Generative design of heart valve prostheses

Customer: Kuzbass Cardiology Center (Kemerovo, Russia)

Summary: This study introduces a novel method for designing prosthetic heart valves (PHVs) by integrating machine learning (ML) and optimization algorithms. The proposed workflow consists of generating PHV designs using parametric modeling and evaluating their mechanical behavior through Finite Element Method (FEM) simulations. This workflow leverages ML to predict key performance metrics, such as lumen opening area and peak stress, based on design parameters. A comprehensive dataset of 11,565 unique PHV designs was used to train and validate the ML models. The best-performing models, based on ensemble methods, were then employed in conjunction with six state-of-the-art optimization algorithms to identify optimal PHV designs. Our approach achieved high accuracy, with mean absolute percentage errors of 11.8% for lumen opening and 10.2% for peak stress, and proved to be computationally efficient. The Tree-structured Parzen Estimator (TPE) and Nondominated Sorting Genetic Algorithm (NSGA) were particularly effective, yielding designs that balanced opening area and stress distribution.

Collaborators: Kirill Klyshnikov (Kuzbass Cardiology Center, Kemerovo, Russia), Evgeny Ovcharenko (Kuzbass Cardiology Center, Kemerovo, Russia)

Project type: Research

Media:  Journal paper,  GitHub repo