Chronic heart failure detection based on long-term RR interval dynamics

Chronic heart failure (CHF) is a condition affecting millions worldwide, characterized by the heart’s reduced ability to pump blood efficiently. Conventional diagnostics, such as imaging and ECG assessments, can be time-consuming and expensive, often identifying CHF only after significant progression. Early detection is crucial for improving treatment options and reducing healthcare costs. Heart rate variability (HRV), which measures the variation in time intervals between heartbeats, is emerging as a non-invasive and cost-effective biomarker for CHF detection. HRV reflects the autonomic nervous system’s regulatory functions, often impaired in CHF patients. This study aims to assess advanced HRV measures for earlier CHF detection. The research involved examining CHF patients (N=934, Age 65±12) compared to healthy controls (N=274, Age 43±17). Data was sourced from Physionet and the Telemetric and Holter ECG Warehouse, with RR interval (RRI) data extracted from 24-hour Holter recordings. The study utilized dynamical detrended fluctuation analysis (DDFA), which considers changes in RRI correlations over time and scale, resulting in scaling exponent α ( t , s ) . This was further aggregated into scale and heart rate (HR)-dependent forms, α ( H R , s ) , classified using XGBoost ensemble method with 10-fold nested cross-validation. The classifier achieved 97% sensitivity and 90% specificity for distinguishing between CHF and control groups. Sensitivity and specificity remained consistent across subgroup analyses based on beta blocker medication and NYHA class. This method demonstrated high classification accuracy, suggesting potential utility for early CHF detection, independent of CHF severity. • 97% sensitivity in chronic heart failure detection using advanced HRV measures. • Method independent of chronic heart failure severity and beta blocker medication. • Potential for integration into wearable and remote monitoring systems.