A forward model-based analysis of cardiovascular system identification methods

Abstract

Cardiovascular system identification is a potentially powerful approach for intelligent patient monitoring of cardiovascular function. Rather than merely recording hemodynamic signals, the signals are mathematically analyzed so as to provide a dynamical characterization of the physiologic mechanisms responsible for generating them. The fundamental aim of this thesis is to develop and evaluate cardiovascular system identification methods based on a test bed of data generated from a forward model of the cardiovascular system whose dynamical properties are known. To this end, we developed a computer model of the human cardiovascular system which includes a lumped parameter model of the heart and circulation and a model of the short-term cardiovascular regulatory system continuously disturbed by resting physiologic perturbations. The short-term regulatory system consists of arterial and cardiopulmonary baroreflex systems and a direct neural coupling mechanism between respiration and heart rate. The resting physiologic perturbations include respiratory activity and stochastic disturbances to total peripheral resistance (TPR) and heart rate representing, for example, autoregulation of local vascular beds and higher brain center activity. We demonstrated that this model emulates experimental data in terms of steady-state pulsatility, limiting static behavior, and low frequency hemodynamic variability. We first evaluated the performance of a previously developed cardiovascular system identification method against the forward model.

(cont.) The method involves the analysis of fluctuations in heart rate, arterial blood pressure (ABP), and instantaneous lung volume in order to characterize quantitatively important physiologic mechanisms including, for example, the heart rate baroreflex. From this analysis, we inferred that the cardiovascular system identification results derived from experimental data are likely to reflect the actual system dynamics of underlying physiologic mechanisms. We then introduced novel identification methods for quantifying TPR baroreflex dynamics from only fluctuations in cardiac output and ABP and for monitoring steady-state changes in TPR from only the ABP waveform. We demonstrated the efficacy f these identification methods with respect to forward model generated data and a preliminary set of experimental data. The results of this forward model-based analysis motivate the experimental validation of the cardiovascular system identification methods considered in this thesis.