Multiscale Correlation Analysis Based Detection/Segmentation of High Amplitude Theta Bursts in Long-Term EEG Records

摘要

The report discusses the results of the original approach we proposed to the EEG paroxysmal activity detection / segmentation. The approach is aimed at processing and analysing essentially non-stationary EEG, containing so-called High Amplitude Theta bursts. The approach analysis pipeline includes local detection of the rhythm, and then an evaluation of its dynamics for the segmentation of episodes with “hyper-synchronous” behavior. A distinctive feature of the proposed approach is its special 2D representations of multiscale correlation analysis, which, unlike usual spectral tools, are formed in the time domain. The two-criteria High Amplitude Theta bursts detection / segmentation method discussed in the report is developed on the basis of these representations.