Dual-Core Beamformer for obtaining highly correlated neuronal networks in MEG

The Dual-Core Beamformer (DCBF) is a new lead-field based MEG inverse-modeling technique designed for localizing highly correlated networks from noisy MEG data. Conventional beamformer techniques are successful in localizing neuronal sources that are uncorrelated under poor signal-to-noise ratio (SNR) conditions. However, they fail to reconstruct multiple highly correlated sources. Though previously published dual-beamformer techniques can successfully localize multiple correlated sources, they are computationally expensive and impractical, requiring a priori information. The DCBF is able to automatically calculate optimal amplitude-weighting and dipole orientation for reconstruction, greatly reducing the computational cost of the dual-beamformer technique. Paired with a modified Powell algorithm, the DCBF can quickly identify multiple sets of correlated sources contributing to the MEG signal. Through computer simulations, we show that the DCBF quickly and accurately reconstructs source locations and their time-courses under widely varying SNR, degrees of correlation, and source strengths. Simulations also show that the DCBF identifies multiple simultaneously active correlated networks. Additionally, DCBF performance was tested using MEG data in humans. In an auditory task, the DCBF localized and reconstructed highly correlated left and right auditory responses. In a median-nerve stimulation task, the DCBF identified multiple meaningful networks of activation without any a priori information. Altogether, our results indicate that the DCBF is an effective and valuable tool for reconstructing correlated networks of neural activity from MEG recordings. Published by Elsevier Inc.