Fast and accurate numerical simulations for the study of coronary artery bypass grafts by artificial neural networks

Abstract

In this work, a non-intrusive data-driven ROM based on a POD–ANN approach is developed for fast and reliable numerical simulation of blood flow patterns occurring in a patient-specific coronary system when an isolated stenosis of the LMCA occurs. A CABG performed with the LITA on the LAD is analyzed. The introduction of a patient-specific configuration is an attractive element of this work because it makes possible to establish personalized clinical treatment. In addition, a FFD technique is used, which gives the opportunity to deform directly the mesh and not only the geometry. Furthermore, the combination of ROM, FV technique and neural networks makes this study mathematically appealing.