Parametric design investigation on geometric orifice area and coaptation area of polymeric heart valves

Valvular diseases are deemed as a common condition in the US. Valve replacement treatments are implemented in severe cases to substitute the native valve with either prosthetic or mechanical heart valves. Prosthetic heart valves are designed after the native valves and can mimic the behavior of the native valves without disturbing the blood flow. However, their lifespan is limited to 10 to 15 years. Therefore, prosthetic valves have been studied and tested widely to obtain a better understanding of the overall performance, the geometric orifice area (GOA), the coaptation area (CA), and exhibited stresses. Three studies were offered in this dissertation. First, a finite element (FE) based parametric design domain investigation of a beating left ventricular (LV) simulator was presented. The LV simulator was shown to be a candidate for testing of polymeric heart valves (PHVs), or any cardiovascular device, by mimicking the heart wall motion. Through 150 FE models, the mechanical tendencies of the LV simulator were identified. Second, the effect of the fundamental curves on the GOA and the CA of the PHVs were investigated using ANOVA. The unique shapes of the valve leaflets have been constructed by researchers using parametric equations, 2 – D splines, 3 – D splines, and or a combination of the mentioned. In the study, the leaflet geometry was defined with the attachment curve, the belly curve, the free edge. A total of 8 different valve geometries were created with different control point combinations. The results showed that the control coordinate of the belly curve had a greater impact on the CA for the valve models with a higher average 100 % modulus. The GOA was affected by a combined contribution of control points of the attachment and the belly curve. Third, a deep learning (DL) framework was presented to estimate the GOA and the CA values, under the time-varying transvalvular pressure, of any proposed coordinate combination to accelerate the design efforts of PHVs. The model space was populated with 175 different valve geometries with different control point combinations. It was shown that the proposed DL framework can accurately predict the GOA and CA values. Also, it reduced the required computation time of the FE process of one coordinate combination by 20 times. This research contributes to the design efforts of PHVs by identifying the relationships between the construction curves on the valve performance and offering a DL driven substitute for the time extensive computational analysis.