[OME2026] Poster and Lightning Talk: OME-Arrow: Unifying Images, Metadata, and Features in an Interoperable Data Model

Abstract: Modern bioimaging workflows increasingly combine images, metadata, and derived measurements across many tools and platforms. Enabling these components to work together seamlessly is key to interoperable and scalable analysis.

OME-Arrow (https://github.com/WayScience/ome-arrow) is a project that applies Open Microscopy Environment (OME) conventions through Apache Arrow to integrate imaging data with modern analytical workflows. By representing images as Arrow-compatible structures alongside metadata and features, OME-Arrow enables programmatic and relational access using a consistent data model across languages while supporting familiar tools such as SQL engines, DuckDB, and Parquet-based pipelines.

The library supports ingestion from TIFF, OME-Zarr, and NumPy, with export to OME-Parquet, OME-Zarr, and OME-TIFF, along with lazy scan-style access for large datasets and tensor pathways for machine learning. OME-Arrow also integrates with napari-ome-arrow for visualization and CytoDataFrame for scalable feature-centric workflows, offering a modular, standards-aligned approach that complements the broader open bioimaging ecosystem.