A Bayesian Spatiotemporal Model for Multi-Subject EEG

• Activity Number: 486
• Type: Invited
• Date/Time: Wednesday, August 3, 2011 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Wesley Thompson*+
• Companies: University of California at San Diego
• Keywords: EEG ; Multi-Subject ; Bayesian ; Functional Neuroimaging ; Source-Localized

Abstract

Electroencephalography (EEG) has millisecond temporal resolution but localization of neural sources inside the brain is a difficult problem. One approach is to apply independent components analysis to scalp EEG recordings to reduce the dimensionality of sources, and then to use a forward head model to determine the best-fitting source dipole location. Applied to different subjects, this typically results in varying numbers and locations of source dipoles across subjects. We develop a Bayesian spatiotemporal model for multi-subject source-localized EEG which exploits time-frequency domain relationships to obtain group inferences on the spatial locations and dynamic inter-relationships among localized sources. Each subject's localized sources is modeled as arising from a mixture distribution of spatial coordinates and time-varying partial coherence. Model inference is obtained via a Markov Chain Monte Carlo algorithm. The utility of this method is demonstrated by simulations and by application to a large dataset of event-related EEG.