def test_head_and_neck_base_with_invalid_summed_probability_rule3() -> None:
"""
Check that an invalid summed probability rule raises a ValueError.
"""
ground_truth_ids = DEFAULT_HEAD_AND_NECK_GROUND_TRUTH_IDS
summed_probability_rules = [
SummedProbabilityRule("spinal_cord", "brainstem", "external2")
]
with pytest.raises(ValueError) as e:
assert HeadAndNeckBase(
ground_truth_ids=ground_truth_ids,
summed_probability_rules=summed_probability_rules)
assert str(
e.value
) == "SummedProbabilityRule.validate: external2 not in ground truth IDs"
def generate_random_summed_probability_rules(
ground_truth_ids: List[str]) -> List[SummedProbabilityRule]:
"""
Generate a list of random summed probability rules, if possible.
"""
if len(ground_truth_ids) < 3:
return []
index0 = random.randint(2, len(ground_truth_ids) - 1)
index1 = random.randint(1, index0 - 1)
index2 = random.randint(0, index1 - 1)
return [
SummedProbabilityRule(ground_truth_ids[index1],
ground_truth_ids[index0],
ground_truth_ids[index2])
]
"""
Check that an invalid slice exclusion rule raises an Exception.
"""
ground_truth_ids = DEFAULT_HEAD_AND_NECK_GROUND_TRUTH_IDS
with pytest.raises(Exception) as e:
assert HeadAndNeckBase(local_dataset=Path("foo"),
ground_truth_ids=ground_truth_ids,
slice_exclusion_rules=slice_exclusion_rules)
assert str(
e.value
) == f"slice_exclusion_rules: {invalid_ground_truth} not in ground truth IDs"
@pytest.mark.parametrize(
["summed_probability_rules", "invalid_ground_truth"],
[([SummedProbabilityRule("spinal_cord2", "brainstem", "external")
], "spinal_cord2"),
([SummedProbabilityRule("spinal_cord", "brainstem2", "external")
], "brainstem2"),
([SummedProbabilityRule("spinal_cord", "brainstem", "external2")
], "external2")])
def test_head_and_neck_base_with_invalid_summed_probability_rule(
summed_probability_rules: List[SummedProbabilityRule],
invalid_ground_truth: str) -> None:
"""
Check that an invalid summed probability rule raises a ValueError.
"""
ground_truth_ids = DEFAULT_HEAD_AND_NECK_GROUND_TRUTH_IDS
with pytest.raises(ValueError) as e:
assert HeadAndNeckBase(
local_dataset=Path("foo"),
def __init__(self, num_structures: int = 0, **kwargs: Any) -> None:
"""
:param num_structures: number of structures from STRUCTURE_LIST to predict (default: all structures)
:param kwargs: other args from subclass
"""
# Number of training epochs
num_epochs = 120
# Number of structures to predict; if positive but less than the length of STRUCTURE_LIST, the relevant prefix
# of STRUCTURE_LIST will be predicted.
if num_structures <= 0 or num_structures > len(STRUCTURE_LIST):
num_structures = len(STRUCTURE_LIST)
ground_truth_ids = STRUCTURE_LIST[:num_structures]
colours = COLOURS[:num_structures]
fill_holes = FILL_HOLES[:num_structures]
ground_truth_ids_display_names = [f"zz_{x}" for x in ground_truth_ids]
