Monitoring the depth of anaesthesia using simplified electroencephalogram (EEG)

Monitoring the depth of anaesthesia using simplified electroencephalogram (EEG)

Anaesthesia is administered routinely every day in hospitals and medical facilities. Numerous methods have been devised and implemented for monitoring the depth of anaesthesia (DoA) in order to guarantee the safety of patients. Monitoring the depth of anaesthesia provides anaesthesia professionals with an additional method to assess anaesthetic effects and patient responses during surgery. The measurement of depth of anaesthesia benefits patients and helps anaesthetists such as 'reduction in primary anaesthetic use, reduction in emergence and recovery time, improved patient satisfaction and decreased incidence of intra-operative awareness and recall' (Kelley S. D.).

Clinical practice uses autonomic signs such as heart rate, blood pressure, pupils, tears, and sweating to determine depth of anaesthesia. However, clinical assessment of DoA is not valuable in predicting the response to a noxious stimulusand may vary depending on disease, drugs and surgical technique. Currently available DoA monitoring devices have been criticised in the literature, such as being redundant (Schneider, 2004), not responsive to some anaesthetic agents (Barr G., 1999), and time delay (Pilge S., 2006).

This research proposes new methods to monitor the depth of anaesthesia (DoA) based on simplified EEG signals. These EEG signals were analysed in both the time domain and the time-frequency domain. In the time domain, the Detrended Fluctuation Analysis (DFA), detrended moving average (DMA) and Chaos methods are modified to study the scaling behaviour of the EEG as a measure of the DoA. In the frequency domain, fast Fourier transform (FFT) and filter bank are used to identify difference states of anaesthesia. In the time-frequency domain, discrete wavelet transforms (DWT) and power spectral density (PSD) function are applied to pre-process EEG data and to monitor the DoA.

Firstly, a new de-noising algorithm is proposed with a threshold TWE, which is a function of wavelet entropy and the window length m for an EEG segment. Secondly, the anaesthesia states are identified into awake, light, moderate, deep and very deep anaesthesia states. Finally, the DoA indices are computed using:
 Modified DFA method (MDFA I),
 Modified DFA-Lagrange method (MDFA II),
 Modified detrended moving average method (MDMA),
 Modified Chaos method, combined Chaos and MDMA method,
 Wavelet-power spectral density.

Simulation results demonstrate that our new methods monitor the DoA in all anaesthesia states accurately. These proposed methods and indices present a good responsive to anaesthetic agent, reduce the time delay when patient’s hypnotic state changes (from 12 to 178 seconds), and can estimate a patient’s hypnotic state when signal quality is poor.