A multivariate time-variant AR method for the analysis of heart rate and arterial blood pressure

This paper approaches the problem of short-term mechanisms that regulate heart rate and blood pressure variability signals, by focusing the evident changes of their frequency content during transients (dynamic situations in which the behaviour of these control mechanisms may vary on a beat-to-beat basis). In this study, we suggest an autoregressive time-variant spectral estimation method, which is able to follow such dynamic changes in the signals. This method has also been extended to a multivariate approach in order to take into account more than one process at a time, and to assess the mutual influences between the different controlling systems. The algorithms successfully tested on simulated series have also been used to analyse series recorded during a vaso-vagal syncope episode in a tilt manoeuvre and a physical exercise stress test protocol. The results show how this method is able to follow the changing dynamics of the signals on the basis of a closed-loop model of their interaction on a beat-to-beat basis. After a proper identification procedure of the blocks forming the model, it is possible, therefore, to obtain the classical spectral parameters and the gain of the transfer function between the signals. Such parameters constitute new time series that describe the physiopathology of the cardiovascular control systems, even during non-stationary epochs.