Technical advances in ElectroEncephaloGraph (EEG) recordings

Using EEG equipment we are now in a position to embark on the exciting journey of exploring our minds. Like sound waves, the brain has its own set of vibrations to communicate with itself and the rest of the body, which can be recorded by EEG equipment. The signals measured are obviously of critical importance when monitoring cases of possible brain injury. Current research shows that artefacts in EEG signals that impeded the use of these monitoring devices can now be detected and rejected from the signal.

Health

ElectroEncephaloGraph could have wide usage as a neuro-monitoring method in clinical patient monitoring. The drawback though with EEG signals is that they often contain artefacts caused either by external factors, most notably electrical interference from the devices, or by physiological sources, electric signals detected from muscle contraction for example. These artefacts are flaws in the signal and may lead to severe deterioration of the results. Until now the lack of any reliable and widely acceptable method to detect these artefacts resulted in the use of empirical knowledge from experts in the field that reject periods of the signal when the corresponding amplitude is not physiologically plausible. Such a procedure is far from accurate and produces additional problems. Differences in the characteristics of the signal under various conditions cannot be accounted for neither do all artefacts manifest themselves as high amplitude signals but sometimes as uncommon signal patterns and cannot therefore be identified and rejected. The current result has successfully addressed these problems by creating algorithms for artefact detection relevant to a specific monitoring environment. The essence of the method lies in the basic and plausible assumption that artefacts are actually associated with outliers in the EEG parameter space. The method allows detection of artefacts in 10-second periods of an EEG recording using a series of threshold detection rules in individual EEG-derived parameters. The need of human expertise is only required for the characterisation of a training subset of the available dataset, in order to assess if artefacts are containing or not. The successful performance of this method can be further improved by the addition of more EEG parameters and by generalising the discriminate analysis to a multivariate technique. The current stage on an independent test set of data, shows there was a 23% of false detections and 71% of all severe artefacts were detected.