Skip to content

Share via the BioImage Archive

Annotations are most useful when others can reuse them. The BioImage Archive accepts annotation datasets described with the MIFA metadata standard. OMERO Annotate AI writes a ready-to-upload bundle: images, masks, file lists and the MIFA metadata files.

This needs the BioImage Archive's metadata models, which are not on PyPI:

pip install "git+https://github.com/BioImage-Archive/bia-mifa-models.git@v1.1.5"

1. Describe the study

The config has three metadata sections used for the archive: study, dataset and annotation_methodology. Fill them in the YAML before annotating, or set them just before export:

study:
  title: "Nuclei annotations in HeLa Cell Painting images"
  description: "Instance masks of nuclei, annotated with micro-SAM and corrected by hand."
  keywords: [nuclei, segmentation, cell painting]
  organism: "Homo sapiens"
  imaging_method: "widefield fluorescence microscopy"
  funding_statement: "Funded by NWO."
dataset:
  source_description: "HeLa cells, Hoechst channel"
  license: "CC-BY-4.0"           # CC0 or CC-BY
annotation_methodology:
  annotation_type: segmentation_mask
  annotation_method: semi_automatic
  annotation_criteria: "Every nucleus fully inside the field; touching nuclei separated."
  annotation_coverage: representative

2. Export from OMERO

from omero_annotate_ai import prepare_bia_data_from_table
from omero_annotate_ai.core.mifa_export import save_bia_package
from omero_annotate_ai.trainingset import read_set_config

config = read_set_config(table_id, conn)  # the set's config, with the study metadata
stats = prepare_bia_data_from_table(conn, table_id, "bia_staging")

summary = save_bia_package(
    config, stats["rows"], "bia_staging", "bia_bundle", move_data=True
)

prepare_bia_data_from_table downloads the finished units in their original bit depth; save_bia_package writes the bundle.

From a local training set

from omero_annotate_ai import TrainingSet
from omero_annotate_ai.core.mifa_export import save_bia_package

ts = TrainingSet.open("~/omero_annotate_ai/local/<id>_hela_nuclei_v1")
save_bia_package(
    ts.config, ts.units(), ts.bundle_dir, "bia_bundle"
)  # annotated units only

What is in the bundle

bia_bundle/
├── metadata/
│   ├── Study_<accession>.yaml
│   ├── Annotations_<accession>.yaml
│   └── Version_<accession>.yaml
├── file_list_images.tsv
├── file_list_annotations.tsv
├── images/<unit id>.tif
└── annotations/<unit id>_mask.tif

Upload it following the BioImage Archive submission guide.

Notebook: OMERO table to BioImage Archive