New online method for removing ocular artefacts from EEG signals

A method for online removal of ocular artefacts from the human electroencephalogram (EEG) is described. It uses numerically stable algorithms based on the efficient recursive least-squares algorithm. The method is shown to give similar results to its offline equivalents from which it has been developed. Compared with the present online methods our approach is superior, requiring no subjective manual adjustment and processing all signals digitally. An automatic online microcomputer-based ocular artefact remover has been built and successfully tested.     

Investigation and comparison of some models for removing ocular artefacts from EEG signals

An investigation of ocular artefacts (OAs) in the human electroencephalogram (EEG) to quantify the effectiveness of OA removal and to find the most effective model for removing OAs online is described. In Part 1, the models used in the investigation are described and the data analysed. The analysis showed that the ‘true’ EEG exhibited a high degree of serial correlation and so the ordinary least-squares (OLS) method employed to remove OA was inefficient. Efficient alternative methods based on autoregressive models of the ‘true’ EEG are discussed. It is also shown that the EOGs are linearly dependent making some of them redundant. In Part 2, the models are compared.     

The removal of ocular artefacts from the electroencephalogram: a review

The causes of ocular artefacts (OAs) in the human electroencephalogram (EEG) are explained and methods for their removal and their effectiveness are discussed. Recommendations for the best procedures to adopt are given together with suggestions for future research. Analogue subtraction techniques are found to be inferior to time domain techniques based on parameter estimation using the method of least squares applied to a linear function of the electro-occulograms (EOGs). Ways of assessing the effectiveness of different models for time domain removal of OAs are discussed. It is concluded that autoregressive modelling of the error terms, or else differenced data, must be used to reduce the effects of correlation in the background EEG. The most generally suitable model for the removal of random eye and blink artefacts should contain terms proportional to the right vertical EOG and the two horizontal EOGs. The EOGs should be linearly filtered to remove noise frequency components in excess of 8 Hz. Adaptive methods are preferred as on line OA removal would be desirable but for the fact that this may result in distortion of stimulus-related responses present. A number of difficulties remain and there are some suggestions for future research.     

Comparison of two methods for correcting ocular artefacts in EEGs

A major problem in the study of brain potentials is the occurrence of ocular artefacts in electro-encephalograms. OAs can be monitored by placing electrodes near the eyes and recording electro-oculograms. In the paper, two OA correction methods based on simulations are compared; the Jervis method and the vandenBerg method. In most simulations, the residual (the difference between the original EEG and the EEG after correction) is smaller in amplitude and variance for the vandenBerg method than for the Jervis method. When eye movements and blinks are given different factors, the blinks are not removed completely. For both methods, the residual of the blinks increases with the differences between the model parameters for the blinks and for eye movements. The occurrence of a slow negative wave greatly disturbs the estimated parameters and thus the residuals of the Jervis method. For the vandenBerg method, there is only a very small effect. The conclusion from correcting a recorded data set, which does not contain a slow negative wave, is that, for these data, there is no evidence that one method is better than the other.     

On-line method for enhancement of electroencephalogram signals in presence of electro-oculogram artefacts using nonlinear recursive least squares technique

A record of the human scalp potentials (EEG) is one of the most commonly used techniques for understanding the brain functions. While recording EEG signals, other undesired signals called artefacts get superimposed on the EEG signals. Of these, electro-oculogram generated by eye movements is found to be the most significant and common. Conventional artefact minimisation schemes use linear estimation models. However, the result in sub-optimal schemes as signal processing in the human brain is highly nonlinear. In this paper therefore an on-line technique is suggested for artefact minimisation using nonlinear recursive least squares. Results confirm the efficacy of the method.     

A comparative study of automatic techniques for ocular artifact reduction in spontaneous EEG signals based on clinical target variables: a simulation case

Eye movement artifacts represent a critical issue for quantitative electroencephalography (EEG) analysis and a number of mathematical approaches have been proposed to reduce their contribution in EEG recordings. The aim of this paper was to objectively and quantitatively evaluate the performance of ocular filtering methods with respect to spectral target variables widely used in clinical and functional EEG studies. In particular the following methods were applied: regression analysis and some blind source separation (BSS) techniques based on second-order statistics (PCA, AMUSE and SOBI) and on higher-order statistics (JADE, INFOMAX and FASTICA). Considering blind source decomposition methods, a completely automatic procedure of BSS based on logical rules related to spectral and topographical information was proposed in order to identify the components related to ocular interference. The automatic procedure was applied in different montages of simulated EEG and electrooculography (EOG) recordings: a full montage with 19 EEG and 2 EOG channels, a reduced one with only 6 EEG leads and a third one where EOG channels were not available. Time and frequency results in all of them indicated that AMUSE and SOBI algorithms preserved and recovered more brain activity than the other methods mainly at anterior regions. In the case of full montage: (i) errors were lower than 5% for all spectral variables at anterior sites; and (ii) the highest improvement in the signal-to-artifact (SAR) ratio was obtained up to 40dB at these anterior sites. Finally, we concluded that second-order BSS-based algorithms (AMUSE and SOBI) provided an effective technique for eye movement removal even when EOG recordings were not available or when data length was short.