Simultaneous extraction and localization of dipolar independent components in evoked potentials

Abstract

EEG stimulus-related responses have been extensively studied to gain insight on the functional behavior of the brain. Traditionally, these responses have been considered as the result of the generation of low-amplitude evoked potentials (EP). When averaged, these low-amplitude potentials come up from the background and can be cleanly observed. Independent component analysis (ICA) is a technique widely used to solve the problem of blind source separation (BSS). When applied to EP, ICA provides a method to obtain activation signals of neural structures responsible for the generation of several components of EP. ICA algorithms may be modified in order to impose some constraints on the independent components (IC) to be extracted or the mixing matrix, resulting in the so-called constrained ICA (cICA). Here, we make use of a cICA approach to get those IC of the EP that can be identified with point-dipolar sources, as well as their position.