Filter Operations on Label2DModel and Shape

src/spatialdata/__init__.py

@@ -53,6 +53,7 @@
     "deepcopy",
     "sanitize_table",
     "sanitize_name",
+    "subset_sdata_by_table_mask",
 ]
 
 from spatialdata import dataloader, datasets, models, transformations
@@ -76,6 +77,7 @@
     match_element_to_table,
     match_sdata_to_table,
     match_table_to_element,
+    subset_sdata_by_table_mask,
 )
 from spatialdata._core.query.spatial_query import bounding_box_query, polygon_query
 from spatialdata._core.spatialdata import SpatialData

src/spatialdata/_core/query/relational_query.py

@@ -9,9 +9,11 @@
 import dask.array as da
 import numpy as np
 import pandas as pd
+import xarray as xr
 from anndata import AnnData
 from dask.dataframe import DataFrame as DaskDataFrame
 from geopandas import GeoDataFrame
+from numpy.typing import NDArray
 from xarray import DataArray, DataTree
 
 from spatialdata._core.spatialdata import SpatialData
@@ -1017,3 +1019,124 @@ def get_values(
         return df
 
     raise ValueError(f"Unknown origin {origin}")
+
+
+def _mask_block(block: xr.DataArray, ids_to_remove: list[int]) -> xr.DataArray:
+    # Use apply_ufunc for efficient processing
+    # Create a copy to avoid modifying read-only array
+    result = block.copy()
+    result[np.isin(result, ids_to_remove)] = 0
+    return result
+
+
+def _set_instance_ids_in_labels_to_zero(image: xr.DataArray, ids_to_remove: list[int]) -> xr.DataArray:
+    processed = xr.apply_ufunc(
+        partial(_mask_block, ids_to_remove=ids_to_remove),
+        image,
+        input_core_dims=[["y", "x"]],
+        output_core_dims=[["y", "x"]],
+        vectorize=True,
+        dask="parallelized",
+        output_dtypes=[image.dtype],
+        dataset_fill_value=0,
+        dask_gufunc_kwargs={"allow_rechunk": True},
+    )
+
+    # Create a new DataArray to ensure persistence
+    return xr.DataArray(
+        data=processed.data,
+        coords=image.coords,
+        dims=image.dims,
+        attrs=image.attrs.copy(),  # Preserve all attributes
+    )
+
+
+def _get_scale_factors(labels_element: DataTree) -> list[tuple[float, float]]:
+    scales = list(labels_element.keys())
+
+    # Calculate relative scale factors between consecutive scales
+    scale_factors = []
+    for i in range(len(scales) - 1):
+        y_size_current = labels_element[scales[i]].image.shape[0]
+        x_size_current = labels_element[scales[i]].image.shape[1]
+        y_size_next = labels_element[scales[i + 1]].image.shape[0]
+        x_size_next = labels_element[scales[i + 1]].image.shape[1]
+        y_factor = y_size_current / y_size_next
+        x_factor = x_size_current / x_size_next
+
+        scale_factors.append((y_factor, x_factor))
+
+    return scale_factors
+
+
+@singledispatch
+def _filter_by_instance_ids(element: Any, ids_to_remove: list[str], instance_key: str) -> Any:
+    raise NotImplementedError(f"Filtering by instance ids is not implemented for {element}")
+
+
+@_filter_by_instance_ids.register(DataArray)
+def _(element: DataArray, ids_to_remove: list[int], instance_key: str) -> DataArray:
+    del instance_key
+    return Labels2DModel.parse(_set_instance_ids_in_labels_to_zero(element, ids_to_remove))
+
+
+@_filter_by_instance_ids.register(DataTree)
+def _(element: DataTree, ids_to_remove: list[int], instance_key: str) -> DataTree:
+    # we extract the info to just reconstruct
+    # the DataTree after filtering the max scale
+    del instance_key
+    max_scale = list(element.keys())[0]
+    scale_factors_temp = _get_scale_factors(element)
+    scale_factors = [int(sf[0]) for sf in scale_factors_temp]
+
+    return Labels2DModel.parse(
+        data=_set_instance_ids_in_labels_to_zero(element[max_scale].image, ids_to_remove),
+        scale_factors=scale_factors,
+    )
+
+
+def subset_sdata_by_table_mask(sdata: SpatialData, table_name: str, mask: NDArray[np.bool_]) -> SpatialData:
+    """Subset the annotated elements of a SpatialData object by a table and a mask.
+
+    The mask is applied to the table and the annotated elements are subsetted
+    by the instance ids in the table.
+    This function returns a new SpatialData object with the subsetted elements.
+    Elements that are not annotated by the table are not included in the returned SpatialData object.
+    The element models that are
+    supported are :class:`spatialdata.models.Labels2DModel`,
+    :class:`spatialdata.models.PointsModel`, and :class:`spatialdata.models.ShapesModel`.
+
+    Parameters
+    ----------
+    sdata
+        The SpatialData object to subset.
+    table_name
+        The name of the table to apply the mask to.
+    mask
+        Boolean mask to apply to the table which is the same length as the number of rows in the table.
+
+    Returns
+    -------
+    The subsetted SpatialData object.
+    """
+    table = sdata.tables.get(table_name)
+    if table is None:
+        raise ValueError(f"Table {table_name} not found in SpatialData object.")
+
+    subset_table = table[mask]
+    sdata.tables[table_name] = subset_table
+    _, _, instance_key = get_table_keys(subset_table)
+    annotated_regions = SpatialData.get_annotated_regions(table)
+    removed_instance_ids = list(np.unique(table.obs[instance_key][~mask]))
+
+    filtered_elements = {}
+    for reg in annotated_regions:
+        elem = sdata.get(reg)
+        model = get_model(elem)
+        if model is Labels2DModel:
+            filtered_elements[reg] = _filter_by_instance_ids(elem, removed_instance_ids, instance_key)
+        elif model in [PointsModel, ShapesModel]:
+            element_dict, _ = match_element_to_table(sdata, element_name=reg, table_name=table_name)
+            filtered_elements[reg] = element_dict[reg]
+
+    return SpatialData.init_from_elements(filtered_elements | {table_name: subset_table})

