Optimizing flow-diverting stent configurations for aneurysm treatment: a computational approach integrating deep learning and differential evolution optimization.

An aneurysm, enlargement of an artery or vein, weakens the surrounding vascular wall, making it susceptible to rupture and the possibility of life-threatening bleeding, ultimately resulting in death. The placement of flow-diverting stents is a highly utilized and effective method for treating aneurysms. This study presents a novel approach combining CFD simulations, deep neural networks (DNN), and differential evolution optimization (DEO) to optimize hemodynamic conditions in aneurysms. Initially, CFD simulations were conducted to generate a comprehensive dataset of 2,700 simulations with various stent configurations. This dataset was then used to train a DNN model, enabling accurate predictions of velocity, vorticity, and wall shear stress for any stent configuration. The model demonstrated consistent and reliable performance across different configurations. DEO was applied to identify the optimal stent, resulting in a configuration with seven struts. The optimal strut sizes were 0.3184, 0.9599, 0.7889, 0.9599, 1.0073, 1.0073, and 2.9283, with gap sizes of 0.2238, 0.5897, 0.3379, 0.2996, 0.2052, 0.0371, and 0.3068 between the struts. This configuration achieved superior performance in reducing velocity, vorticity, and maximum wall shear stress. The study demonstrated that increasing the number of struts, with a concentration at the proximal aneurysm neck, enhanced flow diversion and minimized hemodynamic risks, especially in regions vulnerable to rupture. Validation through additional CFD simulations confirmed the effectiveness of the optimized stent, demonstrating the potential of the proposed methodology to improve stent design and hemodynamic outcomes in aneurysm treatment.