def test_deep_tissue_feature_extractor_aug(self):
"""
Test deep tissue feature extractor with pipeline runner and with augmentation.
"""
config_fname = os.path.join(self.current_path, 'config',
'feature_extraction',
'deep_tissue_feature_extractor_aug.yml')
with open(config_fname, 'r') as file:
config = yaml.safe_load(file)
pipeline = PipelineRunner(output_path=self.out_path, **config)
output = pipeline.run(output_name=self.image_name.replace('.png', ''),
image_path=os.path.join(self.image_path,
self.image_name))
features = output['features']
self.assertTrue(isinstance(features, torch.Tensor)) # check type
# check number of superpixels
self.assertEqual(features.shape[0], 23)
# check number of augmentations
self.assertEqual(features.shape[1], 4)
self.assertEqual(features.shape[2], 1280) # check number features
# Re-run with existing output & ensure equal
output = pipeline.run(output_name=self.image_name.replace('.png', ''),
image_path=os.path.join(self.image_path,
self.image_name))
reload_features = output['features']
self.assertTrue(np.array_equal(features, reload_features))
def test_gaussian_tissue_mask_with_pipeline_runner(self):
"""
Test gaussian tissue mask with pipeline runner.
"""
# 1. Tissue mask detection with saving
config_fname = os.path.join(self.current_path, 'config', 'tissue_mask',
'tissue_mask.yml')
with open(config_fname, 'r') as file:
config = yaml.safe_load(file)
pipeline = PipelineRunner(output_path=self.out_path, **config)
output = pipeline.run(output_name=self.image_name.replace('.jpg', ''),
image_path=os.path.join(self.image_path,
self.image_name))
tissue_mask = output['tissue_mask']
self.assertEqual(list(tissue_mask.shape),
[1024, 1280]) # image HxW = mask HxW
self.assertTrue(list(np.unique(tissue_mask)) == [0,
1]) # mask is binary
# tissue pixel count large enough
self.assertTrue(np.sum(tissue_mask) > 1e6)
self.assertTrue(np.sum(tissue_mask) < 2e6) # but not too large
# 2. Re-run with existing output & ensure equal
output = pipeline.run(output_name=self.image_name.replace('.jpg', ''),
image_path=os.path.join(self.image_path,
self.image_name))
reload_tissue_mask = output['tissue_mask']
self.assertTrue(np.array_equal(tissue_mask, reload_tissue_mask))
def test_color_merged_superpixel_extractor_with_pipeline_runner(self):
"""
Test color merged superpixel extractor with pipeline runner.
"""
config_fname = os.path.join(self.current_path, 'config', 'superpixels',
'color_merged_extractor.yml')
with open(config_fname, 'r') as file:
config = yaml.safe_load(file)
pipeline = PipelineRunner(output_path=self.out_path, **config)
output = pipeline.run(output_name=self.image_name.replace('.png', ''),
image_path=os.path.join(self.image_path,
self.image_name))
superpixels = output['merged_superpixels_map']
self.assertTrue(isinstance(superpixels, np.ndarray)) # check type
self.assertEqual(len(list(superpixels.shape)), 2) # mask is bi-dim
# check number of instances
self.assertEqual(len(np.unique(superpixels)), 23)
# Re-run with existing output & ensure equal
output = pipeline.run(output_name=self.image_name.replace('.jpg', ''),
image_path=os.path.join(self.image_path,
self.image_name))
reload_superpixels = output['merged_superpixels_map']
self.assertTrue(np.array_equal(superpixels, reload_superpixels))
def test_handcrafted_feature_extractor(self):
"""
Test handcrafted feature extractor with pipeline runner.
"""
config_fname = os.path.join(self.current_path, 'config',
'feature_extraction',
'handcrafted_feature_extractor.yml')
with open(config_fname, 'r') as file:
config = yaml.safe_load(file)
pipeline = PipelineRunner(output_path=self.out_path, **config)
output = pipeline.run(output_name=self.image_name.replace('.png', ''),
image_path=os.path.join(self.image_path,
self.image_name))
features = output['features']
self.assertTrue(isinstance(features, torch.Tensor)) # check type
# check number of features per instance
self.assertEqual(features.shape[1], 65)
# check number of instances detected
self.assertEqual(features.shape[0], 23)
# Re-run with existing output & ensure equal
output = pipeline.run(output_name=self.image_name.replace('.png', ''),
image_path=os.path.join(self.image_path,
self.image_name))
reload_features = output['features']
self.assertTrue(np.array_equal(features, reload_features))
def test_rag_builder_annotation_with_pipeline_runner(self):
"""
Test rag builder with annotation and pipeline runner.
