Constructing a Consciousness Meter Based on the Combination of Non-Linear Measurements and Genetic Algorithm-Based Support Vector Machine
IEEE transactions on neural systems and rehabilitation engineering January 8, 2020 Zhenhu Liang, Shuai Shao, Zhe Lv et al.
A framework using electroencephalography (EEG) and nonlinear analysis methods can differentiate four states of consciousness: coma, general anesthesia, minimally conscious state (MCS), and normal wakefulness. Permutation entropy (PE) distinguished all four states. Altered contents of consciousness were best differentiated by sample entropy (SampEn) and permutation Lempel-Ziv complexity (PLZC), while levels of consciousness were best differentiated by relative power of Gamma and PE. A multi-dimensional index combining PE, PLZC, SampEn, and detrended fluctuation analysis (DFA) achieved 92.3% classification accuracy using a genetic algorithm-based support vector machine (GA-SVM), outperforming random forest and neural networks. A multivariable linear regression model constructed coordinate values for level and content dimensions.