Biomagnetic inverse analysis based on decorrelation

In biomagnetic signal source activities, some are correlated, some are uncorrelated, and others are indifferent. If MEG (Magnetoencephalography) data have high signal to noise ratio, then the dominant components of the data are generated from such correlated, uncorrelated, and indifferent signal source activities. In this case, clustering of the signal sources in correlation and uncorrelation senses may be useful. Let the brain be discretized to small lattice points having current dipoles and a certain forward model for calculating magnetic field over the brain from a current dipole be assumed, the brain can be described mathematically by a linear system equation such that the clustering can be treated as simply grouping of the discretized elements. In this paper, we propose a concept of a biomagnetic inverse analysis with care for uncorrelated signal decomposition and for signal source localization simultaneously. The well-known related studies are mostly categorized in the following three: