Comparison of shear stress, residence time and Lagrangian estimates of hemolysis in different ventricular assist devices

Millions of people are diagnosed with heart failure each year and thousands would benefit from a heart transplant if there were enough donor hearts. Ventricular Assist Devices (VADs) are blood pumps which augment the failing heart’s pumping capacity. They are already in clinical use as bridge-to-transplant but could benefit more patients as end-stage therapy. While current devices are more biocompatible than their forerunners they still have problems; device-induced blood damage, including hemolysis, platelet activation, thrombosis and embolization, may still cause serious clinical events such as strokes and renal damage. Reliable computational methods for predicting blood damage, including hemolysis, are desirable since they will aid in selecting devices to be used and the design of new devices. Flow-induced hemolysis is a function of the shear stress on the erythrocytes and the exposure to this shear stress. Computational fluid dynamics (CFD) was used to analyze the flow field in a range of VADs, for a range of operating conditions, and calculate shear stress and residence time parameters. These parameters may give some indication of the hemolysis potential for these devices, however since hemolysis is a function of shear stress and exposure time a model is required. Lagrangian models of hemolysis use flow streamlines as an indication of the paths taken through the device by erythrocytes. The hemolysis at the outlet is then calculated as an incremental function of the shear stress and exposure time along these lines. Various Lagrangian models were used to compute hemolysis indices for the VADs and the results were compared with experimental measurements. Best agreement was found with a model that integrates the temporal derivative of the power law equation along the streamlines. In a centrifugal VAD with a rotational speed of 4000 rpm and flow rate of 5 l/min, the model predicted the HI was 5.0x10− 4 % as compared with the experimental result 5.4x10− 4 %.