Objective To develop a machine learning-based risk prediction model for patients undergoing venous thrombolysis for acute ischemic stroke.Methods A total of 666 patients who received venous thrombolysis for acute ischemic stroke from July 2017 to July 2023 at the Huzhou First People's Hospital were identified.The patients were categorized into two groups according to their postoperative modified Rankin scale:415 patients were in the good prognosis group,and 251 patients were in the poor prognosis group.Data on potential influencing prognostic factors were collected for both groups.Risk factors were identified using group comparisons and LASSO regression analysis.Subsequently,the participants were randomly assigned to either a training set or a test set,maintaining a ratio of 7∶3.Six different machine learning algorithms were employed to construct predictive models within the training set,which were subsequently validated on the test set.The performance of the models was assessed using metrics such as accuracy,the area under the curve(AUC),sensitivity,and specificity.Results Through univariate analysis and LASSO regression analysis,twelve variables were identified as potential predictors for patient outcomes following venous thrombolysis for acute ischemic stroke.These variables include age,body mass index (BMI),smoking,hyp ertension,atrial fibrillation,coronary heart disease,preoperative systolic blood pressure,major arterial occlusion,admission National Center of Health Stroke Scale (NIHSS) score,hemoglobin,INR,and D-dimer.Upon further examination through multivariate logistic regression analysis,hemoglobin,coronary heart disease,and major arterial occlusion were determined to be independent risk factors associated with poor prognosis in these patients.In terms of predictive modeling,the Support Vector Machine model demonstrated the highest overall performance among the machine learning algorithms evaluated,with an AUC value of 0.76.Conclusion Among the risk prediction models developed for acute ischemic stroke venous thrombolysis using machine learning algorithms,the Support Vector Machine model exhibited the most effective predictive performance.This suggests that the SVM model could potentially offer valuable support for clinical decision-making process in the management of patients with acute ischemic stroke.
万方数据
维普
