A GPU parallelization scheme for 3D agent-based simulation of in-stent restenosis

The collective biomechanical and mechanobiological effects of cell behaviors in in-stent restenosis (ISR) can be analyzed numerically using FEM (finite element method)-ABM (agent-based model) coupling approach. Traditional FEM-ABM frameworks, limited to simulations with a small amount of agent cells, are impossible for large-scale analysis, especially 3D simulations. A 3D parallelization scheme for FEM-ABM coupling simulations based on GPU acceleration is proposed to improve the computational efficiency of the coupling framework. Three-dimensional vascular restenosis simulations have been conducted to validate the algorithm and comparisons are completed to test the performance improvement as well as the effects of the GPU core number. The proposed approach is proved to be effective for large-scale FEM-ABM coupling simulations of restenosis and can be used to build real-scale virtual arteries for long-term restenosis prediction.