CFC2023

Shape Optimization of the Total CavoPulmonary Connection: A Challenge between Data-Driven and Model-Driven Techniques

Shah, Imran (Dept Math&CS Emory University)
Ballarin, Francesco (Dep. Mathematics and Physics, Catholic University of Sacred Heart)
Iliescu, Traian (Dep. Mathematics, VA Tech)
San, Omer (Mechanical and Aerospace Engineering, Oklahoma State University)
Dasi, Lakshmi (Dep. Biomedical Engineering, GA Tech and Emory University)
Wei, Alan (Dep. Biomedical Engineering, UMass Lowell)
Veneziani, Alessandro (Dept Math&CS Emory University)

In session: MS1-06A - Modern Model-Driven and Data-Driven Computational Approaches in Fluid Mechanics

Please login to view abstract download link

In the past decade, new methodologies have been developed to reduce computational costs in fluid dynamics modeling. This opens new perspectives on using quantitative tools for surgery optimization. We focus on a shape optimization problem in pediatric surgery, the total cavopulmonary connection (TCPC). TCPC is common palliation for patients with single ventricle (SV) malformation. It consists of the artificial connection of the vena cava and the pulmonary artery in a cross-shaped graft. Some major complications correlated with TCPC hemodynamics stimulated the search for optimal shapes. We will demonstrate that Proper Orthogonal Decomposition \cite{hesthaven2016certified} and trust-region optimization methods \cite{cartis2022escaping} can actually provide rigorous optimal shapes in relatively low computational times. We discuss the comparison with data-driven approaches and Physical-Guided Neural Networks for simulating flow disturbances in TCPC \cite{ahmed2022physics}. Ackn: NSF DMS 2012253, 2012255, 2012286.