Assessment of Multivariate Neural Time Series by Phase Synchrony Clustering in a Time-Frequency-Topography...

Area of Science:

• Neuroscience
• Signal Processing
• Computational Biology

Background:

• Existing electroencephalography (EEG) phase synchrony measures are often bivariate or provide only global multivariate indices.
• There is a need for methods that can describe local spatial and temporal interactions in multivariate neural time series.

Purpose of the Study:

• To present a novel framework for characterizing phase synchrony relationships in multivariate neural time series.
• To develop and apply a clustering algorithm for detailed analysis of synchrony states.
• To visualize and interpret local synchrony dynamics beyond global indices.

Main Methods:

• Development of Multivariate Time Series Clustering by Phase Synchrony (MTS-CPS) algorithm.
• Generation of fuzzy and hard clusters based on phase synchrony and circular variance.
• Creation of Time-Frequency-Topography representations to visualize synchrony.
• Application to EEG data from P300 Speller sessions and analysis using Levenshtein Distance.

Main Results:

• The framework successfully identified synchrony patterns related to event-related potentials (ERPs).
• Steady-state artifacts at 7.6 Hz were detected in the EEG data.
• Contrast maps revealed significant synchrony differences between ERP and non-ERP epochs, particularly in delta and theta bands.

Conclusions:

• The proposed framework provides a powerful tool for observing phase dynamics and channel interactions in EEG data.
• This method offers richer insights into neural synchrony compared to global indices.
• The approach is versatile and applicable to analyzing various cognitive states and neural phenomena.