Modeling and Estimation of Cardiorespiratory Function, with Application to Mechanical Ventilation

Evidence-based medicine is at the heart of current medical practice where clinical decisions are driven by research data. However, most current therapy recommendations follow generalized protocols and guidelines that are based on epidemiological (population) studies and thus not suited for the individual patient's demands. Patient-tailored therapies are considered, hence, an unmet clinical need. We believe that mathematical models of the physiology can attend to such a clinical need, because they can be tuned to the individual patient. Such models provide a sound mathematical framework for personalized clinical decisions. In particular, physiological models in medicine can serve the following two purposes: 1) They can be an efficient tool to quantify cardiopulmonary dynamics, conduct virtual clinical/physiological experiments, and investigate the effects of specific treatments. 2) Model-based estimation techniques can assess physiological parameters or variables, which are otherwise impractical or dangerous to measure; they can effectively tune a generic model to become patient-specific, able to mimic the behavior of a particular patient. In this thesis, we propose a series of modifications to a previously developed cardiopulmonary model (CP Model) in order to better replicate heart-lung interaction phenomena that are typically observed under mechanical ventilation, hence allowing for a more accurate analysis of ventilation-induced changes in cardiac function. The response of this modified model is validated with experimental data collected during mechanical ventilation conditions. Further, as an industrial application of mathematical models, we present a patient emulator system that comprises the modified CP Model, a physical ventilator, and a piston-cylinder arrangement that serves as an electrical-to-hydraulic transducer. The modified CP Model then serves as the virtual patient that is being ventilated, where disease conditions can be instilled. Such a system is designed to offer a well-controlled experimental environment for ventilator manufacturers to efficaciously test and compare ventilation modalities and therapies, thereby enhancing their verification and validation manufacturing processes. Finally, we develop a model-based approach to estimate (noninvasively) the function of the cardiovascular system, in terms of cardiac performance (i.e., cardiac output) and the dynamics of the systemic arterial tree (i.e., time constant). With this technique, we envision to provide continuous and real-time bedside monitoring of changes in cardiovascular function, such as those induced by changes in ventilator settings