Non-invasive measurements based model predictive control of pulsatile flow in an implantable rotary blood pump for heart failure patients

In this paper, firstly, we propose a linear time-variant (LTV) model to estimate the mean pulsatile flow (Q) in an implantable rotary blood pump (RBP). Non-invasive measurement of mean pulse-width modulation (PWM) signal acquired from the pump controller was used as an input to estimate the mean pulsatility index of pump rotational speed (PI) with this subsequently used to estimate the mean Q. Secondly, the proposed LTV model was used to develop a controller to regulate and track the variations in the mean Q and PI. We used a model predictive control (MPC) approach to develop the controller where this allowed us to explicitly apply pre-defined practical constraints to control input PWM as well as the output and the states of the system including; Q and PI . The model and the controller were tested using a parameter-optimized model of the cardiovascular system-rotary blood pump under wide ranges of speed ramp experiments carried out under different operating conditions such as variations in afterload, preload and heart contractility