monospectral

• noise
  • GammaNoise
    • GammaNoise.run()
  • GaussianNoise
    • GaussianNoise.run()
  • PoissonNoise
    • PoissonNoise.run()
  • SaltAndPepperNoise
    • SaltAndPepperNoise.run()
  • UniformNoise
    • UniformNoise.run()

class FieldOfViewCropping(global_settings, settings_key=None)

Bases: ProcessingComponentBase

Initialises the ProcessingComponent object.

Parameters:

component_settings_key – The key where the component settings are stored in

run(device: DigitalDeviceTwinBase)

Executes the respective simulation module

Parameters:

digital_device_twin – The digital twin that can be used by the digital device_twin.

class IterativeqPAI(global_settings, component_settings_key: str)

Bases: ProcessingComponentBase

Applies iterative qPAI Algorithm [1] on simulated initial pressure map and saves the reconstruction result in the hdf5 output file. If a 2-d map of initial_pressure is passed the algorithm saves the reconstructed absorption coefficients as a 2-d map, else a 3-d absorption reconstruction is saved. The reconstruction result is saved as an image processing entry “iterative_qpai_result” in the hdf5 output file. If intended (e.g. for testing) a list of intermediate iteration updates (only 2-d middle slices) can be saved as a npy file. To run the reconstruction the scattering coefficients must be known a priori. Parameters: Tags.DOWNSCALE_FACTOR (default: 0.73) Tags.ITERATIVE_RECONSTRUCTION_CONSTANT_REGULARIZATION (default: False) Tags.ITERATIVE_RECONSTRUCTION_MAX_ITERATION_NUMBER (default: 10) Tags.ITERATIVE_RECONSTRUCTION_REGULARIZATION_SIGMA (default: 0.01) Tags.ITERATIVE_RECONSTRUCTION_SAVE_INTERMEDIATE_RESULTS (default: False) Tags.ITERATIVE_RECONSTRUCTION_STOPPING_LEVEL (default: 0.03) global_settings (required) component_settings_key (required)

[1] B. T. Cox et al. 2006, “Two-dimensional quantitative photoacoustic image reconstruction of absorption distributions in scattering media by use of a simple iterative method”, https://doi.org/10.1364/ao.45.001866

Initialises the ProcessingComponent object.

Parameters:

component_settings_key – The key where the component settings are stored in

convergence_stopping_criterion(errors: list, iteration: int) → bool

Serves as a stopping criterion for the iterative algorithm. If False the iterative algorithm continues.

Parameters:

• errors – List of log (base 10) sum of squared errors.

• iteration – Iteration number.

Returns:

if iteration method should be stopped.

Raises:

AssertionError: if Tags.ITERATIVE_RECONSTRUCTION_STOPPING_LEVEL tag is zero

extract_initial_data_from_hdf5() → Tuple[ndarray, ndarray, ndarray]

Extract necessary information - initial pressure and scattering coefficients - from settings dictionaries. The setting dictionaries is extracted from a hdf5 file.

Returns:

Initial pressure and a priori known scattering coefficients.

forward_model_fluence(absorption: ndarray, scattering: ndarray, anisotropy: ndarray, pa_device) → ndarray

Simulates photon propagation in 3-d volume and returns simulated fluence map in units of J/cm^2.

Parameters:

• absorption – Volume of absorption coefficients in 1/cm for Monte Carlo Simulation.

• scattering – Volume of scattering coefficients in 1/cm for Monte Carlo Simulation.

• anisotropy – Volume of anisotropy data for Monte Carlo Simulation.

• pa_device – The simulation device.

Returns:

Fluence map.

Raises:

AssertionError: if Tags.OPTICAL_MODEL tag was not or incorrectly defined in settings.

iterative_absorption_reconstruction(pa_device) → Tuple[ndarray, list]

Performs quantitative photoacoustic image reconstruction of absorption distribution by use of an iterative method. The distribution of scattering coefficients must be known a priori.

Returns:

Reconstructed absorption coefficients in 1/cm.

Raises:

TypeError: if input data are not passed as a certain type ValueError: if input data cannot be used for reconstruction due to shape or value FileNotFoundError: if paths stored in settings cannot be accessed AssertionError: is Tags.MAX_NUMBER_ITERATIVE_RECONSTRUCTION tag is zero

log_sum_squared_error(image_data: ~numpy.ndarray, absorption: ~numpy.ndarray, fluence: ~numpy.ndarray, sigma: [<class 'numpy.ndarray'>, <class 'int'>, <class 'float'>]) → float

Computes log (base 10) of the sum of squared error between image and reconstructed pressure map in middle slice.

Parameters:

• image_data – Measured image data used for reconstruction.

• absorption – Reconstructed map of absorption coefficients in 1/cm.

• fluence – Simulated fluence map in J/cm^2.

• sigma – Regularization parameter to avoid instability if the fluence is low.

Returns:

sse error.

preprocessing_for_iterative_qpai(intial_pressure: ndarray, scattering: ndarray, anisotropy: ndarray) → Tuple[ndarray, ndarray, ndarray, bool]

Preprocesses image data and scattering distribution for iterative algorithm using mcx. The preprocessing step includes: 1. Stacking the input data from 2-d to 3-d if necessary, since the mcx adapter can only perform

a Monte Carlo Simulation of fluence given 3-d volumes of absorption, scattering, and anisotropy

2. Resampling the input data to mitigate the inverse crime

Parameters:

• intial_pressure – Raw input image of initial pressure.

• scattering – Map of scattering coefficients known a priori.

Returns:

Resampled and (if necessary) stacked volume of noisy initial pressure, scattering and bool indicating if image had to be stacked to 3-d.

regularization_sigma(input_image: ndarray, stacked_to_volume) → [<class 'numpy.ndarray'>, <class 'int'>, <class 'float'>]

Computes spatial (same shape as input image) or constant regularization parameter sigma.

Parameters:

• input_image – Noisy input image.

• stacked_to_volume – If True input data was 2 dimensional and a constant sigma should be used.

Returns:

Regularization parameter.

Raises:

ValueError: if estimated noise is zero, so SNR cannot be computed.

resampling_for_iterative_qpai(initial_pressure: ndarray, scattering: ndarray, anisotropy: ndarray) → Tuple[ndarray, ndarray, ndarray]

Downscales the input image and scattering map by a given scale (downscale factor) to avoid inverse crime.

Parameters:

• initial_pressure – Raw input image of initial pressure.

• scattering – Map of scattering coefficients.

• anisotropy – Map of anisotropies.

Returns:

Downscaled image and scattering map.

run(pa_device)

Executes the respective simulation module

Parameters:

digital_device_twin – The digital twin that can be used by the digital device_twin.

stacking_to_3d(input_data: ndarray) → ndarray

Stacks the input map in sequence along axis=1 to build 3-d volume.

Parameters:

input_data – 2-d image data.

Returns:

Volume of stacked input image.

standard_optical_properties(image_data: ndarray) → dict

Returns a optical properties dictionary containing scattering coefficients and anisotropy.

Parameters:

image_data – Measured image data.

Returns:

Optical properties (scattering and anisotropy).

update_absorption_estimate(image_data: ~numpy.ndarray, fluence: ~numpy.ndarray, sigma: [<class 'numpy.ndarray'>, <class 'int'>, <class 'float'>]) → ndarray

Reconstructs map of absorption coefficients in 1/cm given measured data and simulated fluence.

Parameters:

• image_data – Measured image data (initial pressure) used for reconstruction.

• fluence – Simulated fluence map in J/cm^2.

• sigma – Regularization factor to avoid instability if the fluence is low.

Returns:

Reconstructed absorption.