itasc-tracking¶

Independent ITASC piece for nucleus tracking and interactive correction. It turns 2D+t foreground and contour maps into Ultrack candidate segments, a tracking database, solved tracks, and validated, corrected nucleus labels, all from a napari plugin.

Install¶

pip install itasc-tracking          # correction + browsing
pip install "itasc-tracking[solve]" # + the Ultrack solver (db build / track)

This pulls in itasc-core. Everything installs into the shared itasc.* namespace (PEP 420), so import itasc.tracking_ultrack works whether or not the full ITASC orchestrator is present. The Ultrack engine is only imported lazily when you actually run candidate generation / linking / solving, so correction-only use does not require the [solve] extra.

Use¶

napari plugin: launch napari and add the Ultrack Segment + Track widget. It exposes three path fields:
- Foreground: the foreground probability/intensity .tif (2D+t).
- Contours: the contour/boundary .tif (2D+t).
- Output dir: where every artifact is written (defaults to the foreground file’s folder when left blank).
The two inputs can have any name and live anywhere.

Standalone path contract:

<foreground.tif>            (in, any name/location)
<contours.tif>             (in, any name/location)
output_dir/
  atoms.tif                 (out)
  ultrack_workdir/data.db   (out)
  tracked_labels.tif        (out)
  validated_cells.json      (out)
  corrections.json          (out)

The full itasc orchestrator drives the same widget through its staged <pos>/1_cellpose + <pos>/2_nucleus layout instead; the piece supports both via itasc.napari._paths.NucleusWorkspace (files() / flat() vs staged()).

Backend¶

The headless backend lives under itasc.tracking_ultrack (candidate atoms, database build, linking, solving, export, correction primitives, and the validation_state validation/correction annotation store) and itasc.correction (label-edit helpers). Generic tracked-label-stack IO is shared substrate in itasc.core.label_store.