Handle edge cases when summarising labels layers (#99)

* Prevent errors if napari layers contain non numpy-like types

* Handle empty labels layers

* Add tests for regions edge cases

* apply suggestions from code review

brainreg_segment/regions/analysis.py

@@ -46,6 +46,10 @@ def summarise_brain_regions(label_layers, filename, atlas_resolution):
     for label_layer in label_layers:
         summaries.append(summarise_single_brain_region(label_layer))
 
+    if check_list_only_nones(summaries):
+        print("No regions to summarise")
+        return
+
     result = pd.concat(summaries)
     # TODO: use atlas.space to make these more intuitive
     volume_header = "volume_mm3"
@@ -77,6 +81,10 @@ def summarise_brain_regions(label_layers, filename, atlas_resolution):
     result.to_csv(filename, index=False)
 
 
+def check_list_only_nones(input_list):
+    return all(v is None for v in input_list)
+
+
 def summarise_single_brain_region(
     label_layer,
     ignore_empty=True,
@@ -86,7 +94,7 @@ def summarise_single_brain_region(
         "centroid",
     ],
 ):
-    data = label_layer.data
+    data = np.asarray(label_layer.data)
     if ignore_empty:
         if data.sum() == 0:
             return
@@ -95,7 +103,7 @@ def summarise_single_brain_region(
         data.astype(np.uint16), properties=properties_to_fetch
     )
     df = pd.DataFrame.from_dict(regions_table)
-    df.insert(0, "Region", label_layer.name)
+    df.insert(0, "region", label_layer.name)
     return df
 
 

pyproject.toml

@@ -48,7 +48,8 @@ dev = [
     "pytest",
     "coverage",
     "pytest-cov",
-    "pytest-qt"
+    "pytest-qt",
+    "napari-time-slicer" # to test non ndarray-types
 ]
 
 

tests/tests/test_unit/test_regions/test_region_analysis.py

@@ -0,0 +1,148 @@
+from pathlib import Path
+
+import numpy as np
+import pandas as pd
+import pytest
+import tifffile
+from napari.layers import Labels
+
+from brainreg_segment.regions.analysis import (
+    check_list_only_nones,
+    summarise_brain_regions,
+    summarise_single_brain_region,
+)
+
+region_image_path = Path(
+    "tests/data/brainreg_output/manual_segmentation/"
+    "sample_space/regions/region_0.tiff"
+)
+atlas_resolution = (50, 50, 50)
+label_name = "test_label"
+
+# We run regression tests against a previously known "correct run"
+# that produced these values
+previous_image_space_values = {
+    "region": "test_label",
+    "area": 17842.00000,
+    "bbox-0": 134,
+    "bbox-1": 34,
+    "bbox-5": 65,
+    "centroid-2": 44.284385158614505,
+}
+
+previous_world_space_values = {
+    "region": "test_label",
+    "volume_mm3": 2.23025,
+    "axis_0_min_um": 6700.0,
+    "axis_1_min_um": 1700.0,
+    "axis_2_min_um": 1250.0,
+    "axis_0_max_um": 7850.0,
+    "axis_1_max_um": 3350.0,
+    "axis_2_max_um": 3250.0,
+    "axis_0_center_um": 7250.0,
+    "axis_1_center_um": 2483.154915368232,
+    "axis_2_center_um": 2214.219257930725,
+}
+
+
+@pytest.fixture
+def region_image():
+    return tifffile.imread(region_image_path)
+
+
+@pytest.fixture
+def labels_layer(region_image):
+    labels = Labels(region_image, name=label_name)
+    return labels
+
+
+@pytest.fixture
+def empty_labels_layer(region_image):
+    labels = Labels(np.zeros_like(region_image))
+    return labels
+
+
+def test_check_list_only_ones():
+    assert check_list_only_nones([None, None, None]) is True
+    assert check_list_only_nones([None, 1, None]) is False
+    assert check_list_only_nones([1, "two", 3]) is False
+
+
+def test_summarise_single_brain_region(labels_layer):
+    """
+    Test the summary of a single segmented region.
+    N.B. these results are in "image" space, not "world" space,
+    hence different to the other tests
+    """
+    df = summarise_single_brain_region(labels_layer)
+    assert df.shape == (1, 11)
+    for key, value in previous_image_space_values.items():
+        assert df[key].iloc[0] == value
+
+
+def test_summarise_brain_regions_all_not_empty(
+    labels_layer, empty_labels_layer, tmp_path
+):
+    """
+    Test summarise_brain_regions() when all the regions
+    (usually segmented by the user)
+    contain at least one labelled voxel
+
+    N.B. these results are in "world" space, not "space" space,
+    hence different to the test of a single region
+    """
+    filename = tmp_path / "summary.csv"
+
+    summarise_brain_regions(
+        [labels_layer, labels_layer, labels_layer], filename, atlas_resolution
+    )
+    df = pd.read_csv(filename)
+    assert df.shape == (3, 11)
+    for key, value in previous_world_space_values.items():
+        assert df[key].iloc[0] == value
+
+
+def test_summarise_brain_regions_some_empty(
+    labels_layer, empty_labels_layer, tmp_path
+):
+    """
+    Test summarise_brain_regions() when some of the regions
+    (usually segmented by the user)
+    contain no voxels (i.e. are empty arrays), but
+    others contain labelled voxels.
+
+    N.B. these results are in "world" space, not "space" space,
+    hence different to the test of a single region
+    """
+
+    filename = tmp_path / "summary.csv"
+    summarise_brain_regions(
+        [empty_labels_layer, empty_labels_layer, labels_layer],
+        filename,
+        atlas_resolution,
+    )
+    df = pd.read_csv(filename)
+    assert df.shape == (1, 11)
+    for key, value in previous_world_space_values.items():
+        assert df[key].iloc[0] == value
+
+
+def test_summarise_brain_regions_all_empty(
+    labels_layer, empty_labels_layer, tmp_path
+):
+    """
+    Test summarise_brain_regions() when all the regions
+    (usually segmented by the user)
+    contain no voxels (i.e. are empty arrays)
+    """
+    filename = tmp_path / "summary.csv"
+
+    # check correct behaviour if all regions are empty
+    assert (
+        summarise_brain_regions(
+            [empty_labels_layer, empty_labels_layer],
+            filename,
+            atlas_resolution,
+        )
+        is None
+    )