Constraining minimum-norm inverse by phase synchronization and signal power of the scalp EEG channels.

Abstract

In this paper, the goal is to further improve the output of the scalp EEG source localization by the Euclidean minimum-norm (MN) inverse during single trials. Trials have been selected based on signal power at specific time intervals in specific locations. Then the source localization has been performed by MN. It has been observed that close to a dominant cortical source of EEG, as determined by the MN, both pairwise phase synchronization of a channel with its nearest neighbors and the cumulative signal power of the channels within that neighborhood become high (normalized values remain above certain thresholds). This has also been verified through simulations on the subject's real head model. The conclusion of our study is that only those sources are to be chosen for which MN inverse, and signal power and phase synchronization profile converge. A novel fast Fourier transform (FFT) based phase synchronization measuring algorithm between a pair of signals has been developed whose time complexity is no more than that of the FFT.