Tab Segmentation

Segmentation

Replace the native Xenium cell segmentation with a custom segmentation produced by a separate preprocessing pipeline, enabling re-analysis with alternative cell boundaries. This tab is in the "Tools" control panel group.

Controls

Control	Description
Active segmentation (read-only)	Displays the current segmentation — "Xenium (native)" or "Custom (from cache/file)" — along with the corresponding cell and gene counts.
Load Custom Segmentation...	Opens a file dialog for a custom_segmentation.h5ad file. If a cached version already exists in sdata_cached.zarr, prompts you to load from cache (fast) instead of re-importing the original file.
Revert to Xenium Segmentation	Restores the original Xenium cell labels and AnnData object. Disabled until a custom segmentation is loaded.
Update SpatialData on disk	Saves the current segmentation state (custom or native) back to sdata_cached.zarr.

Workflow

1. Produce a custom_segmentation.h5ad file using the palms-build-custom-segmentation pipeline. This step requires R and Seurat for cell boundary extraction; see the project README for details.
2. In the viewer, open the Segmentation tab and click "Load Custom Segmentation...".
3. Select the custom_segmentation.h5ad file. If a cached copy exists, choose whether to load from cache or re-import.
4. The viewer swaps to the custom segmentation; previously computed cluster-dependent analyses are cleared.
5. Re-run Leiden clustering and any downstream analyses on the new segmentation.
6. To revert, click "Revert to Xenium Segmentation".

Notes

• A custom segmentation is cached in sdata_cached.zarr on first load; subsequent loads use the cached version automatically.
• Building a custom segmentation is a two-stage pipeline. The boundaries are extracted first (typically in R/Seurat), then palms-build-custom-segmentation turns them into the label raster the viewer loads. Its output is a custom_labels.zarr that must stay beside the .h5ad you select here — selecting an .h5ad on its own fails with "not found alongside h5ad".
• Swapping segmentation resets active clusterings. You will need to re-run Leiden clustering on the new segmentation before cluster-dependent visualisations and analyses are available.