The effects of ocular artifacts on (lateralized) broadband power in the EEG

Abstract

Objective: Empirical evidence suggests that blinks and eye movements do not generate substantial activity outside the delta and theta range, and that the propagation of ocular activity to the EEG is rather symmetrical. These observations suggest that an alteration of the alpha and beta asymmetry of the EEG due to ocular artifacts is not likely to occur. The aim of the present study is to examine the effects of ocular artifacts on broadband EEG parameters.

Methods: EEG and EOG were recorded from 31 participants in a resting condition with eyes open and closed, allowing for spontaneous ocular activity. General effects of ocular artifacts were examined with mean comparisons, and differential effects were examined with correlation analysis of data portions that were selected for a presence or absence of artifacts.

Results: At single sites, blinks and eye movements exerted substantial general effects on the whole EEG spectrum, but there were no substantial differential effects of artifacts in the alpha and beta bands, except at the frontopolar sites. The distorting effects of ocular artifacts were smaller in magnitude for asymmetry than for single site measures.

Conclusions: The control of ocular artifacts may be dispensable for correlation analyses of alpha or beta band parameters.

Introduction

The contamination of the electroencephalogram (EEG) with potentials that originate from movements of the eyelid or eyeball are a nuisance for researchers who investigate the electrophysiology of the brain (e.g. Pivik et al., 1993, Picton et al., 2000). The distorting effects of these ocular artifacts on the EEG are very evident in the time domain, and adequate procedures for their control in studies of evoked potentials are indispensable (e.g. Barlow, 1986, Brunia et al., 1989, Gratton, 1998, Picton et al., 2000). However, blinks and eye movements do not generate substantial activity outside the delta and theta range (e.g. Gasser et al., 1985, Gasser et al., 1986). Furthermore, electric field potentials decrease with increasing distance to their source (e.g. Nunez, 1981). In addition, the propagation of ocular activity to the EEG appears to be rather symmetrical (e.g. Gasser et al., 1985, van den Berg-Lenssen and Brunia, 1989). The total of these arguments suggests that ocular artifacts are not likely to distort alpha and beta asymmetry in the EEG. The present article reviews the literature that sustains these arguments, and reports on an empirical examination of the effects of blinks and eye movements on the electrooculogram (EOG) and EEG.

The cornea of the eye has a positive electric charge compared with the retina, and both structures form an electric dipole. This dipole generates an electric field that propagates across the head, and rotating movements of the eyeballs or lid movements result in alterations of this electric field. In particular, in the vicinity of the eyes at anterior scalp regions, the potentials that are generated by ocular activity interfere with the potentials that are generated by the brain. In consequence, the derivations of the EEG reflect neural potentials that are superimposed with ocular potentials or ocular artifacts (for an overview of the generator mechanisms of ocular artifacts, see Elbert et al., 1985, Brunia et al., 1989, Gratton, 1998).

Different ocular artifacts may be distinguished (for an overview, see Barlow (1986) and the literature cited there). Blink artifacts result from a rapid movement of the eyelid and cause a characteristic peak potential in the spontaneous EEG, with an amplitude of up to 800 μV and a duration of about 200–400 ms. Eye movement artifacts (e.g. saccadic eye movements, smooth-pursuit movements, vertical movements) result from a rotation of the eyeball, and are usually of smaller magnitude but longer duration (Brunia et al., 1989). In addition to their different waveforms in the EEG recordings, blink and eye movement artifacts also show differences in the magnitude of propagation across the scalp, in topography, and spectrum (e.g. Corby and Kopell, 1972, Gratton et al., 1983, Gasser et al., 1985).

Since ocular artifacts can be of a much greater amplitude than the neural activity in the EEG, one or another control procedure is obligatory for the analysis of electrical brain activity (Barlow, 1986, Gevins, 1987, Brunia et al., 1989). One common procedure to control for ocular artifacts is the rejection of EEG segments that show contamination with ocular activity (for an overview, see Barlow, 1986). In another approach, ocular activity is removed from the EEG with a regression procedure (for a general overview, see Gratton, 1998).

The spectral composition of different types of eye movements and blinks has been studied with diverse methods for the registration of ocular activity (Whitton et al., 1978, Iacono and Lykken, 1981, Eizenman et al., 1985, Gasser et al., 1985). These studies provided consistent evidence that the mass of power of ocular activity is in the delta (and theta) band, with a sharp and rather monotonous decrease of power with increasing frequency. Based on this evidence, some investigators concluded that the EOG power of eye movements and blinks is negligible in the alpha and beta bands (above 7.5 Hz; e.g. Gasser et al., 1985, 1986).

