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

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Chronic heart failure (CHF) affects millions globally, with current diagnostics often detecting it late.
• Early CHF detection is vital for effective treatment and reduced healthcare burden.
• Heart rate variability (HRV) shows promise as a non-invasive biomarker for autonomic dysfunction in CHF.

Purpose of the Study:

• To evaluate advanced HRV measures using dynamical detrended fluctuation analysis (DDFA) for earlier and more accurate CHF detection.
• To assess the efficacy of DDFA-derived metrics in differentiating CHF patients from healthy controls.
• To determine if the method's accuracy is influenced by CHF severity or medication.

Main Methods:

• Utilized 24-h Holter ECG recordings from 934 CHF patients and 274 controls.
• Extracted RR interval (RRI) data and applied DDFA to derive scaling exponents α(t,s) and α(HR,s).
• Employed an XGBoost ensemble classifier with 10-fold nested cross-validation for group discrimination.

Main Results:

• The classifier achieved high diagnostic accuracy: 97% sensitivity and 90% specificity in distinguishing CHF from controls.
• Classification performance remained robust across subgroups, including those on beta blockers or with varying NYHA classes.
• The DDFA-based HRV analysis demonstrated significant potential for early CHF detection.

Conclusions:

• Advanced HRV analysis via DDFA provides a highly accurate and non-invasive method for early chronic heart failure detection.
• This approach shows promise for identifying CHF independently of disease severity or common treatments.
• The findings support the clinical utility of HRV-based biomarkers for proactive CHF management.