Estimation of multiple sources using spatio-temporal data on a three-dimensional measurement of MEG

This paper describes a computer simulation study for estimating the multiple sources in the brain based on a three-dimensional measurement of magnetoencephalogram (3-D MEG). We propose a new source analysis method using spatio-temporal data of 3-D MEG and singular value decomposition analysis. The number of estimation sources is decided by the number of dominant singular values of spatio-temporal 3-D MEG data. The source localization is done as pattern matching by signal subspace of singular vectors and projection of an estimating source. In order to assess this algorithm, inverse calculation to the MEG model consisting of three sources located in the brain was performed. The results showed that three sources were dearly discriminated, and that the estimation error was less than 1 mm with SNR=5. The proposed method is useful for estimating multiple sources overlapping in time.