Although corneo-retinal dipole rotations and ocular field alterations due to lid movements are the obvious source of EOG activity, the higher frequency components of the EOG appear to be mostly of neural origin, thus constituting the problem of neural artifact in the EOG. For example, Iacono and Lykken (1981) measured frontal EEG and eye movements during a smooth-pursuit tracking task with EOG and infra-red reflectrometry, the latter measuring movements of the eyeball directly. A comparison of EOG and infra-red data revealed a substantial alpha activity (between 7.0 and 13.9 Hz) in the EOG, which was absent in the infra-red data, and thus, suggested a neural origin. This finding is in line with reports of Whitton et al. (1978) and Gasser et al. (1985), who presented EOG spectra that showed a protuberance in the alpha band, which resembled the alpha peak in the corresponding EEG spectra.

Taken together, these findings suggest that blinks and eye movements generate substantial activity in the delta and theta bands, but essential portions of the alpha and beta activity in the EOG might be of neural origin. More evidence for the distinction of ocular and neural activity in the EOG stems in particular from studies of the regression approach of artifact compensation.¹ In a frequency domain regression approach, the EOG and EEG is Fourier-transformed into the frequency domain, a regression is performed separately for each frequency component of the EOG and EEG spectrum, and finally, the residualized EEG is transformed back to the time domain. The regression spectrum is interpreted as a transmission function and yields a separate transmission coefficient for each frequency component (for an overview, see van Driel et al., 1989).

In a study of EOG–EEG propagation, Gasser et al. (1985) applied a frequency domain regression approach to their data and presented EOG–EEG transmission functions for data portions of an eyes-closed resting condition. The EOG–EEG transmission functions showed a peak in the theta/alpha range (approximately between 5 and 15 Hz) which was more pronounced for data portions without artifacts compared with the data with artifacts, and that was interpreted as an indication of coherent neural activity. These findings corroborated earlier observations that suggested neural activity in the EOG (Gasser et al., 1983a, Gasser et al., 1983b), and were replicated by Möcks et al. (1989) and van Driel et al. (1989) who reported on an excessive gain for EOG–EEG transmission in the alpha band. Both author groups suggested that this peak of the transmission functions must be due to coherent prefrontal cerebral activity that was picked up by EOG and EEG electrodes. Commenting on these findings, Berg (1989) also proposed that most of the ocular activity might be below 5 Hz, whereas above 5 Hz, coherent EEG plays the dominant role in the EOG. Similar evidence for a considerable coupling of the EOG and EEG in the alpha range was provided by Whitton et al., 1978, Waterman et al., 1992.

The common conclusion that the EOG activity in the alpha and beta bands is not of ocular origin is also reflected by several procedures for the control of ocular artifacts. To account for artifacts for an analysis of spontaneous EEG, Gasser et al. (1983b) selected only segments of the data that showed a minimal EOG power in a frequency band between 1.5 and 7.5 Hz, and a subsequent study suggested that the introduction of further selection criteria did not improve the results (Möcks and Gasser, 1984). To account for coherent neural activity in the EOG and EEG, Woestenburg et al. (1983) and van Driel et al. (1989) proposed that in the frequency domain regression approach, all transfer coefficients that exceed a threshold value (and which usually comprise the alpha band) should be set to zero to prevent an over-compensation. Similarly, Waterman et al. (1992) suggested that only the delta band should be used for the frequency domain approach. Likewise, Gasser et al. (1992) proposed a filtering of the EOG with a low-pass of 7.5 Hz before the application of a time domain regression approach.

In total, these studies strongly suggest that most of the higher frequency range in the EOG is of neural origin, and thus, it may be assumed that ocular activity does not generate substantial power in the alpha and beta bands. The research on the regression approach not only provided considerable insight into the spectral composition of the EOG–EEG transmission, but also offered many findings on the topography of ocular artifacts.

A simple visual inspection of raw EEG recordings reveals that the size of vertical artifacts like blinks decreases with increasing distance to the frontal poles, and that the size of horizontal artifacts due to eye movements additionally decreases with decreasing distance from the midline (Gasser et al., 1992). This topography is reflected by a decrease of the transmission coefficients/functions from anterior to posterior (e.g. Gratton et al., 1983, Gasser et al., 1985, Gasser et al., 1992, Gratton and Coles, 1989, Lutzenberger and Elbert, 1989, Möcks et al., 1989, van den Berg-Lenssen and Brunia, 1989, van Driel et al., 1989). Although blinks are usually associated with smaller transmission coefficients/functions than other types of eye movements (e.g. Corby and Kopell, 1972, Weerts and Lang, 1973, Gratton et al., 1983, Gasser et al., 1985; but see Gratton and Coles, 1989, Kenemans et al., 1991), it generally appears that ocular potentials are attenuated with increasing distance to the eyes.

In addition, ocular activity appears to propagate along the anterior–posterior axis in a rather symmetrical fashion. In many reports on time domain regression approaches, the transmission coefficients of the vertical or horizontal EOG usually showed a highly similar magnitude for contralateral homologous EEG sites, with a similarity somewhat greater for vertical than for horizontal transmission (e.g. Gratton et al., 1983, Gratton and Coles, 1989, Lutzenberger and Elbert, 1989, van den Berg-Lenssen and Brunia, 1989, Gasser et al., 1992). In parallel, applications of the frequency domain approach also provided highly similar transmission functions for the vertical EOG–EEG transfer (Gasser et al., 1985, Möcks et al., 1989, van Driel et al., 1989; but see Möcks et al., 1989, van Driel et al., 1989). In general, these findings suggest a rather symmetrical propagation of ocular artifacts, which might be explained by the recurrent conjugation of the movements and blinks of both eyes (see Gratton, 1998).

