Instantaneous Bispectral Characterization of the Autonomic Nervous System through a Point-Process Nonlinear Model

Assessment of Heartbeat nonlinear dynamics is an important topic in the study of cardiovascular control physiology. In this paper, we introduce an inverse-Gaussian pointprocess model where an input-output Wiener-Volterra model is linked to a quadratic autoregression within the probability structure in order to estimate the dynamic spectrum and bispectrum of the considered heartbeat dynamics. The proposed framework was tested with an experimental ECG dataset with subjects undergoing a tilt-table procedure. Results show that our model is useful in estimating previously defined instantaneous indices of heart rate (HR) and heart rate variability (HRV). Results demonstrate that the algorithm confirms the characterization of the tilt effect on standard and instantaneous indices of the sympatho-vagal balance, while simultaneously tracking significant changes in the inherent nonlinearity of heartbeat dynamics with tilt.