基于CT影像组学的机器学习模型预测IA期肺腺癌高级别病理成分:一项双中心研究
回顾性纳入两家医院405例经手术病理确诊的IA期肺腺癌患者(高级别成分组156例,非高级别成分组249例),从肿瘤大体体积和瘤周体积提取影像组学特征,采用Lasso-SVM构建影像组学、临床及联合模型。联合模型在训练集、内部验证集和外部验证集的AUC分别为0.860(95%CI:0.809-0.912)、0.849(95%CI:0.764-0.933)和0.822(95%CI:0.730-0.914),提示基于CT影像组学的机器学习模型有助于术前识别IA期肺腺癌的高级别成分。
要点
- 联合模型在外部验证集AUC为0.822(95%CI:0.730-0.914),有助于术前识别IA期肺腺癌高级别成分。
- 共筛选出12个GTV特征、9个PTV特征和14个GPTV特征用于建模。
- 模型评估采用ROC曲线和决策曲线分析,在内部和外部验证集中均进行了效能与临床效用评价。
核心结论
- 基于CT影像组学的机器学习模型有助于术前识别临床IA期肺腺癌中的高级别成分。
研究设计与样本
- 回顾性收集两家医院405例经病理确诊并接受手术切除的IA期肺腺癌患者,其中高级别成分组156例、非高级别成分组249例。
- 从每例患者的肿瘤大体体积(GTV)和瘤周体积(PTV)提取影像组学特征。
- 采用Lasso-SVM在训练集中构建影像组学模型、临床模型和联合模型。
- 在内部验证集和外部验证集中采用受试者工作特征(ROC)曲线和决策曲线分析(DCA)评估模型性能和临床效用。
诊断标准与关键数值
- 共筛选出12个GTV特征、9个PTV特征和14个GPTV特征用于构建模型。
- 联合模型(纳入影像组学特征和临床特征)在训练集的AUC为0.860(95%CI:0.809-0.912)。
- 联合模型在内部验证集的AUC为0.849(95%CI:0.764-0.933)。
- 联合模型在外部验证集的AUC为0.822(95%CI:0.730-0.914)。
临床应用与阅片要点
- 原文结论认为基于CT影像组学的机器学习模型有助于术前识别临床IA期肺腺癌的高级别成分。
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
AIM: To explore the application value of a machine learning model based on CT radiomics in predicting high-grade components in clinical stage IA lung adenocarcinoma.
MATERIALS AND METHODS: A retrospective dataset of 405 patients with pathologically confirmed stage IA lung adenocarcinoma who underwent surgical resection at two hospitals was collected (156 cases in the HGC group and 249 cases in the non-HGC group). Radiomic features were extracted from each patient's gross tumor volume (GTV) and peritumoral volume (PTV). Lasso-SVM was employed to develop radiomics, clinical, and combined models in the training dataset. The models' performance and clinical utility were evaluated in both internal and external validation datasets using receiver operating characteristic (ROC) curves and decision curve analysis (DCA).
RESULTS: A total of 12 GTV features, 9 PTV features, and 14 GPTV features were selected for building the model. The combined model incorporating radiomic features and the clinical feature achieved an area under the curve (AUC) value of 0.860 (95% CI: 0.809-0.912) in the training set, 0.849 (95% CI: 0.764-0.933) in the internal validation set, and 0.822 (95% CI: 0.730-0.914) in the external validation set.
CONCLUSION: Machine learning model based on CT radiomics are helpful in preoperatively identifying high-grade components in clinical stage IA lung adenocarcinoma.
引用
- Pu ZS, He YF, Qiu SL, et al. A CT radiomics-based machine learning model for predicting high-grade pathological components in stage IA lung adenocarcinoma: a two-center study. Clin Radiol. 2025. PMID: 41033287. doi:10.1016/j.crad.2025.107077
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。