Prediction of Recurrence and Rupture Risk of Ruptured and Unruptured Intracranial Aneurysms of the Posterior...

Background: Intracranial aneurysms of the posterior circulation are of particular clinical significance due to their higher risk of rupture-associated morbidity and mortality compared to anterior circulation aneurysms. Moreover, they exhibit an increased tendency for recurrence, posing challenges for long-term management. The purpose of this study is to identify key risk factors and define criteria for the early detection of high-risk aneurysms with a machine learning-based analysis. Methods: This study employs machine learning (ML), which, unlike traditional statistical methods, can detect complex, previously unrecognized patterns without predefined hypotheses to predict recurrence and rupture in patients with intracranial aneurysms of the posterior circulation. A total of 229 patients were retrospectively screened (2008-2020), and the data set was analyzed using ML algorithms. To avoid bias, a 10-fold cross-validation was employed, and the model performing best in terms of the Area Under the Curve (AUC) was selected. In addition, the sensitivity, specificity, and accuracy of the model were computed as secondary metrics. Results: A total of 229 patients were included, with over 70% being female, older than 50 years, and diagnosed with arterial hypertension. The most significant predictors of aneurysm recurrence identified by the ML model (AUC of 0.74 with a sensitivity of 0.76, a specificity of 0.70, and an accuracy of 0.76) were age, aneurysm size, arterial hypertension, and a history of nicotine consumption. The DeLong test confirmed that the ML model performed significantly better than random classification with an AUC of 0.5 (p < 0.001). Further analysis revealed that the presence of multiple aneurysms and localization at the basilar artery were independent risk factors for early recurrence within six months. For aneurysm rupture, key predictive features included advanced age, basilar artery localization, atherosclerosis, irregular aneurysm morphology, and familial predisposition. Conclusions: ML algorithms identified several risk factors for recurrence and rupture of intracranial aneurysms of the posterior circulation, aligning with previously established risk factors. These findings are intended to serve as a basis for further research in clinical use and prospective studies.