Reduction of ocular artefacts in source current density brain mappings by ARX2 filtering

Sweep by sweep analysis of event-related potentials (ERP) of the human scalp represents a reliable tool for both the diagnosis of neurologic diseases and the study of the central nervous system during cognitive tasks. The off-line procedure based on stochastic parametric identification and filtering herewith described, allows an accurate analysis of single-sweep ERP and a drastic reduction of ocular artefacts variously propagating through the skull. Moreover, the spatial distribution of the recorded ERP in bidimensional form was enhanced by using the Laplacian operator in order to get an estimate of the source current density (SCD) flow from the skull into the scalp. Complete single-trial signals were filtered according to an autoregressive model of signal generation with 2 exogenous inputs (ARX2). The ARX2 procedure models the recorded signal as the sum of three signals: (a) the background EEG activity, modelled as an autoregressive process driven by a white noise; (b) a filtered version of a reference signal carrying the average information contained in each sweep; (c) a signal due to the ocular artefact propagation. The evaluation of the effect of artefact suppression on those channels close to the eyes was compared with standard ordinary least squares method (OLS) based on a linear model of the influence of EOG on ERP. Finally, the better results obtainable through ARX filtering on sweep-by-sweep brain mappings are also presented.