文献速递乳腺MR

基于MRI的瘤内及瘤周生境放射组学用于乳腺癌病理完全缓解的早期预测:一项多中心研究

MRI-based Intratumoral and Peritumoral Habitat Radiomics for Early Prediction of Pathologic Complete Response in Breast Cancer: A Multicenter Study.
PubMed 文献 · Acad Radiol · 2025-12-27

这项回顾性多中心研究纳入301例接受新辅助治疗并手术的乳腺癌患者,基于治疗前MRI构建生境放射组学模型预测病理完全缓解。生境模型在训练集、内部验证集和两个外部测试集的AUC分别为0.931、0.850、0.811和0.802,优于全局及瘤周放射组学模型;联合3 mm瘤周特征与临床因素的联合模型AUC更高,分别为0.957、0.871、0.842和0.853。

要点

  • 基于治疗前MRI的瘤内及瘤周生境放射组学可无创预测乳腺癌新辅助治疗后病理完全缓解。
  • 生境放射组学模型外部测试AUC为0.811和0.802,优于全局及瘤周放射组学模型。
  • 联合3 mm瘤周特征与临床因素的联合模型外部测试AUC达0.842和0.853。

核心结论

  • 基于MRI的生境放射组学通过捕捉肿瘤内及瘤周空间异质性,可无创预测病理完全缓解,该方法可能改善乳腺癌个体化治疗规划。

研究设计与样本

  • 回顾性多中心研究,纳入301例接受新辅助治疗后手术的乳腺癌患者,分为训练集143例、内部验证集62例和两个外部测试集共96例。
  • 基于治疗前MRI,通过K-means聚类生成生境亚区,并勾画3 mm、5 mm和7 mm同心瘤周区域。
  • 构建临床模型、瘤内放射组学模型、生境放射组学模型、三个瘤周模型和联合模型,以AUC评价,并用校准曲线和决策曲线评估可靠性和临床适用性,以SHAP评估模型可解释性。

诊断标准与关键数值

  • 生境放射组学模型AUC:训练集0.931、内部验证集0.850、外部测试集1为0.811、外部测试集2为0.802。
  • 联合3 mm瘤周特征与临床因素的联合模型AUC:训练集0.957、内部验证集0.871、外部测试集1为0.842、外部测试集2为0.853。
  • SHAP分析显示排名前五的贡献特征中有四个来自一个主要的瘤内生境亚区。

临床应用与阅片要点

  • 阅片时可关注治疗前MRI瘤内生境亚区和3 mm瘤周区域的特征提取,用于预测新辅助治疗后病理完全缓解。
  • 生境放射组学模型在外部测试集AUC为0.811和0.802,联合模型外部测试AUC为0.842和0.853,可辅助个体化治疗规划。

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

RATIONALE AND OBJECTIVES: Tumor heterogeneity is a main driver of varied treatment responses for neoadjuvant therapy (NAT) in breast cancer, yet conventional radiomics often overlooks intratumoral heterogeneity. This study aims to develop and validate a habitat radiomics model derived from pretreatment MRI for noninvasive prediction of pathologic complete response (pCR) in breast cancer patients undergoing NAT.

MATERIALS AND METHODS: This retrospective multicenter study included 301 patients with breast cancer who underwent NAT followed by surgery. Patients were assigned to a training set (n=143), internal validation set (n=62), and two external testing sets (n=96). Habitat subregions were generated via K-means clustering, and concentric peritumoral regions (3 mm, 5 mm, and 7 mm) were delineated. Several models were developed: a clinical model, an intratumoral radiomics model, a habitat radiomics model, three peritumoral models, and a combined model. Models were evaluated using the area under the receiver operating characteristic curve (AUC), while calibration and decision curves were used to assess model reliability and clinical applicability. Model interpretability was assessed using Shapley Additive exPlanations (SHAP).

RESULTS: The habitat radiomics model achieved AUCs of 0.931 (training), 0.850 (internal validation), 0.811 (external test 1), and 0.802 (external test 2), outperforming both global and peritumoral radiomics models. The combined model integrating 3 mm peritumoral features and clinical factors achieved higher AUCs of 0.957, 0.871, 0.842, and 0.853, respectively. SHAP analysis revealed that four of the top five contributing features originated from one dominant intratumoral habitat, highlighting habitat subregions as an important predictive marker of pCR.

CONCLUSION: MRI-based habitat radiomics enables noninvasive prediction of pCR by capturing spatial heterogeneity within and around tumors. This approach may improve individualized treatment planning in breast cancer.

来源:Acad Radiol,查看原文。版权归原作者所有。

引用

  1. Zhang B, Zhang S, Tian G, et al. MRI-based Intratumoral and Peritumoral Habitat Radiomics for Early Prediction of Pathologic Complete Response in Breast Cancer: A Multicenter Study. Acad Radiol. 2025. PMID: 41456995. doi:10.1016/j.acra.2025.12.012

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。