🐛 Fix DICOM Reading (Issue #933)

tests/conftest.py

@@ -162,6 +162,16 @@ def sample_jp2(remote_sample: Callable) -> Path:
     return remote_sample("jp2-omnyx-small")
 
 
+@pytest.fixture(scope="session")
+def sample_dicom(remote_sample: Callable) -> Path:
+    """Sample pytest fixture for dicom images.
+
+    Download dicom image for pytest.
+
+    """
+    return remote_sample("dicom-1")
+
+
 @pytest.fixture(scope="session")
 def sample_all_wsis(
     sample_ndpi: Path,

tests/test_wsireader.py

@@ -2979,3 +2979,22 @@ def test_fsspec_reader_open_pass_empty_json(tmp_path: Path) -> None:
     json_path.write_text("{}")
 
     assert not FsspecJsonWSIReader.is_valid_zarr_fsspec(str(json_path))
+
+
+def test_oob_read_dicom(sample_dicom: Path) -> None:
+    """Test that out of bounds returns background value.
+
+    For consistency with openslide, our readers should return a
+    background tile when reading out of bounds.
+
+    """
+    wsi = DICOMWSIReader(sample_dicom)
+    # Read a region that is out of bounds
+    region = wsi.read_rect(
+        location=(200000, 200),
+        size=(100, 100),
+    )
+    # Check that the region is the same size as the requested size
+    assert region.shape == (100, 100, 3)
+    # Check that the region is white (255)
+    assert np.all(region == 255)

tiatoolbox/visualization/bokeh_app/main.py

@@ -725,7 +725,16 @@ def populate_slide_list(slide_folder: Path, search_txt: str | None = None) -> No
     """Populate the slide list with the available slides."""
     file_list = []
     len_slidepath = len(slide_folder.parts)
-    for ext in ["*.svs", "*ndpi", "*.tiff", "*.mrxs", "*.jpg", "*.png", "*.tif"]:
+    for ext in [
+        "*.svs",
+        "*ndpi",
+        "*.tiff",
+        "*.mrxs",
+        "*.jpg",
+        "*.png",
+        "*.tif",
+        "*.dcm",
+    ]:
         file_list.extend(list(Path(slide_folder).glob(str(Path("*") / ext))))
         file_list.extend(list(Path(slide_folder).glob(ext)))
     if search_txt is None:
@@ -2086,7 +2095,16 @@ def setup_doc(self: DocConfig, base_doc: Document) -> tuple[Row, Tabs]:
 
         # Set initial slide to first one in base folder
         slide_list = []
-        for ext in ["*.svs", "*ndpi", "*.tiff", "*.tif", "*.mrxs", "*.png", "*.jpg"]:
+        for ext in [
+            "*.svs",
+            "*ndpi",
+            "*.tiff",
+            "*.tif",
+            "*.mrxs",
+            "*.png",
+            "*.jpg",
+            "*.dcm",
+        ]:
             slide_list.extend(list(doc_config["slide_folder"].glob(ext)))
             slide_list.extend(
                 list(doc_config["slide_folder"].glob(str(Path("*") / ext))),

tiatoolbox/wsicore/wsireader.py

@@ -4595,6 +4595,7 @@ def _info(self: DICOMWSIReader) -> WSIMeta:
             mpp=mpp,
             level_count=len(level_dimensions),
             vendor=dataset.Manufacturer,
+            file_path=self.input_path,
         )
 
     def read_rect(
@@ -4826,8 +4827,19 @@ def read_rect(
         _, constrained_read_size = utils.transforms.bounds2locsize(
             constrained_read_bounds,
         )
+
+        # if out of bounds, return empty image consistent with openslide
+        if np.any(np.array(constrained_read_size) <= 0):
+            return (
+                np.ones(
+                    shape=(int(size[1]), int(size[0]), 3),
+                    dtype=np.uint8,
+                )
+                * 255
+            )
+
         dicom_level = wsi.levels[read_level].level
-        im_region = wsi.read_region(location, dicom_level, constrained_read_size)
+        im_region = wsi.read_region(level_location, dicom_level, constrained_read_size)
         im_region = np.array(im_region)
 
         # Apply padding outside the slide area
@@ -5003,7 +5015,6 @@ class docstrings for more information.
         wsi = self.wsi
 
         # Read at optimal level and corrected read size
-        location_at_baseline = bounds_at_baseline[:2]
         level_location, size_at_read_level = utils.transforms.bounds2locsize(
             bounds_at_read_level,
         )
@@ -5016,7 +5027,7 @@ class docstrings for more information.
         _, read_size = utils.transforms.bounds2locsize(read_bounds)
         dicom_level = wsi.levels[read_level].level
         im_region = wsi.read_region(
-            location=location_at_baseline,
+            location=level_location,
             level=dicom_level,
             size=read_size,
         )