itasc-aggregate 0.2.0.dev0

ITASC aggregate quantification: per-position quantifiers (density, shape, dynamics, contacts: cell-cell edges, T1 events) and napari visualization

itasc-aggregate

Independent ITASC piece for aggregate quantification. It hosts per-position quantifiers; the bundled one is contacts: extract cell-cell edges, border edges, and T1 events from 2D+t cell-label stacks (optionally validated against nucleus labels) into a self-describing HDF5 file, and visualize the result in napari.

Install

pip install itasc-aggregate

This pulls in itasc-core. Both install into the shared itasc.* namespace (PEP 420), so import itasc.contact_analysis works whether or not the full ITASC orchestrator is present.

Use

The napari plugin is the full ITASC catalog app restricted to the contact step: it does not run segmentation or tracking, so it treats the committed cell_labels.tif (required) and nucleus_labels.tif (optional) as inputs and produces the contact-analysis .h5.

• napari plugin: add the ITASC Aggregate widget.

  • Data folders: point Find at a parent directory to discover every position (a folder containing cell_labels.tif) beneath it; the columns are derived from each position's nesting under that directory. Each row carries a three-dot status rail (cell labels, nucleus labels, contact analysis) that shows, at a glance, which inputs are present and whether the .h5 is built. Click a rail dot to load that input into the viewer (or, for the contact dot, open the overlays).
  • Contact Analysis: select a row to retarget the stage to that position; run or re-run its contact analysis and visualize the result.
  • Aggregate: the project-level capstone pools across the whole catalog, grouped by your columns, once positions have been added.

• Headless / scripting:

  from itasc.contact_analysis import (
      ensure_contacts,        # build only if missing (or overwrite=True)
      discover_contact_batch_jobs,
      run_contact_batch,
  )
  
  # Single position (missing-only):
  ensure_contacts(
      cell_labels_path="cells.tif",
      nucleus_labels_path=None,            # optional
      output_path="contact_analysis.h5",
  )
  
  # Batch by name-based autodiscovery:
  jobs = discover_contact_batch_jobs(
      "/data/study",
      cell_name="cell_labels.tif",
      nucleus_name="nucleus_labels.tif",  # optional
      h5_name="contact_analysis.h5",
  )
  results = run_contact_batch(jobs, overwrite=False)
  

  build_contacts(...) remains available for an unconditional build.

I/O contract

• Input: cell_labels (2D+t TIFF, required); nucleus_labels (2D+t TIFF, optional: when given, the cell_id == nucleus_id invariant is enforced).
• Output: an HDF5 file with cells, edges, t1_events, and provenance.