Representation of multi-channel evoked potential data using a dipole component model of intracranial generators: application to the auditory P300.

A number of procedures have been employed to decompose recorded scalp potential wave forms into their hypothesized constituent elements. The shortcomings of the various decomposition methods (principal components analysis, topographic components modeling, inverse dipole localization and spatio-temporal dipole modeling) are reviewed and a new dipole components model, which incorporates the strengths of the topographic components model and the spatio-temporal dipole model, is presented. This model decomposes ERPs into subcomponents reflecting the activity of dipole sources with location and orientation fixed across subjects and with the temporal activity of each dipole modeled as a decaying sinusoid. The requirement that the equivalent dipole generators be the same across subjects and experimental conditions permits analysis of inter-group differences and of the effects of experimental variables. An application of the model to data from a 3-tone auditory target detection task is presented, and equivalent dipole sources of the components of the auditory evoked potential are described. Assumptions inherent in the model, as well as practical obstacles to its widespread implementation, are discussed.