Analyzing autocorrelation fluctuation of EEG signal for estimating depth of anesthesia

In each surgical operation, monitoring the depth of anesthesia is vital for anesthesiologists to control this depth during the surgery. Several methods have been suggested to determine quantitative indices for depth of anesthesia but most of these methods suffer from high sensitivity of their indices during the surgery. In this paper, to make the index more robust, a beneficial Electroencephalogram (EEG) signal preprocessing method is proposed. Additionally, an efficient method is proposed to estimate the depth index during the surgery. In the signal preprocessing the signal amplitude is normalized by the signal energy in each epoch and the effect of signal amplitude is declined. After this preprocessing, EEG signal is analyzed by autocorrelation to evaluate amount of self-similarity. Then fractal dimensions are used to interpret autocorrelation content. Experimental results have shown that applying the proposed preprocessing and method to EEG signals of 1870 epochs during the surgery can precisely classify the awake, moderate and deep anesthesia levels. Moreover, our real-time approach leads to increase the depth index robustness and provides similar results to the Bispectral index (BIS).