spatialdata-io 0.1.4__tar.gz → 0.1.5__tar.gz

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/CHANGELOG.md

@@ -8,7 +8,18 @@ and this project adheres to [Semantic Versioning][].
 [keep a changelog]: https://keepachangelog.com/en/1.0.0/
 [semantic versioning]: https://semver.org/spec/v2.0.0.html
 
-## [0.1.4] - xxxx-xx-xx
+## [0.1.5] - 2024-09-25
+
+### Added
+
+-   (Xenium) added `dims` parameter for more control in `xenium_aligned_image()`
+
+### Fixed
+
+-   Passing `rgb=None` to image model parser for both visium and visiumhd, leading to 3-4 channel images being
+    interpreted as RGB(A)
+-   Fix header bug Visium data #200
+-   (Visium HD) Fix path parsing when images are missing #204 #206
 
 ## [0.1.4] - 2024-08-07
 

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: spatialdata-io
-Version: 0.1.4
+Version: 0.1.5
 Summary: SpatialData IO for common techs
 Project-URL: Documentation, https://spatialdata-io.readthedocs.io/
 Project-URL: Source, https://github.com/scverse/spatialdata-io
@@ -48,7 +48,7 @@ Requires-Dist: pyarrow
 Requires-Dist: readfcs
 Requires-Dist: scanpy
 Requires-Dist: scikit-image
-Requires-Dist: spatialdata>=0.2.1
+Requires-Dist: spatialdata>=0.2.2
 Requires-Dist: tifffile>=2023.8.12
 Provides-Extra: dev
 Requires-Dist: bump2version; extra == 'dev'

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/_version.py

@@ -12,5 +12,5 @@ __version__: str
 __version_tuple__: VERSION_TUPLE
 version_tuple: VERSION_TUPLE
 
-__version__ = version = '0.1.4'
-__version_tuple__ = version_tuple = (0, 1, 4)
+__version__ = version = '0.1.5'
+__version_tuple__ = version_tuple = (0, 1, 5)

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/pyproject.toml

@@ -25,7 +25,7 @@ dependencies = [
     "anndata",
     "numpy",
     "scanpy",
-    "spatialdata>=0.2.1",
+    "spatialdata>=0.2.2",
     "scikit-image",
     "h5py",
     "joblib",

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/src/spatialdata_io/_constants/_constants.py

@@ -353,9 +353,9 @@ class VisiumHDKeys(ModeEnum):
     TISSUE_POSITIONS_FILE = "tissue_positions.parquet"
 
     # images
-    IMAGE_HIRES_FILE = "spatial/tissue_hires_image.png"
-    IMAGE_LOWRES_FILE = "spatial/tissue_lowres_image.png"
-    IMAGE_CYTASSIST = "spatial/cytassist_image.tiff"
+    IMAGE_HIRES_FILE = "tissue_hires_image.png"
+    IMAGE_LOWRES_FILE = "tissue_lowres_image.png"
+    IMAGE_CYTASSIST = "cytassist_image.tiff"
 
     # scalefactors
     SCALEFACTORS_FILE = "scalefactors_json.json"

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/src/spatialdata_io/readers/_utils/_utils.py

@@ -18,8 +18,8 @@ try:
 
     NDArrayA = NDArray[Any]
 except (ImportError, TypeError):
-    NDArray = np.ndarray  # type: ignore[misc]
-    NDArrayA = np.ndarray  # type: ignore[misc]
+    NDArray = np.ndarray
+    NDArrayA = np.ndarray
 
 
 def _read_counts(

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/src/spatialdata_io/readers/codex.py

@@ -80,6 +80,7 @@ def codex(
             "images": Image2DModel.parse(
                 image,
                 scale_factors=[2, 2],
+                rgb=None,
             )
         }
         sdata = SpatialData(images=images, shapes={str(region): shapes}, table=table)

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/src/spatialdata_io/readers/cosmx.py

@@ -200,6 +200,7 @@ def cosmx(
                         "global_only_image": aff,
                     },
                     dims=("y", "x", "c"),
+                    rgb=None,
                     **image_models_kwargs,
                 )
                 images[f"{fov}_image"] = parsed_im

