Modeling human left ventricle and aortic function using VisSim

Accurate prediction of cardiac output via non-invasive techniques has multiple applications in clinical care and research. Models developed for this purpose have typically not been able to function in real time on commonly available platforms. A real time model that can predict cardiac output for individual patients from signature data, such as the arterial pressure pulse, and can also accommodate cardiac instabilities may prove to be a useful tool. The initial goal of this project, reported here, was to develop a model of the left ventricle and aorta (in the immediate vicinity of the left ventricle) which responds appropriately to variations in significant physical variables. Model development was based upon standard fluid statics and dynamics principles for predicting flow and pressure; the pressure-volume characteristics of the cardiac cycle; and previously published empirical relationships for elasticity and pressure decay. The phases of the cardiac cycle provided both the physical and logical model structure, with Boolean algebra used to link model sections. The model was implemented using VisSim, a modeling software package which uses a block diagram language which is easily learned and highly intuitive. Model predictions of left ventricular pressure, aortic pressure, ventricular volume, and cardiac output were compared to published curves for healthy individuals. These comparisons suggest that the prototype model closely approximates expected behavior of the variables examined. In its present form, the model responds in a plausible manner to variations of independent variables. Detailed comparisons to clinical data have, however, not yet been performed. The present model represents only a first step toward a reliable real time model of left ventricle output. Although largely mechanistic, there is a significant empirical component which may ultimately create difficulty. Certain of the mechanistic relationships used may also prove simplistic. A detailed comparison with clinical data must be undertaken as a second step to evaluate the model's sensitivity to individual variation. Further model refinement will likely require use of more complex computational solutions. As hardware capability continues to improve, there is reason to hope that this approach may ultimately yield reliable predictions in real time.