AnalyticSimulation

class AnalyticSimulation(name=None, building_blocks=None, build_kwargs=None)

This module provides the entry point to build and call the simulations defined within the cmrsim framework. It is meant to be subclassed, where all subclasses need to implement the abstract method _build defining the building blocks of the simulation to run.

Creates an instance of class either by using the given building Blocks or by calling the _build function to set up the simulation.

Parameters:

• name (str) –

• building_blocks (Tuple[CompositeSignalModel, BaseSampling, BaseRecon]) – Tuple containing one instance each of ‘CompositeSignalModel’, ‘BaseSampling’, ‘BaseRecon’ defining the actual simulation configuration if it is None, the memberfunction _build is called

• build_kwargs (dict) –

Methods:

__call__(dataset[, voxel_batch_size, ...])	Wrapper for the decorated (@tf.function) call of the simulation loop.
build(**kwargs)	Abstract method that needs to be defined in subclasses to configure specific simulations
from_checkpoint(checkpoint_path)	Creates instance of the class and loads variables from the specified checkpoint :type checkpoint_path: str :param checkpoint_path: str :return: Instance of class
get_k_space_shape()	Calculates the expected result shape of the simulated k-space, given the configuration
save(checkpoint_path)	Saves a tf checkpoint of the current simulation configuration :type checkpoint_path: str :param checkpoint_path: str
write_graph(dataset, graph_log_dir)	Traces the graph of the simulation for one batch and saves it as graphdef to the specified folder

Attributes:

configuration_summary

Creates a summary of the configured variables in the submodules as dictionaries :return: dict

__call__(dataset, voxel_batch_size=1000, unstack_repetitions=True, trajectory_module=None, trajectory_signatures=None, additional_kwargs=None)

Wrapper for the decorated (@tf.function) call of the simulation loop.

Parameters:

• dataset (AnalyticDataset) –

• voxel_batch_size (int) – (int) see documentation of _simulate_segmented_k_space.

• unstack_repetitions (bool) – If False the returned shape will be (#images, #Reps, #noise_levels, #samples). If True refer to ‘return’ in docstring.

• trajectory_signatures (dict) –

• additional_kwargs (dict) –

Return type:

Tensor

Returns:

(tf.Tensor) | Stack of images with shape (#images, …, #noise_levels, #X, #Y, #Z), or k-spaces with shape (#images, …, #noise_levels, #samples) where the ellipsis (…) depends on the specific forward model. The order in which the CompositeSignalModel calls the submodules determines the order of dimensions and number of axis in the ellipsis.

abstract build(**kwargs)

Abstract method that needs to be defined in subclasses to configure specific simulations

Return type:

(CompositeSignalModel, BaseSampling, BaseRecon)

classmethod from_checkpoint(checkpoint_path)

Creates instance of the class and loads variables from the specified checkpoint :type checkpoint_path: str :param checkpoint_path: str :return: Instance of class

Parameters:

checkpoint_path (str) –

get_k_space_shape()

Calculates the expected result shape of the simulated k-space, given the configuration

of encoding and forward-model modules

Returns:

tf.Tensor specifying the shape (#repetitions, #k_space_samples)

save(checkpoint_path)

Saves a tf checkpoint of the current simulation configuration :type checkpoint_path: str :param checkpoint_path: str

Parameters:

checkpoint_path (str) –

write_graph(dataset, graph_log_dir)

Traces the graph of the simulation for one batch and saves it as graphdef to the specified folder

Parameters:

graph_log_dir (str) –

property configuration_summary: OrderedDict

Creates a summary of the configured variables in the submodules as dictionaries :return: dict