Estimation of hypnosis susceptibility based on electroencephalogram signal features

Quantitative estimation of hypnosis susceptibility is a crucial factor for psychotherapists. WaterlooStanford is the gold-standard qualitative index of measuring the hypnosis dept but still is not as correct as hypnotizers expect. In this way, a robust criterion is presented uses electroencephalogram (EEG) signal features to quantitatively estimate the hypnosis depth. Thirty two subjects were voluntarily participated in our study and their EEG signals from 19 channels were recorded during hypnosis induction. Several features, such as fractal dimension, autoregressive (AR) coefficients, wavelet entropy, and band power were extracted from the signals. Regarding high dimensionality of the extracted features, Sequential Forward Selection (SFS) is employed to reduce the size of input features. To categorize the hypnosis susceptibility of the participants based on their EEG features, Nearest Neighbor (NN), Fuzzy NN (FNN), and a Fuzzy Rule-Based Classification System (FRCBS) were utilized. Subjects were classified into three hypnosis ability classes including lows, mediums and highs. Leave-one(subject)-out cross validation method was utilized for validation of our results. Experimental results are completely matched to that of WaterlooStanford, such that degrees of hypnotic susceptibility for 32 (out of 32) subjects were correctly determined.