Society for Neuroscience, San Diego, Nov 2007

New methods for analyzing trial-to-trial EEG spectral variability

Scott Makeig* and Julie Onton

Event-related potential (ERP) and spectral perturbation (ERSP) measures are typically assumed to reveal EEG activities specifically associated with brain responses to selected task events while canceling out other concurrent ('random') brain variability. By contrast, we assume that all changes in EEG brain activity index the actions and interactions of meaningful brain processes. To reveal these, we first spatially decompose EEG data by independent component analysis (ICA) into maximally distinct or temporally independent component (IC) activities. Under favorable circumstances, many separated ICs can be localized to specific cortical areas. However, even trial averages of IC activities or activity spectra still ignore the significant variability in these activities across single experimental trials or time windows. Here, we propose new methods for studying trial-to-trial or moment-to-moment variability in the log power spectra of IC processes. Our aim is to identify characteristic modes of spectral modulation of single or multiple EEG processes, and the relationship of these modes to specific cognitive demands. We compare and contrast three new ICA decomposition methods: 1) Template-Weight (TW) decompositions into frequency template and independent trial weight matrices, respectively (see Onton et al., Neuroimage, 2005).; 2) Weight-Template (WT) decompositions into time weight and independent frequency template matrices; 3) Context ICA (XICA) decomposition. TW or WT decompositions may be performed on single trials from one or more components from one or more subjects. XICA, a TW log spectral decomposition that incorporates behavioral and event information for each trial and previous trials, identifies modes of spectral activity associated with specific types of event context, e.g., with specific 'brain challenges'. We believe the new methods may help clarify the functions of EEG activity.