文献速递神经头颈MR

基于脑肿瘤界面的MRI影像组学模型预测脑膜瘤Ki-67增殖状态:多中心研究

Brain-tumor interface-based MRI radiomics models to predict Ki-67 proliferation status of meningiomas: A multi-center study.
PubMed 文献 · Eur J Radiol · 2025-11-25

该多中心回顾性研究纳入两个中心共300例脑膜瘤患者,按免疫组化Ki-67表达分为低表达组(<5%)和高表达组(≥5%),从T1-CE和T2WI图像提取肿瘤实质及4 mm脑肿瘤界面区域的影像组学特征构建模型,并结合临床独立预测因子(肿瘤体积和最大径)建立列线图;列线图在内外部验证队列中AUC分别达0.846和0.752,优于所有其他评估模型。

要点

  • 联合肿瘤实质与4 mm脑肿瘤界面区域的影像组学模型优于任一单区域模型。
  • 肿瘤体积和最大径是Ki-67增殖状态的独立临床预测因子。
  • 整合影像组学评分与临床预测因子的列线图诊断性能最高。

核心结论

  • 脑肿瘤界面区域有预测脑膜瘤Ki-67增殖的潜力,所构建的列线图提供了有效的预测工具,支持个体化治疗策略。

研究设计与样本

  • 多中心回顾性研究,纳入两个中心共300例脑膜瘤患者,以免疫组化Ki-67表达(<5%为低表达组,≥5%为高表达组)为参考标准,从T1-CE和T2WI图像提取肿瘤实质及4 mm脑肿瘤界面区域的影像组学特征。

诊断标准与关键数值

  • 内部验证队列:肿瘤影像组学模型AUC为0.754,脑肿瘤界面影像组学模型AUC为0.702,联合区域影像组学模型AUC为0.780,显著优于两个单区域模型。
  • 外部测试队列:肿瘤影像组学模型AUC为0.681,脑肿瘤界面影像组学模型AUC为0.682,联合区域影像组学模型AUC为0.709,联合模型同样更优。
  • 临床模型(含肿瘤体积和最大径)内部验证队列AUC为0.741,外部测试队列AUC为0.688。
  • 整合影像组学评分与临床预测因子的列线图内部验证队列AUC为0.846,外部测试队列AUC为0.752,显著优于所有其他评估模型。
  • Ki-67表达分组标准:低表达<5%,高表达≥5%。

临床应用与阅片要点

  • 阅片时可关注肿瘤体积和最大径,原文多因素分析显示二者为Ki-67增殖状态的独立临床预测因子。
  • 脑肿瘤界面区域(4 mm)的影像组学特征对Ki-67预测有附加价值,联合肿瘤实质区域可提升模型性能。
  • 列线图整合影像组学评分与临床预测因子,原文报告其在内外部验证中诊断性能最高,作者认为可作为预测工具支持个体化治疗策略。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

英文原文(摘要)· 点击展开

OBJECTIVE: To evaluate the potential of brain-tumor interface (BTI) radiomics for predicting the Ki-67 proliferation index in meningiomas and to construct and validate a nomogram integrating radiomics with clinical features for this purpose.

MATERIALS AND METHODS: This multicenter retrospective study enrolled 300 patients from two distinct centers. Patients diagnosed with meningioma were stratified into low (<5%) and high (≥5%) Ki-67 expression groups based on immunohistochemistry. Clinical data were collected, and independent predictors were identified via univariate and multivariate logistic regression analyses. Radiomics features from the tumor parenchyma and the 4 mm BTI region were extracted from T1-CE and T2WI images to construct single-region and combined-region radiomics models. A nomogram integrating the radiomics scores (Radscore) from the optimal radiomics model and clinical predictors was developed and validated.

RESULTS: In the internal validation cohort, the tumor radiomics model achieved an AUC of 0.754, the BTI radiomics model achieved an AUC of 0.702, and the combined Radiomics model performed better with an AUC of 0.780, significantly outperforming both single-region models.In the external test cohort, corresponding AUCs were 0.681 for the tumor radiomics model, 0.682 for the BTI radiomics model, and 0.709 for the combined Radiomics model, again showing the combined model's superiority. Univariate and multivariate analyses identified tumor volume and maximum diameter as independent clinical predictors of Ki-67 proliferation status. The clinical model incorporating these features reached an AUC of 0.741 in the internal validation cohort and 0.688 in the external test cohort. The nomogram integrating radiomics scores with clinical predictors achieved the highest diagnostic performance (AUC of 0.846 in the internal validation cohort, 0.752 in external test cohort), significantly outperforming all other evaluated models.

CONCLUSIONS: The BTI region shows the potential for predicting meningioma Ki-67 proliferation. The nomogram developed in this study provides an effective tool for this prediction, supporting personalized treatment strategies.

来源:Eur J Radiol,查看原文。版权归原作者所有。

引用

  1. Gao P, Wei J, Wang W, et al. Brain-tumor interface-based MRI radiomics models to predict Ki-67 proliferation status of meningiomas: A multi-center study. Eur J Radiol. 2025. PMID: 41325728. doi:10.1016/j.ejrad.2025.112577

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。