Autoregressive modeling of EEG signals for monitoring anesthetic levels

Changes in electroencephalogram (EEG) were analyzed at different levels of halothane anesthesia. Five experiments were carried out on mongrel dogs. Four channels of EEG data were recorded, at different anesthetic levels. A tenth-order autoregressive (AR) model was used to represent the EEG signal. The AR model parameters were used as input to a three-layer perceptron feedforward neural network, and the network was trained and tested on different sets of data. The network was able to correctly classify the anesthetic levels in 83% of the cases with a testing tolerance of 0.1. The results indicate that the changes in AR model parameters representing the EEG signal can be used for decision-making during administration of general anesthetics.