Forward Library (`FWDLIB`)

The Forward library computes MEG and EEG forward solutions — the mapping from source currents in the brain to sensor measurements. It supports BEM (Boundary Element Method) models with constant or linear collocation, multi-layer concentric-sphere EEG head models, and arbitrary coil/electrode configurations. The library also provides sensor-to-surface field mapping for topographic interpolation. All classes reside in the FWDLIB namespace.

Architecture

Class Inventory

Forward Solution

Class	Description	MNE-Python	MNE-C
`Fwd`	Static wrapper for reading forward solutions from FIFF files	`mne.read_forward_solution()`	`mne_read_forward_solution`
`ComputeFwd`	Main worker that orchestrates forward solution computation: initialises source space, BEM/sphere model, coils, and dispatches computation to threads	`mne.make_forward_solution()`	Forward solver kernel
`ComputeFwdSettings`	Container for all forward computation parameters: source space path, BEM path, EEG model, coordinate frame	—	Configuration struct

BEM (Boundary Element Method)

Class	Description	MNE-Python	MNE-C
`FwdBemModel`	BEM model holding tissue surfaces (scalp, outer skull, inner skull), conductivity parameters, and the BEM solution matrix	`mne.make_bem_model()` / `mne.read_bem_solution()`	`fwdBemModel`
`FwdBemSolution`	Precomputed BEM solution matrix (n_coil × n_potential_points) mapping infinite-medium potentials to sensor readings	Internal	`fwdBemSolution`

Key BEM Methods:

Method	Description
`fwd_bem_constant_collocation_solution()`	Compute BEM solution using constant-basis collocation on triangle centroids
`fwd_bem_linear_collocation_solution()`	Compute BEM solution using linear-basis collocation on triangle vertices
`fwd_bem_inf_field()` / `fwd_bem_inf_pot()`	Compute the magnetic field / electric potential in an infinite homogeneous medium
`fwd_bem_solid_angles()`	Compute solid-angle matrix for the BEM geometry
`fwd_bem_pot_calc()` / `fwd_bem_lin_pot_calc()`	Compute potentials at electrodes given BEM solution (constant / linear collocation)
`fwd_bem_pot_grad_calc()`	Compute potentials and spatial gradients at electrodes
`make_guesses()`	Generate dipole guess-point grid within the inner-skull boundary

Coils & Sensors

Class	Description	MNE-Python	MNE-C
`FwdCoil`	Single MEG coil or EEG electrode: integration points, weights, orientation, coordinate frame	Internal (via `pick_info`)	`fwdCoil`
`FwdCoilSet`	Collection of `FwdCoil` objects representing the full sensor array; supports creation from channel info and transformation between coordinate frames	`fwd['info']`	`fwdCoilSet`
`FwdCompData`	CTF compensation coils and field computation routines for gradiometer-to-magnetometer compensation	`mne.io.CompensationGrade`	`fwdCompData`

Key Coil Methods:

Method	Description
`create_meg_coils()`	Create MEG coil set from measurement info and coordinate transform
`create_eeg_els()`	Create EEG electrode set from measurement info
`read_coil_defs()`	Read coil definitions from `coil_def.dat`
`is_axial_coil()` / `is_planar_coil()` / `is_magnetometer_coil()`	Query coil geometry type
`dup_coil_set(transform)`	Duplicate coil set and apply a coordinate transform

EEG Sphere Model

Class	Description	MNE-Python	MNE-C
`FwdEegSphereLayer`	Single concentric sphere layer with radius (relative to outermost) and conductivity	Layer in `eeg_sphere_model`	`fwdEegSphereLayer`
`FwdEegSphereModel`	Multi-layer spherical head model for EEG forward computation using Legendre polynomial series expansion	`mne.make_sphere_model()`	`fwdEegSphereModel`
`FwdEegSphereModelSet`	Collection of EEG sphere models (e.g., default four-shell plus user-defined)	—	`fwdEegSphereModelSet`

Key Sphere Methods:

Method	Description
`setup_eeg_sphere_model()`	Initialise Berg-Scherg equivalent parameters via optimisation
`eeg_multi_sphere_pot()`	Compute scalp potential for a current dipole using Legendre series
`next_legen()`	Legendre polynomial recursion helper (P, P') for series evaluation
`fwd_eeg_get_multi_sphere_model_coeff()`	Compute series expansion coefficients for the multi-layer geometry

Field Mapping / Interpolation

Class	Description	MNE-Python	MNE-C
`FwdFieldMap`	Sphere-model-based sensor-to-surface field mapping using Legendre polynomials and SVD pseudo-inverse; supports MEG and EEG with optional SSP projection	`mne.forward._field_interpolation._make_surface_mapping()`	— (ported from Python)

Key Methods:

Method	Description
`computeMegMapping()`	Compute dense sensor → surface mapping matrix for MEG
`computeEegMapping()`	Compute dense sensor → surface mapping matrix for EEG (with average reference option)

Parallelisation

Class	Description	MNE-Python	MNE-C
`FwdThreadArg`	Thread-local workspace for parallel forward computation: source range, coils, result buffer, field function pointers	Implicit in parallel loops	`fwdThreadArg`

Function pointer types for field callbacks:

Type	Signature
`fwdFieldFunc`	`(rd, Q, coils, result, client) → int` — single dipole, single orientation
`fwdVecFieldFunc`	`(rd, coils, result_3col, client) → int` — single dipole, all 3 orientations
`fwdFieldGradFunc`	`(rd, Q, coils, result, xgrad, ygrad, zgrad, client) → int` — field + spatial gradients

Usage Example

#include <fwd/fwd.h>
#include <fwd/compute_fwd/compute_fwd.h>
#include <fwd/compute_fwd/compute_fwd_settings.h>

using namespace FWDLIB;

// Option A: Read an existing forward solution
MNEForwardSolution fwd = Fwd::read_forward_solution(
    "sample_audvis-meg-eeg-oct-6-fwd.fif");

// Option B: Compute from scratch
ComputeFwdSettings settings;
settings.srcname   = "sample-oct-6-src.fif";
settings.bemname   = "sample-5120-5120-5120-bem-sol.fif";
settings.measname  = "sample_audvis_raw.fif";
settings.eeg       = true;
settings.meg       = true;

ComputeFwd compute(&settings);
compute.calculateFwd();  // result stored in compute.fwdSolution

// Option C: Sensor-to-surface field mapping
Eigen::MatrixXd megMapping =
    FwdFieldMap::computeMegMapping(info, trans, surface, sphereModel);

Doxygen Reference

For method signatures, inheritance diagrams, and source-level documentation see the auto-generated FWDLIB namespace in the Doxygen API reference.

Architecture​

Class Inventory​

Forward Solution​

BEM (Boundary Element Method)​

Coils & Sensors​

EEG Sphere Model​

Field Mapping / Interpolation​

Parallelisation​

Usage Example​

Doxygen Reference​

See Also​