Particle-Based Simulation & Machine Learning of Thrombosis Models in Brain Circulation

A cerebral aneurysm is a life-threatening vascular distension. A minimally invasive treatment for
selected aneurysms involves catheter insertion of an endovascular flow diverter that redirects blood flow away from the aneurysm. This triggers thrombus formation pathways and causes the aneurysm to fill with a structured fibrous thrombus, eventually leading to the complete elimination of the aneurysm. However, the role of certain types of thrombus within large aneurysms is also thought to be related to delayed ruptures. The project will focus on:(i) particle-based methods (LBM) that provide a biologically motivated way to track platelet activation under shear stresses (ii) acceleration of the computations by modelling the biochemical reactions with physics-informed neural networks coupled with the particle-resolved flow field. Recent approaches will be explored using level-set functions to define the computational region of interest and solve the flow problem using the immersed boundary method. The particle-based models will be validated against reference solutions from traditional CFD solvers and clinical flow imaging data provided by Prof Tufail Patankar. To predict the stable clot configuration, a set of biochemical transport-reaction equations will be defined and solved in a comparative study both with particle-based methods and physics-informed neural networks.