Variability Analysis for Noisy Physiological Signals: A Simulation Study

Physiological monitoring is prone to artifacts originating from various sources such as motion, device malfunction, and interference. The artifact occurrence not only elevates false alarm rates in clinics but also complicates data analysis in research. When techniques to characterize signal dynamics and the underlying physiology are applied (e.g., heart rate variability), noise and artifacts can produce misleading results that describe the signal artifacts more than the physiology. Signal quality metrics can be applied to identify signal segments with noise and artifacts that would otherwise lead analyses to produce non-physiologic or misleading results. In this study we utilized simulated electrocardiogram signals and artifacts to demonstrate effects of noise on heart rate variability frequency domain methods. We then used these simulations to assess an automated artifact correction algorithm that included a signal quality index comparing electrocardiogram beats to a beat template. Simulation results show that the proposed algorithm can significantly improve estimation of signal spectra in presence of various artifacts. This algorithm can be applied to automatically clean real world physiological time series before conducting variability analysis.