Signal Space Separation method with block matrix inversion for extraction of signals in magnetoencephalography

摘要

Signal Space Separation (SSS) method is a technique utilized to eliminate the contribution of unwanted signals due to external magnetic noise that inevitably get recorded along with the actual magnetic signal due to neuronal currents in the brain. In this paper, the SSS method has been implemented using block matrix inversion. By implementing block matrix inversion along with regrouping of radial terms, it has been possible to extract the true brain signal from the measured magnetic signal which includes the contribution from an external magnetic dipole artifact for measurements from as few as 64 channels. We observe that the minimum root mean square deviation (RMSD) of the signal inferred from block matrix inversion with regrouped terms is around 6 fT when the truncation order is set at L_1 = 11 for signals of interest and L_2 = 2 for external noise sources and saturates to similar values for higher L,. The RMSD of extracted signal is 50% smaller than the minimum RMSD when the magnetoencephalography signal is extracted by direct pseudoinverse technique and does not have a deep minimum in the truncation order L_1 as observed when direct pseudoinverse technique is used.