# The amount of GPU memory required increases with both the number of structures and the
# number of feature channels. The following is a sensible default to avoid out-of-memory,
# but you can override is by passing in another (singleton list) value for feature_channels
# from a subclass.
num_feature_channels = 32 if num_structures <= 20 else 26
bg_weight = 0.02 if len(ground_truth_ids) > 1 else 0.25
# In case of vertical overlap between brainstem and spinal_cord, we separate them
# by converting brainstem voxels to cord, as the latter is clinically more sensitive.
# We do the same to separate SPC and MPC; in this case, the direction of change is unimportant,
# so we choose SPC-to-MPC arbitrarily.
slice_exclusion_rules = []
summed_probability_rules = []
if "brainstem" in ground_truth_ids and "spinal_cord" in ground_truth_ids:
slice_exclusion_rules.append(
SliceExclusionRule("brainstem", "spinal_cord", False))
if "external" in ground_truth_ids:
summed_probability_rules.append(
SummedProbabilityRule("spinal_cord", "brainstem",
"external"))
if "spc_muscle" in ground_truth_ids and "mpc_muscle" in ground_truth_ids:
slice_exclusion_rules.append(
SliceExclusionRule("spc_muscle", "mpc_muscle", False))
if "external" in ground_truth_ids:
summed_probability_rules.append(
SummedProbabilityRule("mpc_muscle", "spc_muscle",
"external"))
if "optic_chiasm" in ground_truth_ids and "pituitary_gland" in ground_truth_ids:
slice_exclusion_rules.append(
SliceExclusionRule("optic_chiasm", "pituitary_gland", True))
if "external" in ground_truth_ids:
summed_probability_rules.append(
SummedProbabilityRule("optic_chiasm", "pituitary_gland",
"external"))
super().__init__(
should_validate=False, # we'll validate after kwargs are added
num_gpus=4,
num_epochs=num_epochs,
save_start_epoch=num_epochs,
save_step_epoch=num_epochs,
architecture="UNet3D",
kernel_size=3,
train_batch_size=4,
inference_batch_size=1,
feature_channels=[num_feature_channels],
crop_size=(96, 288, 288),
test_crop_size=(144, 512, 512),
inference_stride_size=(72, 256, 256),
image_channels=["ct"],
norm_method=PhotometricNormalizationMethod.CtWindow,
level=50,
window=600,
start_epoch=0,
l_rate=1e-3,
min_l_rate=1e-5,
l_rate_polynomial_gamma=0.9,
optimizer_type=OptimizerType.Adam,
opt_eps=1e-4,
adam_betas=(0.9, 0.999),
momentum=0.9,
test_diff_epochs=1,
test_step_epochs=1,
use_mixed_precision=True,
use_model_parallel=True,
monitoring_interval_seconds=0,
num_dataload_workers=4,
loss_type=SegmentationLoss.Mixture,
mixture_loss_components=[
MixtureLossComponent(0.5, SegmentationLoss.Focal, 0.2),
MixtureLossComponent(0.5, SegmentationLoss.SoftDice, 0.1)
],
ground_truth_ids=ground_truth_ids,
ground_truth_ids_display_names=ground_truth_ids_display_names,
largest_connected_component_foreground_classes=ground_truth_ids,
colours=colours,
fill_holes=fill_holes,
class_weights=equally_weighted_classes(
ground_truth_ids, background_weight=bg_weight),
slice_exclusion_rules=slice_exclusion_rules,
summed_probability_rules=summed_probability_rules,
)
self.add_and_validate(kwargs)
[([], [])])
def test_slice_exclusion_rules_none(
model_config: SegmentationModelBase) -> None:
"""
Test `apply_slice_exclusion_rules` if no rule is provided
"""
# create a copy as apply_slice_exclusion_rules modifies in place
segmentation_copy = np.copy(segmentation_single_overlap)
image_util.apply_slice_exclusion_rules(model_config, segmentation_copy)
assert np.array_equal(segmentation_copy, segmentation_single_overlap)
# noinspection PyTestParametrized
@pytest.mark.parametrize(
"slice_exclusion_rules, summed_probability_rules",
[([], [SummedProbabilityRule("region1", "region2", "region3")])])
def test_apply_summed_probability_rules_incorrect_input(
model_config: SegmentationModelBase) -> None:
"""
Test `apply_summed_probability_rules` with invalid inputs
"""
posteriors = np.zeros(shape=(2, 4, 3, 3, 3))
segmentation = image_util.posteriors_to_segmentation(posteriors)
with pytest.raises(ValueError):
# noinspection PyTypeChecker
image_util.apply_summed_probability_rules(model_config,
posteriors=posteriors,
segmentation=None)
with pytest.raises(ValueError):
def __init__(self, num_structures: Optional[int] = None, **kwargs: Any) -> None:
"""
Creates a new instance of the class.
:param num_structures: number of structures from STRUCTURE_LIST to predict (default: all structures)
:param kwargs: Additional arguments that will be passed through to the SegmentationModelBase constructor.
