direct.algorithms package#

Submodules#

direct.algorithms.mri_algorithms module#

This module contains mathematical optimization techniques specific to MRI.

class direct.algorithms.mri_algorithms.EspiritCalibration(backward_operator, threshold=0.05, kernel_size=6, crop=0.95, max_iter=100, kspace_key=KspaceKey.MASKED_KSPACE)[source][source]#

Bases: DirectModule

Estimates sensitivity maps estimated with the ESPIRIT calibration method as described in [1].

We adapted code for ESPIRIT method adapted from [2].

References

[1]

Uecker M, Lai P, Murphy MJ, Virtue P, Elad M, Pauly JM, Vasanawala SS, Lustig M. ESPIRiT–an eigenvalue approach to autocalibrating parallel MRI: where SENSE meets GRAPPA. Magn Reson Med. 2014 Mar;71(3):990-1001. doi: 10.1002/mrm.24751. PMID: 23649942; PMCID: PMC4142121.

calculate_sensitivity_map(acs_mask, kspace)[source][source]#

Calculates sensitivity map given as input the acs_mask and the k-space.

Parameters:
acs_masktorch.Tensor

Autocalibration mask.

kspacetorch.Tensor

K-space.

Returns:
sensitivity_maptorch.Tensor
Return type:

Tensor

forward(sample)[source][source]#

Forward method of EspiritCalibration.

Parameters:
sample: Dict[str, Any]

Contains key kspace_key.

Returns:
sample: Dict[str, Any]

Contains key ‘sampling_mask’.

Return type:

Tensor

training: bool#

direct.algorithms.optimization module#

General mathematical optimization techniques.

class direct.algorithms.optimization.Algorithm(max_iter=30)[source][source]#

Bases: ABC

Base class for implementing mathematical optimization algorithms.

done()[source][source]#

Check if the algorithm has converged.

Returns:
bool

Whether the algorithm has converged or not.

Return type:

bool

fit(*args, **kwargs)[source][source]#

Fit the algorithm.

Parameters:
*argstuple

Tuple of arguments for _fit method.

**kwargsdict

Keyword arguments for _fit method.

Return type:

None

update()[source][source]#

Update the algorithm’s parameters and increment the iteration count.

Return type:

None

class direct.algorithms.optimization.MaximumEigenvaluePowerMethod(forward_operator, norm_func=None, max_iter=30)[source][source]#

Bases: Algorithm

A class for solving the maximum eigenvalue problem using the Power Method.

Module contents#

Direct module for traditional mathematical optimization techniques, general or mri-specific.