INDEPENDENT COMPONENT ANALYSIS OF ELECTROENCEPHALOGRAPHIC AhJD EVENT-RELATED POTENTIAL DATA

The electroencephalogram (EEG) is a non-invasive measure of brain electrical activity recorded as changes in potential difference between points on the human scalp. Because of volume conduction through cerebrospinal fluid, skull and scalp, EEG data collected from any point on the scalp includes activity from processes occurring within a large brain volume. Event-related potentials (ERP) are the portion of the EEG both timeand phase-locked to experimental events. Through timellocked averaging across a set of equivalent experixnental events, EEG activity, but not both timeand phase-locked to event onsets, is removed by phase cancellation. What remains is called the event-related potential. In general, the problem of determining brain electrical sources from potential patterns recorded on the scalp surface is mathematically underdetermined. For several decades, ERP researchers have proposed a number of techniques to localize the sources of stimulus-evoked potentials, either by assuming that they have a known or simple spatial configuration (Scherg and Von Cramon, 1986), or that generators are restricted to a small subset of possible locations and orientations (Dale and Sereno, 1993). Recently, we reported a method for separating joint problems of source identification and source localization by applying an information-theoretic neural network algorithm, Independent Component Analysis (ICA), to multiple-channel EEG and ERP activity recorded from the scalp (Makeig et al., 1996a; S. Makeig, T-P Jung, A.J. Bell, and T.J. Sejnowski, unpublished observations). The algorithm tells what temporally independent activations compose the observed scalp recording, without specifying where in the brain these activations arise. Here, we report an application of the ICA algorithm to the analysis of 14-channel EEG and ERP data recorded during eyes-closed performance of a sustained auditory detection task (Makeig and Inlow, 1993).