Combination of data approaches to heuristic control and fault detection [heart assist devices

Abstract

Heart assist devices based on turbo-hydrodynamic methods of pumping are smaller, more efficient, and more reliable than the reciprocating, pulsatile methods used in current devices. However, they pose a more difficult control problem because the appropriate setpoint for the pump speed depends on preload and afterload, which vary with time. The pump should be run as fast as possible to maximize cardiac output, but ventricular suction can occur when the device attempts to pump more blood than is available. Suction must be detected quickly and the pump speed reduced before suction damages the heart. In many cases, control algorithms must rely on limited or no measurement of hemodynamic variables. We describe a method for suction detection and pump speed control based only on measurements of pump current, speed, and, when available, pump flow. Four heuristic indices that can detect suction are used, and the data are fused to decide whether suction is present If suction is present, speed is immediately reduced. If suction is not present, the indices are used as inputs to a fuzzy controller to adjust the pump speed so that the pump is operating just below the speed that induces suction. Data fusion techniques for suction detection based on Bayesian, fuzzy logic and Dempster-Shafer theory were evaluated. Fusion techniques based on fuzzy logic and Dempster-Shafer theory provide a measure of uncertainty in the fused result. This uncertainty measure can be used in the control process, and it can also be used to identify faults in pump operation.