Principal component based system identification and its application to the study of cardiovascular regulation

Abstract

(cont.) Our methods analyze the coupling between instantaneous lung volume and heart rate and, subsequently, derive representative indices of parasympathetic and sympathetic control based on physiological and experimental findings. The validity of each method is evaluated via experimental data collected following interventions with known effect on the parasympathetic or sympathetic control. With the above techniques, this thesis explores an important topic in the field of space medicine: effects of simulated microgravity on cardiac autonomic control and orthostatic intolerance (OI). Experimental data from a prolonged bed rest study (simulation of microgravity condition) are analyzed and the conclusions are: 1) prolonged bed rest may impair autonomic control of heart rate; 2) orthostatic intolerance after bed rest is associated with impaired sympathetic responsiveness; 3) there may be a pre-bed rest predisposition to the development of OI after bed rest. These findings may have significance for studying Earth-bound orthostatic hypotension as well as for designing effective countermeasures to post-flight OI. In addition, they also indicate the efficacy of our proposed methods for autonomic function quantification.

System identification is an effective approach for the quantitative study of physiologic systems. It deals with the problem of building mathematical models based on observed data and enables a dynamical characterization of the underlying physiologic mechanisms specific to the individual being studied. In this thesis, we develop and validate a new linear time-invariant system identification approach which is based on a weighted-principal component regression (WPCR) method. An important feature of this approach is its asymptotic frequency-selective property in solving time-domain parametric system identification problems. Owing to this property, data-specific candidate models can be built by considering the dominant frequency components inherent in the input (and output) signals, which is advantageous when the signals are colored, as are most physiologic signals. The efficacy of this method in modeling open-loop and closed-loop systems is demonstrated with respect to simulated and experimental data. In conjunction with the WPCR-based system identification approach, we propose new methods to noninvasively quantify cardiac autonomic control. Such quantification is important in understanding basic pathophysiological mechanisms or in patient monitoring, treatment design and follow-up.