Automatic Sleep Spindle Detection with EEG Based on Complex Demodulation Method and Decision Tree Model

Sleep spindle is the characteristic waveform of electroencephalogram (EEG) which is important for clinical diagnosis. In this study, an automatic sleep spindle detection method was developed. The EEG signals were recorded based on the standard polysomnogram (PSG) measurement. A preprocessing procedure is introduced to exclude the unnecessary data segments and normalized the necessary data segments. Complex demodulation method is adopted to detect the candidate sleep spindle waveforms and calculate the features. The sleep spindles are recognized based on a decision tree model. Finally, the detected sleep spindles were utilized to amend the sleep stage recognition results. The sleep EEG data from 3 patients with sleep disorders were analyzed. The obtained results showed that the detected sleep spindles in EEG signal improved the accuracy of sleep stage recognition.