Model-Based Classification of Heart Rate Variability.

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Heart Rate Variability (HRV) analysis is crucial for risk assessment.
• Existing linear HRV features may not capture complex heart rate dynamics.
• Novel methodologies are needed to describe non-linear characteristics of HRV.

Purpose of the Study:

• To apply Autoregressive Fractionally Integrated Moving Average-Exponential Generalized Autoregressive Conditional Heteroskedasticity (ARFIMA-EGARCH) modeling to HRV data.
• To investigate the discrimination power of novel HRV features derived from ARFIMA-EGARCH modeling.
• To compare the efficacy of new HRV features against traditional linear measures.

Main Methods:

• ARFIMA-EGARCH modeling was applied to HRV recordings from 30 patients.
• A feature set combining linear HRV metrics and new measures (long memory, volatility persistence/asymmetry) was created.
• Principal Components Analysis (PCA) was used for feature selection.
• Quadratic discriminant analysis was employed for classification.

Main Results:

• The study identified long memory in the mean and persistence/asymmetry in volatility as key HRV features.
• PCA effectively reduced the dimensionality of the HRV feature set.
• Quadratic discriminant analysis achieved 93.3% discrimination accuracy using the new feature set.

Conclusions:

• ARFIMA-EGARCH modeling provides valuable insights into complex HRV dynamics.
• The novel HRV features derived from this modeling significantly improve discrimination accuracy.
• This methodology holds promise for enhanced risk assessment in cardiac patients.