Continuous estimation of cardiac output and arterial resistance from arterial blood pressure using a third-order Windkessel model

Abstract

Intensive Care Units (ICUs) have high impact on the survival of critically-ill patients in hospitals. Recent statistics have shown that only 10% of the 5 million patients admitted to ICUs in the United States die each year. In modern ICUs, the heart's electrical and mechanical activity is routinely monitored using various sensors. Arterial blood pressure (ABP) and heart rate (HR) are the most commonly recorded waveforms which provide key information to the ICU clinical staff. However, clinicians find themselves in many cases unable to determine the causes behind abnormal behavior of the cardiovascular system because they lack frequent measures of cardiac output (CO), the average blood flow out of the left ventricle. CO is monitored via intermittent thermodilution measurements which are highly invasive and only applied to the sickest ICU patients. The lack of frequent CO measurements has encouraged researchers to develop estimation methods for cardiac output from routinely measured arterial blood pressure waveforms. The prospects of estimating cardiac output from minimally-invasive blood pressure measurements has resulted in numerous estimation algorithms, however, there is no consensus on the performance of the algorithms that have been proposed. In this thesis, we investigate the use of a third-order variation of the Windkessel model, which is referred to as the modified Windkessel model. We validate its ability to generate well-behaved proximal and distal pressure waveforms for a given flow waveform and thus characterize the arterial tree. We also develop a model-based CO estimation algorithm which uses central and peripheral blood pressure waveforms to obtain reliable estimates of CO and the total peripheral resistance (TPR). We applied the estimation algorithm to a porcine data set.

(cont.) The results of our estimation algorithm are promising: the weighted-mean root-mean-squared-normalized-error (RMSNE) is about 13.8% over four porcine records. In each porcine experiment, intravenous drug infusions were used to vary CO, ABP, and HR over wide ranges. Our results suggest that the modified Windkessel model is a good representation of the arterial tree and that the estimation algorithm yields reliable estimates of CO and TPR under various hemodynamic conditions.