Stochastic Event Synchrony Based on a Modified Sparse Bump Modeling: Application to PTSD EEG Signals

Abstract

Stochastic event synchrony (SES) is an approach of quantifying synchrony between two time series. First of all, the time-frequency transform of each signal is approximated as a sum of half-ellipsoid basis functions, referred to as “bumps.” Each bump is considered as an event on the time-frequency plane. After pruning according to a threshold, the resulting bump model is considered as a two-dimensional point process, which represents the most prominent oscillatory activity. SES computes the alignment of two point processes based on five parameters: time delay, timing jitter variance, ratio of spurious events, variance of the frequency offset (frequency jitter), and the average frequency offset between events. In this chapter, we propose a modified version of bump modeling using the second-order wavelet-based synchrosqueezing transform (WSST2), and a modified SES as a result. WSST is an EMD-like approach, which is a combination of wavelet transform and reallocation methods. WSST provides a time-frequency representation with less blurring compared to the wavelet transform. Our proposed method is applied to combat-related posttraumatic stress disorder (PTSD) EEG signals, in addition to healthy controls, and trauma-exposed non-PTSD veterans. EEG signals are collected in two resting states (eyes open and closed). ANOVA (analysis of variance) at 99% confidence level is used to assess the results. Ratio of spurious events reveals the best performance in two methods. Moreover, the modified SES outperforms the original one.