Left Ventricular Assist Device Flow Pattern Analysis Utilizing a Novel Computational Fluid Dynamics Model: Using Physics and Physiology to Better Define the Pump-Patient Interaction

Abstract

Introduction Blood flow patterns through the left ventricular assist device (LVAD) and aorta remain incompletely understood. Computational fluid dynamics (CFD) allows for analysis of flow in the cardiovascular system although current LVAD models have failed to fully incorporate the interplay between the pulsatile left ventricle and continuous-flow generated by the LVAD. Hypothesis Flow through the LVAD is dependent on the interaction between device and patient specific factors with suboptimal flow patterns evoking increased risk of LVAD-related complications. CFD can be used to analyze how different pump and patient factors affect flow through the system. Methods CFD simulations were carried out on a patient with a HeartMate II (HMII) (Fig A). Simulations were also done substituting HeartWare HVAD (Fig B), HeartMate 3 (HM3) in continuous mode (Fig C) and HM3 with Artificial Pulse (Fig D). An anatomical model of the patient was reconstructed from CT images. The LVAD flow was calculated separately using a lumped-parameter-model (LPM), which was calibrated to the patient based on the patient-specific ventricular volume change reconstructed from 4 dimensional CT and hemodynamic tracings. The LVADs were implemented in the LPM via published H-Q curves. In order to quantify the flushing effect, virtual contrast agent was released in the ascending aorta and its flushing over the cycles was quantified. Results Under standard operation conditions of the LVADs (9200 RPM for HMII, 5500 RPM for HM3 and 2600 RPM for HVAD), the cardiac outputs were 5.92 L/min, 6.14 L/min and 6.90 L/min, for HMII, HM3 and HVAD, respectively. The velocity of blood flow in the outflow cannula was higher in the HVAD than in the two HeartMate pumps with a cycle average (range) of 0.72 m/s (0.57-0.93 m/s), 0.70 m/s (0.64-0.77 m/s) and 1.61 m/s (1.21-1.92 m/s) for HMII, HM3 and HVAD, respectively. Artificial Pulse increased the peak flow rate to 9.84 L/min for the HM3 but the overall cardiac output was 5.96 L/min, which was similar to the continuous mode (Fig E). Artificial pulse markedly accelerated flushing of the blood from the ascending aorta; after 6 cardiac cycles, 43% of the blood was flushed out from the ascending aorta under the continuous operation mode while 61% was flushed under artificial pulse (Fig F). Conclusions Pump-specific factors such as LVAD type and programmed flow algorithms lead to unique flow patterns that can be studied using a novel CFD model to better understand and potentially mitigate the risk of downstream LVAD complications.