# Curate in Mastodon

Automatic tracking is never perfect, so tracks are corrected by hand in [Mastodon](https://mastodon.readthedocs.io/), a Fiji tool for editing tracks. The image comes straight from the [local OME-Zarr cache](../user-guide/caching.md); the tracks come from a Cell Tracking Challenge (CTC) bundle exported by the Tracks widget.

> **Important: Important**
>
> **Use the CTC importer, not the CSV importer.** Mastodon's CSV importer creates spots but **no links**, so it silently drops every track and lineage. Its label-image importer cannot create division links. Only the [Cell Tracking Challenge](https://celltrackingchallenge.net/) importer rebuilds full lineages, divisions included, because it reads the parent of each track from `res_track.txt`.


# 1. Export the bundle

Cache the plate and open the well in napari, then click **Export well for Mastodon (CTC)** in the [Tracks widget](../napari/user-guide/tracks_widget.md#export-well-for-mastodon-ctc). It writes `~/mastodon_exports/plate_<id>_<well>_ctc/` containing:

- `mask000.tif` … `mask<T-1>.tif`: one nucleus label image per frame, read from the cache. Each cell's pixel value is its CTC track label.
- `res_track.txt`: the lineage table, one line `L B E P` per track (label, first frame, last frame, parent label; `0` for a founder). Labels are renumbered `1..N` in order of first frame, matching the masks.
- `manifest.json`: maps each CTC label back to its CellView `track_id`, with per-frame centroids. Keep it and don't edit it.
- `README.txt`: these import steps with your paths filled in.

Before a long curation session, **Pin** the plate in the Tracks widget so the cache does not [evict](../user-guide/caching.md#eviction-and-pinning) it, and **Unpin** it afterwards.


# 2. Open the image

In Fiji:

1.  **Plugins → Tracking → Mastodon → new from OME-NGFF…**

2.  Paste the path to the well's image group, the multiscale `0` group inside the plate store:

    ``` text
    ~/omero-cache/zarr/plate_<id>.zarr/<row>/<col>/0
    ```

    for example `~/omero-cache/zarr/plate_4155.zarr/B/2/0`. Don't point at the plate root or the well folder.

3.  Click **Detect datasets**, select the listed row, click **OK**, and save the BDV XML when asked.

4.  The image opens in BigDataViewer. `P` opens the side panel for per-channel contrast, `1` and `2` switch channels, and `F` toggles a fused overlay.


# 3. Import the tracks

In the main Mastodon window choose **File → Import → Import from CellTrackingChallenge**. Select the bundle folder and set the file pattern to `mask%03d.tif` (three digits, starting at 0). Mastodon reads the parent and division links from `res_track.txt`.


# 4. Curate

Click the same group-lock number (e.g. `1`) in both the BigDataViewer and TrackScheme windows; double-clicking a spot then moves both views to it. Save your work as a `.mastodon` project **outside** the cache, so it survives the plate being evicted.


# Bringing corrections back (planned)

Export the corrected tracks with **File → Export → CellTrackingChallenge**. Writing them back into CellView is not implemented yet. The plan is to map each corrected spot onto its CellView row by per-frame centroid, using `manifest.json` as the anchor, and write the result to `track_id` / `parent_track_id`, leaving the `_raw` columns as the audit trail.

Until then, the corrected CTC export (kept with its `manifest.json`) and the `.mastodon` project are the record of your corrections. On PIP-FUCCI plates, the [napari review workflow](review.md) already writes corrections back.