voidspace 0.1.4__tar.gz → 0.1.5__tar.gz

{voidspace-0.1.4/src/voidspace.egg-info → voidspace-0.1.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: voidspace
-Version: 0.1.4
+Version: 0.1.5
 Summary: Spacing-aware voidspace analysis for HR-pQCT segmentations
 Author: Matthias Walle
 License-Expression: MIT

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

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "voidspace"
-version = "0.1.4"
+version = "0.1.5"
 description = "Spacing-aware voidspace analysis for HR-pQCT segmentations"
 readme = "README.md"
 requires-python = ">=3.11"

{voidspace-0.1.4 → voidspace-0.1.5}/src/voidspace/workflows.py

@@ -69,8 +69,14 @@ def intersect_masks(
     intersection, spacing, reference = read_mask(paths[0])
     for path in paths[1:]:
         mask, mask_spacing, mask_reference = read_mask(path)
-        if mask_spacing != spacing:
-            raise ValueError("all masks must have the same spacing")
+        # AIM header quantization and NIfTI float storage can differ below 1 nm.
+        # Tolerate that rounding, but do not silently mix scan resolutions.
+        if not np.allclose(mask_spacing, spacing, rtol=0, atol=1e-6):
+            raise ValueError(
+                f"mask spacing mismatch (z, y, x; mm): {paths[0]} has {spacing}; "
+                f"{path} has {mask_spacing}. Expected the matching native-space "
+                "common region at the same scan resolution."
+            )
         mask = _array_on_reference_grid(mask, mask_reference, reference)
         if mask.shape != intersection.shape:
             raise ValueError("all masks must overlap the reference image space")

{voidspace-0.1.4 → voidspace-0.1.5/src/voidspace.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: voidspace
-Version: 0.1.4
+Version: 0.1.5
 Summary: Spacing-aware voidspace analysis for HR-pQCT segmentations
 Author: Matthias Walle
 License-Expression: MIT

{voidspace-0.1.4 → voidspace-0.1.5}/tests/test_workflows.py

@@ -116,6 +116,47 @@ def test_intersect_masks_writes_boolean_intersection(tmp_path):
     assert np.array_equal(result, full & common)
 
 
+@pytest.mark.parametrize("shift", [0.0, 0.0607])
+@pytest.mark.parametrize("source_spacing, common_spacing", [
+    ((0.0607,) * 3, (float(np.float32(0.0607)),) * 3),
+    ((0.06069900095462799, 0.06069900095462799, 0.06069599837064743),
+     (0.06069965288043022, 0.06069965288043022, 0.06069643050432205)),
+])
+def test_intersect_masks_accepts_header_rounding_and_preserves_physical_placement(
+    tmp_path, shift, source_spacing, common_spacing,
+):
+    full = sitk.GetImageFromArray(np.ones((5, 5, 5), dtype=np.uint8))
+    full.SetSpacing(source_spacing)
+    sitk.WriteImage(full, str(tmp_path / "full.mha"))
+    common = sitk.GetImageFromArray(np.ones((3, 3, 3), dtype=np.uint8))
+    common.SetSpacing(common_spacing)
+    common.SetOrigin((shift,) * 3)
+    sitk.WriteImage(common, str(tmp_path / "common.nii.gz"))
+
+    output = intersect_masks(
+        [tmp_path / "full.mha", tmp_path / "common.nii.gz"],
+        output_path=tmp_path / "intersection.nii.gz",
+    )
+    image = sitk.ReadImage(str(output))
+    expected = np.zeros((5, 5, 5), dtype=bool)
+    start = 0 if shift == 0 else 1
+    expected[start:start + 3, start:start + 3, start:start + 3] = True
+    assert np.array_equal(sitk.GetArrayFromImage(image).astype(bool), expected)
+    assert np.allclose(image.GetSpacing(), source_spacing, rtol=0, atol=1e-8)
+
+
+def test_intersect_masks_reports_genuinely_different_resolution(tmp_path):
+    for name, spacing in [("full", 0.0607), ("common", 0.082)]:
+        image = sitk.GetImageFromArray(np.ones((3, 3, 3), dtype=np.uint8))
+        image.SetSpacing((spacing,) * 3)
+        sitk.WriteImage(image, str(tmp_path / f"{name}.mha"))
+    with pytest.raises(ValueError, match="spacing.*full.*common"):
+        intersect_masks(
+            [tmp_path / "full.mha", tmp_path / "common.mha"],
+            output_path=tmp_path / "intersection.nii.gz",
+        )
+
+
 def test_intersect_masks_resamples_cropped_mask_to_reference_grid(tmp_path):
     full = np.ones((5, 5, 5), dtype=bool)
     common = np.ones((3, 3, 3), dtype=bool)