Integrating Proper Orthogonal Decomposition and Transformers for Flow Predictions inside cerebral Aneurysm in various body physiologies: A Computational-Based Approach

Abstract

Aneurysm rupture remains a critical challenge in medical diagnostics, and understanding the underlying hemodynamic factors is essential for assessing rupture risk. This study combines computational fluid dynamics (CFD) simulations and machine learning models to analyze the hemodynamics of aneurysms in real patient-specific models. The focus was on key hemodynamic parameters such as Wall Shear Stress (WSS), Oscillatory Shear Index (OSI), and Pressure, which play a crucial role in predicting aneurysm rupture. High-fidelity CFD simulations were conducted for aneurysm models under both rest and exercise conditions, revealing detailed flow patterns and localized variations in shear stress. A machine learning approach, combining Proper Orthogonal Decomposition (POD) and Transformer models, was employed to predict these hemodynamic parameters efficiently, offering a reduced-order model for real-time applications. The results showed that while the machine learning predictions accurately captured the overall trends in OSI, WSS, and Pressure, some local discrepancies were observed, particularly in regions of high shear stress. These findings highlight the potential of integrating CFD with machine learning for improved aneurysm risk assessment and management, though further refinement is necessary to enhance prediction accuracy, particularly in complex regions of interest.