Parameter estimation of cardiovascular dynamics for artificial heart control

In order to realize control tasks of the artificial heart or monitoring of physiological state of the patient supported by the artificial heart, it is very important to measure parameters of cardiovascular dynamics such as cardiac output and blood pressure. However, we should avoid direct measurement of cardiac output or blood pressure by means of sensors positioned in the body, which may cause many problems such as biocompatibility, durability, drift, and so on. The present study tried to estimate, without direct measurement, cardiac output or blood pressure which may be used in controlling a total artificial heart (a motor driven centrifugal pump) by means of a neural network (NN) or a linear time series model (ARMA model) on the basis of the data obtained from a mock circulatory system. It was revealed that NN may be superior to ARMA model from the view point of versatility which will be important in the application stage of the total artificial heart.