中国中西医结合影像学杂志2025，Vol.23Issue(4)：420-424,5.DOI:10.3969/j.issn.1672-0512.2025.04.004
术前MRI影像组学预测模型在非增强成人弥漫性胶质瘤分级诊断中的应用
Application of MRI radiomics predictive model in grading non-enhancing adult diffuse gliomas
摘要
Abstract
Objective:To evaluate the value of preoperative MRI radiomics models in grading non-enhancing adult diffuse gliomas.Methods:A retrospective analysis of 153 patients(160 lesions)with adult diffuse gliomas was constructed.All lesions had no enhancement on T1WI contrast-enhancement(T1WI-CE)image.Among the 160 lesions,103 lesions were classified into the low-grade glioma group and 57 lesions in the high-grade glioma group,then randomly divided into the training set(128 lesions)and the validation set(32 lesions)at a ratio of 8∶2.T2WI and ADC images were co-registered with T1WI-CE images,and ROIs were delineated on T2WI iamges.A total of 1 132 features were extracted form T2WI,T1WI-CE and ADC images,comprising shape,first-order histogram,second-order texture and wavelet features.The features with stability(ICC≥0.85)were retained,and the optimal feature subset was selected via elastic network regression(ENR)combined with recursive feature elimination(RFE).Six machine learning classifiers were adopted to construct the model,including logistic regression,support vector machine(radial basis function kernel,linear kernel),K-nearest neighbor(KNN),decision tree,Naive Bayes.The predictive performance of the models was evaluated using the AUC,accuracy,sensitivity,specificity and F1-score.Results:Radiomics models effectively predicted high-grade glioma.The multiparametric model(T2WI+T1WI-CE+ADC)with KNN classifier achieved AUCs of 0.892 and 0.805,accuracies of 0.805 and 0.719,sensitivities of 0.915 and 0.800,specificities of 0.753 and 0.727,F1-scores of 0.782 and 0.506 in the training set and the validation set.Conclusion:Preoperative MRI radiomics models enable accurate grading of non-enhancing adult diffuse gliomas,offering critical support for clinical decision-making and patient management.关键词
弥漫性胶质瘤/高级别胶质瘤/磁共振成像/影像组学/机器学习Key words
Diffuse glioma/High-grade glioma/Magnetic resonance imaging/Radiomics/Machine learning引用本文复制引用
刘彦宏,梁瑜晗,陈泽龙,王明霄,孙嘉翊,张雪梦,刘嘉霖,王玉林..术前MRI影像组学预测模型在非增强成人弥漫性胶质瘤分级诊断中的应用[J].中国中西医结合影像学杂志,2025,23(4):420-424,5.基金项目
国家自然科学基金项目(62136004). （62136004）
