Novel method of fast automated discrimination of sleep stages

Abstract

A new approach to sleep quantification analysis based on the mutual information (MI) of brain cortex is described. The mutual information time series between four leads were first computed using the electroencephalogram (EEG). The Lempel-Ziv complexity measure, C(n)s, were extracted from the mutual information time series. Sleep staging was then made by a three-layer artificial neural network (ANN) using the C(n)s. The combination of these three different approaches enables the system to address the non-analytical, non-stationary, non-linear and dynamical properties of EEG. From 6 subject experiments, 720 distinct EEG epochs were used to test the results of sleep stage classification. The accuracy rate obtained for the system is 90.83%. Comparisons with other methods show that the proposed system has a certain advantage. Furthermore, the new method was computationally fast and well suited for real-time clinical implementation.