"""
config_fname = os.path.join(self.current_path, 'config',
'graph_builder',
'rag_graph_builder_annotation.yml')
with open(config_fname, 'r') as file:
config = yaml.safe_load(file)
pipeline = PipelineRunner(output_path=self.out_path, **config)
output = pipeline.run(
output_name=self.image_name.replace('.png', '_annotation'),
image_path=os.path.join(self.image_path, self.image_name),
annotation_path=os.path.join(self.image_path,
self.annotation_name))
graph = output['graph']
labels, counts = np.unique(graph.ndata['label'].numpy(),
return_counts=True)
self.assertTrue(isinstance(graph, dgl.DGLGraph)) # check type
self.assertEqual(graph.number_of_nodes(), 23) # check number of nodes
self.assertEqual(graph.number_of_edges(), 112) # check number of edges
# check assigned features
self.assertEqual(graph.ndata["feat"].shape, (23, 1, 1282))
# check assigned centroid
self.assertEqual(graph.ndata["centroid"].shape, (23, 2))
# check assigned node labels
self.assertTrue(np.array_equal(labels, np.array([0, 1, 2, 4])))
# check assigned node label counts
self.assertTrue(np.array_equal(counts, np.array([2, 9, 2, 10])))
def test_rag_builder_augmentation_with_pipeline_runner(self):
"""
Test rag builder with augmentation and pipeline runner.
"""
config_fname = os.path.join(self.current_path, 'config',
'graph_builder',
'rag_graph_builder_aug.yml')
with open(config_fname, 'r') as file:
config = yaml.safe_load(file)
pipeline = PipelineRunner(output_path=self.out_path, **config)
output = pipeline.run(output_name=self.image_name.replace(
'.png', '_aug'),
image_path=os.path.join(self.image_path,
self.image_name))
graph = output['graph']
self.assertTrue(isinstance(graph, dgl.DGLGraph)) # check type
self.assertEqual(graph.number_of_nodes(), 23) # check number of nodes
self.assertEqual(graph.number_of_edges(), 112) # check number of edges
# check assigned features
self.assertEqual(graph.ndata["feat"].shape, (23, 4, 1282))
# check assigned centroid
self.assertEqual(graph.ndata["centroid"].shape, (23, 2))
def test_graph_diameter_with_pipeline_runner(self):
"""
Test graph diameter with pipeline runner.
"""
config_fname = os.path.join(self.current_path, 'config', 'stats',
'graph_diameter.yml')
with open(config_fname, 'r') as file:
config = yaml.safe_load(file)
pipeline = PipelineRunner(output_path=None, **config)
output = pipeline.run(output_name=None,
graph_path=os.path.join(self.graph_path,
self.graph_name))
diameter = output['diameter']
self.assertEqual(diameter, 6) # check true diameter
def test_superpixel_counter_with_pipeline_runner(self):
"""
Test superpixel counter with pipeline runner.
"""
config_fname = os.path.join(self.current_path, 'config', 'stats',
'superpixel_counter.yml')
with open(config_fname, 'r') as file:
config = yaml.safe_load(file)
pipeline = PipelineRunner(output_path=None, **config)
output = pipeline.run(output_name=None,
image_path=os.path.join(self.image_path,
self.image_name))
count = output['counter']
self.assertEqual(count, 95) # check true count
def test_nuclei_extractor_with_pipeline_runner(self):
"""Test nuclei extraction with local model."""
config_fname = os.path.join(self.current_path, 'config',
'nuclei_extraction',
'nuclei_extractor.yml')
with open(config_fname, 'r') as file:
config = yaml.safe_load(file)
pipeline = PipelineRunner(output_path=self.out_path, **config)
output = pipeline.run(output_name=self.image_name.replace('.png', ''),
image_path=os.path.join(self.image_path,
self.image_name))
instance_map = output['nuclei_map']
instance_centroids = output['nuclei_centroids']
# 3. run tests
self.assertTrue(isinstance(instance_map, np.ndarray))
self.assertTrue(isinstance(instance_centroids, np.ndarray))
self.assertEqual(len(instance_centroids), 331)
def test_deep_nuclei_feature_extractor_aug(self):
"""Test deep nuclei feature extractor with pipeline runner and with augmentation."""
config_fname = os.path.join(self.current_path, 'config',
'feature_extraction',
'deep_nuclei_feature_extractor_aug.yml')
with open(config_fname, 'r') as file:
config = yaml.safe_load(file)
pipeline = PipelineRunner(output_path=self.out_path, **config)
output = pipeline.run(output_name=self.image_name.replace('.png', ''),
image_path=os.path.join(self.image_path,
self.image_name),
nuclei_map_path=os.path.join(
self.nuclei_map_path, self.nuclei_map_name))
features = output['features']
self.assertTrue(isinstance(features, torch.Tensor)) # check type
self.assertEqual(features.shape[0], 331) # check number of nuclei
# check number of augmentations
self.assertEqual(features.shape[1], 4)
self.assertEqual(features.shape[2], 1280) # check number features
def test_knn_builder_with_pipeline_runner(self):
"""
Test knn builder with pipeline runner.
"""
config_fname = os.path.join(self.current_path, 'config',
'graph_builder', 'knn_graph_builder.yml')
with open(config_fname, 'r') as file:
config = yaml.safe_load(file)
pipeline = PipelineRunner(output_path=self.out_path, **config)
output = pipeline.run(output_name=self.image_name.replace('.png', ''),
image_path=os.path.join(self.image_path,
self.image_name),
nuclei_map_path=os.path.join(
self.nuclei_map_path, self.nuclei_map_name))
graph = output['graph']
self.assertTrue(isinstance(graph, dgl.DGLGraph)) # check type
self.assertEqual(graph.number_of_nodes(), 331) # check number of nodes
self.assertEqual(graph.number_of_edges(),
1655) # check number of edges