An axial-flow ventricular assist device (VAD) under development at the authors' facility is intended for use as a long-term implantable device. At high speeds axial-flow VADs can collapse the native ventricle and damage the heart muscle, lung tissue, and blood. A prototype algorithm was developed to maintain physiologic perfusion to the vital organs while preventing ventricular collapse, through analysis of the electrical current waveform of the motor. The premise of the control algorithm is that the hemodynamics of the patient are reflected in the shape of this waveform. This approach is intended to eliminate the need for invasive sensors, thus effectively using the pump itself as a transducer. The control algorithm regulates the speed of the pump by comparing the motor-current waveform with reference waveforms using a matched filter. The matched filter was evaluated by its classification and differentiation performance. Thus far, the authors have been able to classify the waveforms into one of the four physiologic regions (below, within, or above the optimal range, and ventricular suction) with over 90% reliability. Ongoing work is directed toward improving the detection of ventricular suction, as this condition must be strictly avoided.