多参数双能CT影像组学用于鉴别喉部良恶性病变
回顾性纳入282例接受双能CT检查的患者构建影像组学模型,多参数模型鉴别喉部良恶性病变的AUC在训练集为0.938、测试集为0.920;结合影像组学特征与双能CT报告的病变大小的影像组学-临床模型在全队列AUC为0.945,优于影像组学模型(AUC 0.933,P<.05)和临床模型(AUC 0.855,P<.05)。结论认为该模型可作为无创鉴别喉部良恶性病变的重要方法。
要点
- 多参数双能CT影像组学模型鉴别喉部良恶性病变训练集AUC 0.938、测试集AUC 0.920。
- 影像组学-临床模型含影像组学特征与双能CT报告的病变大小,全队列AUC 0.945。
- 影像组学-临床模型优于影像组学模型(AUC 0.933,P<.05)和临床模型(AUC 0.855,P<.05)。
核心结论
- 多参数双能CT影像组学-临床模型可作为无创鉴别喉部良恶性病变的重要方法。
研究设计与样本
- 回顾性研究,基于282例接受双能CT检查患者的全队列构建影像组学模型。
- 采用最小绝对收缩和选择算子法筛选可重复特征构建影像组学标签。
- 采用多变量逻辑回归分析结合影像组学特征、临床特征及双能CT表现建立影像组学-临床模型。
- 通过区分度、校准度和临床使用评估列线图性能。
诊断标准与关键数值
- 多参数模型鉴别喉部良恶性病变的AUC训练集为0.938、测试集为0.920。
- 影像组学-临床模型(含影像组学标签和双能CT报告的病变大小)全队列AUC为0.945。
- 影像组学模型全队列AUC为0.933,影像组学-临床模型与之比较P<.05。
- 临床模型全队列AUC为0.855,影像组学-临床模型与之比较P<.05。
临床应用与阅片要点
- 原文结论认为多参数双能CT影像组学-临床模型可作为无创鉴别喉部良恶性病变的重要方法。
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
OBJECTIVES: To evaluate a multi-parametric dual-energy CT (DECT) radiomics model combined with clinical factors and DECT findings for differentiating between laryngeal benign and malignant lesions.
METHODS: For this retrospective study, the radiomic models were constructed based on a whole cohort of 282 patients underwent DECT examination. The radiomics signature was selected based on reproducible features by using the least absolute shrinkage and selection operator method, and highly related radiomic, clinical features and DECT findings were selected to establish the radiomics-clinical model by using multivariable logistic regression analysis. Nomogram performance was evaluated by its discrimination, calibration, and clinical use.
RESULTS: Compared with the single-parametric DECT radiomic models, the multi-parametric model can improve the diagnostic efficacy for differentiating laryngeal benign from malignant lesions, with area under the curves (AUCs) of 0.938 in training set and 0.920 in test set, respectively. Moreover, multi-parametric DECT-based radiomics-clinical model incorporates the radiomics signature and DECT-reported lesion size showed the best diagnostic performance (AUC of 0.945) in the whole cohort, when compared with the radiomics model (AUC of 0.933, P <.05) and clinical model (AUC of 0.855, P <.05).
CONCLUSIONS: The multi-parametric DECT-based radiomics-clinical model can serve as an important modality for non-invasively differentiating laryngeal benign lesions from malignant ones.
ADVANCES IN KNOWLEDGE: This study highlights the value of DECT-based radiomics model for differentiating laryngeal benign from malignant lesions. Moreover, the combination of radiomic, clinical features and DECT findings demonstrated good performance for differential diagnosing in laryngeal lesions.
引用
- Zhan Y, Zhang C, Wang Y, et al. Multi-parametric dual-energy CT radiomics for differentiating laryngeal benign and malignant lesions. Br J Radiol. 2026. PMID: 41408085. doi:10.1093/bjr/tqaf311
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。