T2-FLAIR影像组学对高级别脑胶质瘤术后1年复发的预测价值

Radiomics Analysis Based on T2-Fluid Attenuated Inversion Recovery for Evaluating the Recurrence within 1 Year After Surgery in High-Grade Glioma

目的:探讨T2-液体衰减反转恢复序列(T2-FLAIR)影像组学对高级别脑胶质瘤(HGG)术后1年复发的预测价值。方法:回顾性收集我院2019年1月—2021年12月收治的81例HGG(WHOⅢ~Ⅳ级)患者,所有患者均接受标准的肿瘤切除手术及术后辅助治疗。获取患者术后4周内的MRI T2-FLAIR图像,根据术后1年内的随访数据将患者分为未复发组(51例)和复发组(30例),将所有患者按照7∶3比例随机分配为训练组(57例)和验证组(24例)。采用3D-Slicer软件对残腔周围T2-FLAIR高信号区进行图像分割,并提取851个影像组学特征,包括一阶特征、形状特征、纹理特征、小波变换后的特征。采用观察者一致性检验、最大相关最小冗余算法、最小绝对收缩和选择算子算法逐步进行组学特征选择;采用logistic回归模型构建疗效评估模型;采用受试者工作特征(ROC)曲线分析模型的诊断效能。结果:最终筛选出12个影像组学特征纳入影像组学模型,训练组和验证组ROC曲线下面积(AUC)分别为0.789和0.757。进一步将临床参数IDH1基因突变状态、1p19q杂合缺失情况纳入模型后,联合模型中的AUC在训练组和验证组分别为0.880(95%CI 0.788~0.902)、0.929(95%CI 0.830~1.000),灵敏度分别为0.864、0.857,特异度分别为0.834、0.910。结论:基于T2-FLAIR建立的临床-影像组学联合模型具备早期预测HGG术后1年复发的价值。

Purpose:To explore the value of radiomics analysis based on T2-fluid attenuated inversion recovery(T2-FLAIR)for evaluating the recurrence within 1 year after surgery in high-grade glioma(HGG).Methods:We retrospectively collected 81 HGG patients(WHO gradeⅢ-Ⅳ)admitted to our hospital from January 2019 to December 2021.All patients had received standard postoperative adjuvant therapy.The MRI T2-FLAIR images within 4 weeks after surgery were collected.All patients were divided into a non-recurrence group(51 cases)and a recurrence group(30 cases)according to the follow-up data within 1 year after surgery.All patients were randomly assigned to training group(57 cases)and testing group(24 cases)according to a ratio of 7:3.The 3D-Slicer software was used to segment the tissues with T2-FLAIR hyperintense in the surrounding area of residual cavity.The 851 radiomics features were extracted,including first-order features,shape features,texture features and wavelet transformed features.The intraclass correlation coefficient test,minimum redundancy maximum correlation(MRMR),and least absolute shrinkage and selection operator(LASSO)algorithms were used to gradually select radiomics features.Logistic regression was used to construct evaluation model.Receiver operating characteristic(ROC)curve was used to analyze the diagnostic efficacy of the model.Results:In this study,the 12 radiomics features were selected and included in the radiomics model.The areas under the ROC curve(AUCs)were 0.789 and 0.757 in the training and testing group,respectively.The clinical features including IDH1 and 1p19q were further included into the combined model,and the AUCs in the training and testing group were 0.880(95%CI 0.788-0.902)and 0.929(95%CI 0.830-1.000),respectively.The sensitivity was 0.864 and 0.857,and the specificity was 0.834 and 0.910,respectively.Conclusions:The radiomics analysis based on T2-FLAIR images around residual cavity can be used to evaluate the recurrence within 1 year after surgery in HGG patients.