ALIBY

End-to-end processing for high-throughput microscopy.

ALIBY (pronounced alib-bee) orchestrates various tools for processing large-scale imaging data. It quantifies microscopy experiments, fluorescence and/or time series, through segmentation (e.g., Cellpose) and feature extraction or via deep learning models.

Main Features

• End-to-End: Extract morphological profiles from a set of images.
• Multimodal: Handles Time-series, Fluorescence (Cell Painting, Fluorophores), and both 2D and 3D data.
• Interpretable & Deep Learning: Leverages cp_measure for classical features or nahual for deep learning embeddings.
• Local-First, but also Distributed: Can be run on a single machine, but also supports GPU-based processing across machines.
• Standardized Output: Object-level profiles (e.g., cells, nuclei) stored as efficient Parquet files.
• Zarr support: Zarr support for compressed, scalable datasets.

Installation

Using pip

pip install aliby

Using uv

For development installations, we recommend uv.

git clone git@github.com:afermg/aliby.git
cd aliby
uv sync --all-groups

Nix

For a fully reproducible, self-contained shell:

nix develop .

Quick Start: Basic Pipeline for Local TIFFs

Process a local dataset of TIFF files using Cellpose for segmentation and cp_measure for feature extraction.

from pathlib import Path
from aliby.io.dataset import DatasetDir
from aliby.pipe import run_pipeline_and_post
from aliby.pipe_builder import build_pipeline_steps

# 1. Setup paths and metadata identification
input_path = Path("data/my_experiment")

# Regex to capture Well, Field-of-view, and Channel from filenames
# example: `testdir/testimage__A01__1__DNA.tif` -> A01, 1, DNA
regex = r".*__([A-Z][0-9]{2})__([0-9])__([A-Za-z]+).tif"
capture_order = "WFC" 

# 2. Identify positions in the dataset
dataset = DatasetDir(input_path, regex=regex, capture_order=capture_order)
positions = dataset.get_position_ids()

# Take the first position for this example
key, path = positions[0]["key"], positions[0]["path"]

# 3. Build pipeline steps
pipeline = build_pipeline_steps(
    channels_to_segment={"nuclei": 1},
    channels_to_extract=[0, 1], 
    features_to_extract=["intensity", "sizeshape"], 
)

# 4. Configure the tiling step
pipeline["steps"]["tile"]["image_kwargs"] = {
    "source": {"key": key, "path": path},
    "regex": regex,
    "capture_order": capture_order,
}

# 5. Run the pipeline
result = run_pipeline_and_post(
    pipeline=pipeline, 
    pipeline_name=key, 
    output_path="results"
)

See the examples/ directory for more advanced use cases.

Core Concepts & Glossary

Microscopy terminology can vary. Here’s how ALIBY defines these concepts:

• Position / Field of View (FOV) / Site: A set of images (all time points, z-stacks, channels) captured at a single physical location.
• Time point / Frame: A single moment in a time-series experiment.
• Regex & Capture Groups: ALIBY uses Regular Expressions to map filenames into internal arrays. Capture groups () in your regex define metadata like Well (W), Channel (C), or Time (T).
• Capture Order: The order of metadata groups in your regex (e.g., "WFC").
• TCZYX Dimensions: Internally, all data is handled as a 5-D array: Time, Channel, Z-stack, Y, and X.
• Cell Painting: A high-content screening assay using five dyes to highlight various organelles, typically used to generate phenotypic vectors.

About the project

• Under Construction: Detailed documentation is currently being
• We aim to keep the dependency tree minimal. Further downstream processing is often relegated to specialized libraries.
• ALIBY (Analyser of Live-Cell Imaging for Budding Yeast) originated in 2021 at the Swain Lab to quantify yeast signalling. It has since evolved into a general-purpose tool for high-throughput imaging, including mammalian Cell Painting assays.