Inverse source estimation (MNE, dSPM, sLORETA, dipole fitting). More...

Classes
class	InvBeamformer
	Computed beamformer spatial filter container. More...
class	InvBeamformerCompute
	Core beamformer math. More...
class	InvDICS
	DICS beamformer (frequency-domain). More...
class	InvLCMV
	LCMV beamformer (time-domain). More...
class	InvDipoleFit
	High-level driver for sequential dipole fitting. More...
struct	dipoleFitFuncsRec
	Forward field computation function pointers and client data for MEG and EEG dipole fitting. More...
struct	FitDipUserRec
	Workspace for the dipole fitting objective function, holding forward model, measured field, and fit limits. More...
class	InvDipoleFitData
	Dipole fit workspace holding sensor geometry, forward model, noise covariance, and projection data. More...
class	InvDipoleFitSettings
	Dipole Fit setting implementation. More...
class	InvDipoleForward
	Stores forward field matrices and SVD decomposition for magnetic dipole fitting. More...
class	InvEcd
	Single equivalent current dipole with position, orientation, amplitude, and goodness-of-fit. More...
struct	bdipEcd
	Binary-format dipole record for file I/O, storing fitted dipole parameters and error estimates. More...
class	InvEcdSet
	Holds a set of Electric Current Dipoles. More...
class	InvGuessData
	Precomputed guess point grid with forward fields for initial dipole position candidates. More...
class	InvHpiDataUpdater
	Preprocesses raw HPI coil data (SSP projection, compensation, sinusoidal model fitting) before dipole localization. More...
struct	CoilParam
	Estimated dipole parameters (position, moment, goodness-of-fit) for a single HPI coil. More...
struct	HpiFitResult
	Complete HPI fit output: per-coil dipole parameters, head-to-device transform, fit error, and head movement distance. More...
class	InvHpiFit
	HPI Fit algorithms. More...
struct	DipFitError
	Residual error and moment vector from a single magnetic dipole fit iteration. More...
struct	HPISortStruct
	Helper for sorting HPI coil dipole fits by matching each fit to the nearest expected coil position. More...
class	InvHpiFitData
	HPI Fit algorithm data structure. More...
class	InvHpiModelParameters
	Configuration parameters for the HPI signal model (line frequency, coil frequencies, sample rate, buffer size). More...
class	InvSensorSet
	Stores MEG sensor geometry (positions, orientations, weights, coil count) for a single sensor type. More...
class	InvSensorSetCreator
	Builds InvSensorSet objects from FiffInfo channel definitions, applying SSP projections and compensation. More...
class	InvSignalModel
	Generates the forward sinusoidal model matrix for HPI coil signals at known drive frequencies. More...
struct	InvConnectivity
	Pairwise source connectivity matrix with measure, directionality, and frequency/time metadata. More...
struct	InvFocalDipole
	Single focal dipole with free 3D position, moment, and fit-quality metrics. More...
struct	InvSourceCoupling
	N-tuple of correlated grid sources with orientations and correlation matrix. More...
class	InvSourceEstimate
	Source estimation. More...
struct	InvToken
	One element of a tokenised neural-source representation. More...
struct	InvTokenizeOptions
	Tokenization options controlling layer inclusion and sub-sampling. More...
class	InvMinimumNorm
	Minimum norm estimation. More...
class	InvDipole
	Stores position, orientation, and correlation of a single current dipole estimated by RAP MUSIC. More...
struct	InvDipolePair
	Pair of correlated dipole indices and orientations found by the RAP MUSIC scanning step. More...
class	InvPwlRapMusic
	Powell-accelerated RAP MUSIC variant using gradient-free optimization for refined dipole fitting. More...
struct	Pair
	Index pair representing two grid points evaluated together in the RAP MUSIC subspace scan. More...
class	InvRapMusic
	RAP MUSIC (Recursively Applied and Projected Multiple Signal Classification) source localization algorithm. More...

Typedefs
using	dipoleFitFuncs = dipoleFitFuncsRec*
	Pointer alias for dipoleFitFuncsRec, used throughout the dipole fitting module.
typedef struct INVLIB::bdipEcd	bdipEcdRec
typedef struct INVLIB::Pair	Pair
	Index pair representing two grid points evaluated together in the RAP MUSIC subspace scan.

Enumerations
enum class	BeamformerWeightNorm { None = 0 , UnitNoiseGain , NAI , UnitNoiseGainInv }
enum class	BeamformerPickOri { None = 0 , Normal , MaxPower , Vector }
enum class	BeamformerInversion { Matrix = 0 , Single }
enum class	Accuracy : int { high = 2 , medium = 1 , low = 0 }
enum class	InvTokenId : int32_t { Pad = 0 , Bos = 1 , Eos = 2 , Sep = 3 , MetaBegin = 10 , MetaEnd = 11 , GridBegin = 20 , GridEnd = 21 , GridRow = 22 , CouplingBegin = 30 , CouplingEnd = 31 , GroupBegin = 32 , GroupEnd = 33 , FocalBegin = 40 , FocalEnd = 41 , DipoleBegin = 42 , DipoleEnd = 43 , ConnBegin = 50 , ConnEnd = 51 , ConnEntryBegin = 52 , ConnEntryEnd = 53 , PosBegin = 60 , PosEnd = 61 , MethodUnknown = 100 , MethodMNE = 101 , MethodDSPM = 102 , MethodSLORETA = 103 , MethodELORETA = 104 , MethodLCMV = 105 , MethodDICS = 106 , MethodSAM = 107 , MethodMixedNorm = 108 , MethodGammaMAP = 109 , MethodDipoleFit = 110 , MethodRapMusic = 111 , MethodPwlRapMusic = 112 , SpaceUnknown = 150 , SpaceSurface = 151 , SpaceVolume = 152 , SpaceMixed = 153 , SpaceDiscrete = 154 , OrientUnknown = 170 , OrientFixed = 171 , OrientFree = 172 , OrientLoose = 173 , Amplitude = 200 , Vertex = 201 , TimeVal = 202 , FreqVal = 203 , PosX = 204 , PosY = 205 , PosZ = 206 , MomX = 207 , MomY = 208 , MomZ = 209 , Goodness = 210 , ChiSquared = 211 , Correlation = 212 , ConnValue = 213 , GridIndex = 214 , ValidTrue = 250 , ValidFalse = 251 , DirectedTrue = 252 , DirectedFalse = 253 , NSources = 280 , NTimes = 281 , NGroups = 282 , NDipoles = 283 , NMeasures = 284 , NIndices = 285 , NFreeDof = 286 , TStep = 287 , MeasCoh = 300 , MeasImCoh = 301 , MeasPlv = 302 , MeasPli = 303 , MeasWpli = 304 , MeasGranger = 305 , MeasPdc = 306 , MeasDtf = 307 , MeasCorrelation = 308 , MeasCrossCorr = 309 , MeasOther = 310 , QuantBinBase = 1000 }
enum class	InvEstimateMethod { Unknown = 0 , MNE , dSPM , sLORETA , eLORETA , LCMV , DICS , SAM , MixedNorm , GammaMAP , DipoleFit , RapMusic , PwlRapMusic }
enum class	InvSourceSpaceType { Unknown = 0 , Surface , Volume , Mixed , Discrete }
enum class	InvOrientationType { Unknown = 0 , Fixed , Free , Loose }

Functions
INVSHARED_EXPORT const char *	buildDateTime ()
INVSHARED_EXPORT const char *	buildHash ()
INVSHARED_EXPORT const char *	buildHashLong ()
INVSHARED_EXPORT std::vector< InvToken >	tokenize (const InvSourceEstimate &estimate, const InvTokenizeOptions &options=InvTokenizeOptions())
	Serialise an InvSourceEstimate into a flat token sequence.
INVSHARED_EXPORT InvSourceEstimate	fromTokens (const std::vector< InvToken > &tokens)
	Reconstruct an InvSourceEstimate from a token sequence.
std::vector< int32_t >	tokenIds (const std::vector< InvToken > &tokens)
std::vector< float >	tokenValues (const std::vector< InvToken > &tokens)

Detailed Description

Inverse source estimation (MNE, dSPM, sLORETA, dipole fitting).

Typedef Documentation

◆ bdipEcdRec

typedef struct INVLIB::bdipEcd INVLIB::bdipEcdRec

◆ dipoleFitFuncs

using INVLIB::dipoleFitFuncs = dipoleFitFuncsRec*

Pointer alias for dipoleFitFuncsRec, used throughout the dipole fitting module.

Definition at line 106 of file inv_dipole_fit_data.h.

◆ Pair

typedef struct INVLIB::Pair INVLIB::Pair

Index pair representing two grid points evaluated together in the RAP MUSIC subspace scan.

Declares a pair structure for index combinations used in RAP MUSIC algorithm.

Enumeration Type Documentation

◆ Accuracy

enum class INVLIB::Accuracy : int

strong

Enumerator
high
medium
low

Definition at line 78 of file inv_sensor_set.h.

◆ BeamformerInversion

enum class INVLIB::BeamformerInversion

strong

Method for inverting the beamformer denominator matrix G^H Cm^{-1} G.

Enumerator
Matrix	Full matrix inversion per source.
Single	Scalar (diagonal) inversion per source.

Definition at line 77 of file inv_beamformer_settings.h.

◆ BeamformerPickOri

enum class INVLIB::BeamformerPickOri

strong

Orientation picking mode for beamformer filters.

Enumerator
None	Keep all orientations (3 per source for free, 1 for fixed).
Normal	Extract surface normal component only (Z in local coords).
MaxPower	Optimal orientation via max eigenvalue of power matrix.
Vector	Return the full 3-component vector solution.

Definition at line 65 of file inv_beamformer_settings.h.

◆ BeamformerWeightNorm

enum class INVLIB::BeamformerWeightNorm

strong

Weight normalization strategy for beamformer spatial filters.

References:

Sekihara & Nagarajan, "Adaptive Spatial Filters for Electromagnetic Brain Imaging", 2008.
Van Veen et al., IEEE Trans. Biomed. Eng. 44(9), 867-880, 1997.

Enumerator
None	No normalization (raw unit-gain filter).
UnitNoiseGain	Normalize by sqrt(diag(W @ W^H)), Sekihara 2008.
NAI	Neural Activity Index: unit-noise-gain + noise subspace normalization.
UnitNoiseGainInv	Rotation-invariant unit-noise-gain via sqrtm(inner)^{-0.5}.

Definition at line 53 of file inv_beamformer_settings.h.

◆ InvEstimateMethod

enum class INVLIB::InvEstimateMethod

strong

Enumerates the inverse estimation method that produced a source estimate.

Enumerator
Unknown
MNE
dSPM
sLORETA
eLORETA
LCMV
DICS
SAM
MixedNorm
GammaMAP
DipoleFit
RapMusic
PwlRapMusic

Definition at line 49 of file inv_types.h.

◆ InvOrientationType

enum class INVLIB::InvOrientationType

strong

Enumerates the orientation constraint used during source estimation.

Enumerator
Unknown
Fixed
Free
Loose

Definition at line 83 of file inv_types.h.

◆ InvSourceSpaceType

enum class INVLIB::InvSourceSpaceType

strong

Enumerates the source space type underlying a source estimate.

Enumerator
Unknown
Surface
Volume
Mixed
Discrete

Definition at line 70 of file inv_types.h.

◆ InvTokenId

enum class INVLIB::InvTokenId : int32_t

strong

Token vocabulary for the InvSourceEstimate tokenization format.

IDs are organised into ranges: 0..9 Special tokens (PAD, BOS, EOS, SEP) 10..19 Metadata section markers 20..29 Grid section markers 30..39 Coupling section markers 40..49 Focal-dipole section markers 50..59 Connectivity section markers 60..69 Positions section markers 100..149 Method labels 150..169 Source-space labels 170..189 Orientation labels 200..249 Value-carrier tokens (id + continuous float) 250..279 Boolean tokens 280..299 Dimension/size tokens (id + int-as-float) 300..349 Connectivity measure name tokens 1000+ Reserved for external quantisation bins

Enumerator
Pad
Bos	Beginning of sequence.
Eos	End of sequence.
Sep	Separator (between modalities, optional).
MetaBegin
MetaEnd
GridBegin
GridEnd
GridRow	Marks the start of one source's time-course.
CouplingBegin
CouplingEnd
GroupBegin	Start of one N-tuple coupling group.
GroupEnd
FocalBegin
FocalEnd
DipoleBegin	Start of one focal dipole.
DipoleEnd
ConnBegin
ConnEnd
ConnEntryBegin	Start of one connectivity measure.
ConnEntryEnd
PosBegin
PosEnd
MethodUnknown
MethodMNE
MethodDSPM
MethodSLORETA
MethodELORETA
MethodLCMV
MethodDICS
MethodSAM
MethodMixedNorm
MethodGammaMAP
MethodDipoleFit
MethodRapMusic
MethodPwlRapMusic
SpaceUnknown
SpaceSurface
SpaceVolume
SpaceMixed
SpaceDiscrete
OrientUnknown
OrientFixed
OrientFree
OrientLoose
Amplitude	Source amplitude at one (source, time) sample.
Vertex	Vertex / grid index.
TimeVal	Time in seconds.
FreqVal	Frequency in Hz.
PosX	3-D position x component (m).
PosY	3-D position y component (m).
PosZ	3-D position z component (m).
MomX	Dipole moment x component (Am).
MomY	Dipole moment y component (Am).
MomZ	Dipole moment z component (Am).
Goodness	Goodness-of-fit (0..1).
ChiSquared	Chi-squared fit quality.
Correlation	Coupling correlation value.
ConnValue	Pairwise connectivity value.
GridIndex	Grid reference index inside a coupling group.
ValidTrue
ValidFalse
DirectedTrue
DirectedFalse
NSources	Number of grid sources.
NTimes	Number of time samples.
NGroups	Number of coupling groups.
NDipoles	Number of focal dipoles.
NMeasures	Number of connectivity measures.
NIndices	Number of indices in a coupling group.
NFreeDof	Degrees of freedom.
TStep	Time step (s) — grid sampling interval.
MeasCoh
MeasImCoh
MeasPlv
MeasPli
MeasWpli
MeasGranger
MeasPdc
MeasDtf
MeasCorrelation
MeasCrossCorr
MeasOther	Unrecognised metric (name not preserved in token stream).
QuantBinBase	Bins occupy IDs [QuantBinBase .. QuantBinBase + numBins - 1].

