Brain Connectivity 2006 Sendai, Japan
May 17-20, 2006
Scott Makeig , Swartz Center for Computational Neuroscience, Institute for Neural Computation, University of California San Diego, La Jolla, CA
Identifying Patterns of Distributed Dynamics in EEG Data
A leading-edge goal of systems neuroscience is to model cooperative brain processes that couple very high-order microscale brain dynamics to macroscopic dynamic patterns to select and guide appropriate behavior. Traiditonal dynamic imaging methods including single-cell spike counting and single-channel EEG averaging long have limited the possibility of identifying and modeling distributed brain dynamics. Our work in human EEG centers on identifying the activity of functionally distinct brain areas and it discovering how they cooperate to guide 'top-down' behavioral choices in light of the anticipated consequences of events. Our approach begins with spatial filtering for information sources in high-density EEG data using independent component analysis (ICA). Next, we apply event-related time/frequency analysis to the recovered cortical source activity time courses. Trial-by-trial analysis and visualization methods easily isolate a variety of (additive) 'evoked' and (multiplicative) 'induced' event-related phenomena. Detailed examination of trial statistics beyond means and mean differences show that most endogenous EEG activities vary strongly from trial to trial in both power and frequency contents. ICA methods for isolating independent spectral modulations and co-modulations show promise for revealing event-related actions of cortical modulatory systems. They demonstrate that EEG rhythms are only rarely sinusoidal, despite their typical measurement using Fourier spectra. Spectral comodulation does not require phase-locking of the affected EEG source activities. Event-related coherence or phase coherence between maximally independent EEG componsnts, while rare, reveals a class of distributed, brief (half-second), theta band events that appear at moments suggesting that they express corticolimbic dynamics associated with phasic release of dopamine. These and related methods may soon reveal much more about how distributed macroscopic brain dynamics support active cognition.