Machine Learning analysis of functional outcome and hemodynamic parameters during stroke thrombectomy under general anesthesia

Abstract

Background: Intraoperative hemodynamic management may influence cerebral perfusion and neurological recovery in patients undergoing mechanical thrombectomy (MT) for acute ischemic stroke (AIS). This study aimed to identify intraoperative blood pressure and heart rate patterns associated with three-month functional outcomes in AIS patients treated under general anesthesia, using a machine learning approach capable of capturing nonlinear relationships and variable interactions.

Methods: We conducted a post hoc analysis of a prospectively maintained cohort of consecutive patients with anterior circulation stroke who underwent MT under general anesthesia at a tertiary stroke center between March 2014 and June 2019. The primary outcome was functional independence at three months (modified Rankin Scale [mRS] score 0-2). Intraoperative hemodynamic variables were analyzed across pre- and post-reperfusion phases. An eXtreme Gradient Boosting (XGBoost) model was trained to predict outcomes, and SHapley Additive exPlanations (SHAP) were used to identify the most influential predictors.

Results: Among the 229 patients included, 101 (44.1%) achieved an mRS score of ≤2. Increased time spent with mean arterial pressure <80 mmHg and systolic blood pressure <140 mmHg before reperfusion was associated with unfavorable outcomes. Elevated intraoperative heart rate in the same phase also emerged as a strong marker of poor prognosis. The model demonstrated good predictive performance (AUC ROC = 0.85) with internal validation.

Conclusions: These findings highlight distinct intraoperative hemodynamic patterns associated with outcome after thrombectomy under general anesthesia and support further investigation into heart rate as a relevant physiological indicator during anesthetic care.

Keywords: Functional Outcome; General Anesthesia; Hemodynamic Management; Machine Learning; Thrombectomy.