Source-Space Beamforming (Spatial Filter)

A beamformer is an adaptive spatial filter that estimates the activity at one chosen source location while suppressing contributions from all others. The linearly constrained minimum-variance (LCMV) beamformer builds, for each candidate location, a channel-weight vector that minimizes total output variance subject to a unit-gain constraint on that location's lead-field vector, so signals from elsewhere are cancelled by the data-driven covariance. Scanning the weights across the source grid yields a time series for each voxel, converting sensor data into source-space activity used for localization or connectivity.

Its defining and often overlooked failure mode is that two sources with strongly correlated time courses are treated by the minimum-variance criterion as one interfering pattern and are partially or fully cancelled. This makes vanilla beamforming unreliable precisely for the bilateral, tightly coupled networks that connectivity studies most want to measure, motivating paired-dipole and array-gain variants.