Systemic modeling of chaotic EEG during human sleep

The use of chaos theory in modeling biological signals is becoming increasingly common, especially in EEG signal processing for predicting and detecting different brain states. However, the chaotic nature of the sleep EEG signal remains a challenging debate, and the use of nonlinear techniques such as Correlation Dimension, which is a common method for measuring a system's complexity, may lead to erroneous results in non-chaotic systems. To address this issue, in the present study, we attempt to provide an analysis to demonstrate the chaoticity of sleep EEG. The goals of this study are to model and detect the strange attractor of sleep and its stages. We model changes in the sleep attractor dynamics in phase space by exponential regression. Our model indicates that the sleep attractor is the sleep cycle attractor, whose size shrinks during successive cycles, by presenting a new definition of the sleep cycle. We study the EEG dynamics of different sleep stages by introducing two new features based on phase space properties and identify unique chaotic attractors for each sleep stage. We model the geometric changes of these attractors during successive sleep cycles. Our model achieves an accuracy, sensitivity, and specificity of 88.75%, 85.34%, and 83.64% in classifying sleep stages.