Monitoring Depth of Anaesthesia Based on Electroencephalogram Extracted Features and Artificial Neural Network

Monitoring Depth of Anaesthesia (DoA) is one of the current challenges in medical research. Anaesthetic drugs affects mainly the central nervous system and therefore Electroencephalogram (EEG) signal analysis during anaesthesia will help to quantify the Depth of Anaesthesia. This paper proposes a novel method to quantify DoA based on the analysis of EEG signals during anaesthesia. Dimensionality of EEG signals are reduced by extracting the time and frequency domain features Approximate entropy (ApEn), Spectral Entropy (SEN) and Wavelet Entropy (WE). A comparison is performed on extracted EEG feature variations during different phases of anaesthesia awake, induction, maintenance and recovery. Validation is succeeded by calculating the correlation of extracted EEG features with BIS index (commercially available DoA monitor). The transition from different phases of anaethesia shows characteristic changes on extracted features than with BIS. Finally, the extracted EEG features are fed to an Artificial Neural Network (ANN) to classify the different anesthetic states as awake, light anaesthesia, moderate anaesthesia and deep anaesthesia. The classification accuracy attained through training and validation is 90.5 percentage.