A MULTIPLE SOURCE APPROACH TO THE CORRECTION OF EYE ARTIFACTS

Previously published methods correct eye artifacts by subtracting proportions of the EOG from EEG electrodes. The implicit assumption made by these methods is that the EOG signals are a good measure of eye activity and contain no EEG. In this paper a new multiple source eye correction (MSEC) method of eye artifact treatment based on multiple source analysis is presented, which incorporates a model of brain activity. An accurate, head model-independent estimate of the spatial distribution of eye activity can be obtained empirically from calibration data containing systematic eye movements and blinks. Using the resulting spatial vectors together with the brain model, eye activity in EEG and event-related response data can be estimated in the presence of overlapping brain activity and corrected. A consequence of the MSEC approach is that data at EOG electrodes can be included in analyses of brain activity. In addition, by suitable selection of the spatial vectors, the eye activity can be split into signals which identify vertical and horizontal movements and eyeblinks. Using auditory ERP data sets with and without large eye artifacts, the MSEC method is compared with a ''traditional'' method in which brain activity is not modelled, particularly with respect to the spatial distribution of the corrected EEG. Traditional eye correction methods are shown to alter the spatial distribution of the EEG, resulting, for example, in changes in location and orientation of modelled equivalent Sources. Such distortion is much reduced in the MSEC method, thus enhancing the precision of topographical EEG analyses.