Estimating the depth of anesthesia by applying sub parameters to an artificial neural network during general anesthesia

This paper presents two artificial neural network (ANN) structures to estimate the depth of anesthesia (DOA). First, a clinical study involved on 33 patients is proposed to construct reference data and also to compare the results with BIS monitor (Aspect Medical, Vista), which represents satisfactory correlation with clinical assessments. Secondly, to extract features from electroencephalogram (EEG) signals, we extract some features in frequency and time domain as well as in wavelet (Daubechies) domain. Finally, to integrate EEG features to estimate DOA, ANNs based on back propagation (BP) algorithm are proposed. Since each of the proposed features may has good performance only for a specific range of DOA, this model proved to have good prediction properties, and the output of the proposed ANN has a high correlation with the output of the BIS index.