Fractal dimensions, Lyapunov exponents, and predictability of human EEG signals

The author presents several developments in the application of nonlinear science to diagnostic problems. These include estimation of Lyapunov characteristic exponents, the temporal evolution of local dimensional complexity and divergence rates, and their relationship to the temporal predictability of the EEG. These measures allow the estimation of state changes based on alteration of the complexity of the EEG due to loss of coherence or due to bifurcations to higher chaotic states. Results are presented which show the variability of the nonlinear complexity of the EEG during different stages of anesthesia and sleep. Questions of robustness of the method and data requirements are discussed.<>