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/src/spatialdata_io/readers/mcmicro.py

@@ -100,7 +100,7 @@ def mcmicro(
         assert len(samples) == 1
         data = imread(samples[0], **imread_kwargs)
         # , scale_factors=[2, 2]
-        data = Image2DModel.parse(data, transformations=_get_transformation(), **image_models_kwargs)
+        data = Image2DModel.parse(data, transformations=_get_transformation(), rgb=None, **image_models_kwargs)
         images["tma_map"] = data
 
         image_dir = path / McmicroKeys.IMAGES_DIR_TMA
@@ -116,7 +116,7 @@ def mcmicro(
             image_id = core_id
 
         data = imread(sample, **imread_kwargs)
-        data = Image2DModel.parse(data, c_coords=marker_names, **image_models_kwargs)
+        data = Image2DModel.parse(data, c_coords=marker_names, rgb=None, **image_models_kwargs)
         transformations = _get_transformation(
             tma=int(core_id) if tma else None, tma_centroids=tma_centroids, raster_data=data
         )
@@ -131,7 +131,9 @@ def mcmicro(
             raw_name = raw_image.with_name(raw_image.stem).with_suffix("").stem
 
             data = imread(raw_image, **imread_kwargs)
-            images[raw_name] = Image2DModel.parse(data, transformations={raw_name: Identity()}, **image_models_kwargs)
+            images[raw_name] = Image2DModel.parse(
+                data, transformations={raw_name: Identity()}, rgb=None, **image_models_kwargs
+            )
 
     illumination_dir = path / McmicroKeys.ILLUMINATION_DIR
     if illumination_dir.exists():
@@ -143,7 +145,7 @@ def mcmicro(
 
             data = imread(illumination_image, **imread_kwargs)
             images[illumination_name] = Image2DModel.parse(
-                data, transformations={raw_name: Identity()}, **image_models_kwargs
+                data, transformations={raw_name: Identity()}, rgb=None, **image_models_kwargs
             )
 
     samples_labels = list((path / McmicroKeys.LABELS_DIR).glob("*/*" + McmicroKeys.IMAGE_SUFFIX))

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/src/spatialdata_io/readers/merscope.py

@@ -273,7 +273,7 @@ def _rioxarray_load_merscope(
         dim="c",
     )
 
-    return Image2DModel.parse(im, c_coords=stainings, **image_models_kwargs)
+    return Image2DModel.parse(im, c_coords=stainings, rgb=None, **image_models_kwargs)
 
 
 def _dask_image_load_merscope(
@@ -292,6 +292,7 @@ def _dask_image_load_merscope(
         im,
         dims=("c", "y", "x"),
         c_coords=stainings,
+        rgb=None,
         **image_models_kwargs,
     )
 

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/src/spatialdata_io/readers/steinbock.py

@@ -97,7 +97,7 @@ def _get_images(
     image_models_kwargs: Mapping[str, Any] = MappingProxyType({}),
 ) -> Union[SpatialImage, MultiscaleSpatialImage]:
     image = imread(path / SteinbockKeys.IMAGES_DIR / f"{sample}{SteinbockKeys.IMAGE_SUFFIX}", **imread_kwargs)
-    return Image2DModel.parse(data=image, transformations={sample: Identity()}, **image_models_kwargs)
+    return Image2DModel.parse(data=image, transformations={sample: Identity()}, rgb=None, **image_models_kwargs)
 
 
 def _get_labels(

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/src/spatialdata_io/readers/visium.py

