Automated Machine Learning for Predicting Perioperative Ischemia Stroke in Endovascularly Treated Ruptured Intracranial Aneurysm Patients

Yuhang peng¹Ke Bi²Xiaolin Zhang¹Ning Huang¹Xiang Ji¹Weifu Chen¹Ying Ma¹Yue Jianhe¹Jianhe Yue^1*Yongxiang Jiang^1*

• ¹Department of Neurosurgery, Second Affiliated Hospital, Chongqing Medical University, Chongqing, China
• ²Department of Emergency Medicine, Second Affiliated Hospital, Chongqing Medical University, Chongqing, China

Objective: This study constructs and validates an automated machine learning to predict perioperative ischemic stroke (PIS) risk in endovascularly treated patients with ruptured intracranial aneurysms (RIAs), establishing a clinical decision-support tool for PIS. Methods: This retrospective cohort study analyzed RIAs patients undergoing endovascular treatment at our neurosurgical center (December 2013 -February 2024). The least absolute shrinkage and selection operator (LASSO) method was used to screen essential features associated with PIS. Evaluating feature-related variables, we built nine machine learning models on the training set (75% of participants) and assessed them on the test set (25% of participants). Through comparative analysis, employing metrics such as Area Under the Receiver Operating Characteristic Curve (ROCAUC) and Brier score, we identified the optimal model-random forest (RF)-for predicting PIS. To interpret the RF models, we utilized the Shapley Additive package (SHAP) explanation. Results: The final cohort comprised 647 consecutive RIAs patients undergoing endovascular intervention, with LASSO regression identifying 13 clinically actionable predictors of PIS from initial variables. These predictors encompassed: vascular risk factors (hyperlipidemia, arteriosclerosis); neuroimaging severity (modified Fisher scale, aneurysm location, neck-to-diameter ratio); clinical status (Glasgow Coma Scale, Hunt-Hess grade, age, sex); procedural complications (intraprocedural rupture, periprocedural re-rupture); therapeutic determinants (therapy method, ischemic comorbidity history). Nine machine learning architectures were evaluated through stratified ten-fold cross-validation, among these models, the RF model demonstrated the best performance, with the ROCAUC of 92.11% (95%CI: 89.74%-94.48%) on the test set and 87.08% (95%CI: 81.23%-92.93%) on the training set. Finally, the RF predictive model demonstrated an accuracy of 88.23% in forecasting the incidence of PIS in a prospective cohort. Additionally, based on this predictive model, this study developed a highly convenient web calculator. Clinicians only need to input the patient' s key factors into this calculator to predict the postoperative incidence of PIS and provide individualized treatment plans for the patient. Conclusion: We successfully developed and validated an interpretable machine learning framework with an integrated clinical decision-support system for predicting postprocedural PIS in endovascularly treated RIAs patients. These tools effectively predicted the likelihood of PIS, enabling high-risk patients to promptly take specific preventive and therapeutic measures.