Source Reconstruction of Resting-State MEG and EEG Activity: A Technical Note on the Choice of Noise Covariance

This technical note proposes a source-level method to define the noise covariance for minimum variance beamforming of resting-state MEG and EEG data. The approach models a maximum-entropy "ground state" of brain activity as a uniform distribution of uncorrelated, randomly oriented neural dipoles, which is then projected to the sensor level. This captures realistic spatial correlations, unlike conventional solutions like empty-room recordings or diagonal white noise. Applied to real human data, the method produces structured, non-uniform sensor covariance and smooth, plausible source reconstructions, free from artificial peaks. This source-level approach provides a principled and physiologically grounded baseline for beamforming, improving the reliability of resting-state analyses and interpretation.