@@ -4,7 +4,6 @@ import json
 import os
 import re
 from collections.abc import Mapping
-from functools import partial
 from pathlib import Path
 from types import MappingProxyType
 from typing import Any
@@ -138,7 +137,6 @@ def visium(
     if (path / "spatial" / VisiumKeys.SPOTS_FILE_1).exists() or (
         tissue_positions_file is not None and str(VisiumKeys.SPOTS_FILE_1) in str(tissue_positions_file)
     ):
-        read_coords = partial(pd.read_csv, header=None, index_col=0)
         tissue_positions_file = (
             path / "spatial" / VisiumKeys.SPOTS_FILE_1
             if tissue_positions_file is None
@@ -147,7 +145,6 @@ def visium(
     elif (path / "spatial" / VisiumKeys.SPOTS_FILE_2).exists() or (
         tissue_positions_file is not None and str(VisiumKeys.SPOTS_FILE_2) in str(tissue_positions_file)
     ):
-        read_coords = partial(pd.read_csv, header=0, index_col=0)
         tissue_positions_file = (
             path / "spatial" / VisiumKeys.SPOTS_FILE_2
             if tissue_positions_file is None
@@ -155,11 +152,18 @@ def visium(
         )
     else:
         raise ValueError(f"Cannot find `tissue_positions` file in `{path}`.")
-    coords = read_coords(tissue_positions_file)
+    coords = pd.read_csv(tissue_positions_file, header=None, index_col=0)
 
-    # to handle spaceranger_version < 2.0.0, where no column names are provided
-    if "in_tissue" not in coords.columns:
-        coords.columns = ["in_tissue", "array_row", "array_col", "pxl_row_in_fullres", "pxl_col_in_fullres"]
+    if tissue_positions_file.name == VisiumKeys.SPOTS_FILE_1:
+        # spaceranger_version < 2.0.0 but header detected: https://github.com/scverse/spatialdata-io/issues/146
+        if "in_tissue" in coords.iloc[0].tolist():
+            coords = pd.read_csv(tissue_positions_file, header=0, index_col=0)
+        else:
+            assert "in_tissue" not in coords.columns
+            coords.columns = ["in_tissue", "array_row", "array_col", "pxl_row_in_fullres", "pxl_col_in_fullres"]
+    else:
+        assert tissue_positions_file.name == VisiumKeys.SPOTS_FILE_2
+        coords = pd.read_csv(tissue_positions_file, header=0, index_col=0)
 
     adata.obs = pd.merge(adata.obs, coords, how="left", left_index=True, right_index=True)
     coords = adata.obs[[VisiumKeys.SPOTS_X, VisiumKeys.SPOTS_Y]].values
@@ -217,6 +221,7 @@ def visium(
                 full_image,
                 scale_factors=[2, 2, 2, 2],
                 transformations={"global": transform_original},
+                rgb=None,
                 **image_models_kwargs,
             )
         else:
@@ -226,13 +231,15 @@ def visium(
         image_hires = imread(path / VisiumKeys.IMAGE_HIRES_FILE, **imread_kwargs).squeeze().transpose(2, 0, 1)
         image_hires = DataArray(image_hires, dims=("c", "y", "x"))
         images[dataset_id + "_hires_image"] = Image2DModel.parse(
-            image_hires, transformations={"downscaled_hires": Identity()}
+            image_hires, transformations={"downscaled_hires": Identity()}, rgb=None
         )
     if (path / VisiumKeys.IMAGE_LOWRES_FILE).exists():
         image_lowres = imread(path / VisiumKeys.IMAGE_LOWRES_FILE, **imread_kwargs).squeeze().transpose(2, 0, 1)
         image_lowres = DataArray(image_lowres, dims=("c", "y", "x"))
         images[dataset_id + "_lowres_image"] = Image2DModel.parse(
-            image_lowres, transformations={"downscaled_lowres": Identity()}
+            image_lowres,
+            transformations={"downscaled_lowres": Identity()},
+            rgb=None,
         )
 
     return SpatialData(images=images, shapes=shapes, table=table)

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/src/spatialdata_io/readers/visium_hd.py

@@ -86,15 +86,14 @@ def visium_hd(
     SpatialData object for the Visium HD data.
     """
     path = Path(path)
+    all_files = [file for file in path.rglob("*") if file.is_file()]
     tables = {}
     shapes = {}
     images: dict[str, Any] = {}
 
     if dataset_id is None:
         dataset_id = _infer_dataset_id(path)
-        filename_prefix = f"{dataset_id}_"
-    else:
-        filename_prefix = ""
+    filename_prefix = f"{dataset_id}_"
 
     def load_image(path: Path, suffix: str, scale_factors: list[int] | None = None) -> None:
         _load_image(
@@ -127,8 +126,12 @@ def visium_hd(
             ]
         )
 
