The study of DWI-based machine learning model for predicting the prognosis in patients with acute ischemic stroke undergoing mechanical embolectomy

WANG Youshan; DIAO Siyi

doi:10.20258/j.cnki.1006-9011.2025.08.002

Abstract：Objective To explore the prognostic value of machine learning model based on diffusion weighted imaging(DWI)in patients with acute ischemic stroke(AIS)undergoing mechanical embolectomy.Methods The 128 newly diagnosed AIS patients who underwent mechanical embolectomy in our hospital were retrospectively collected,and divided into two groups according to the mRS score at 3 months after mechanical embolectomy.The 72 patients with mRS scores of 0-2 were included into good prognosis group,and 56 patients with mRS score of 3-5 were included into poor prognosis group.The DWI images were imported into 3D-Slicer software,and the ROIs were delineated by two radiologists along the acute stroke area with high DWI sig-nal.The 830 radiomics features were extracted.The intraclass correlation coeficient test,minimum redundancy maximum correla-tion test,and logistic regression analysis were used to select radiomics features.Five classifiers including logistic regression,support vector machine,naive Bayes,decision tree,and random forest were used to construct prognostic models.The receiver operating characteristic(ROC)curve and calibration curve were used to analyze the diagnostic and calibration efficiency of the model.Results In present study,eight imaging features were selected.The logistic regression and support vector machine model showed optimal prognostic value.The area under the ROC curve was 0.92 and 0.84 in the training group,and 0.91 and 0.83 in the verification group,respectively.Conclusion DWI-based machine learning model can be used to predict the early prognosis of AIS patients undergoing mechanical embolectomy,and logistic regression and support vector machine model shows optimal prediction efficiency.