Abstract
Abstract
Aortic valve disease is a common valvular heart disorder that often progresses to heart failure. Aortic valve replacement with artificial heart valves is the standard treatment, making valve performance critical to patient outcomes and driving growing interest in optimized polymeric heart valves. Although prosthetic valve replacement is widely used, current artificial valves still face durability–hemodynamics tradeoffs. In this study, we sought to explore a computational design framework for improving the flow area, leakage, and stress distribution of a polyurethane-based synthetic polymer heart valve by simultaneously optimizing leaflet thickness and height through a data-driven workflow. A three-dimensional fluid–structure interaction model of a tri-leaflet valve made from siloxane-modified polyurethane was constructed and meshed. Twelve design points and three verification points covering the clinical design space were analyzed for effective orifice area (EOA), regurgitant fraction (RF), and peak von Mises stress. Response-surface surrogate models were fitted and interrogated using a multi-objective genetic algorithm. The leading Pareto solution was subsequently resimulated for half a cardiac cycle to evaluate its predicted performance. The surrogate prediction error was quantified as the absolute difference between surrogate and simulation results, and the ability to identify designs meeting ISO 5840-3 limits (EOA ≥ 125 mm2, RF ≤ 20%) was assessed. The optimized geometry (thickness 0.197 mm; height 15.49 mm) increased EOA from 140.5 mm2 to 164.5 mm2, reduced RF from 3.10% to 2.88%, and lowered peak stress from 1.88 MPa to 1.59 MPa. Surrogate predictions differed from the underlying simulation results by ≤5.3 mm2 for EOA, 0.59 percentage-point for RF, and 0.081 MPa for stress. Coupled optimization of leaflet thickness and height improved simulated hemodynamic and mechanical performance while satisfying ISO 5840-3 requirements. The presented workflow—combining fluid–structure interaction simulation, response-surface modeling, and multi-objective optimization—illustrates a computational methodology for exploring polymeric heart valve design spaces. Because the computational model has not yet been validated against independent experimental measurements, the reported performance improvements should be interpreted as preliminary computational findings pending future validation and comprehensive verification, validation, and uncertainty quantification (VVUQ).