Reducing the effect of correlated brain sources in MEG using a linearly constrained spatial filter based on Minimum Norm

Magnetoencephalogram (MEG) studies rely on the use of spatial filters to find and extract the brain activity generated by neuronal currents. Two of the most used filters are the Linearly Constrained Minimum Variance beamformer (LCMV) and the Minimum Norm Estimates (MNE) non-adaptive spatial filter. These filters have different properties that can increase or decrease their performances, especially in the presence of correlated brain activity for the LCMV case, or in the presence of a poor signal to noise ratio (SNR) for the MNE case. This study introduces a filter based on the least-squares method to be used as a benchmark to decide when to use the MNE or the LCMV to increase the accuracy of the finding and estimation of the brain activity.