This postdoctoral project is part of the ERC CURE project, which aims to develop fast patient-specific computational models to assist the treatment of cerebral aneurysms. A central component of this effort is the development of robust and efficient meshing technologies capable of handling highly complex vascular geometries and embedded medical devices. In a context where computational pipelines are expected to deliver results within minutes rather than hours, revisiting and accelerating our meshing technology represents a major step toward real-time or near-real-time hemodynamic simulation.
The objective of this project is to redesign and accelerate the MTC meshing framework, by enabling GPU-based execution and hybrid CPU–GPU workflows. The research will focus on three main aspects.
- GPU Acceleration of the MTC Mesh Generator. MTC is a simplex-based mesh generator and remesher (triangles, tetrahedra, and higher-dimensional simplices) widely used in mesh generation, mesh adaptation, and anisotropic remeshing procedures in both academic libraries and industrial simulation codes. The parallel version of MTC exists under the MPI paradigm and the first objective of the project is to develop a version of MTC compatible with GPUs.
- Anisotropic Mesh Adaptation and Parallel Mesh Operations. MTC can be coupled with metric fields to generate highly anisotropic meshes adapted to physical solution features. These metric fields may be defined analytically or derived from a posteriori error estimate. The project will explore GPU-enabled implementations of anisotropic mesh adaptation techniques, including dynamic remeshing and field interpolation within the CimLib_CFD parallel computing framework, which provides dynamic mesh partitioning and transparent data redistribution for large-scale simulations.
- Application to Patient-Specific Stent Modeling. The final objective of the project is the application of these technologies to the simulation of intracranial aneurysm treatment, within the ERC CURE framework. Patient-specific vascular geometries combined with detailed stent wire structures lead to extremely challenging meshing problems due to: large differences in spatial scales, highly complex braided wire geometries and tight geometric constraints near the arterial wall.
