A broadband method of quantifying phase synchronization for discriminating seizure EEG signals. (July 2019)

• Record Type: Journal Article
• Authors: Wang, Lei; Long, Xi; Aarts, Ronald M.; van Dijk, Johannes P.; Arends, Johan B.A.M.
• Is Part Of: Biomedical signal processing and control. Volume 52(2019)
• Page Start: 371
• Page End: 383
• Publication Date: 2019-07
• Subjects: Phase synchronization (PS) -- Correlation between probabilities of recurrence (CPR) -- Phase lock index (PLI) -- Epilepsy, intellectual disability -- EEG discharge patterns; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2018.10.019
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: CPR captures the nonlinear phase coupling of a pair of broadband EEG signals and provides a new method for generating brain functional network. How PS changes in seizures depends on the EEG discharge patterns. CPR-based PS can discriminate four major types of seizures and show potential for predicting two major types of seizures in an early manner. We optimized the computation of CPR to reduce the computation costs and variance across EEG epochs. We compared CPR with a traditional method PLI on the same dataset. Abstract: The nonlinear nature of phase coupling enables rich and context-sensitive interactions that characterize real brain dynamics, playing an important role in brain dysfunction such as epileptic disorders. Numerous phase synchronization (PS) measurements have been developed for seizure detection and prediction. However, the performance remains low for minor seizures in epileptic patients with an intellectual disability (ID), who are characterized by complex EEG signals associated with brain development disorders. The traditional PS measurements, e.g., phase locking index (PLI), are limited by the inability in detecting the nonlinear coupling of EEG signals and are sensitive to the background noise. This study focuses on developing a new EEG feature that can measure the nonlinear coupling, which thus would help improve seizure detection performance. We employ the correlation between probabilities of recurrence (CPR) to measure the PS on broadband EEG signals. CPR can capture the underlying nonlinear coupling of EEG signals and is robust to signal frequency and amplitude variance. The effectiveness of CPR-based features on identifying seizure EEG was evaluated on 26 epileptic patients with ID. Results show that the PS changes in seizures depend on the EEG discharge patterns including fast spike (SP), spike-wave (SPWA), wave (WA) and discharge with EMG activity (EMG). CPR-based PS decreased significantly in seizures with SP and EMG, (-0.1845 and -0.4278, with 95% CI [-0.1850, -0.1839] and [-0.4283, -0.4273], respectively), while it increases significantly in the SPWA seizures (+0.0746, with 95% CI [0.0744, 0.0749]). In addition, CPR-based PS shows potential for predicting SPWA and EMG seizures in an early manner. We conclude that CPR measurement is promising to improve seizure detection in ID patients and provides a promising method for modeling epilepsy-related brain functional networks.