-    if VisiumHDKeys.BINNED_OUTPUTS in os.listdir(path):
-        path_bins = path / VisiumHDKeys.BINNED_OUTPUTS
+    all_path_bins = [path_bin for path_bin in all_files if VisiumHDKeys.BINNED_OUTPUTS in str(path_bin)]
+    if len(all_path_bins) != 0:
+        path_bins_parts = all_path_bins[
+            -1
+        ].parts  # just choosing last one here as users might have tar file which would be first
+        path_bins = Path(*path_bins_parts[: path_bins_parts.index(VisiumHDKeys.BINNED_OUTPUTS) + 1])
     else:
         path_bins = path
     all_bin_sizes = _get_bins(path_bins)
@@ -266,15 +269,26 @@ def visium_hd(
             fullres_image_filenames = [
                 f for f in os.listdir(path_fullres) if os.path.isfile(os.path.join(path_fullres, f))
             ]
-            if len(fullres_image_filenames) > 1:
-                warnings.warn(
-                    f"Multiple files found in {path_fullres}, using the first one: {fullres_image_filenames[0]}. Please"
-                    " specify the path to the full resolution image manually using the `fullres_image_file` argument.",
-                    UserWarning,
-                    stacklevel=2,
-                )
-            fullres_image_filename = fullres_image_filenames[0]
-            fullres_image_file = path_fullres / fullres_image_filename
+            fullres_image_paths = [path_fullres / image_filename for image_filename in fullres_image_filenames]
+        elif list((path_fullres := (path / f"{filename_prefix}tissue_image")).parent.glob(f"{path_fullres.name}.*")):
+            fullres_image_paths = list(path_fullres.parent.glob(f"{path_fullres.name}.*"))
+        else:
+            fullres_image_paths = []
+        if len(fullres_image_paths) > 1:
+            warnings.warn(
+                f"Multiple files found in {path_fullres}, using the first one: {fullres_image_paths[0].stem}. Please"
+                " specify the path to the full resolution image manually using the `fullres_image_file` argument.",
+                UserWarning,
+                stacklevel=2,
+            )
+        if len(fullres_image_paths) == 0:
+            warnings.warn(
+                "No full resolution image found. If incorrect, please specify the path in the "
+                "`fullres_image_file` parameter when calling the `visium_hd` reader function.",
+                UserWarning,
+                stacklevel=2,
+            )
+        fullres_image_file = fullres_image_paths[0] if len(fullres_image_paths) > 0 else None
 
     if fullres_image_file is not None:
         load_image(
@@ -284,8 +298,15 @@ def visium_hd(
         )
 
     # hires image
+    hires_image_path = [path for path in all_files if VisiumHDKeys.IMAGE_HIRES_FILE in str(path)]
+    if len(hires_image_path) == 0:
+        warnings.warn(
+            f"No image path found containing the hires image: {VisiumHDKeys.IMAGE_HIRES_FILE}",
+            UserWarning,
+            stacklevel=2,
+        )
     load_image(
-        path=path / VisiumHDKeys.IMAGE_HIRES_FILE,
+        path=hires_image_path[0],
         suffix="_hires_image",
     )
     set_transformation(
@@ -295,8 +316,15 @@ def visium_hd(
     )
 
     # lowres image
+    lowres_image_path = [path for path in all_files if VisiumHDKeys.IMAGE_LOWRES_FILE in str(path)]
+    if len(lowres_image_path) == 0:
+        warnings.warn(
+            f"No image path found containing the lowres image: {VisiumHDKeys.IMAGE_LOWRES_FILE}",
+            UserWarning,
+            stacklevel=2,
+        )
     load_image(
-        path=path / VisiumHDKeys.IMAGE_LOWRES_FILE,
+        path=lowres_image_path[0],
         suffix="_lowres_image",
     )
     set_transformation(
@@ -306,9 +334,16 @@ def visium_hd(
     )
 
