Hidden markov models and mutual information analysis to characterize nonlinear dynamics in heart rate variability

Area of Science:

• Cardiology
• Nonlinear Dynamics
• Biomedical Engineering

Background:

• Heart Rate Variability (HRV) analysis is crucial for assessing cardiac health.
• Nonlinear dynamics offer deeper insights into complex physiological systems like the heart.
• Idiopathic Dilated Cardiomyopathy (IDC) poses significant risks, necessitating improved diagnostic tools.

Purpose of the Study:

• To develop and validate a Hidden Markov Model (HMM) based methodology for analyzing nonlinear heart rate variability (HRV).
• To differentiate between healthy subjects (NRM) and patients with Idiopathic Dilated Cardiomyopathy (IDC), including high-risk (HR) and low-risk (LR) groups.
• To explore the utility of HMM-derived measures in classifying cardiac conditions based on RR time series.

Main Methods:

• Cardiac RR series data from 37 IDC patients and 46 NRM subjects were analyzed.
• Time series were decomposed into Very Low Frequency (VLF), Low Frequency (LF), and High Frequency (HF) bands.
• A Hidden Markov Model (HMM) with ergodic topology and 10 states was employed to model the HRV data, optimizing model parameters to maximize observation probability.

Main Results:

• HMM-based measures demonstrated high classification accuracy, reaching up to 100% in distinguishing between subject groups.
• The methodology effectively captured nonlinear dynamics within HRV across different frequency bands.
• Significant differences in HMM parameters were observed between IDC patients and healthy controls, and between HR-IDC and LR-IDC subgroups.

Conclusions:

• HMM provides a robust framework for analyzing nonlinear HRV dynamics.
• The developed HMM methodology shows significant potential as a non-invasive tool for diagnosing and risk-stratifying Idiopathic Dilated Cardiomyopathy (IDC).
• Further research can refine HMM parameters for enhanced predictive capabilities in cardiovascular disease management.