A Noise-Robust Approach to Estimate Dimension of Sleep EEG in Mice Using Permutation Entropy

Generally, an electroencephalogram (EEG) follows nonlinear dynamics because it shows complex regularities. The analysis of nonlinear dynamics using the ‘embedding’ method requires the determination of parameters such as embedding dimension and time delay. However, we cannot obtain correct parameters from a conventional method, such as correlation integration, due to noise. Therefore, we use permutation entropy as a robust method against noise. This paper estimates the EEG parameters of mice during different sleep stages by using permutation entropy. The study results suggest that the embedding dimension and time delay are 5 and 8 ms, respectively, regardless of the sleep stage. Furthermore, we analyze the EEG of a sleeping mouse from the nonlinear dynamics. Permutation entropy can distinguish between rapid eye movement sleep and wakefulness, unlike linear methods such as power spectrum density.