Radiomics models based on fluid attenuated inversion recovery and contrast enhanced MRI for noninvasive prediction of isocitrate dehydrogenase mutation status in glioma
10.3969/j.issn.1002-1671.2025.06.005
- VernacularTitle:基于液体衰减反转恢复和对比增强MRI的影像组学模型无创预测胶质瘤IDH突变状态
- Author:
Qian'ang MA
1
;
Jun LU
;
Qi YAO
;
Yafeng DONG
;
Xuejun CHEN
;
Jinrong QU
Author Information
1. 郑州大学附属肿瘤医院 河南省肿瘤医院放射科,河南 郑州 450000
- Publication Type:Journal Article
- Keywords:
glioma;
isocitrate dehydrogenase mutation;
magnetic resonance imaging;
radiomics;
nomogram
- From:
Journal of Practical Radiology
2025;41(6):915-919
- CountryChina
- Language:Chinese
-
Abstract:
Objective To investigate the value of MRI radiomics for the preoperative noninvasive prediction of isocitrate dehydrogenase(IDH)mutation status in glioma.Methods Totally,306 glioma patients were retrospectively selected.All patients were randomly assigned into training group(n=214)and validation group(n=92)at a ratio of 7∶3.Region of interest(ROI)was manually delineated by two radiologists independently on the fluid attenuated inversion recovery(FLAIR)and contrast enhanced(CE)MRI images for obtaining whole volume of interest(VOI)of lesion.A total of 851 radiomics features were extracted from the VOI,respectively.The least absolute shrinkage and selection operator(LASSO)method was used for features dimension reduction combing 10-fold cross validation.Three Radiomics score(Radscore)were calculated by linear combination of retained features and their corresponding coefficients.The optimal Radscore and clinical characteristics were incorporated to perform logistic regression analysis for establishing the IDH mutation status noninvasive prediction model.A nomogram was plotted for realizing the visualization of model.The receiver operating characteristic(ROC)curve was plotted to evaluate the prediction performance of model.The calibration and clinical utility of the model were evaluated by calibration curve and decision curve.Results The area under the curve(AUC)of Radscore-combined based on combination of two sequences was 0.856 in the training group,which was superior to the Radscore-CE(AUC=0.821),Radscore-FLAIR(AUC=0.766)from single sequence,with consistent result in the validation group.The addition of clinical characteristics to the model improved predictive value with AUC,sensitivity and specificity of 0.898,79.59％,90.52％in the training group.Conclusion The radiomics model based on FLAIR and CE MRI contributes to preoperative noninvasive prediction of IDH mutation status in glioma.The combination of multi-sequence and the addition of clinical characteristics can improve the prediction performance.
