Multiscale analysis of biological data by scale-dependent lyapunov exponent

Area of Science:

• Physiological signal analysis
• Biomedical engineering
• Complexity science

Background:

• Physiological signals are often non-stationary and multiscaled, exhibiting complex behaviors like chaos and long memory.
• Accurate analysis of these signals is crucial for quality testing, physician assessment, and patient diagnosis in health sciences.
• Existing methods struggle to simultaneously characterize signal behaviors across a wide range of scales.

Purpose of the Study:

• To introduce and explain the Scale-Dependent Lyapunov Exponent (SDLE) for multiscale analysis of physiological signals.
• To develop a unifying theoretical framework for SDLE-based multiscale analysis.
• To demonstrate the practical application of SDLE in diagnosing medical conditions using real-world data.

Main Methods:

• Description of the Scale-Dependent Lyapunov Exponent (SDLE) for accessibility to non-mathematical researchers.
• Development of a consistent, unifying theory for multiscale analysis using SDLE.
• Application of SDLE to analyze heart-rate variability (HRV) and electroencephalography (EEG) data.

Main Results:

• SDLE effectively characterizes diverse signal types, including deterministic chaos, noisy chaos, 1/f(α) processes, and stochastic limit cycles.
• SDLE uniquely detects fractal structures in non-stationary data and identifies intermittent chaos.
• SDLE successfully identified congestive heart failure from HRV data and seizures from EEG data.

Conclusions:

• The Scale-Dependent Lyapunov Exponent (SDLE) provides a powerful and versatile tool for multiscale analysis of complex physiological signals.
• SDLE offers unique capabilities for characterizing non-stationary and fractal data, advancing signal analysis in health sciences.
• SDLE demonstrates significant potential for clinical applications, including the early detection of diseases like heart failure and neurological disorders.