Dynamic Prediction of Mechanical Thrombectomy Outcome for Acute Ischemic Stroke Patients Using Machine Learning

• DOI: 10.3390/brainsci12070938
• Publication Date: 2022-07-18T15:32:23Z
• Authors: Yixing Hu; Tongtong Yang; Juan Zhang; Xixi Wang; Xiaoli Cui; Nihong Chen; Junshan Zhou; Fuping Jiang; Junrong Zhu; Jianjun Zou
• Keywords: Brier score; Stroke

The unfavorable outcome of acute ischemic stroke (AIS) with large vessel occlusion (LVO) is related to clinical factors at multiple time points. However, predictive models used for dynamically predicting outcomes using clinically relevant preoperative and postoperative point variables have not been developed. Our goal was develop a machine learning (ML) model the dynamic prediction outcomes. We retrospectively reviewed patients AIS who underwent consecutive mechanical thrombectomy (MT) from three centers in China between January 2014 December 2018. Based on eXtreme gradient boosting (XGBoost) algorithm, we characteristics admission ("Admission" Model) additional regarding intraoperative management National Institute Health scale (NIHSS) score ("24-Hour" Model, "3-Day" Model "Discharge" Model). an three-month mark (modified Rankin scale, mRS 3-6: unfavorable). area under receiver operating characteristic curve Brier scores were main evaluating indexes. observed 156 (62.0%) 238 patients. These four had high accuracy range 75.0% 87.5% good discrimination AUC 0.824 0.945 testing set. ranged 0.122 0.083 showed ability This first dynamic, constructed MT, which more accurate than previous model. could be predict before MT support decision perform would further improve after timely adjust therapeutic strategies.