Outputs

Using the polygons generated during post-segmentation (see Post-Segmentation), we can now create the cells x genes mapping table and save the results into a new SpatialData Zarr file.

Cell x Gene Matrix

Assign transcript points to segmented cells and build an AnnData table:

• gdf_points – a GeoDataFrame of transcript points with a cell_ID column linking each point to a segmented cell.

• vdata – an AnnData object representing gene counts per cell.

def process_points_and_tables(self):
 """Assign transcript points to segmented cells and build AnnData table."""
 points = self.sdata.points[f"{self.fov}_points"]
 points_df = points.compute()

 self.gdf_points = gpd.GeoDataFrame(
     points_df,
     geometry=gpd.points_from_xy(points_df["x"], points_df["y"]),
     crs=self.gdf_polygons.crs,
 )

 joined = gpd.sjoin(
     self.gdf_points,
     self.gdf_polygons[["cell_id", "geometry"]],
     how="left",
     predicate="within",
 )
 self.gdf_points["cell_ID"] = joined["cell_id"].values

 all_genes = self.gdf_points["target"].unique().tolist()
 cell_gene_counts = {}

 for cell in self.gdf_points["cell_ID"].dropna().unique():
     cell_data = self.gdf_points[self.gdf_points["cell_ID"] == cell]
     gene_count = Counter(cell_data["target"])
     ordered_counts = OrderedDict((g, gene_count.get(g, 0)) for g in all_genes)
     cell_gene_counts[str(cell)] = ordered_counts

 df = pd.DataFrame.from_dict(cell_gene_counts, orient="index", columns=all_genes)
 df = df.reindex(columns=all_genes, fill_value=0)

 self.vdata = ad.AnnData(X=df.values)
 self.vdata.obs_names = df.index
 self.vdata.var_names = df.columns
 self.vdata.obs["cells"] = self.vdata.obs_names
 self.vdata.var["genes"] = self.vdata.var_names
 return self.gdf_points, self.vdata

Update SpatialData

Processed results are saved into a new SpatialData Zarr file. This includes:

• Original images (images)

• Labels (labels)

• Shapes (shapes)

• Original transcript points (points)

• Tables (vdata)

def update_spatialdata(self):
 """Save processed SpatialData into a new Zarr file."""
 self.process_masks_to_shapes()
 self.process_labels()
 self.process_points_and_tables()

 new_sdata = sd.SpatialData()
 dst_path = os.path.join(self.zarr_dir, f"{self.fov}.zarr")

 new_sdata.images[f"{self.fov}_image"] = self.sdata.images[f"{self.fov}_image"]
 new_sdata.labels[f"{self.fov}_labels"] = self.labels_model
 new_sdata.shapes[f"{self.fov}_shapes"] = self.shapes_model
 new_sdata.points[f"{self.fov}_points"] = self.sdata.points[f"{self.fov}_points"]
 new_sdata.tables = {'vdata': self.vdata}

 try:
     client = Client.current()
     client.close()
 except ValueError:
     pass

 new_sdata.write(dst_path, overwrite=True)

 for var in (self.masks, self.flows, self.styles):
     try:
         del var
     except Exception:
         pass
 gc.collect()

 return new_sdata