A comparison of five different algorithms for EEG signal analysis in artifacts rejection for monitoring depth of anesthesia

Electroencephalography (EEG) signals have been commonly used for assessing the level of anesthesia during surgery. However, the collected EEG signals are usually corrupted with artifacts which can seriously reduce the accuracy of the depth of anesthesia (DOA) monitors. In this paper, the main purpose is to compare five different EEG based anesthesia indices, namely median frequency (MF), 95% spectral edge frequency (SEF), approximate entropy (ApEn), sample entropy (SampEn) and permutation entropy (PeEn), for their artifacts rejection ability in order to measure the DOA accurately. The current analysis is based on synthesized EEG corrupted with four different types of artificial artifacts and real data collected from patients undergoing general anesthesia during surgery. The experimental results demonstrate that all indices could discriminate awake from anesthesia state (p < 0.05), however PeEn is superior to other indices. Furthermore, a combined index is obtained by applying these five indices as inputs to train, validate and test a feed-forward back-propagation artificial neural network (ANN) model with bispectral index (BIS) as target. The combined index via ANN offers more advantages with higher correlation of 0.80 ± 0.01 for real time DOA monitoring in comparison with single indices.