Sleep Spindle Detection in EEG Signal for Investigating Sleep Disturbances

The sleep spindles are discriminant patterns of the sleep stage 2, whose detection is of significant importance for studying memory consolidation and sleep disorders. Because of the non-linear nature of the EEG signal, sleep spindles detection by visual inspection is time-consuming and prone to human error. For this purpose, we proposed a new automatic method for sleep spindles detection. The EEG signal was first divided into one-second segments using a sliding window with an overlapping of 0.8s; as an effective time-frequency method, the Empirical Wavelet Transform (EWT) was used to extract Intrinsic Mode Function (IMF). In the next step, some non-linear features such as Shannon Entropy, Renyi Entropy, Tsallis Entropy, Katz s and Petrosian Fractal Dimension extracted for the first three IMFs. Finally, to classify the extracted features, Support Vector Machines, K-Nearest Neighbor, Probabilistic Neural Network, and AdaBoost were employed. The results of this research show that the proposed method for sleep spindles detection has a better performance than the existing methods.