On the Robustness of Multiscale Indices for Long-Term Monitoring in Cardiac Signals

Area of Science:

• Cardiology and Biomedical Engineering
• Nonlinear Dynamics and Time Series Analysis
• Medical Signal Processing

Background:

• Sudden Cardiac Death (SCD) risk identification is crucial, with Heart Rate Variability (HRV) analysis from electrocardiogram (ECG) recordings showing predictive value.
• Multiscale nonlinear methods for HRV analysis offer more comprehensive insights into cardiac dynamics than single-scale measures.
• Current knowledge on the long-term robustness of these nonlinear HRV measures in multi-day monitoring is limited.

Purpose of the Study:

• To scrutinize the long-term robustness of three nonlinear HRV methods: Multiscale Entropy (MSE), Multiscale Time Irreversibility (MTI), and Multifractal Spectrum (MFS).
• To assess the suitability of these methods for characterizing cardiac dynamics in extended Holter monitoring (7-day recordings).
• To propose a new statistical procedure for comparing multiscale representations across different patient groups or processing conditions.

Main Methods:

• Analysis of 7-day Holter recordings from patients with Atrial Fibrillation and Congestive Heart Failure.
• Application of Multiscale Entropy (MSE), Multiscale Time Irreversibility (MTI), and Multifractal Spectrum (MFS) to assess cardiac dynamics up to 100 time scales.
• Development of a non-parametric statistical comparison procedure using confidence intervals for averaged median differences.

Main Results:

• Multiscale estimators demonstrated variance reduction in long-term recordings.
• MSE and MTI showed the lowest bias and variance at large scales, respectively, with consistent large-scale process descriptions across methods.
• MSE exhibited greater standard error reduction over several days compared to one-day recordings, while MFS showed more apparent bias.

Conclusions:

• Nonlinear HRV methods, particularly MSE and MTI, are robust for characterizing cardiac dynamics in long-term Holter monitoring.
• These techniques, with improved algorithms and statistical tests, can enhance risk stratification for Sudden Cardiac Death (SCD).
• Further research into algorithmic inconsistencies at unclear origins is warranted for optimal application in clinical settings.