NIH Workshop, "Neuronal Variability and Noise: Challenges and Promises"
Session, "Network Dynamics Evaluated with Interdependence Measures and the Latent Variable Problem"
Washington DC
September 20-21, 2002
Scott Makeig ,
Swartz Center for Computational Neuroscience,
Institute for Neural Computation, UCSD
Event-related EEG dynamics: Time-varying interdependence among cortical domains
High-dimensional recordings from the human scalp contain rich information about human cortical event-related brain dynamics. Until recently, interpretation of theses data has been severely limited by the use of simple response averaging methods originally introduced for fitting simplistic 'bottom-up' models of sensory perceptual processing. Recent neurophysiological and EEG evidence suggests that brain field dynamics are intimately related to dynamic reallocation of attention, to memory-related processing, and to self-evaluation of the consequences of actions, all more 'top-down' than 'bottom-up' processes. A more adequate model of human EEG dynamics must involve, first, spatial unmixing of contributions from quasi-independent domains of synchrony, for which Independent Component Analysis (ICA) is proving useful, followed by modeling of dynamic interrelations between EEG domains, which may be approached using time/frequency analysis. First results of the combined approach suggest new conceptions of macroscopic brain dynamics supporting cognitive events. More adequate models and analysis methods should also permit closer integration of scalp field, local field and single-unit spike information, pointing toward the emergence of a more unified science of brain electrophysiology.