Interpretable machine learning model for predicting rupture risk in anterior communicating artery aneurysms.

Early prediction of brain aneurysm rupture is critical, as it enables timely clinical interventions and avoid the significant morbidity and mortality associated with subarachnoid hemorrhage. Anterior communicating artery aneurysm (ACoA) is the most common site of ruptured intracranial aneurysms. In this study, we developed a machine learning (ML) pipeline that uses demographic, clinical data and morphological measurements to predict rupture risk of ACoA, providing a non-invasive and cost-effective decision-support approach based on documented clinical and imaging-derived morphologic variables. The dataset consisted of 170 patient records, each including 24 demographic, morphological and clinical variables. An XGBoost-based feature ranking algorithm was employed, followed by feature incrementation approach to identify the top eight features for model development. Among several evaluated classifiers, the Random Forest model demonstrated the highest performance, achieving an accuracy of 88.82%, precision of 88.96%, recall of 88.82%, specificity of 88.82%, F1-score of 88.81%, and an area under the curve (AUC) of 94.80%. SHapley Additive Explanations (SHAP) were used to interpret how each feature influenced the prediction, providing a clear picture of model’s decision-making rationale. Overall, the proposed ML pipeline showed promising internal performance, supporting further external and prospective validation for clinical decision support tool in the early identification and management of ACoA at risk of rupture.