Development of an Optimal Flow Diverter Stent Prediction Model Based on Parent Artery Morphology Analysis

Intracranial aneurysms (IA) are weak spots in brain arteries that can bulge and rupture, causing life-threatening bleeding. Although Flow Diverter Stents (FDS) are increasingly used to treat these aneurysms, selecting the optimal device remains challenging due to complex vascular geometries and reliance on surgeon expertise. In this study, we developed a machine learning decision support system to predict the appropriate FDS size and length by integrating clinical data with detailed measurements of aneurysms and their parent arteries. We analyzed data from 94 internal carotid artery aneurysm cases treated between October 2016 and April 2024, using 61 features (7 clinical parameters and 54 vessel diameter measurements). Six regression algorithms were compared using Bayesian hyperparameter optimization and five-fold cross validation. The best models achieved a size prediction accuracy of 94.7% (18 out of 19 cases) and a length prediction accuracy of 78.9% (15 out of 19 cases), with the latter improving to 89% through SHAP-based feature selection. These findings suggest that our machine learning system can support FDS selection, potentially enhancing treatment planning and patient outcomes.