Removal of eye activity artifacts from visual event-related potentials in normal and clinical subjects

Abstract

Objectives: Electrical potentials produced by blinks and eye movements present serious problems for electroencephalographic (EEG) and event-related potential (ERP) data interpretation and analysis, particularly for analysis of data from some clinical populations. Often, all epochs contaminated by large eye artifacts are rejected as unusable, though this may prove unacceptable when blinks and eye movements occur frequently.

Methods: Frontal channels are often used as reference signals to regress out eye artifacts, but inevitably portions of relevant EEG signals also appearing in EOG channels are thereby eliminated or mixed into other scalp channels. A generally applicable adaptive method for removing artifacts from EEG records based on blind source separation by independent component analysis (ICA) (Neural Computation 7 (1995) 1129; Neural Computation 10(8) (1998) 2103; Neural Computation 11(2) (1999) 606) overcomes these limitations.

Results: Results on EEG data collected from 28 normal controls and 22 clinical subjects performing a visual selective attention task show that ICA can be used to effectively detect, separate and remove ocular artifacts from even strongly contaminated EEG recordings. The results compare favorably to those obtained using rejection or regression methods.

Conclusions: The ICA method can preserve ERP contributions from all of the recorded trials and all the recorded data channels, even when none of the single trials are artifact-free.

Introduction

Single-trial event-related potentials (ERPs) consist of brief epochs of electroencephalographic (EEG) activity time-locked to experimental events of interest. These recordings are usually averaged prior to analysis to increase their signal/noise ratio. Here, ‘noise’ includes non-phase-locked EEG signals and non-neural artifacts such as eye blinks and eye movements. However, ERP averaging may not cancel some artifacts induced by eye movements or blinks if they are time-locked to experimental events. These artifacts may seriously interfere with correct ERP analysis and interpretation. In addition, data from frontal and temporal electrodes located near the eyes or scalp muscles are often discarded since these are more heavily contaminated by artifacts than central scalp channels. Another common strategy is to reject all EEG epochs containing artifacts larger than some arbitrarily selected EEG voltage value. However, when limited data are available, or when blinks and muscle movements occur too frequently as in children and some patient groups, the amount of data lost to artifact rejection may be unacceptable. For example, Small (1971) reported a visual ERP experiment conducted on autistic children who produced electrooculographic (EOG) artifacts in nearly 100% of the trials. In this case, the presence of large background EEG signals not time- and phase-locked to experimental events may make ERP averages of the few artifact-free trials too unstable to permit useful analysis.

One approach to reducing contamination from eye movement artifacts is to regress out reference signals collected near the eyes. Regression methods have been proposed using both time domain (Hillyard and Galambos, 1970, Verleger et al., 1982) and frequency domain techniques (Whitton et al., 1978, Woestenburg et al., 1983). All regression methods, whether in time or frequency domains, depend on having one or more clean reference channels (e.g. one or more ‘EOG’ channels) which cannot be further analyzed after regression. However, these methods share an inherent weakness, in that both eye movements and EEG signals propagate to periocular (‘EOG’) sites. Therefore, regression-based artifact removal procedures also eliminate neural activity common to the reference electrodes and to other frontal electrodes. Because the regression coefficients are determined largely by the typically large EOG signals, regression methods may also introduce neural activity projecting to the reference channel into other sites (Jung et al., 2000).

Principal component analysis (PCA) has been proposed as a method to remove eye artifacts from multichannel EEG (Berg and Scherg, 1991). However, PCA cannot completely separate eye artifacts from brain signals, especially when they have comparable amplitudes (Lagerlund et al., 1997, Jung et al., 1998b, Jung et al., 2000). By combining PCA, multiple source models for EOG and EEG, and an artifact-aligned averaging method (Lins et al., 1993), Berg and Scherg (1994) demonstrated a more effective PCA-based approach to correct eye artifacts. However, the accuracy of their method depends on the availability of separate and accurate inverse source solutions for EEG and EOG. Building an accurate EEG source model requires a priori knowledge of event-related brain activity following not only stimuli, but also blinks and saccades. The method also relies on the amount and quality of separately recorded calibration data which are needed to provide estimates of the source vectors and transmission coefficients in the EOG model.

Makeig et al. (1996) proposed an approach to the analysis of EEG and ERP data based on an unsupervised neural network learning algorithm that takes a logistic infomax approach to performing independent component analysis (ICA) (Bell and Sejnowski, 1995). They showed that this ICA algorithm can be used to separate neural activity from muscle and blink artifacts in spontaneous EEG data and reported its use for finding independent components of EEG and ERP data and for tracking changes in alertness (Makeig et al., 1997, Jung et al., 1998c). Subsequent independent work (Vigário, 1997) based on a related approach verified that different artifacts can also be detected in multichannel magnetoencephalographic (MEG) recordings. However, this study did not attempt to remove the identified artifacts. Jung et al., 1998a, Jung et al., 1998b, Jung et al., 2000 introduced an ICA-based method based on an extended infomax ICA algorithm (Girolami, 1998, Lee et al., 1999). This method can be used to detect and remove a wide variety of artifacts (including eye blinks, muscle noise, heart signal, and line noise) from spontaneous EEG data.

This study demonstrates, through analysis of sample data sets collected in a visual spatial selective attention task, that the ICA-based method can also be used to remove stimulus-induced eye artifacts from single-trial ERP records. The method uses spatial filters derived by the ICA algorithm, avoiding the need for separate reference channels for each artifact source, and allowing analysis of neural ERP activity projecting to periocular (EOG) channels. Here, we analyze experimental data collected from 28 normal controls and from 22 clinical subjects (10 autistic and 12 brain lesion subjects) who had difficulty in inhibiting unwanted eye movements toward peripheral target stimuli.