How To

Run segger segmentation

segger segment \
    -i /path/to/your/ist/data/ \
    -o /path/to/save/outputs/

-i is a standardized IST dataset directory (Xenium, CosMx, …); -o is where outputs are written. The main output is segger_segmentation.parquet, the per-transcript assignment table (see Outputs).

See all available options:

segger segment --help

Export to other formats

segger export turns segger_segmentation.parquet into other formats:

• anndata — cell x gene AnnData table with spatial coordinates

• boundaries — one polygon per cell, as GeoParquet

• transcripts — per-transcript table with assigned cell IDs

• spatialdata — adds segger’s outputs into an existing SpatialData store

For example, to write an AnnData table and cell boundaries:

segger export anndata boundaries \
    -s /path/to/save/outputs/segger_segmentation.parquet \
    -o /path/to/export/

See Outputs for a description of each file, or:

segger export --help

Use with SpatialData

If you want a SpatialData object instead of plain files (e.g. for squidpy, SOPA, or napari-spatialdata), the recommended workflow is:

1. Run segmentation as above.

2. Build a SpatialData object from your raw data. For Xenium, use spatialdata-io:

   import spatialdata_io
   
   sdata = spatialdata_io.xenium("/path/to/your/ist/data/")
   sdata.write("/path/to/sdata.zarr")
   

3. Add segger’s elements to that store:

   segger export spatialdata \
       -s /path/to/save/outputs/segger_segmentation.parquet \
       --sdata /path/to/sdata.zarr
   

This edits sdata.zarr in place: it appends segger’s per-transcript columns to the existing transcripts points element, and adds cell_boundaries_segger (shapes) and table_segger (table) elements. See Outputs for details.

Full walkthrough notebook

Runs through segmentation, export, loading the outputs, GPU clustering, and visualizing a spatial crop end to end.

• Notebook