Robust approach to depth of anaesthesia assessment based on hybrid transform and statistical features

Abstract

To develop an accurate and efficient depth of anaesthesia (DoA) assessment technique that could help anaesthesiologists to trace the patient's anaesthetic state during surgery, a new automated DoA approach was proposed. It applied wavelet-Fourier analysis (WFA) to extract the statistical characteristics from an anaesthetic electroencephalogram (EEG) signal and to design a new DoA index. In this proposed method, firstly, the wavelet transform was applied to a denoised EEG signal, and a fast Fourier transform was then applied to the wavelet detail coefficient D3. Ten statistical features were extracted and analysed, and from these, five features were selected for designing a new index for the DoA assessment. Finally, a new DoA (WFADoA) was developed and compared with the most popular bispectral index (BIS) monitor. The results from the testing set showed that there were very high correlations between the WFADoA and the BIS index during the awake, light and deep anaesthetic stages. In the case of poor signal quality, the BIS index and the WFADoA were also tested, and the obtained results demonstrated that the WFADoA could indicate the DoA values, while the BIS failed to show valid outputs for those situations.