In 2001 a University of Bristol team patented a novel data reduction method of the EEG for characterising categorical changes in consciousness. After pre-whitening the EEG signal with Gaussian white noise a parametric spectral estimation technique was applied. Two frequency domain indices were then proposed: the relative power found between 8Hz to 12Hz and 0.5Hz to 32Hz termed the 'alpha index', and the relative power between 0.5Hz to 4Hz and 0.5Hz to 32Hz termed the 'delta index'. The research and development of a precision EEG monitoring device designed to embody the novel algorithm is described in this thesis. The efficacy of the technique was evaluated using simulated and real EEG data recorded during Propofol anaesthesia. The simulated data showed improvements could be made to the patented method. Real EEG data collected whilst patients were wakeful and data from patients unresponsive to noxious stimuli were cleaned of obvious artefacts and analysed using the proposed algorithm. A Bayesian diagnostic test showed the alpha index had 65% sensitivity and selectivity to patient state. The delta index showed 72% sensitivity and selectivity. Taking a pragmatic approach, the literature is reviewed in this thesis to evaluate the use of EEG in depth of anaesthesia monitoring. Pertinent aspects of the sciences are profiled to identify physiological links to the characteristics of the EEG signal. Methods of data reduction are also reviewed to identify useful features and possible sources of error. In conclusion it is shown that the proposed indices do not provide a robust measure of depth of anaesthesia. An approach for further research is proposed based on the review work.
|Date of Award||2008|