Independent Components Analysis (ICA) in the study of electroencephalographic signals

Independent Component Analysis (ICA) is a mathematical tool able to separate complex signals in statistically independent components. It solves the blind source separation problem (BSS). The EEG satisfies most of the assumptions of ICA, so it may be an adequate signal for ICA for its use. In this paper we review the method and the applications of ICA in EEG. The studied applications are: a) artifacts removal; b) source estimation of spikes; c) analysis of seizures. Several studies have demonstrated that ICA is useful to remove artifacts from contaminated EEG records without distorting cerebral activity. It is able to decompose epileptic discharges and seizures in independent spatio-temporal components. In combination with techniques of source localisation, as dipole modelling, ICA improves the localization of the epileptic focus. Finally, we discuss the future role of ICA in the study of epileptic patients.