A spatial filtering technique to detect and localize multiple sources in the brain

An algorithm for localization of electromagnetic activity in the central nervous system is explored. This algorithm generates a neural activity index map within the brain by passing surface recordings through a set of spatial filters. The covariance matrix of the surface recordings is used to optimize the spatial filters' responses. This approach is studied in simulated situations and in real data. The simulations show the method's capability to detect areas of activity without prior knowledge of the number of sources. The resolving power of the method increases with number of electrodes and signal-to-noise ratio, and it decreases with depth The analysis of the electrophysiological data indicates that the method can distinguish simultaneously active areas in a realistic fashion. The analyzed recordings are bilateral median SEP responses, an epoch of spike activity showing several active regions and a recording with eye movement superimposed on spike activity. The method and the results are discussed in relation to current localization techniques.