Independent Component Analysis reveals the spatial sources of the visual P300
8th Joint Symposium on Neural Computation, Salk Institute, La Jolla, May 19, 2001 Marissa Westerfield , Scott Makeig, T-P. Jung, J. Townsend, E. Courchesne, T. J. Sejnowski
The Salk Institute and Swartz Center for Computational Neuroscience Institute for Neural Computation University of California San Diego
We demonstrate a decomposition of the P300 feature of the target-evoked averaged event-related potential (ERP) from a sustained visual spatial attention task into functionally independent subcomponents that may represent contributions from brain networks producing the ongoing EEG. EEG was recorded from 15 subjects during responding to visual target stimuli presented interleaved with nontarget stimuli at five attended and non-attended screen locations. Independent Component Analysis was performed on single trial EEG data: 1-sec epochs extracted from each target-centered EEG epoch were concatenated and decomposed. The resulting components were clustered into groups according to similarity of the component scalp projections and frequency distributions. One central-posterior cluster with dominant frequencies in the delta, theta and alpha bands accounted for the majority of the scalp-recorded P300 (P3b). Other contributing clusters included a fronto-central, theta-dominant cluster and several posterior alpha-dominated clusters. These results model scalp-recorded ERPs as summing stimulus-induced perturbations in ongoing EEG processes rather than as a series of monophasic stimulus-evoked potentials.