Cardiac and circulatory dysfunction are leading causes of admission, cost, and mortality in the intensive care unit (ICU). However, choosing a suitable treatment is extremely difficult, as a wide range of complex and patient-specific dysfunction types are found. Furthermore, due to the limits and constraints on the currently obtainable data, a full, clear picture of patient state cannot be precisely delineated, which can result in misdiagnosis and incorrect treatment choices.

To overcome this problem, cardiovascular parameters essential for correct diagnosis and treatment must be accurately estimated from clinically available measured data. Specifically, the volume of blood ejected from the heart per beat, known as stroke volume (SV), needs to be estimated from easily accessible measurements, such as blood pressure, as it is an important hemodynamic parameter for assessment of cardiovascular performance. This goal can be accomplished by “adding value” to existing clinical data using physiological models of the cardiovascular system.

This research develops a novel aortic model and patient-specific hemodynamic parameter identification method for continuous and accurate estimation of SV using measurements commonly available in the ICU. Thus, SV can be acquired in a non-additionally invasive fashion. In addition, use of this SV measurement can enhance the diagnosis, treatment and therapeutic decision support of bedside clinicians.

The aortic model developed in this thesis uses continuous aortic pressure waveform and pulse wave velocity (PWV) as inputs to estimate SV. The parameters within the aortic model are aortic characteristic impedance, aortic compliance, and systemic resistance, and are identified beat-to-beat. These parameters are used to compute blood flow and thus to estimate SV for every heartbeat.

The SV estimation method is validated with two series of pig experiments involving administration of dobutamine and inducing septic shock, where direct and invasive measurement of SV is obtained for a gold standard comparator. In addition, SV is significantly changed throughout the experiment by modifying preload using various levels of positive end expiratory pressure, as well as fluid administration. The method developed is also compared against the PiCCO system from PULSION, a commercially available pressure-based SV estimation device that is currently considered the most accurate in the critical care.

The Bland-Altman results from the porcine study showed clinically acceptable accuracy within approximately ± 30% by the developed model. The PiCCO system also showed similar accuracy compared with the direct SV measurement. However, the PiCCO system required multiple calibrations during the pig study, while the developed method only required one. This result suggests that the developed model and methods are more accurate and clinically useful, particularly when hemodynamic instability is present.

Overall, the model developed in this research shows great potential for improving patient care in the ICU. The model offers key hemodynamic parameters for optimizing cardiovascular treatment. In particular, accurate titration of fluid, inotropes, and vasoactive drugs to patient specific responses are now possible. The overall methods and model can be generalized to outpatient management.

The overall outcome provide new opportunities to reduce the cost of care, while improving quality. Adding value to existing measurements has not previously been proven in circulartory management. Hence, this research provides a template for further advances, particularly in the highly monitored critical environment.