An AI-enhanced SPH Framework for Whole-Process Dynamic Simulation of Cardiovascular Diseases
Prof Moubin Liu
Peking University
Abstract: Cardiovascular diseases (CVDs) remain the leading cause of death globally, with nearly 20 million deaths annually, and represent a major health burden affecting all populations. Among them, acute events caused by vascular rupture and thrombosis are the most fatal, accounting for the majority of CVD-related mortality. Despite advances in imaging and computational CVD diagnostics, most studies remain limited to the intact vessel state and cross-sectional risk assessments, lacking longitudinal resolution and predictive capacity for rupture and thrombotic evolution. Consequently, key dynamic mechanisms such as fatigue and flow pulsatility remain insufficiently resolved. Moreover, conventional cardiovascular simulations are often computationally intensive, limiting their applicability in time-sensitive clinical decision-making. To address these challenges, we present an AI-enhanced Smoothed Particle Hydrodynamics (SPH) framework for whole-process dynamic simulation of CVDs. Firstly, we develop a robust and accurate SPH fluid-structure interaction (FSI) solver that captures nonlinear hemodynamics, large wall deformations, and rapidly evolving interfaces. Physiological fidelity is further enhanced by incorporating a novel fiber orientation method and biomechanical details. Crucially, we extend the solver with phase-field fracture and coagulation cascade models, yielding a unified fluid-structure-fracture-thrombosis coupling algorithm that, for the first time, achieves dynamic prediction from vessel failure to thrombus formation. To reduce computational cost, a 0D Windkessel model and a 2D solid-shell SPH formulation are tightly coupled with the 3D solver. Additionally, leveraging the SPH simulation datasets, a highly efficient surrogate model based on Fast Function Extraction is constructed for fast or nearly real-time predictions. The framework has been rigorously validated and successfully applied to patient-specific cases of aortic dissection, cerebral aneurysm, and acute coronary syndrome, achieving clinically consistent predictions while reducing time cost from several days to seconds. It represents the first approach capable of dynamically predicting the full progression of CVDs, offering a powerful tool for longitudinal study, precision prevention, and decision support.
Bio: Moubin Liu is a Tenured Full Professor, the Vice Dean of the School of Mechanics and Engineering Science, and the Director of the Center of Industrial Software Research, Peking University. His research interests including computational mechanics, fluid-structure interaction, multiphysics modeling and AI-enhanced engineering intelligence. He authored two popular monographs, including the first-ever book on the Smoothed Particle Hydrodynamics. He published over 180 SCI indexed papers with 10 ESI highly cited papers, and more than 16800 Google citations. He has received a number of awards from universities and scientific organizations worldwide including the International Computational Award (2019), ICACM Computational Mechanics Award (2018), First Prize in Natural Sciences from the Ministry of Education (2017), the 100 Talent Program Award from CAS (2009), and the prestigious Lee Kuan Yew Fellowship Award (2005). He is the associate editor of EABE and IJCM and the editorial Board Member of four other international journals, and has been consecutively listed by Elsevier as a “Most Cited Chinese Researcher” in Computational Mechanics from 2015 and the World's Top 2% Scientists (both career and single year) by Stanford.
