Nonlinear, data-driven modeling of cardiorespiratory control mechanisms

We present applications of recently developed algorithms for data-driven nonlinear systems identification to the study of cardiovascular and respiratory control mechanisms on an integrated systems level, utilizing experimental data obtained during resting conditions. Specifically, we consider cerebrovascular regulation during normal conditions, orthostatic stress and autonomic blockade in a two-input context, as well as respiratory control during a model opioid drug (remifentanil) infusion in a closed-loop context. The results illustrate the potential of using data-driven modeling approaches, which do not rely on prior assumptions about model structure, for modeling physiological systems, as they are well-suited to their complexity. They also illustrate the potential of utilizing spontaneous physiological variability, which can be monitored noninvasively and does not require experimental interventions, to extract rich information about the function of the underlying mechanisms. We also discuss some important practical issues, such as the presence of nonstationarities and model order selection, related to the application of similar approaches to the analysis of physiological systems.