A Novel Multivariate Sample Entropy Algorithm for Modeling Time Series Synchronization.

Area of Science:

• Complexity and Information Theory
• Biomedical Signal Processing
• Time Series Analysis

Background:

• Approximate entropy (AE) and sample entropy (SE) quantify time series regularity and complexity.
• Existing multivariate sample entropy (mSE) algorithms have limitations in analyzing within- and cross-channel dynamics and can misinterpret correlations.

Purpose of the Study:

• To develop a novel multivariate sample entropy (mSE) algorithm addressing shortcomings of existing methods.
• To improve the analysis of multivariate time series, including physiological data.

Main Methods:

• Revisiting the embedding of multivariate delay vectors (DVs) for physically meaningful analysis.
• Proposing and validating a novel mSE algorithm using synthetic and real-world physiological data.
• Developing a model for the novel mSE algorithm's operation in white Gaussian noise.

Main Results:

• The novel mSE algorithm demonstrates improved performance over existing methods for synthetic and real-world data.
• Synchronized regularity dynamics are uniquely identified by the novel mSE, unlike other methods.
• The new model reveals that increased correlation between variates reduces entropy, a finding contrary to existing algorithms.

Conclusions:

• The proposed mSE algorithm offers a more accurate and robust method for analyzing multivariate time series complexity and regularity.
• This advancement has significant implications for biomedical signal processing and understanding physiological states.
• The novel mSE uniquely captures synchronized dynamics and clarifies the impact of correlation on entropy.