     # cytassist image
+    cytassist_path = [path for path in all_files if VisiumHDKeys.IMAGE_CYTASSIST in str(path)]
+    if len(cytassist_path) == 0:
+        warnings.warn(
+            f"No image path found containing the cytassist image: {VisiumHDKeys.IMAGE_CYTASSIST}",
+            UserWarning,
+            stacklevel=2,
+        )
     if load_all_images:
         load_image(
-            path=path / VisiumHDKeys.IMAGE_CYTASSIST,
+            path=cytassist_path[0],
             suffix="_cytassist_image",
         )
         image = images[dataset_id + "_cytassist_image"]
@@ -352,7 +387,7 @@ def _load_image(
             # dask_image doesn't recognize .btf automatically and imageio v3 throws error due to pixel limit -> use imageio v2
             data = imread2(path, **imread_kwargs).squeeze().transpose(2, 0, 1)
         image = DataArray(data, dims=("c", "y", "x"))
-        parsed = Image2DModel.parse(image, scale_factors=scale_factors, **image_models_kwargs)
+        parsed = Image2DModel.parse(image, scale_factors=scale_factors, rgb=None, **image_models_kwargs)
         images[dataset_id + suffix] = parsed
     else:
         warnings.warn(f"File {path} does not exist, skipping it.", UserWarning, stacklevel=2)

{spatialdata_io-0.1.4 → spatialdata_io-0.1.5}/src/spatialdata_io/readers/xenium.py

@@ -22,12 +22,11 @@ import zarr
 from anndata import AnnData
 from dask.dataframe import read_parquet
 from dask_image.imread import imread
+from datatree.datatree import DataTree
 from geopandas import GeoDataFrame
 from joblib import Parallel, delayed
-from multiscale_spatial_image.multiscale_spatial_image import MultiscaleSpatialImage
 from pyarrow import Table
 from shapely import Polygon
-from spatial_image import SpatialImage
 from spatialdata import SpatialData
 from spatialdata._core.query.relational_query import get_element_instances
 from spatialdata._types import ArrayLike
@@ -39,6 +38,7 @@ from spatialdata.models import (
     TableModel,
 )
 from spatialdata.transformations.transformations import Affine, Identity, Scale
+from xarray import DataArray
 
 from spatialdata_io._constants._constants import XeniumKeys
 from spatialdata_io._docs import inject_docs
@@ -111,7 +111,9 @@ def xenium(
     morphology_focus
         Whether to read the morphology focus image.
     aligned_images
-        Whether to also parse, when available, additional H&E or IF aligned images.
+        Whether to also parse, when available, additional H&E or IF aligned images. For more control over the aligned
+        images being read, in particular, to specify the axes of the aligned images, please set this parameter to
+        `False` and use the `xenium_aligned_image` function directly.
     cells_table
         Whether to read the cell annotations in the `AnnData` table.
     n_jobs
@@ -513,6 +515,7 @@ def _get_points(path: Path, specs: dict[str, Any]) -> Table:
         feature_key=XeniumKeys.FEATURE_NAME,
         instance_key=XeniumKeys.CELL_ID,
         transformations={"global": transform},
+        sort=True,
     )
     return points
 
@@ -550,7 +553,7 @@ def _get_images(
     file: str,
     imread_kwargs: Mapping[str, Any] = MappingProxyType({}),
     image_models_kwargs: Mapping[str, Any] = MappingProxyType({}),
-) -> SpatialImage | MultiscaleSpatialImage:
+) -> DataArray | DataTree:
     image = imread(path / file, **imread_kwargs)
     if "c_coords" in image_models_kwargs and "dummy" in image_models_kwargs["c_coords"]:
         # Napari currently interprets 4 channel images as RGB; a series of PRs to fix this is almost ready but they will
@@ -559,7 +562,7 @@ def _get_images(
         # let's add a dummy channel as a temporary workaround.
         image = da.concatenate([image, da.zeros_like(image[0:1])], axis=0)
     return Image2DModel.parse(
-        image, transformations={"global": Identity()}, dims=("c", "y", "x"), **image_models_kwargs
+        image, transformations={"global": Identity()}, dims=("c", "y", "x"), rgb=None, **image_models_kwargs
     )
 
