Physiological control of left ventricular assist devices based on gradient of flow

The new generation of left ventricular assist devices (LVADs) is based on rotary pumps used to augment the blood flow from the failing heart of a patient. A controller is necessary for these devices to properly control the blood flow, to detect failure modes, and to avoid adverse effects of the pump operation. In particular, it is essential that the pump accommodate changes to preload and afterload of the patient's circulatory system. The pump must increase blood flow in response to additional venous return, and maintain pressure in response to changes in systemic vascular resistance (SVR) of the patient. However, the limited availability of observable variables makes it difficult to explicitly close the feedback loop to generate the control. In this paper, we describe a feedback mechanism which uses the gradients of mean pump flow and minimum (diastolic) pump flow to control the pump speed. The objective of this controller is to optimize flow while avoiding suction. The optimum speed is obtained by minimizing in real time the gradient of blood flow with respect to speed. Simulations were carried out on a state-space model of the cardiovascular system combined with a rotary pump to demonstrate the responsiveness and robustness of this algorithm. The results show that the gradient methods can continuously track the optimal flow point in response to step perturbations of preload and SVR. Our results also show that a feedback controller based on the gradient method applied on a diastolic pump flow demonstrates better performance than when applied on mean pump flow.