Identifying physiologically significant pumping state transitions in implantable rotary blood pumps used as left ventricular assist devices: an in-vivo study P.J. Ayre 1,2 , N.H. Lovell 1 , R. W. Morris 3 , M. L. Wilson 2 and J.C. Woodard 2 1 Graduate School of Biomedical Engineering, University of New South Wales, Sydney NSW 2052, Australia. 2 Ventrassist, Sydney NSW 2067, Australia 3 Sydney Medical Simulation Centre, Royal North Shore Hospital, Sydney NSW, Australia Abstract - The VentrAssist implantable rotary blood pump (IRBP) is a centrifugal pump that uses a hydrodynamic bearing to support its impeller. The pump is to be used as a left ventricular assist device (LVAD). Varying pump speed can control the degree of left ventricular assistance. By increasing impeller speed, it is possible to transition from the normal physiological state of ventricular ejection (VE) to a state where the aortic valve remains closed (AC) throughout the cardiac cycle. Using the non-invasive parameter of instantaneous impeller speed in an ovine experimental model (N=3), we investigated state transitions. The cardiovascular system of the animal was perturbed by pharmacological intervention or by exsanguination. A total of six pump speed set point changes that caused physiological state transitions (VE to AC) were examined. A state transition index (STI) derived originally from data obtained in an in-vitro mock loop setup was found to be directly applicable in the in-vivo studies and showed statistically significant (p<0.0005) reliability in differentiating between no change in state and change in state. These data indicate that the STI may be a valuable mechanism to in optimal LVAD control. Keywords - Implantable rotary blood pump, pumping states, control strategy, left ventricular assist device. I. INTRODUCTION The VentrAssist (Micromedical Industries, Sydney) implantable rotary blood pump (IRBP) has a novel hydrodynamic bearing that produces a characteristically flat pump-head versus pump-flow curve [1,2,3,4]. The pump is to be used as a left ventricular assist device (LVAD) with both bridge-to-transplant and long-term implantation anticipated. Current commercially used rotary pumps make no attempt to automatically control pump speed to optimize ventricular assistance, with fixed speed pumps being in common usage. This level of control may not be satisfactory for long-term implantation of these devices as optimal pump speed will be influenced by a myriad of factors, including variation in venous compliance, arterial resistance, heart rate, ventricular contractility and metabolic demand. There is obviously a need to detect pumping states that cause such deleterious affects as ventricular collapse due to over- pumping or pump back flow (regurgitation) as a result of under-pumping [5]. Other physiological heart states exist within these extremes. The normal state would be where left ventricular ejection is occurring and there is a net positive aortic and pump flow (state VE). A state that would be of long-term concern to an implant recipient would occur at a higher pump speed where there is insufficient blood in the ventricle to sustain normal left ventricular ejection and the aortic valve remains closed throughout the entire cardiac cycle (state AC). In this instance there is no net positive aortic flow. Stasis of blood distal to the aortic valve could lead to significant patient complications due to clotting. While it is a routine matter to determine aortic valve closure (and aortic flow) from invasive implantation of flow and pressure transducers, such transducers cause significant patient complications in long-term implant [5] and thus should be avoided if possible. In-vivo experimentation in the transition of states has previously been investigated [6,7,8,9,10]. These studies concentrated primarily on the detection of ventricular collapse using motor current analysis in the time and frequency domains. Amin et al. [6] used an axial blood pump to examine the condition of complete aortic valve closure with increasing pump speed. Pump current was mentioned as a method of analysis but the authors made no attempt at devising an algorithm to detect this transition. The hypothesis that we propose is that by using only the non- invasive measure of instantaneous pump impeller speed, it is possible to detect the afore-mentioned state transition from VE to AC. II. MATERIALS AND METHODS In-vivo Experiments Three acute ovine experiments were conducted with the heart instrumented as shown in Fig. 1 to record left ventricular pressure (LVP), aortic pressure (AoP) pump differential pressure (Pump dP), aortic flow, pump flow. The transition between states was induced by changes in pump speed set point for hypertensive, normovolemic and hypovolemic