Dipole modelling of MEG rhythms in time and frequency domains.

Abstract

Local generators of spontaneous brain rhythms can be identified from the measured magnetic field pattern both in the time and frequency domains when the sources are dipolar. The dipole assumption is most efficient when only a few sources contribute to the signal at one time or frequency. We have quantified the effects of different filters and spectral transformation sizes in the analysis of spontaneous magnetic activity measured simultaneously over the whole head to determine the optimal values for maximum identification probability of localized (dipolar) activity. The criteria employed were (i) the percentage of dipolar sources, (ii) the goodness-of-fit of the dipole model, (iii) the confidence volumes and (iv) spatial distributions of the source sites, and (v) the effective spectral resolving power of the various Fast Fourier transform (FFT) sizes. The systematic changes of (i-v) suggested the use of 3-5 Hz passbands around the major spectral peaks and FFT lengths of 2-3 s for extraction of the maximum amount of information from this data set. The most successful choices of filter passband and spectral transformation length for source localization simultaneously provide estimates for the inherent spectral fluctuation of cortical rhythms (0.5 Hz) and for the activation lifetime of individual sources (0.3 s).