Skip to main content

InvPwlRapMusic

Namespace: INVLIB  ·  Library: Inverse Library

#include <inv/inv_pwl_rap_music.h>

class INVLIB::InvPwlRapMusic

Powell-accelerated RAP MUSIC variant using gradient-free optimization for refined dipole fitting.

ToDo Detailed description

src/examples/ex_inv_beamformer/main.cpp
// Noise-free data from the two sources with different time courses; RAP-MUSIC scans source pairs.
const int nSamples = 60;
MatrixXd measured(G.rows(), nSamples);
for (int t = 0; t < nSamples; ++t)
    measured.col(t) = g1 * std::sin(0.21 * t) + g2 * std::sin(0.53 * t + 0.4);
QList<InvDipolePair<double>> pairs;
InvRapMusic rap(fwd, false, 1, 0.5); // one correlated pair, subspace correlation >= 0.5
rap.calculateInverse(measured, pairs);
QList<InvDipolePair<double>> powellPairs;
InvPwlRapMusic pwlRap(fwd, false, 1, 0.5); // Powell coordinate search: far fewer pairs, may stop at a local optimum
pwlRap.calculateInverse(measured, powellPairs);

Inheritance​

Public Methods​

InvPwlRapMusic()​

Default constructor creates an empty POWELL RAP MUSIC algorithm which still needs to be initialized.

InvPwlRapMusic(p_pFwd, p_bSparsed, p_iN, p_dThr)​

Constructor which initializes the POWELL RAP MUSIC algorithm with the given model.

Parameters:

  • p_pFwd : MNEForwardSolution & The model which contains the gain matrix and its corresponding grid matrix.

  • p_bSparsed : bool True when sparse matrices should be used.

  • p_iN : int The number (default 2) of uncorrelated sources, which should be found. Starting with. the strongest.

  • p_dThr : double The correlation threshold (default 0.5) at which the search for sources stops.

~InvPwlRapMusic()​

calculateInverse(p_fiffEvoked, pick_normal)​

Note: Since they are virtual they have to be implemented to be called.

Even so the base class RAP MUSIC implementation is called.

Parameters:

  • p_fiffEvoked : const FiffEvoked & Evoked data to localize.

  • pick_normal : bool Keep only the component normal to the cortex.

calculateInverse(data, tmin, tstep, pick_normal)​

Note: Since they are virtual they have to be implemented to be called.

Even so the base class RAP MUSIC implementation is called.

Parameters:

  • data : const Eigen::MatrixXd & Measurement data (channels x samples).

  • tmin : float Time of the first sample in seconds.

  • tstep : float Time between samples in seconds.

  • pick_normal : bool Keep only the component normal to the cortex.

calculateInverse(p_matMeasurement, p_RapDipoles)​

getName()​

Static Methods​

PowellOffset(p_iRow, p_iNumPoints)​

PowellIdxVec(p_iRow, p_iNumPoints, p_pVecElements)​

Example​

Used in examples​

ex_inv_beamformer​

Inverse library: LCMV and DICS beamformers and the RAP-MUSIC family of dipole scanners.

src/examples/ex_inv_beamformer/main.cpp, line 157
InvRapMusic rap(fwd, false, 1, 0.5); // one correlated pair, subspace correlation >= 0.5
rap.calculateInverse(measured, pairs);
QList<InvDipolePair<double>> powellPairs;
InvPwlRapMusic pwlRap(fwd, false, 1, 0.5); // Powell coordinate search: far fewer pairs, may stop at a local optimum
pwlRap.calculateInverse(measured, powellPairs);
ok &= expect(pairs.size() == 1 && isPair(pairs[0], 10, 50) && pairs[0].m_vCorrelation > 1.0 - 1e-6 && powellPairs.size() == 1 && powellPairs[0].m_vCorrelation > 0.99,
             QString("InvRapMusic/InvPwlRapMusic/InvDipolePair: the exhaustive scan finds sources 10 and 50 (correlation 1), Powell a pair at %1")
                 .arg(powellPairs.isEmpty() ? 0.0 : powellPairs[0].m_vCorrelation, 0, 'f', 4));

ex_clustered_inverse_pwl_rap_music_raw​

Example of the computation of a raw clustered inverse Powell RAP-MUSIC.

src/examples/ex_clustered_inverse_pwl_rap_music_raw/main.cpp, line 433
//
// Compute inverse solution
//
InvPwlRapMusic t_pwlRapMusic(t_clusteredFwd, false, numDipolePairs);

ex_inverse_pwl_rap_music​

Example of MNEForwardSolution and Powell Rap Music application.

src/examples/ex_inverse_pwl_rap_music/main.cpp, line 152
InvPwlRapMusic t_pwlRapMusic(t_clusteredFwd, false, numDipolePairs);

Also used in ex_roi_clustered_inverse_pwl_rap_music, ex_st_clustered_inverse_pwl_rap_music.

Authors of this file​