"""
# Number of structures to predict; if positive but less than the length of STRUCTURE_LIST, the relevant prefix
# of STRUCTURE_LIST will be predicted.
if (num_structures is not None) and \
(num_structures <= 0 or num_structures > len(STRUCTURE_LIST)):
raise ValueError(f"num structures must be between 0 and {len(STRUCTURE_LIST)}")
if num_structures is None:
logging.info(f'Setting num_structures to: {len(STRUCTURE_LIST)}')
num_structures = len(STRUCTURE_LIST)
ground_truth_ids = STRUCTURE_LIST[:num_structures]
if "ground_truth_ids_display_names" in kwargs:
ground_truth_ids_display_names = kwargs.pop("ground_truth_ids_display_names")
else:
logging.info('Using default ground_truth_ids_display_names')
ground_truth_ids_display_names = [f"zz_{x}" for x in ground_truth_ids]
if "colours" in kwargs:
colours = kwargs.pop("colours")
else:
logging.info('Using default colours')
colours = COLOURS[:num_structures]
if "fill_holes" in kwargs:
fill_holes = kwargs.pop("fill_holes")
else:
logging.info('Using default fill_holes')
fill_holes = [True] * num_structures
# The amount of GPU memory required increases with both the number of structures and the
# number of feature channels. The following is a sensible default to avoid out-of-memory,
# but you can override is by passing in another (singleton list) value for feature_channels
# from a subclass.
if "num_feature_channels" in kwargs:
num_feature_channels = kwargs.pop("num_feature_channels")
else:
logging.info('Using default num_feature_channels')
num_feature_channels = 32 if num_structures <= 20 else 26
bg_weight = 0.02 if len(ground_truth_ids) > 1 else 0.25
if "class_weights" in kwargs:
class_weights = kwargs.pop("class_weights")
else:
logging.info('Using default class_weights')
class_weights = equally_weighted_classes(ground_truth_ids, background_weight=bg_weight)
# In case of vertical overlap between brainstem and spinal_cord, we separate them
# by converting brainstem voxels to cord, as the latter is clinically more sensitive.
# We do the same to separate SPC and MPC; in this case, the direction of change is unimportant,
# so we choose SPC-to-MPC arbitrarily.
if "slice_exclusion_rules" in kwargs:
slice_exclusion_rules = kwargs.pop("slice_exclusion_rules")
else:
logging.info('Using default slice_exclusion_rules')
slice_exclusion_rules = []
if "brainstem" in ground_truth_ids and "spinal_cord" in ground_truth_ids:
slice_exclusion_rules.append(SliceExclusionRule("brainstem", "spinal_cord", False))
if "spc_muscle" in ground_truth_ids and "mpc_muscle" in ground_truth_ids:
slice_exclusion_rules.append(SliceExclusionRule("spc_muscle", "mpc_muscle", False))
if "optic_chiasm" in ground_truth_ids and "pituitary_gland" in ground_truth_ids:
slice_exclusion_rules.append(SliceExclusionRule("optic_chiasm", "pituitary_gland", True))
if "summed_probability_rules" in kwargs:
summed_probability_rules = kwargs.pop("summed_probability_rules")
else:
logging.info('Using default summed_probability_rules')
summed_probability_rules = []
if "brainstem" in ground_truth_ids and "spinal_cord" in ground_truth_ids and \
"external" in ground_truth_ids:
summed_probability_rules.append(SummedProbabilityRule("spinal_cord", "brainstem", "external"))
if "spc_muscle" in ground_truth_ids and "mpc_muscle" in ground_truth_ids and \
"external" in ground_truth_ids:
summed_probability_rules.append(SummedProbabilityRule("mpc_muscle", "spc_muscle", "external"))
if "optic_chiasm" in ground_truth_ids and "pituitary_gland" in ground_truth_ids and \
"external" in ground_truth_ids:
summed_probability_rules.append(SummedProbabilityRule("optic_chiasm", "pituitary_gland", "external"))
super().__init__(
ground_truth_ids=ground_truth_ids,
ground_truth_ids_display_names=ground_truth_ids_display_names,
colours=colours,
fill_holes=fill_holes,
class_weights=class_weights,
slice_exclusion_rules=slice_exclusion_rules,
summed_probability_rules=summed_probability_rules,
num_feature_channels=num_feature_channels,
**kwargs)