CIBC:Documentation:SCIRun:Reference:BioPSE:SolveInverseProblemWithTikhonov
From NCRR Biomedical Software Development, Engineering, and Dissemination Wiki
Contents |
SolveInverseProblemWithTikhonov
Information
- Package: BioPSE
- Catagory: Inverse
- Author(s): Yesim Serinagaoglu
- Status: Supported in latest version
- Version: 3.0
Description
Summary
This module applies SolveInverseProblemWithTikhonov regularization to an existing forward model, with flexible control of regularization type and parameters. Currently, the module solves the inverse problem in a single time-instant. This module requires a forward model matrix, geometries for the associated surfaces, and some remote boundary conditions, i.e., torso or head surface potentials.
Detailed Description
The SolveInverseProblemWithTikhonov module has three inputs, two outputs, and a user interface (UI) for selecting the regularization parameter.
Forward problem matrix, A: This matrix could be created using the MODULE REFERENCE module, or created elsewhere and saved as SCIRun matrix format, which loads via the MODULE REFERENCE module.
A Regularization matrix will contain the transform that constrains the regularized inverse solution. This is typically the output of the MODULE REFERENCE module, but as in the Forward matrix, can be created externally and loaded via the MODULE REFERENCE module. If nothing is connected to this input, the identity matrix is used as default.
The Neumann Boundary conditions, y, are values of potential on the outer boundary of the geometry, typically the body surface or head surface potentials. This is the "attributes" part of a field data type.
The MODULE REFERENCE module can be used to extract the vector of attributes (i.e., the potentials) from the original field data.
Output 1 from the left is a vector of potentials without any geometry information. This vector can be combined with the appropriate geometry file using the ManageFieldData module to create a field data format.
The UI of the SolveInverseProblemWithTikhonov module selects a method to choose lambda. Thus far, we have implemented three options: 1)enter a single value: the user can type any value in the UI and the solution is implemented for that lambda value only, 2) choose from a slider: the user can select a value by moving the slider (the range of the slider and the increments are pre-defined inside the code), and 3) determine the value using the L-curve method. The range of regularization parameters used for the slider and to obtain the L-curve is user defined.
Frequently Asked Questions
Known Bugs
Recent Changes
Go back to Documentation:SCIRun:Reference:BioPSE