Given the extensive research on procedures for the control of ocular artifacts in the EEG, surprisingly few studies reported on the effects of ocular actions on the power spectrum of the EEG. In an early study on the occipital alpha rhythm, Verbaten et al. (1975) manipulated the amount of eye movements by a rapid visual search vs. a fixation task, but did not observe a significant change in parieto-occipital alpha activity between these conditions. Unfortunately, EEG was only recorded from the right hemisphere, no anterior EEG sites were included, no other frequency bands were analyzed, and the sample size was limited to 8 participants. Nearly one decade later, Möcks and Gasser (1984) recorded EEG in an eyes-closed condition at rest and compared broadband EEG power of the whole, unselected data, including all artifacts, with epochs of 20 s that were individually selected for a minor artifact contamination as indicated by minimal delta and theta power in the simultaneously recorded EOG. A descriptive comparison revealed lesser delta and theta power in the EEG that was selected for the absence of ocular artifacts, in particular, for delta activity at frontal sites. However, there were no substantial differences in alpha and beta powers between these two data selections. The limitations of these findings are due to the procedure for selecting epochs with low artifact contamination, since the 20 s data segments with minimal EOG activity might still contain transient portions with gross artifacts (see the critique of Gasser et al., 1992). Thus, the comparison was probably based on two EEG selections, each containing data with and without gross artifacts, and the effects of ocular activity on EEG broadband power may have been underestimated. In addition, no attempts were made to evaluate the significance of these effects of data selection, and only the data of right hemisphere electrodes were reported.

In a similar approach, Gasser et al. (1986) recorded EOG and EEG in a resting state with eyes closed. Among other findings, they presented the EEG broadband power separately for 20 s data segments with minor and major ocular artifact contamination as indicated by minimal and maximal EOG power. On a descriptive level, the EEG portions with minimal EOG power had lower broadband power in the delta and theta bands than the EEG portions with maximal EOG power, in particular, for delta activity at frontal and central sites. At the frontal sites, the power differences in the alpha and beta bands were in the same direction, but rather small in magnitude. Most remarkably, the correlation between the medial frontal EEG power of data selected for minimal and maximal ocular artifacts was low in the delta (0.58), but higher in the theta (0.77), alpha (0.88 and 0.72), and beta (0.86 and 0.88) bands, and these relations increased for posterior sites (all r≥0.89 for all bands at the occipital region). The limitations of these findings are again due to the procedure for selecting epochs with low vs. high artifact contamination (see Gasser et al., 1992), which may have yielded an underestimation of the effects of ocular activity on EEG broadband power, and yet again, no attempts were made to evaluate the significance of these effects of data selection, and no data of contralateral homologous sites were reported. In total, these studies do not provide a firm conclusion on the effects of ocular artifacts on alpha and beta band parameters of the EEG, and no cogent inference on the effects of ocular artifacts on EEG asymmetry can be derived from these findings.

A bulk of evidence implied that ocular artifacts do not generate substantial power in the alpha and beta bands, and that the activity in these broadbands of the EOG is rather of neural origin. However, reliable power in the alpha and beta bands due to ocular activity is a necessary condition for a distortion of spectral parameters in the same bands of the EEG.² Furthermore, electric field potentials decrease with increasing distance to their source, thus minor ocular activity in the alpha and beta bands may be attenuated to an insignificant magnitude after propagation to the scalp electrodes. In addition, ocular activity is propagated to the EEG in a rather symmetrical fashion, and may result in rather symmetrical distortions of the EEG that leave the true EEG asymmetry unchanged. The main conclusion from the literature is that a distortion of (lateralized) alpha and beta band parameters of the EEG due to ocular artifacts is not likely to occur.

On the other hand, all authors commonly concluded that a control of ocular artifacts is indispensable for any analysis of spontaneous background EEG, which is contradictory to the assumed invariance of EEG alpha and beta band parameters. Rather, a lack of distorting effects of ocular artifacts on (lateralized) EEG alpha and beta activity implies that the control of ocular artifacts may be omitted, in particular, if alpha and beta asymmetry is the target of research (for recent examples of such research, see Hagemann et al., 1999, Nitschke et al., 1999, Pauli et al., 1999, Debener et al., 2000). The aim of the present study is to examine the effects of blinks and eye movements on single site and asymmetry broadband spectral parameters of the EEG, and to evaluate the necessity of a control/correction of ocular artifacts for EEG alpha and beta band parameters. Since ocular artifacts (and their potential effects on EEG alpha and beta activity) are greatest at the anterior electrodes, and since previous studies gave rise to rather paradoxical implications, we included more anterior sites in the present study and conducted more detailed analyses.