tests/core/query/test_relational_query_subset_sdata_by_table_mask.py

@@ -0,0 +1,80 @@
+import numpy as np
+import pytest
+
+from spatialdata import concatenate, subset_sdata_by_table_mask
+from spatialdata._core.query.relational_query import _filter_by_instance_ids
+from spatialdata.datasets import blobs_annotating_element
+
+
+def test_filter_labels2dmodel_by_instance_ids() -> None:
+    sdata = blobs_annotating_element("blobs_labels")
+    labels_element = sdata["blobs_labels"]
+    all_instance_ids = sdata.tables["table"].obs["instance_id"].unique()
+    filtered_labels_element = _filter_by_instance_ids(labels_element, [2, 3], "instance_id")
+
+    # because 0 is the background, we expect the filtered ids to be the instance ids that are not 0
+    filtered_ids = set(np.unique(filtered_labels_element.data.compute())) - {
+        0,
+    }
+    preserved_ids = np.unique(labels_element.data.compute())
+    assert filtered_ids == (set(all_instance_ids) - {2, 3})
+    # check if there is modification of the original labels
+    assert set(preserved_ids) == set(all_instance_ids) | {0}
+
+    sdata.tables["table"].uns["spatialdata_attrs"]["region"] = "blobs_multiscale_labels"
+    sdata.tables["table"].obs.region = "blobs_multiscale_labels"
+    labels_element = sdata["blobs_multiscale_labels"]
+    filtered_labels_element = _filter_by_instance_ids(labels_element, [2, 3], "instance_id")
+
+    for scale in labels_element:
+        filtered_ids = set(np.unique(filtered_labels_element[scale].image.compute())) - {
+            0,
+        }
+        preserved_ids = np.unique(labels_element[scale].image.compute())
+        assert filtered_ids == (set(all_instance_ids) - {2, 3})
+        # check if there is modification of the original labels
+        assert set(preserved_ids) == set(all_instance_ids) | {0}
+
+
+def test_subset_sdata_by_table_mask() -> None:
+    sdata = concatenate(
+        {
+            "labels": blobs_annotating_element("blobs_labels"),
+            "shapes": blobs_annotating_element("blobs_circles"),
+            "points": blobs_annotating_element("blobs_points"),
+            "multiscale_labels": blobs_annotating_element("blobs_multiscale_labels"),
+        },
+        concatenate_tables=True,
+    )
+    third_elems = sdata.tables["table"].obs["instance_id"] == 3
+    subset_sdata = subset_sdata_by_table_mask(sdata, "table", third_elems)
+
+    assert set(subset_sdata.labels.keys()) == {"blobs_labels-labels", "blobs_multiscale_labels-multiscale_labels"}
+    assert set(subset_sdata.points.keys()) == {"blobs_points-points"}
+    assert set(subset_sdata.shapes.keys()) == {"blobs_circles-shapes"}
+
+    labels_remaining_ids = set(np.unique(subset_sdata.labels["blobs_labels-labels"].data.compute())) - {0}
+    assert labels_remaining_ids == {3}
+
+    for scale in subset_sdata.labels["blobs_multiscale_labels-multiscale_labels"]:
+        ms_labels_remaining_ids = set(
+            np.unique(subset_sdata.labels["blobs_multiscale_labels-multiscale_labels"][scale].image.compute())
+        ) - {0}
+        assert ms_labels_remaining_ids == {3}
+
+    points_remaining_ids = set(np.unique(subset_sdata.points["blobs_points-points"].index)) - {0}
+    assert points_remaining_ids == {3}
+
+    shapes_remaining_ids = set(np.unique(subset_sdata.shapes["blobs_circles-shapes"].index)) - {0}
+    assert shapes_remaining_ids == {3}
+
+
+def test_subset_sdata_by_table_mask_with_no_annotated_elements() -> None:
+    with pytest.raises(ValueError, match="Table table_not_found not found in SpatialData object."):
+        sdata = blobs_annotating_element("blobs_labels")
+        _ = subset_sdata_by_table_mask(sdata, "table_not_found", sdata.tables["table"].obs["instance_id"] == 3)
+
+
+def test_filter_by_instance_ids_fails_for_unsupported_element_models() -> None:
+    with pytest.raises(NotImplementedError, match="Filtering by instance ids is not implemented for"):
+        _filter_by_instance_ids([1, 1, 1, 2], [1], "instance_id")