Signal-Space Projection (SSP) Method for Separating MEG or EEG into Components

摘要

We examine the inverse problem of estimating current configurations on the basisof measured magnetic and/or electric fields. We consider the measured signals as a linearly weighted sum of time-varying, but spatially fixed components. First, the most evidence components are determined. Then, the measured signals are divided into two parts, the first giving the contribution of the determined sources and the other the rest of the signals. Both parts can be used for further analysis, e.g., for successive decomposition or source localization. The method can be used in analysing signals from multiple sources, encountered, e.g., in MEG (magnetoencephalography) and EEG (electroencephalography). Errors in localization and in other current model parameters of the unknown components are shown to be inversely proportional to signal-to-noise ratio and sin theta, where theta is the angle between the signal produced by the determined activity and the modelled unknown activity. Based on this observation, the separability of sources is analysed, using Meg signals generated by current dipoles as a specific example. It is possible to use the method without any conductivity or source models. (Copyright (c) 1996 Helsinki University of Technology.)