Segmentation Configs

Basic Models

class InnerEye.ML.configs.segmentation.BasicModel2Epochs.BasicModel2Epochs(**kwargs: Any)[source]

get_model_train_test_dataset_splits(dataset_df: DataFrame) → DatasetSplits[source]

Computes the training, validation and test splits for the model, from a dataframe that contains the full dataset.

Parameters:: dataset_df – A dataframe that contains the full dataset that the model is using.
Returns:: An instance of DatasetSplits with dataframes for training, validation and testing.

name = 'BasicModel2Epochs'

param = <param.parameterized.Parameters object>

class InnerEye.ML.configs.segmentation.BasicModel2Epochs.BasicModelForEnsembleTest[source]

A copy of the basic model for PR builds, to use for running in a cross validation job.

name = 'BasicModelForEnsembleTest'

param = <param.parameterized.Parameters object>

pretraining_run_checkpoints: Optional[Any]

class InnerEye.ML.configs.segmentation.BasicModel2Epochs1Channel.BasicModel2Epochs1Channel[source]

name = 'BasicModel2Epochs1Channel'

param = <param.parameterized.Parameters object>

pretraining_run_checkpoints: Optional[Any]

class InnerEye.ML.configs.segmentation.BasicModel2EpochsMoreData.BasicModel2EpochsMoreData[source]

A clone of the basic PR build model, that has more training data, to avoid PyTorch throwing failures because each rank does not have enough data to train on.

get_model_train_test_dataset_splits(dataset_df: DataFrame) → DatasetSplits[source]

Computes the training, validation and test splits for the model, from a dataframe that contains the full dataset.

Parameters:: dataset_df – A dataframe that contains the full dataset that the model is using.
Returns:: An instance of DatasetSplits with dataframes for training, validation and testing.

name = 'BasicModel2EpochsMoreData'

param = <param.parameterized.Parameters object>

pretraining_run_checkpoints: Optional[Any]

HelloWorld

class InnerEye.ML.configs.segmentation.HelloWorld.HelloWorld(**kwargs: Any)[source]

This is a very basic model that is pre-configured to train on the CPU for 2 epochs on a dummy dataset ../Tests/ML/test_data/dataset.csv

The aim of this config is to demonstrate how to: 1) Subclass SegmentationModelBase which is the base config for all segmentation model configs 2) Configure the UNet3D implemented in this package 3) Configure Azure HyperDrive based parameter search

This model can be trained from the commandline: python InnerEye/runner.py –model=HelloWorld
If you want to test that your AzureML workspace is working, please follow the instructions in

<repo_root>/docs/hello_world_model.md.

In this example, the model is trained on 1 input image channels “channel1”, and predicts 2 foreground classes region, region_1.

get_model_train_test_dataset_splits(dataset_df: DataFrame) → DatasetSplits[source]

Computes the training, validation and test splits for the model, from a dataframe that contains the full dataset.

Parameters:: dataset_df – A dataframe that contains the full dataset that the model is using.
Returns:: An instance of DatasetSplits with dataframes for training, validation and testing.

get_parameter_search_hyperdrive_config(run_config: ScriptRunConfig) → HyperDriveConfig[source]

Specify an Azure HyperDrive configuration. Further details are described in the tutorial https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-tune-hyperparameters A reference is provided at https://docs.microsoft.com/en-us/python/api/azureml-train-core/azureml.train .hyperdrive?view=azure-ml-py

Parameters:: run_config – The configuration for running an individual experiment.
Returns:: An Azure HyperDrive run configuration (configured PyTorch environment).

name = 'HelloWorld'

param = <param.parameterized.Parameters object>

Head and Neck

class InnerEye.ML.configs.segmentation.HeadAndNeckBase.HeadAndNeckBase(ground_truth_ids: List[str], ground_truth_ids_display_names: Optional[List[str]] = None, colours: Optional[List[Tuple[int, int, int]]] = None, fill_holes: Optional[List[bool]] = None, roi_interpreted_types: Optional[List[str]] = None, class_weights: Optional[List[float]] = None, slice_exclusion_rules: Optional[List[SliceExclusionRule]] = None, summed_probability_rules: Optional[List[SummedProbabilityRule]] = None, num_feature_channels: Optional[int] = None, **kwargs: Any)[source]

Head and Neck radiotherapy image segmentation model.

This configuration needs to be supplied with a value for azure_dataset_id that refers to your dataset. You may also supply a value for num_structures, feature_channels or any other feature. For example, with the appropriate dataset, this would build the model whose results are reported in the InnerEye team’s paper:

class HeadAndNeckPaper(HeadAndNeckBase):
    def __init__(self):
        super().__init__(
            azure_dataset_id="foo_bar_baz",
            num_structures=10
        )

get_model_train_test_dataset_splits(dataset_df: DataFrame) → DatasetSplits[source]

Computes the training, validation and test splits for the model, from a dataframe that contains the full dataset.

Parameters:: dataset_df – A dataframe that contains the full dataset that the model is using.
Returns:: An instance of DatasetSplits with dataframes for training, validation and testing.

name = 'HeadAndNeckBase'

param = <param.parameterized.Parameters object>

class InnerEye.ML.configs.segmentation.HeadAndNeckPaper.HeadAndNeckPaper(num_structures: Optional[int] = None, **kwargs: Any)[source]

Head and Neck model, as used in the paper.

name = 'HeadAndNeckPaper'

param = <param.parameterized.Parameters object>

pretraining_run_checkpoints: Optional[Any]

Prostate

class InnerEye.ML.configs.segmentation.ProstateBase.ProstateBase(ground_truth_ids: List[str], ground_truth_ids_display_names: Optional[List[str]] = None, colours: Optional[List[Tuple[int, int, int]]] = None, fill_holes: Optional[List[bool]] = None, roi_interpreted_types: Optional[List[str]] = None, class_weights: Optional[List[float]] = None, largest_connected_component_foreground_classes: Optional[List[str]] = None, **kwargs: Any)[source]

Prostate radiotherapy image segmentation model.

get_model_train_test_dataset_splits(dataset_df: DataFrame) → DatasetSplits[source]: Return an adjusted split

name = 'ProstateBase'

param = <param.parameterized.Parameters object>

class InnerEye.ML.configs.segmentation.ProstatePaper.ProstatePaper(**kwargs: Any)[source]

Prostate radiotherapy image segmentation model, as in the paper.

name = 'ProstatePaper'

param = <param.parameterized.Parameters object>

pretraining_run_checkpoints: Optional[Any]

Hippocampus

class InnerEye.ML.configs.segmentation.Hippocampus.Hippocampus(**kwargs: Any)[source]

Hippocampus segmentation in MR scans.

get_model_train_test_dataset_splits(dataset_df: DataFrame) → DatasetSplits[source]

Computes the training, validation and test splits for the model, from a dataframe that contains the full dataset.

Parameters:: dataset_df – A dataframe that contains the full dataset that the model is using.
Returns:: An instance of DatasetSplits with dataframes for training, validation and testing.

name = 'Hippocampus'

param = <param.parameterized.Parameters object>

Lung

class InnerEye.ML.configs.segmentation.Lung.Lung(**kwargs: Any)[source]

Lung CT image segmentation model Target structures are: Esophagus, Heart, Left and Right Lungs, and Spinal cord

get_model_train_test_dataset_splits(dataset_df: DataFrame) → DatasetSplits[source]

Computes the training, validation and test splits for the model, from a dataframe that contains the full dataset.

Parameters:: dataset_df – A dataframe that contains the full dataset that the model is using.
Returns:: An instance of DatasetSplits with dataframes for training, validation and testing.

name = 'Lung'

param = <param.parameterized.Parameters object>

Glioblastoma

class InnerEye.ML.configs.segmentation.GbmBase.GbmBase(**kwargs: Any)[source]

Glioblastoma malignant brain tumour segmentation in MR scans.

get_model_train_test_dataset_splits(dataset_df: DataFrame) → DatasetSplits[source]

Computes the training, validation and test splits for the model, from a dataframe that contains the full dataset.

Parameters:: dataset_df – A dataframe that contains the full dataset that the model is using.
Returns:: An instance of DatasetSplits with dataframes for training, validation and testing.

name = 'GbmBase'

param = <param.parameterized.Parameters object>