Establishment the prediction model of cerebral infarction based on the ultrasound radiomics features of ca-rotid plaque and clinical features
Objective This study aimed to explore the application value of clinical model based on clinical features,ra-diomics model based on gray-scale ultrasound characteristics of carotid artery plaques,and the combined model to predict cerebral in-farction.Methods Carotid plaque cases were independently collected by two sonographers as a training set and a validation set.The patients were divided into cerebral infarction group and non-cerebral infarction group according to head CT or MR examination for ipsilateral cerebral infarction lesions.In the training set,a clinical model,a radiomics model and a joint clinics-radiomics model were constructed to predict cerebral infarction based on clinical features and/or plaque grayscale ultrasonic features.In the training set and val-idation set,receiver operating characteristic(ROC)curve and cali-bration curve were used to evaluate the prediction efficiency of the model.And the area under the curve(AUC),sensitivity,specificity and accuracy were calculated.Decision curve was used to evaluate the clinical application value of the model.Results There was no significant difference in clinical characteristics between the train-ing set and the validation set(P>0.05).A total of 5 key clinical features,6 key radiomics features and 9 clinical and radiomics features were selected in the training set to build the clinical model,the radiomics model,and the joint models,respectively.In the training set,the AUC of the clinical model was 0.724,the accuracy was 66.42%.The AUC of the radiomics model was 0.886,the accuracy was 80.60%.The AUC of the joint model was 0.908,the accuracy was 83.58%.The calibration curve and decision curve of the joint model were superior to the radiomics model and clinical model.The validation set also showed that the predictive performance of the ra-diomics model and the joint model reached a higher level.The joint model had the highest predictive efficacy and the net benefit value of the decision curve,followed by the radiomics model and the clinical model.Conclusion Both the radiomics model based on gray-scale ultrasound and the joint model based on clinics-radiomics can be used as the predictive models of carotid plaque stability,and the joint model has higher predictive efficiency and benefits.
NETL
NSTL
万方数据
