OMERO-Screen

End-to-end high-content bio-image analysis powered by AI.

OMERO-Screen takes a 96-well immunofluorescence plate from an OMERO server to a publication figure: segmentation, feature extraction, cell-cycle assignment, storage, classification and plotting as one reproducible workflow. It is built for cell biologists running high-content screens.

📖 Read the documentation →

Warning

Active development. OMERO-Screen is developed alongside ongoing research, so expect frequent changes and occasional breaking ones. A stable release is planned for later in 2026. Pin a commit if you need an analysis to stay reproducible.

omero-screen 1234 --stitch          # segment a plate, write results back to OMERO
cellview import plate 1234          # pull the measurements into a local database

What makes it different

• Rapid model adaptation — builds on Cellpose and Trackastra, with a fast generation-and-deployment loop for classifiers. Models are selected from the command line, so an improved model swaps in without touching analysis code.
• 4i / cyclic IF and automated tracking, without the coding — registration and tracking pipelines produce highly multiplexed single-cell data from a few commands.
• Agentic AI integration — the whole platform is driven by terminal commands, so AI agents can run it. Bundled agent skills execute the pipeline from plate analysis to figure production from natural-language prompts.
• Data integration and management — raw data, masks and analysis files stay together on an OMERO server; single-cell measurements are queried from a local DuckDB database. Terabytes stay organised, queryable and reproducible.

Install

Requires Python 3.12, 3.13 or 3.14 and uv. Not tested on Windows.

git clone https://github.com/HocheggerLab/omero-screen.git
cd omero-screen
uv sync --dev

Then create a .env.development file with your OMERO credentials and database path. See Install & configure for the full list and the hardware requirements — in short, Cellpose 3 runs on Apple silicon or NVIDIA, while Cellpose 4 and tracking want an NVIDIA GPU.

Verify the connection:

uv run omero-integration-test e2e_connection

The packages

Six packages in one uv workspace. Each has its own documentation site with generated API and CLI reference.

Package	What it is
omero-screen	The analysis pipeline: segmentation, feature extraction, cell-cycle phases, flatfield correction, stitching, tracking
cellview	DuckDB-backed single-cell database with a CLI and a Python API
omero-screen-plots	Statistical plots built for publication figures
omero-screen-napari	napari widgets for browsing plates, inspecting galleries and labelling training data
cellclass	CNN classifier training, sweeps and inference
omero-utils	Connection handling, attachments, annotations and image I/O

Documentation

Analyse your first plate	One continuous journey from a plate ID to a figure
How it works	Pipeline stages, data model, benchmarks
CLI reference	Generated from the command definitions
API reference	Public Python API

Development

uv sync --dev
pre-commit install          # ruff, mypy and formatting run on every commit
pytest tests/unit_tests     # the full unit suite

A parallel OMERO server for integration tests runs at 127.0.0.2:4064:

./scripts/manage_test_server.sh start|stop|status

To populate it, scripts/load_plates.sh imports plates into a "Screens" project. It expects the Operetta layout, with */Images/Index.idx.xml under each plate directory. -x executes the import (without it the script is a dry run), -d sets the source directory, and -s prompts for host, port and username instead of using the defaults.

./scripts/load_plates.sh                        # show options
./scripts/load_plates.sh -d /path/to/plates -x  # import

Versioning

Semantic Versioning with Conventional Commits; versions are bumped by CI. feat: → minor, fix: → patch, ! or BREAKING CHANGE: → major. Only the root package is bumped by default — name the scope to bump another, e.g. feat(omero-utils): ....

Use cz commit rather than git commit to get the format right.

Contributing

Fork, branch, commit with cz commit, and open a pull request. Please run the unit suite and let the pre-commit hooks pass before pushing.

Citing

If OMERO-Screen contributes to work you publish, please cite it. A paper is in preparation; until then, cite this repository.

Authors

Built by Helfrid Hochegger and Alex Herbert in the Hochegger lab, University of Sussex. Contact: hh65@sussex.ac.uk

License

MIT — see LICENSE.

OMERO-Screen is an independent project built on OMERO. OMERO is a trademark of the University of Dundee.