基于增强CT影像组学预测浸润性肺腺癌EGFR突变状态的价值  

The value of predicting EGFR mutation status in invasive lung adenocarcinoma based on enhanced CT imaging omics

作　　者：王杰 王振羽 马良勇 张金华[3] 林玉芬 林吉征[3] 张濬韬 WANG Jie;WANG Zhenyu;MA Liangyong;ZHANG Jinhua;LIN Yufen;LIN Jizheng;ZHANG Juntao(Department of Radiology,Lanshan District people’s Hospital,Rizhao 276807,China;School of Micro-Electronics South China Universrry of Technology,Guangzhou 511442,China;Department of Radiology,Affilliated Hospital of Qingdao University,Qingdao 266071,China;GE Healthcare,Precision Health Institution,Shanghai 210000,China)

机构地区：[1]山东省日照市岚山区人民医院放射科,山东日照276807 [2]华南理工大学微电子学院,广东广州511442 [3]青岛大学医学院附属医院放射科,山东青岛266071 [4]通用电气医疗精准医学研究院,上海210000

出　　处：《医学影像学杂志》2024年第5期61-66,共6页Journal of Medical Imaging

基　　金：青岛大学医疗集团科研专项一般项目(编号:YLJT20202027)。

摘　　要：目的探讨建立并验证影像组学诺模图预测浸润性肺腺癌表皮生长因子受体(EGFR)突变的能力。方法选取231例经病理证实为浸润性肺腺癌患者,并有EGFR突变资料,按照7∶3的比例分为训练集(n=162)和测试集(n=69)。利用最大相关最小冗余(mRMR)及最小绝对收缩与选择子算法(LASSO)筛选最佳影像组学特征,构建预测肺腺癌EGFR突变状态的模型;以单因素及多因素logistic回归筛选浸润性肺腺癌EGFR突变相关的临床、病理及CT特征,构建临床模型;联合影像组学特征及临床、病理、CT特征构建影像组学诺模图,ROC曲线及AUC值评估模型的预测效能,并采用DeLong检验比较模型间AUC值差异是否有统计学意义,校正曲线及DCA曲线评估影像组学诺模图的临床价值。结果最终基于增强扫描动脉期及静脉期筛选出13项影像组学特征建立影像组学模型,该模型在训练集及测试集中的AUC值为0.79、0.76;临床模型由性别、病理分期及血管集束征构成,该模型在训练集及测试集中的AUC值分别为0.75、0.75;影像组学诺模图在训练集及测试集中的AUC值分别为0.82、0.80;DeLong检验显示,训练集中影像组学诺模图AUC值大于临床模型,差异有统计学意义(P<0.05),影像组学模型与临床模型AUC值差异无统计学意义(P>0.05);测试集中3种模型AUC值差异均无统计学意义(P>0.05);影像组学诺模图在训练集及测试集中的拟合优度良好,DCA曲线显示影像组学诺模图临床效用优于临床模型。结论基于增强CT影像组学诺模图对浸润性肺腺癌EGFR突变状态有良好的预测能力。Objective To develop and validate a radiomics nomogram for the prediction of the epidermal growth factor recep⁃tor(EGFR)mutation status in invasive lung adenocarcinoma.Methods 231 patients with pathologicallyconfirmed invasive lung adenocarcinoma with EGFR mutation data were retrospectively analyzed and divided into training set(n=162)and test set(n=69)at a ratio of 7:3.The minimum redundancy maximum relevance(mRMR)and the least absolute shrinkage and selection operator(LASSO)were used to select best radiomics features for constructing radiomics model to indentify the mutation status of EGFR,Univariate and multivariate logistic regression analysis were performed to screen the clinical,pathological and CT fea⁃tures associated with mutation status of EGFR for constructing clinical model and radiomics nomogram based on clinical,patho⁃logical,CT features and radiomics features.Receiver operating characteristic(ROC)curve and area under the curve(AUC)were used to evaluate the prediction efficiency of models and the AUC differences were compared by DeLong test.And calibration curve and decision curve were used to analyze to assess the clinical value of radiomics nomogram.Results 19 radiomics fea⁃tures were selected to build radiomics model and the AUC value of radiomics model were 0.79 and 0.76.The clinical model was composed of sex,pathological stage and vessel convergence sign and the AUC value of clinical model was 0.75 and 0.75.The AUC value of radiomics nomogram was 0.82 and 0.80.DeLong test showed that in training set,the AUC value of radiomics nomo⁃gram had a better performance than clinical model(P<0.05),there was no significant difference in the prediction performance of radiomics model and clinical model,and in test dataset,there was no significant difference in the prediction performance of three models.Radiomics nomogram had a better performance goodness of fit and DCA showed radiomics nomogram had a better performance than clinical model.Conclusion Radiomics based on enhanced CT had a better perfo

关 键 词：影像组学 浸润性肺腺癌 表皮生长因子受体 体层摄影术 X线计算机 

分 类 号：R734.2[医药卫生—肿瘤] R814.42[医药卫生—临床医学]