Application of machine learning algorithms to predict and assess factors related to internal carotid siphon aneurysm occlusion treated with flow diversion

PURPOSE: Numerous clinical indicators have been identified as risk factors for aneurysm occlusion and used to development of various Flow Diversion Predictive Score grading scales. The objective of this study is to predict and assess factors related to Internal carotid siphon aneurysm occlusion, by leveraging Machine Learning(ML) Algorithms. METHODS: We conducted a retrospective analysis of patients treated with flow diversion from January 2020 to December 2023 in our department. The samples were randomly divided into derivation and validation groups in a 70/30 ratio. We employed six ML algorithms, including LR, DT, SVM, RF, GBM, and XGBoost, these algorithms were utilized alongside preoperative and intraoperative clinicopathological characteristics to develop predictive models. RESULTS: A total of 297 patients with 342 target aneurysms were included. Multivariate logistic regression analyses identified aneurysm orientation, W, BNF, IA, incorporated branch vessels, and adjunct coil deployment as independent predictors. Additionally, a random forest feature selection analysis was conducted to identify potentially significant factors based on importance scores, IA performed the most outstandingly. Combining with previously reported literature, we ultimately determined the variables for inclusion in six ML models: IA, BNF, incorporated branch vessels, adjunct coil deployment, ostium area, CND, NR, and Dmax. Among these models, the GBM showed superior performance with an AUROC of 0.766. CONCLUSION: By integrating preoperative and intraoperative factors, ML algorithms can achieve acceptable predictions. If widely implemented, this approach could serve as a valuable reference for selecting surgical methods for specific aneurysms prior to operation in clinical practice.