 pre-trained CNN models for S. aureus cell cycle determination, a pre-trained model for E. coli antibiotic phenotyping plus support for custom models.
- Morphological Analysis: Comprehensive measurements using scikit-image regionprops
- Interactive Filtering: Real-time cell selection based on computed statistics
๐ For Quantitative Research
- Colocalization Analysis: Multi-channel fluorescence quantification
- Timelapse Analysis: 2D+t workflows with optional registration and label reassignment
- Batch Processing: Analyze one folder containing multiple FoVs with merged outputs
- Automated Reports: HTML reports with visualizations and statistics
- Data Export: CSV export for downstream statistical analysis
๐ Installation
Standard Installation:
We recommend using an environment manager like conda to handle dependencies and assure reproducibility.
Regardless of environment, you can install via pip in Python 3.10 or 3.11. This should handle all dependencies and might take a couple of minutes depending on your internet connection.
pip install napari-mAIcrobe
Development Installation:
git clone https://github.com/HenriquesLab/mAIcrobe.git
cd mAIcrobe
pip install -e .
๐ฏ Detailed Installation Instructions โ
๐ Key Features
๐จ Cell Segmentation
- Thresholding: Isodata and Local Average methods with watershed
- StarDist2D: custom models (pretrained available for S. aureus)
- Cellpose: cyto3 model
- Custom U-Net Models: custom models (pretrained available for S. aureus, B. subtilis, and S. pneumoniae)
๐ง Single cell Classification
- Pre-trained Models: S. aureus cell cycle and E. coli antibiotic phenotyping
- Custom Model Support: Build your training dataset in napari with out custom widget, train using our jupyter notebook and load your own TensorFlow models,
๐ Comprehensive Morphometry
- Shape Analysis: Area, perimeter, eccentricity
- Intensity Measurements: Fluorescence statistics
- Custom Measurements: Septum detection, colocalization, and more
๐ Integration with other tools
mAIcrobe is designed to fit into broader image-analysis workflows.
- napari ecosystem: Keep working in napari with complementary plugins by reusing the layers generated in mAIcrobe (images, labels, and measurements).
- Fiji / TrackMate: Export labels and measurement tables for downstream tracking and curation workflows in Fiji, including TrackMate-compatible analysis pipelines.
- Open outputs: Use exported CSV tables in Python, R, Prism, or spreadsheet tools for custom analysis and visualization.
๐ Documentation
For programmatic usage:
| โ๏ธ API Reference
๐ Available Jupyter Notebooks
Explore advanced functionality with included notebooks:
๐ค Community
๐๏ธ Contributing
We welcome contributions! Whether it's:
- ๐ Bug reports and fixes
- โจ New segmentation algorithms
- ๐ Documentation improvements
- ๐งช Additional test datasets
- ๐ค New AI models for classification
Quick contributor setup:
git clone https://github.com/HenriquesLab/mAIcrobe.git
cd mAIcrobe
pip install -e .[testing]
pre-commit install
Testing:
# Run tests
pytest -v
# Run tests with coverage
pytest --cov=napari_mAIcrobe
# Run tests across Python versions
tox
๐ Full Contributing Guide โ
๐ License
Distributed under the terms of the BSD-3 license, mAIcrobe is free and open source software.
๐ Acknowledgments
mAIcrobe is developed in the Henriques and Pinho Labs with contributions from the napari and scientific Python communities.
Built with: