Shared math kernels (filter derivation, source power, regularised pseudo-inverse) used by both the LCMV and DICS beamformers. More...


Go to the source code of this file.
Classes | |
| class | INVLIB::InvBeamformerCompute |
| Core beamformer math. More... | |
Namespaces | |
| namespace | INVLIB |
| Inverse source estimation (MNE, dSPM, sLORETA, dipole fitting). | |
Shared math kernels (filter derivation, source power, regularised pseudo-inverse) used by both the LCMV and DICS beamformers.
SPDX-License-Identifier: BSD-3-Clause Copyright (c) 2026 MNE-CPP Authors
INVLIB::InvBeamformerCompute is a stateless helper that holds the numerical core of every beamformer in INVLIB. It implements the unit-gain filter \(W_{ug} = (G^{T} C_m^{-1} G)^{-1} G^{T} C_m^{-1}\), the Sekihara unit-noise-gain normalisation, the NAI variant, the per-source max-power orientation search, the regularised symmetric pseudo-inverse used to invert the data covariance / CSD, and the rank-reduction step for radially-deficient MEG sphere models. Pulling this code out keeps InvLCMV and InvDICS as thin orchestration layers and ensures both algorithms share the same numerically-robust kernel.
References:
Definition in file inv_beamformer_compute.h.