| 翟晓阳,谷一冰,高宇,钱伟军,李立,韩东明.基于MRI时序性肿瘤内异质性特征预测高级别胶质瘤放化疗反应及患者总生存期[J].中国医学影像技术,2025,41(12):1980~1984 |
| 基于MRI时序性肿瘤内异质性特征预测高级别胶质瘤放化疗反应及患者总生存期 |
| MRI-based temporal intratumoral heterogeneity features for predicting chemoradiotherapy response of high-grade glioma and patients' overall survival |
| 投稿时间：2025-05-23 修订日期：2025-10-12 |
| DOI：10.13929/j.issn.1003-3289.2025.12.009 |
| 中文关键词: 胶质瘤 肿瘤微环境 放化疗 磁共振成像 总生存期 |
| 英文关键词:glioma tumor microenvironment chemoradiotherapy magnetic resonance imaging overall survival |
| 基金项目:河南省科技攻关计划(LHGJ20230528)。 |
|
| 摘要点击次数: 124 |
| 全文下载次数: 25 |
| 中文摘要: |
| 目的 观察基于MRI时序性肿瘤内异质性(ITH)特征预测高级别胶质瘤(HGG)放化疗反应及患者总生存期(OS)的价值。方法 回顾性纳入72例于肿瘤次全切除术后接受同步放化疗的HGG患者,根据其治疗反应分为有效组(n=40)与无效组(n=32)。基于组间差异有统计学意义的临床资料构建预测HGG放化疗反应的临床模型;采用nnUnet模型分割肿瘤,以K-means聚类算法生成ITH区域,提取并筛选治疗前(T1)、后(T2)ITH特征,据以构建T1模型、T2模型及时序性模型(即T1+T2模型);结合时序性模型与临床模型构建联合模型,以受试者工作特征曲线(ROC)的曲线下面积(AUC)评估模型效能。采用单及多因素Cox回归分析筛选可用于预测HGG患者OS特征,并以时间依赖型ROC(tROC)曲线评估其效能。结果 基于异柠檬酸脱氢酶(IDH)伴1p/19q状态及O6-甲基鸟嘌呤-DNA甲基转移酶(MGMT)启动子甲基化状态构建临床模型,基于T1 ITH特征gldm_LowGrayLevelEmphasis构建T1模型,基于T2 ITH特征ngtdm_Strength构建T2模型,基于T1、T2 ITH特征构建时序性模型;上述模型预测HGG放化疗反应的AUC为0.687~0.758,均低于联合模型(AUC=0.853)。T2 ITH特征glszm_SizeZoneNonUniformity和治疗反应均为预测HGG患者放化疗后OS的最优特征,其联合预测放化疗后HGG患者12、24及36个月OS的AUC分别为0.724、0.704及0.802。结论 基于MRI时序性ITH特征可有效预测HGG放化疗反应及患者OS。 |
| 英文摘要: |
| Objective To observe the value of MRI-based temporal intratumoral heterogeneity (ITH) features for predicting chemoradiotherapy response of high-grade glioma (HGG) and patients’ overall survival (OS). Methods Totally 72 HGG patients who underwent concurrent chemoradiotherapy after subtotal resection were retrospectively enrolled and divided into effective group (n=40) and ineffective group (n=32) according to chemoradiotherapy response of HGG. Based on clinical data being statistically different between groups, a clinical model for predicting chemoradiotherapy response of HGG was constructed. Then the tumors were segmented using a nnUnet model. Tumors’ ITH region were generated based on K-means clustering algorithm, and ITH features of HGG before (T1) and after (T2) treatment were extracted and screened to construct T1 model, T2 model and temporal model (T1+T2 model), respectively. A combined model was constructed based on temporal model and clinical model, and the efficacy of the above models were evaluated with the area under the curve (AUC) of receiver operating characteristic (ROC) curve. Univariate and multivariate Cox regression analyses were used to screen the characteristics able to be used to predict OS of patients with HGG, and their efficacy were evaluated with time-independent ROC (tROC). Results Clinical model was constructed based on isocitrate dehydrogenase (IDH) with 1p/19q status and methylation status of O6-methylguanine-DNA methyltransferase (MGMT) promoter, T1 model and T2 model was constructed based on characteristics of T1 ITH gldm_LowGrayLevelEmphasis and T2 ITH feature ngtdm_Strength, respectively, while the temporal model was constructed based on T1 and T2 ITH features. AUC of the above models for predicting chemoradiotherapy response of HGG ranged from 0.687 to 0.758, all lower than that of combined model (0.853). T2 ITH feature glszm_SizeZoneNonUniformity and chemoradiotherapy response of HGG were both optimal features for predicting HGG patients’ OS after chemoradiotherapy, and AUC of their combination for predicting OS of HGG patients 12, 24 and 36 months after chemoradiotherapy was 0.724, 0.704 and 0.802, respectively. Conclusion MRI-based temporal ITH features of HGG could be sued to effectively predict tumor chemoradiotherapy response and patients' OS. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
