Percutaneous pulmonary valve implantation is an improved technique that is used to treat narrowed pulmonary valves or leaky pulmonary valves in patients with congenital heart disease. This technique represents a promising strategy to reduce surgical risk and operation. In clinical cases, commercial valve stents are sometimes not available for children or special subjects due to restrictions in stent size. Hence, the handmade pulmonary valved conduit provides a strategy to design stents with customized size for valve replacement. In this paper, we propose a meta-learning-based intelligent model to train an estimator (including two sub-estimators) to determine optimal trileaflet parameters for customized trileaflet valve reconstruction. This estimation model overcomes the problem of empirical parameter determination. The meta-learning model possesses learning-to-optimization capability for training generalized regression neural network by particle swarm optimization algorithm. Through incremental training patterns, this scheme can gradually enhance optimization to provide refined parameters for customized designs that can be applicable to individuals of all age groups. The customized handmade pulmonary valved conduit was validated by assessing the regurgitation fraction and the heart pump efficiency using an experimental cardiopulmonary circulation loop system.
All Science Journal Classification (ASJC) codes
- Computer Science(all)
- Materials Science(all)