多参数MRI影像组学预测IDH野生型胶质母细胞瘤的MGMT启动子甲基化表达  

Prediction of the Expression of MGMT Promoter Methylation in IDH Wild-Type Glioblastoma via Multiparametric MRI Radiomics

作　　者：张馨月 刘国莉 张君[3] 马林 ZHANG Xinyue;LIU Guoli;ZHANG Jun;MA Lin(Chinese PLA Medical School,Beijing 100853,China;Department of Radiology,the First Center of Chinese PLA General Hospital,Beijing 100853,China;不详)

机构地区：[1]解放军医学院,北京100853 [2]解放军总医院第一医学中心放射诊断科,北京100853 [3]解放军总医院第六医学中心核医学科,北京100037

出　　处：《中国医学影像学杂志》2023年第5期433-441,共9页Chinese Journal of Medical Imaging

摘　　要：目的 结合多参数MRI影像组学与机器学习,无创预测异柠檬酸脱氢酶(IDH)野生型胶质母细胞瘤的O6-甲基鸟嘌呤-DNA甲基转移酶(MGMT)启动子甲基化表达。资料与方法 回顾性收集2016年10月—2022年8月解放军总医院第一医学中心112例IDH野生型胶质母细胞瘤的术前多参数MRI数据及患者年龄、性别,按7∶3随机分为训练集80例与验证集32例。选取表观扩散系数、增强T1WI及T2WI序列进行感兴趣区分割,并从每个感兴趣区中提取1 013个影像特征。采用组内相关系数(ICC)进行一致性检验并过滤ICC<0.85的影像特征。依次通过ICC、t检验及最小绝对收缩和选择算子筛选出最佳特征。在7种分类器中对3个单序列特征组与1个多序列特征组进行训练及验证。绘制受试者工作特征曲线评估模型的诊断效能。结果随机森林分类器与多序列特征组的组合模型为预测IDH野生型胶质母细胞瘤MGMT启动子甲基化表达的最优模型,训练集与验证集中的曲线下面积分别为0.933与0.915,准确度分别为0.868、0.822。结论 无创提取多参数MRI特征结合机器学习算法可以有效预测IDH野生型胶质母细胞瘤MGMT启动子甲基化的表达,为制订个体化治疗方案提供支持。Purpose To noninvasively predict the expression of O6-methylguanine-DNA methyltransferase promoter(pMGMT)methylation status in isocitrate dehydrogenase(IDH)wild-type glioblastoma by combining multiparametric MRI radiomics and machine learning algorithms.Materials and Methods The preoperative imaging and clinical data of 112 IDH wild-type glioblastoma patients in the First Center of Chinese PLA General Hospital from October 2016 to August 2022 were retrospectively collected,and all patients were randomly divided into a training set and a validation set according to 7∶3.The apparent diffusion coefficient(ADC),T1-weighted contrast-enhanced(T1CE),and T2WI sequences were selected for tumor segmentation,and 1013 radiomics features were extracted from each region of interest.The intraclass correlation coefficient(ICC)was used for the consistency test,and the image features with ICC<0.85 were filtered.ICC,student's t test,and least absolute shrinkage and selection operator were used to screen the best radiomics features.Seven classifiers were used to train and verify three single sequence feature sets and one multi-sequence set.The receiver operating characteristic curve was plotted to evaluate the diagnostic efficacy of the model.Results The combination model of the Random Forest classifier and multi-sequence feature set was the best model for predicting the expression of pMGMT methylation status.The area under the curve of the training set and the validation set were 0.933 and 0.915,and the accuracy were 0.868 and 0.822,respectively.Conclusion Noninvasive extraction of multiparametricMRI features combined with machine learning algorithms can effectively predict the expression of pMGMT methylation status in IDH wild-type glioblastoma,which provides support for the development of individualized treatment plans for patients.

关 键 词：胶质母细胞瘤 磁共振成像 O6-甲基鸟嘌呤-DNA甲基转移酶 异柠檬酸脱氢酶 影像组学 机器学习 

分 类 号：R739.41[医药卫生—肿瘤] R445.2[医药卫生—临床医学]