摘要
目的探讨基于术前T2-FLAIR图像的影像组学模型预测弥漫性较低级别胶质瘤1p/19q缺失状态的价值。资料与方法回顾性分析经术后病理证实、具有完整的术前头颅MRI图像的弥漫性较低级别胶质瘤(WHOⅡ~Ⅲ级)98例,其中1p/19q联合缺失45例(WHOⅡ级31例,WHOⅢ级14例),1p/19q非联合缺失53例(WHOⅡ级27例,WHOⅢ级26例),以7∶3随机分为训练组和验证组。使用3DSlicer软件在T2-FLAIR图像上对病灶所有轴向层面手动勾画3D-ROI,包含瘤体及周围水肿区。使用3DSlicer软件内Radiomics模块提取纹理特征,使用t检验或Welch t检验、LASSO进行特征筛选和降维,得到对预测1p/19q缺失状态最有价值的特征参数。建立支持向量机(SVM)预测模型,并采用受试者工作特征(ROC)曲线对模型进行效能评价。结果最后筛选出8个影像组学特征参数,预测模型在训练组中ROC曲线下面积(AUC)为0.839,准确率为88.2%,敏感度为81.5%,特异度为89.2%;在验证组中AUC为0.820,准确率为83.3%,敏感度为77.2%,特异度为89.5%。结论基于术前T2-FLAIR图像建立的影像组学模型可以有效地预测弥漫性较低级别胶质瘤1p/19q缺失状态。
Purpose To explore the value of a radiomics model based on preoperative T2-FLAIR images in predicting the 1p/19q deficiency status in diffuse lower grade gliomas.Materials and Methods A total of 98 patients with diffuse lower grade glioma(WHO gradeⅡ-Ⅲ)confirmed by pathology were retrospectively included and analyzed and all patients underwent brain MRI examination before operation.Forty-five cases with 1p/19q codeletion(31 cases with WHO gradeⅡ,14 cases with WHO gradeⅢ)and 53 cases with 1p/19q non-codeletion(27 cases with WHO gradeⅡ,26 cases with WHO gradeⅢ)were enrolled in this study and these cases were randomly divided into a training dataset and a validation dataset at the ratio of 7∶3.All of the lesions,including tumor and peripheral edema area,on the T2-FLAIR axial images,were manually delineated slice-by-slice via 3D region-of-interest(ROI).Texture features were extracted using radiomics module in the 3DSlicer software.The feature screening and dimension reduction were performed to obtain the most valuable feature parameters for predicting the 1p/19q status via t test or Welch's t test and LASSO.A support vector machine(SVM)prediction model was constructed and receiver operating characteristic(ROC)curve analysis was performed to evaluate the efficiency performance evaluation of the prediction model.Results A total of eight radiomics feature parameters were finally selected,showing good performances in the training dataset area under the curve(AUC)was 0.839,accuracy rate was 88.2%,sensitivity was 81.5%,specificity was 89.2%;in the validation dataset,AUC was 0.820,accuracy rate was 83.3%,sensitivity was 77.2%,specificity was 89.5%.Conclusion The 1p/19q deficiency status in diffuse lower grade gliomas could be effectively predicted via radiomics model based on the preoperative T2-FLAIR images.
作者
樊建坤
程勇
王腾
胡龙翔凤
唐光才
FAN Jiankun;CHENG Yong;WANG Teng;HU Longxiangfeng;TANG Guangcai(Department of Radiology,the Affiliated Hospital of Southwest Medical University,Luzhou 646000,China)
出处
《中国医学影像学杂志》
CSCD
北大核心
2021年第5期425-429,共5页
Chinese Journal of Medical Imaging
作者简介
通信作者:唐光才.168345315@qq.com。
