Modeling of a New Sensorless Suction Detection System for the Rotary Left Ventricular Assist Device

The rotary left ventricular assist device (LVAD) has been successfully and clinically used for congestive heart failure patients. However, patients may be at significant risk for ventricular suction while using a LVAD. Suction may lead to ventricular collapse when the blood pump of the LVAD is working at higher rotational speeds to draw more blood than available. Some suction detection systems typically required the external implantation of sensors, which were not reliable for long-term use because of failure, baseline drift, and short lifespan of sensors. In this study, a new sensorless suction detection system only using the intrinsic pump parameter (pump speed) to eliminate all external sensors is proposed. Three features (suction indicators) are extracted from the pump speed signals without and with measurement noise and adopted as inputs to different classifiers. Two pumping states as no suction or suction are classified with each classifier. The in-silico results with a combined human circulatory system and LVAD model show that the proposed strategy can detect ventricular suction effectively with high accuracy and stability. It could provide theoretical basis and technology support for developing the LVAD controller with high reliability.