Shape Optimization of a NURBS Modelled Coronary Stent Using Kriging and Genetic Algorithm

Abstract

In this paper, structural shape of stent has been optimized using NURBS for parameterization of stent structure and target those objectives which are critical for vascular injury. NURBS modeling is done using python coding in RHINO 3D software. For later part of the design, Solidworks is used. The objectives considered in our study are dogboning, foreshortening and arterial wall stresses, all of which are strongly linked to vascular injury leading to restenosis. We use control point weights, strut thickness and strut width as design variables for Latin Hypercube sampling (LHS) in order to generate dataset for Stent deployment simulations. In our study, we generate 80 design data points using LHS in Matlab R2014a. Finite element analysis of stent deployment process is then carried out using ANSYS for all 80 designs of stent generated using LHS. Thereafter, we use Kriging for surrogate modeling and non-dominated sorting genetic algorithm (NSGA-II) in MATLAB for multi-objective design optimization so as to minimize dogboning, foreshortening and arterial wall stresses. As a result, we obtain a range of pareto optimal design parameter values which can be used in clinical design guides so as to accommodate variations observed across different patients.