Design Method Using Statistical Models for Miniature Left Ventricular Assist Device Hydraulics

Left ventricular assist devices (LVADs) are increasingly used to treat heart failure patients. These devices' impeller blades and diffuser vanes must be designed for hydraulic performance and hemocompatibility. The traditional design method, applying mean-line theory, is not applicable to the design of small-scale pumps such as miniature LVADs. Furthermore, iterative experimental testing to determine how each geometric variable affects hydraulic performance is time and labor intensive. In this study, we tested a design method wherein empirical hydraulic results are used to establish a statistical model to predict pump hydraulic performance. This method was used to design an intra-atrial blood pump. Five geometric variables were chosen, and each was assigned two values to define the variable space. The experimental results were then analyzed with both correlation analysis and linear regression modeling. To validate the linear regression models, 2 test pumps were designed: mean value of each geometric variable within the boundaries, and random value of each geometric variable within the boundaries. The statistical model accurately predicted the hydraulic performance of both pump designs within the boundary space. This method could be expanded to include more geometric variables and broader boundary conditions, thus accelerating the design process for miniature LVADs.