New tests, improving code coverage.

.github/workflows/main-ci.yml

@@ -37,7 +37,5 @@ jobs:
         python -c "import imgtools"
     - name: Run pytest
       run: |
-        pytest tests/test_components.py
-        pytest tests/test_modalities.py
-        pytest tests/test_pipeline.py
+        pytest tests
 

.gitignore

@@ -1,6 +1,7 @@
 # test scripts
 examples/autotest.py
 .idea
+temp_outputs
 
 #vscode files
 /.idea

tests/test_ops.py

@@ -0,0 +1,95 @@
+import os, pathlib
+import pytest
+import SimpleITK as sitk
+import numpy as np
+import h5py
+from imgtools.ops import *
+
+@pytest.fixture(scope="session")
+def output_path():
+    curr_path = pathlib.Path(__file__).parent.parent.resolve()
+    out_path = pathlib.Path(curr_path, "temp_outputs").as_posix()
+    if not os.path.exists(out_path):
+        os.makedirs(out_path)
+    return out_path
+
+img_shape   = (100, 100, 100)
+direction   = (1, 0, 0, 0, 1, 0, 0, 0, 1)
+origin      = (37, 37, 37)
+spacing     = (.37, .37, .37)
+
+# build blank image sample
+blank = sitk.Image(img_shape, sitk.sitkInt16)#np.zeros((100,100,100))
+blank.SetDirection(direction)
+blank.SetOrigin(origin)
+blank.SetSpacing(spacing)
+
+class TestOutput:
+    @pytest.mark.parametrize("op", [NumpyOutput, HDF5Output])#, "CT,RTDOSE,PT"])
+    def test_output(self, op, output_path):
+        # get class name
+        class_name = op.__name__
+
+        # save output
+        saver = op(output_path, create_dirs=False)
+        saver(class_name, blank)
+        saved_path = pathlib.Path(output_path, saver.filename_format.format(subject_id=class_name)).as_posix()
+
+        # check output
+        if class_name == "HDF5Output":
+            f = h5py.File(saved_path, "r")
+            img = f['image']
+            assert tuple(img.attrs['origin'])    == origin
+            assert tuple(img.attrs['direction']) == direction
+            assert tuple(img.attrs['spacing'])   == spacing
+        elif class_name == "NumpyOutput":
+            img = np.load(saved_path)
+
+        # class-agnostic
+        assert img.shape == img_shape
+
+
+class TestTransform:
+    @pytest.mark.parametrize("op,params", [(Resample, {"spacing": 3.7}), 
+                                           (Resize, {"size": 10}), 
+                                           (Zoom, {"scale_factor": .1}), 
+                                           (Crop, {"crop_centre": (20, 20, 20), "size": 10}), 
+                                           (CentreCrop, {"size": 10})])
+    def test_transform(self, op, params):
+        transform = op(**params)
+        new_img   = transform(blank)
+
+        # check output
+        # resample
+        if isinstance(transform, Resample):
+            assert new_img.GetSpacing() == (3.7, 3.7, 3.7)
+
+        # not zoom
+        if not isinstance(transform, Zoom):
+            assert new_img.GetSize() == (10, 10, 10)
+
+        # zoom
+        if isinstance(transform, Zoom):
+            assert new_img.GetSize() == (100, 100, 100) 
+            assert new_img.GetSpacing() == (.37, .37, .37)
+
+class TestIntensity:
+    @pytest.mark.parametrize("op,params", [(ClipIntensity, {"lower": 0, "upper": 500}),
+                                           (WindowIntensity, {"window": 500, "level": 250}),
+                                           (StandardScale, {}),
+                                           (MinMaxScale, {"minimum": 0, "maximum": 1000})])
+    def test_intesity(self, op, params):
+        img_cube = copy.deepcopy(blank)
+        img_cube[5:15,5:15,5:15] = 1000
+
+        intensify = op(**params)
+        new_img   = intensify(img_cube)
+        stats = ImageStatistics()(new_img)
+
+        # check output
+        if isinstance(intensify, (ClipIntensity, WindowIntensity)):
+            assert stats.sum == 5e5
+        elif isinstance(intensify, StandardScale):
+            assert np.allclose(stats.mean, 0.)
+            assert np.allclose(stats.standard_deviation, 1., rtol=1e-1)     
+            print(stats.minimum, stats.maximum)