Predictive model of stroke after transient ischemic attack based on multimodal magnetic resonance imaging

GE Yujie; ZHANG Xianfeng; MAO Xiaohua; YAN Xue; TANG Yanliang; LI Zhiping

doi:10.20258/j.cnki.1006-9011.2025.03.002

Abstract：Objective To establish a predictive model of stroke risk after transient ischemic attack(TIA)based on multi-modal magnetic resonance imaging(MRI)to reduce the risk of stroke progression after TIA.Methods A total of 103 patients who experienced TIA for the first time were prospectively collected and followed up for 90 days(the follow-up was terminated if stroke occurred during the period).Baseline MRI features and clinical data of patients with TIA were analyzed.Univariate and multivariate logistic regression analysis was used to determine the independent risk factors of stroke after TIA,and a nomogram model was established.Receiver operating characteristic(ROC)curve was used to evaluate the diagnostic efficiency.Results The incidence of stroke after TIA was 10.7%,which was divided into 11 cases in the post-TIA stroke group and 92 cases in the non-stroke group.There were statistically significant differences in age,periventricular high white matter signal(PWMH),deep white matter high signal(DWMH)and cerebral arteriosclerosis between the two groups(P<0.05).Multivariate analysis showed that PWMH and cerebral arteriosclerosis were independent risk factors for stroke after TIA(OR=5.696,3.183,P=0.001,0.049).PWMH and cerebral arteriosclerosis were included to establish a histogram model for stroke after TIA.The area under the curve(AUC)of the predictive model for diagnosing stroke after TIA was 0.880,which was higher than the AUC of PWMH,DWMH,cerebral arteriosclerosis and age alone(0.838,0.797,0.787 0.733).Conclusion A nomogram model based on multi-modal MRI features at baseline after TIA may provide a potential,non-invasive way to predict the risk of stroke after TIA.