Neural net classification of REM sleep based on spectral measures as compared to nonlinear measures

In various studies the implementation of nonlinear and nonconventional meas ures has significantly improved EEG (electroencephalogram) analyses as comp ared to using conventional parameters alone. A neural network algorithm wel l approved in our laboratory for the automatic recognition of rapid eye mov ement (REM) sleep was investigated in this regard. Originally based on a br oad range of spectral power inputs, we additionally supplied the nonlinear measures of the largest Lyapunov exponent and correlation dimension as well as the nonconventional stochastic measures of spectral entropy and entropy of amplitudes. No improvement in the detection of REM sleep could be achie ved by the inclusion of the new measures. The accuracy of the classificatio n was significantly worse, however, when supplied with these variables alon e. In view of results demonstrating the efficiency of nonconventional measu res in EEG analysis, the benefit appears to depend on the nature of the pro blem.