The value of contrast-enhanced CT-based radiomics for predicting lymphovascular invasion of esophageal squamous cell carcinoma
李扬 1王向明 1谷霄龙 1杨丽 1王琦 1时高峰 1随义 2徐校胜 1岳萌 3王明博 4任嘉梁5
扫码查看
点击上方二维码区域，可以放大扫码查看
作者信息
1. 050011 石家庄,河北医科大学第四医院CT磁共振科
2. 053099 河北,衡水市第四人民医院
3. 050011 石家庄,河北医科大学第四医院病理科
4. 050011 石家庄,河北医科大学第四医院胸外科
5. 100176 北京,GE中国
折叠
摘要
目的:探讨基于增强CT影像组学预测食管鳞癌(ESCC)淋巴血管侵犯(LVI)的价值.方法:回顾性搜集行根治性切除术并经术后病理证实的224例食管鳞癌患者,其中包括66例LVI阳性和158例LVI阴性患者.所有患者均在术前2周内进行胸部增强CT扫描.将入组的患者按照7∶3的比例随机分为训练集和测试集.使用3D Slicer软件逐层勾画全肿瘤感兴趣区(ROI),采用Python软件的Pyradiomics包提取肿瘤组织的影像组学特征,建立影像组学模型用于预测食管鳞癌的LVI状态并进行验证.采用受试者工作特征(ROC)曲线的曲线下面积(AUC)、敏感度、特异度、准确度、阳性预测值和阴性预测值来评价影像组学模型的诊断效能,使用校准曲线评价影像组学模型在训练集和测试集中的拟合程度.使用决策曲线分析(DCA)评价影像组学模型的临床应用价值.结果:从全肿瘤ROI中提取了 1130个组学特征,经过筛选最终保留了 7个影像组学特征,并使用多因素logistic回归建立影像组学预测模型.在训练集中,影像组学模型预测LVI的AUC值为0.930,敏感度为0.851,特异度为0.919,准确度为0.899,阳性预测值为0.816,阴性预测值为0.936;在测试集中,AUC值为0.897,敏感度为0.789,特异度为0.787,准确度为0.788,阳性预测值为0.600,阴性预测值为0.902.校准曲线显示影像组学模型在训练集及测试集中的预测概率与实际概率的一致性良好.D C A曲线显示影像组学模型具有良好的临床应用价值.结论:基于增强CT构建的影像组学模型,能够在术前有效预测食管鳞癌的LVI状态.
Abstract
Objective:To explore the value of predicting lymphovascular invasion(LVI)of e-sophageal squamous cell carcinoma(ESCC)based on enhanced CT radiomics.Methods:A total of 224 patients with ESCC who underwent radical resection and were confirmed by postoperative pathology were retrospectively collected,including 66 LVI-positive and 158 LVI-negative patients.All patients underwent contrast-enhanced chest CT scan within 2 weeks before surgery.The enrolled patients were randomly divided into training and test set in the ratio of 7∶3.The whole tumor region of interest(ROI)was outlined layer by layer using 3D slicer software,and the radiomics features of the tumor tissues were extracted using the Pyradiomics package of Python software.Then,a radiomics model was built to predict the LVI status of ESCC and to be validated.The area under the curve(AUC)of receiv-er operating characteristic(ROC),sensitivity,specificity,accuracy,positive predictive value(PPV),and negative predictive value(NPV)were used to evaluate the diagnostic efficiency of radiomics mod-el.The calibration curves were used to evaluate the degree of fit of the radiomics model in the training and test sets.Evaluation of clinical applications of CT radiomics models using decision curve analysis(DC A).Results:A total of 1130 texture features were extracted from the whole-tumor ROIs,and 7 ra-diomics features were finally retained after screening to build a prediction model using multivariate lo-gistic regression.In the training set,the AUC,sensitivity,specificity,accuracy,PPV,and NPV of ra-diomics model for predicting LVI were 0.930,0.851,0.919,0.899,0.816,and 0.939,respectively;in the test set,the AUC,sensitivity,specificity,accuracy,PPV and NPV of radiomics model for predicting LVI were 0.897,0.789,0.787,0.788,0.600,and 0.902,respectively.The calibration curves showed good consistency of the radiomics model between the predicted and the actual probability in training and test sets.The DCA curve showed that the radiomics model had good clinical applications.Conclusion:The radiomics model based on the contrast-enhanced CT can effectively predict the LVI status of ESCC be-fore surgery.
维普
万方数据