 
@@ -567,7 +570,7 @@ def _add_aligned_images(
     path: Path,
     imread_kwargs: Mapping[str, Any] = MappingProxyType({}),
     image_models_kwargs: Mapping[str, Any] = MappingProxyType({}),
-) -> dict[str, MultiscaleSpatialImage]:
+) -> dict[str, DataTree]:
     """Discover and parse aligned images."""
     images = {}
     ome_tif_files = list(path.glob("*.ome.tif"))
@@ -598,7 +601,8 @@ def xenium_aligned_image(
     alignment_file: str | Path | None,
     imread_kwargs: Mapping[str, Any] = MappingProxyType({}),
     image_models_kwargs: Mapping[str, Any] = MappingProxyType({}),
-) -> MultiscaleSpatialImage:
+    dims: tuple[str, ...] | None = None,
+) -> DataTree:
     """
     Read an image aligned to a Xenium dataset, with an optional alignment file.
 
@@ -610,6 +614,12 @@ def xenium_aligned_image(
         Path to the alignment file, if not passed it is assumed that the image is aligned.
     image_models_kwargs
         Keyword arguments to pass to the image models.
+    dims
+        Dimensions of the image (tuple of axes names); valid strings are "c", "x" and "y". If not passed, the function
+        will try to infer the dimensions from the image shape. Please use this argument when the default behavior fails.
+        Example: for an image with shape (1, y, 1, x, 3), use dims=("anystring", "y", "dummy", "x", "c"). Values that
+        are not "c", "x" or "y" are considered dummy dimensions and will be squeezed (the data must have len 1 for
+        those axes).
 
     Returns
     -------
@@ -627,19 +637,28 @@ def xenium_aligned_image(
     # In fact, it could be that the len(image.shape) == 4 has actually dimes (1, x, y, c) and not (1, y, x, c). This is
     # not a problem because the transformation is constructed to be consistent, but if is the case, the data orientation
     # would be transposed compared to the original image, not ideal.
-    if len(image.shape) == 4:
-        assert image.shape[0] == 1
-        assert image.shape[-1] == 3
-        image = image.squeeze(0)
-        dims = ("y", "x", "c")
+    if dims is None:
+        if len(image.shape) == 4:
+            assert image.shape[0] == 1
+            assert image.shape[-1] == 3
+            image = image.squeeze(0)
+            dims = ("y", "x", "c")
+        else:
+            assert len(image.shape) == 3
+            assert image.shape[0] in [3, 4]
+            if image.shape[0] == 4:
+                # as explained before in _get_images(), we need to add a dummy channel until we support 4-channel images as
+                # non-RGBA images in napari
+                image = da.concatenate([image, da.zeros_like(image[0:1])], axis=0)
+            dims = ("c", "y", "x")
     else:
-        assert len(image.shape) == 3
-        assert image.shape[0] in [3, 4]
-        if image.shape[0] == 4:
-            # as explained before in _get_images(), we need to add a dummy channel until we support 4-channel images as
-            # non-RGBA images in napari
-            image = da.concatenate([image, da.zeros_like(image[0:1])], axis=0)
-        dims = ("c", "y", "x")
+        logging.info(f"Image has shape {image.shape}, parsing with dims={dims}.")
+        image = DataArray(image, dims=dims)
+        # squeeze spurious dimensions away
+        to_squeeze = [dim for dim in dims if dim not in ["c", "x", "y"]]
+        dims = tuple(dim for dim in dims if dim in ["c", "x", "y"])
+        for dim in to_squeeze:
+            image = image.squeeze(dim)
 
     if alignment_file is None:
         transformation = Identity()
@@ -762,3 +781,6 @@ def prefix_suffix_uint32_from_cell_id_str(cell_id_str: ArrayLike) -> tuple[Array
     cell_id_prefix = [int(x, 16) for x in cell_id_prefix_hex]
 
     return np.array(cell_id_prefix, dtype=np.uint32), np.array(dataset_suffix_int)
+
+
+##