API Reference#

Model#

class deep_quality_estimation.model.DQE(device: torch.device | None = None, cuda_devices: str | None = '0')#

Bases: object

predict(t1c: Path | ndarray[Any, dtype[_ScalarType_co]], t1: Path | ndarray[Any, dtype[_ScalarType_co]], t2: Path | ndarray[Any, dtype[_ScalarType_co]], flair: Path | ndarray[Any, dtype[_ScalarType_co]], segmentation: Path | ndarray[Any, dtype[_ScalarType_co]]) → Tuple[float, Dict[View, float]]#

Predict the quality of the given Segmentation

Parameters:

  • t1c (Union[Path, NDArray]) – Numpy NDArray or Path to the T1c NIfTI file

  • t1 (Union[Path, NDArray]) – Numpy NDArray or Path to the T1 NIfTI file

  • t2 (Union[Path, NDArray]) – Numpy NDArray or Path to the T2 NIfTI file

  • flair (Union[Path, NDArray]) – Numpy NDArray or Path to the FLAIR NIfTI file

  • segmentation (Union[Path, NDArray]) – Numpy NDArray or Path to the segmentation NIfTI file (In BraTS style)

Returns:

The predicted mean score and a dict with the scores per view

Return type:

Tuple[float, Dict[View, float]]

Data Handling#

class deep_quality_estimation.data_handler.DataHandler(t1c: Path | ndarray[Any, dtype[_ScalarType_co]], t1: Path | ndarray[Any, dtype[_ScalarType_co]], t2: Path | ndarray[Any, dtype[_ScalarType_co]], flair: Path | ndarray[Any, dtype[_ScalarType_co]], segmentation: Path | ndarray[Any, dtype[_ScalarType_co]])#

Bases: object

ALL_CHANNELS = ['images', 'labels']#
ONLY_IMAGES = ['images']#
ONLY_LABELS = ['labels']#
get_dataloader() → torch.utils.data.DataLoader#

Get the dataloader for the dataset

Returns:

The dataloader

Return type:

DataLoader

class deep_quality_estimation.transforms.CustomConvertToMultiChannelBasedOnBratsClasses(*args: Any, **kwargs: Any)#

Bases: Transform

Adapted from Monai’s ConvertToMultiChannelBasedOnBratsClasses since label 4 changed to 3 in newer datasets

Convert labels to multi channels based on brats18 classes: label 1 is the necrotic and non-enhancing tumor core label 2 is the peritumoral edema label 3 is the GD-enhancing tumor (used to be label 4 in older datasets) The possible classes are TC (Tumor core), WT (Whole tumor) and ET (Enhancing tumor).

backend: list[TransformBackends] = [monai.utils.enums.TransformBackends.TORCH, monai.utils.enums.TransformBackends.NUMPY]#
class deep_quality_estimation.transforms.CustomConvertToMultiChannelBasedOnBratsClassesd(*args: Any, **kwargs: Any)#

Bases: MapTransform

Adapted from Monai’s ConvertToMultiChannelBasedOnBratsClassesd

Dictionary-based wrapper of monai.transforms.ConvertToMultiChannelBasedOnBratsClasses. Convert labels to multi channels based on brats18 classes: label 1 is the necrotic and non-enhancing tumor core label 2 is the peritumoral edema label 3 is the GD-enhancing tumor (used to be label 4 in older datasets) The possible classes are TC (Tumor core), WT (Whole tumor) and ET (Enhancing tumor).

backend = [monai.utils.enums.TransformBackends.TORCH, monai.utils.enums.TransformBackends.NUMPY]#

Utilities#

deep_quality_estimation.center_of_mass.compute_center_of_mass(segmentation: Path | ndarray[Any, dtype[_ScalarType_co]], fallback_to_edema: bool = True) → List[int]#

Compute the center of mass of the tumor core in the given segmentation. If no tumor core is found and the @fallback_to_edema is true, the center of mass of the edema is used as a fallback.

Parameters:

  • segmentation (Union[Path, NDArray]) – Path to the segmentation file or the segmentation data as numpy NDArray.

  • fallback_to_edema (bool, optional) – Use Edema CoM as fallback if no tumor core is found. Defaults to True.

Returns:

List of the center of mass coordinates

Return type:

List[int]

deep_quality_estimation.center_of_mass.get_center_of_mass_slices(image: Path | ndarray[Any, dtype[_ScalarType_co]], center_of_mass: List[int]) → dict[View, ndarray[Any, dtype[_ScalarType_co]]]#

Get the slices of the given image that contain the center of mass in axial, coronal and sagittal view.

Parameters:

  • image (Path) – Path to the NiFTI image file

  • center_of_mass (List[int]) – List of the center of mass coordinates

Returns:

Dictionary with the views as keys and the 2D image slices as values

Return type:

Dict[View, NDArray]

class deep_quality_estimation.enums.View(value)#

Bases: Enum

Enum for the different views of the brain

AXIAL = 1#
CORONAL = 2#
SAGITTAL = 3#