Objective:To explore the value of the model based on multi-sequence MRI radiomics features combined with visually accessible Rembrandt images(VASARI)features in predicting the isocitrate dehydrogenase 1(IDH1)mutation status of gliomas.Methods:The clinical and pathological information and preoperative MRI data of 452 gliomas patients from two institutions were retrospec-tively analyzed.The patients were randomly divided into training set(n=271)or validation set(n=181)at the ratio of 3:2.Twenty-two VASARI features were extracted and analyzed.The independent predictors for IDH1 status were selected using univariate and multivariate logistic regression(LR),and VASARI model was constructed.The optimal radiomics features based on T2WI,T2 fluid attenua-tion inversion recovery(T2-FLAIR)and contrast enhanced T1WI(CE-T1WI)were extracted and screened,and the radiomics score(Rad-score)was calculated.The radiomics model was constructed by classifier eXtreme Gradient Boosting(XGBoost).Filtered VASARI features and Rad-score were incor-porated into the multivariate LR to construct a combined model.The efficacy of models was evaluated and compared through receiver operating characteristic(ROC)curves and DeLong test,while their clinical utility and calibration were evaluated by decision curve analysis(DCA)and calibration curves.Results:The F1,F4,F7,and F11 in the VASARI feature set were independent predictors of the IDH1 mutation status.Totally 11 optimal radiomics features were screened and the Rad-score was obtained.Then,the radiomics model was constructed.The AUC of combined model were higher than that of VASARI model and radiomics model in both training set and validation set(training set:0.952 vs.0.872,0.882;validation set:0.938 vs.0.890,0.836),and the difference was statistically significant(De-long test,P＜0.05).There was no significant difference in AUC between radiomics model and VASA-RI model(P＞0.05).DCA showed that the combined model had the largest net benefit and the best clinical practicality within a certain hazard threshold.The calibration curve showed that the three mod-els were well calibrated,and the combined model had the best calibration among them.Conclusions:Multi-sequence MRI radiomics model and VASARI model can effectively predict the IDH1 mutation status in gliomas,and their combination can help to improve the diagnostic performance.
关键词
胶质瘤/异柠檬酸脱氢酶/突变状态/磁共振成像/影像组学/伦勃朗视觉感受图像
Key words
Gliomas/Isocitrate dehydrogenase/Mutation status/Magnetic resonance ima-ging/Radiomics/Visually accessible Rembrandt images
维普
万方数据