文献速递神经头颈MR

基于MRI的影像组学列线图预测鼻咽癌放疗后颞叶损伤

MRI-based radiomics nomogram for predicting temporal lobe injury after radiotherapy in nasopharyngeal carcinoma.
PubMed 文献 · Eur Radiol · 2021-08-31

训练集中列线图模型预测放疗后颞叶损伤的AUC为0.87(95%CI 0.82-0.91),优于影像组学模型(AUC 0.71)和临床模型(AUC 0.73),验证集结果一致,校准良好且临床实用。

要点

  • 基于T2WI的影像组学列线图可无创预测鼻咽癌放疗后颞叶损伤。
  • 列线图模型预测效能优于单纯影像组学模型和临床模型。
  • 剂量学因素可提升影像组学模型的预测性能。

核心结论

  • 结合影像组学标签与临床因素的列线图模型能有效无创预测鼻咽癌患者放疗后颞叶损伤。

研究设计与样本

  • 回顾性分析203例接受放疗的鼻咽癌患者,共128侧RTLI阳性和278侧RTLI阴性颞叶,随机分为训练集285侧和验证集121侧。
  • 将临床危险因素与影像组学标签通过多变量逻辑回归分析构建影像组学列线图模型。

诊断标准与关键数值

  • 影像组学标签由14个与RTLI显著相关的特征组成。
  • 训练集中列线图模型AUC为0.87(95%CI 0.82-0.91),影像组学模型AUC为0.71(95%CI 0.65-0.78),临床模型AUC为0.73(95%CI 0.67-0.79)。
  • 验证集结果与训练集一致,列线图模型校准良好,决策曲线分析显示临床有用。

临床应用与阅片要点

  • 在鼻咽癌放疗后随访中,可利用基于T2WI的影像组学列线图辅助预测颞叶损伤风险。
  • 列线图模型整合了影像组学特征与临床因素,预测效能优于单一模型。
  • 剂量学因素可纳入模型以提升预测性能。

局限性

  • 研究为回顾性设计,可能存在选择偏倚。
  • 样本来自单中心,外部泛化性需进一步验证。

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

OBJECTIVES: To develop and validate a magnetic resonance imaging (MRI)-based radiomics nomogram model combining radiomic features and clinical factors for the prediction of radiotherapy-induced temporal lobe injury (RTLI) in patients with nasopharyngeal carcinoma (NPC).

METHODS: From 203 NPC cases receiving radiotherapy, 128 RTLI-positive and 278 RTLI-negative lobes were retrospectively analyzed. They were randomly divided into training (n = 285) and validation (n = 121) sets. Three hundred ninety-six texture features based on T2WI images were extracted from each temporal lobe. The minimum redundancy maximum relevance (mRMR) and the least absolute shrinkage and selection operator (LASSO) were used to reduce the dimension of the features and establish a radiomics signature model. Clinical risk factors and the radiomics signature were combined by multivariable logistic regression analysis to construct a radiomics nomogram model. We assessed the performance of the radiomics nomogram on discrimination, calibration, and clinical utility.

RESULTS: The radiomics signature consisted of 14 selected features that were significantly associated with RTLI. In the training set, the radiomics nomogram model demonstrated a better predictive performance (AUC, 0.87; 95% CI, 0.82-0.91) than the radiomics model (AUC, 0.71; 95% CI, 0.65-0.78) and clinical model (AUC, 0.73; 95% CI, 0.67-0.79). These results were confirmed in the validation set. The radiomics nomogram model demonstrated good calibration and was clinically useful by decision curve analysis.

CONCLUSION: The radiomics nomogram model combining radiomics signatures and clinical factors is an effective method for the noninvasive prediction of RTLI in NPC patients after radiotherapy.

KEY POINTS: • The radiomics model based on T2WI images at the end of intensity-modulated radiotherapy can predict radiotherapy-induced temporal lobe injury in patients with NPC. • Dosimetric factors can improve the prediction performance of the radiomics model in predicting radiotherapy-induced temporal lobe injury. • An MRI-based radiomics nomogram combining radiomics signatures and clinical factors had better prediction performance than both radiomics and clinical model for the prediction of radiotherapy-induced temporal lobe injury in patients with NPC.

来源:Eur Radiol,查看原文。版权归原作者所有。

引用

  1. Hou J, Li H, Zeng B, et al. MRI-based radiomics nomogram for predicting temporal lobe injury after radiotherapy in nasopharyngeal carcinoma. Eur Radiol. 2021. PMID: 34467454. doi:10.1007/s00330-021-08254-5

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。