Definition at line 99 of file inv_token.h.

Function Documentation

◆ buildDateTime()

const char * INVLIB::buildDateTime ( )

Returns build date and time.

Definition at line 46 of file inv_global.cpp.

◆ buildHash()

const char * INVLIB::buildHash ( )

Returns abbreviated build git hash.

Definition at line 50 of file inv_global.cpp.

◆ buildHashLong()

const char * INVLIB::buildHashLong ( )

Returns full build git hash.

Definition at line 54 of file inv_global.cpp.

◆ fromTokens()

INVSHARED_EXPORT InvSourceEstimate INVLIB::fromTokens ( const std::vector< InvToken > & tokens )

Reconstruct an InvSourceEstimate from a token sequence.

Parses a token stream previously produced by tokenize() and rebuilds the InvSourceEstimate fields (metadata, grid data, positions, couplings, focal dipoles, connectivity). Unknown or out-of-order tokens are silently skipped.

Parameters

[in] tokens The token sequence to decode.

Returns: The reconstructed source estimate.

See also: tokenize, InvToken

◆ tokenIds()

std::vector< int32_t > INVLIB::tokenIds ( const std::vector< InvToken > & tokens )

inline

Extract just the integer token IDs from a token sequence (discarding continuous values). Useful for feeding into a purely discrete embedding table.

Parameters

[in] tokens The token sequence.

Returns: Vector of int32_t token IDs.

Definition at line 261 of file inv_token.h.

◆ tokenize()

INVSHARED_EXPORT std::vector< InvToken > INVLIB::tokenize	(	const InvSourceEstimate &	estimate,
		const InvTokenizeOptions &	options = InvTokenizeOptions() )

Serialise an InvSourceEstimate into a flat token sequence.

Converts the full multimodal content of an InvSourceEstimate (grid amplitudes, positions, couplings, focal dipoles, connectivity matrices) into a linear sequence of InvToken elements. Each token carries a vocabulary ID and an optional continuous float value.

The options parameter controls which data layers are emitted and allows sub-sampling of dense grids to fit context-limited attention windows.

Parameters

[in]	estimate	The source estimate to tokenize.
[in]	options	Controls layer inclusion and sub-sampling (default: all layers, no sub-sampling).

Returns: A vector of InvToken representing the serialised estimate.

See also: fromTokens, InvTokenizeOptions, InvToken

◆ tokenValues()

std::vector< float > INVLIB::tokenValues ( const std::vector< InvToken > & tokens )

inline

Extract just the continuous values from a token sequence (discarding IDs). Useful as the real-valued branch in a hybrid discrete-continuous embedding.

Parameters

[in] tokens The token sequence.

Returns: Vector of float values.

Definition at line 277 of file inv_token.h.

Classes

Typedefs

Enumerations

Functions

Detailed Description

Typedef Documentation

◆ bdipEcdRec

◆ dipoleFitFuncs

◆ Pair

Enumeration Type Documentation

◆ Accuracy

◆ BeamformerInversion

◆ BeamformerPickOri

◆ BeamformerWeightNorm

◆ InvEstimateMethod

◆ InvOrientationType

◆ InvSourceSpaceType

◆ InvTokenId

Function Documentation

◆ buildDateTime()

◆ buildHash()

◆ buildHashLong()

◆ fromTokens()

◆ tokenIds()

◆ tokenize()

◆ tokenValues()