Prospective Evaluation of Wearable ECG Monitoring for Heart Failure Assessment: Signal Processing and HRV Analysis in NYHA Classification

The swift progress in wearable device technology has permitted electrocardiogram (ECG) monitoring to be conducted beyond the confines of traditional hospital settings, offering heart failure (HF) patients the benefit of affordable and prolonged home monitoring. This study addresses the challenges inherent in wearable ECG monitoring, including lower accuracy and increased susceptibility to motion artifacts and environmental noise, by employing a series of signal processing strategies to enhance ECG signal quality. We prospectively collected ECG signals from a cohort of hospitalized HF patients at West China Hospital using wearable devices, preprocessed the data to extract high-quality RR intervals, and developed a model using a random forest classifier combined with recursive feature elimination (RFE). The model manifested robust predictive accuracy, achieving an average AUC of 0.85 for both Micro- and Macro-Average ROC across New York Heart Association (NYHA) functional classifications. This research indicates the potential of wearable ECG monitoring in condition assessment and disease management for HF patients, pointing towards promising avenues for the future of personalized medicine and remote health monitoring strategies.