Real time depth of anaesthesia monitoring through electroencephalogram (EEG) signal analysis based on Bayesian method and analytical technique

Real time depth of anaesthesia monitoring through electroencephalogram (EEG) signal analysis based on Bayesian method and analytical technique

The electroencephalogram (EEG) signal from the brain is used for analysing brain abnormality, diseases, and monitoring patient conditions during surgery. One of the applications of the EEG signals analysis is real-time anaesthesia monitoring, as the anaesthetic drugs normally targeted the central nervous system.

Depth of anaesthesia has been clinically assessed through breathing pattern, heart rate, arterial blood pressure, pupil dilation, sweating and the presence of movement. Those assessments are useful but are an indirect-measurement of anaesthetic drug effects. A direct method of assessment is through EEG signals because most anaesthetic drugs affect neuronal activity and cause a changed pattern in EEG signals.

The aim of this research is to improve real-time anaesthesia assessment through EEG signal analysis which includes the filtering process, EEG features extraction and signal analysis for depth of anaesthesia assessment. The first phase of the research is EEG signal acquisition. When EEG signal is recorded, noises are also recorded along with the brain waves. Therefore, the filtering is necessary for EEG signal analysis.

The filtering method introduced in this dissertation is Bayesian adaptive least mean square (LMS) filter which applies the Bayesian based method to find the best filter weight step for filter adaptation. The results show that the filtering technique is able to remove the unwanted signals from the EEG signals.

This dissertation proposed three methods for EEG signal features extraction and analysing. The first is the strong analytical signal analysis which is based on the Hilbert transform for EEG signal features' extraction and analysis. The second is to extract EEG signal features using the Bayesian spike accumulation technique. The third is to apply the robust Bayesian Student-t distribution for real-time anaesthesia assessment.

Computational results from the three methods are analysed and compared with the recorded BIS index which is the most popular and widely accepted depth of anaesthesia monitor. The outcomes show that computation times from the three methods are leading the BIS index approximately 18-120 seconds. Furthermore, the responses to anaesthetic drugs are verified with the anaesthetist's documentation and then compared with the BIS index to evaluate the performance. The results indicate that the three methods are able to extract EEG signal features efficiently, improve computation time, and respond faster to anaesthetic drugs compared to the existing BIS index.