Model-Free Physiological Denoising Using State-Space Reconstruction and Time Reversal

Physiological systems exhibit nonlinear deterministic behavior driven by physiological noise. Since the exact deterministic functions governing these systems are unknown, estimating physiological noise and, in particular, performing effective denoising is challenging. This study introduces a model-free denoising method for biomedical signals based on state-space reconstruction and time-reversed forecasting. The method's effectiveness is demonstrated on discrete-time noisy synthetic data, where it outperforms existing techniques. Applying the method to real Heart Rate Variability (HRV) series (10 subjects from each cohort: healthy, heart failure, and atrial fibrillation) resulted in reduced physiological noise and decreased Sample Entropy (SampEn) across all groups. While a standard HRV analysis associated the highest SampEn with atrial fibrillation, denoising revealed that healthy individuals actually exhibit the highest cardiac complexity. Additionally, the method effectively enhances the distinction between healthy and heart failure conditions.Clinical relevance- The proposed physiological noise reduction technique offers new unbiased insights into the complexity and dynamical analysis of the cardiovascular system.