摘要
目的通过基于胸部CT影像组学的列线图对非小细胞肺癌(NSCLC)表皮生长因子受体(EGFR)基因突变进行鉴别和预测。方法回顾杭州市第一人民医院2019年1月至2020年8月经病理检查证实为NSCLC的153例患者胸部CT图像及EGFR基因检测结果,将所有患者分为基因突变组90例及野生组63例,通过7︰3比例的分层抽样将所有患者分为训练组108例和验证组45例,提取所有CT图像影像组学特征并筛选,得到影像组学特征参数分数(Rad-score),同时建立影像组学特征模型。通过纳入Rad-score、图像语义特征及患者的临床资料,用多因素二元logistic回归建立联合模型,得到联合模型的列线图,实现模型可视化,并进行模型验证。绘制ROC曲线评价影像组学特征模型、临床-语义特征模型及联合模型对NSCLC EGFR基因突变的预测效能。结果联合模型对于鉴别NSCLC EGFR基因突变具有较好的预测效能,训练组AUC=0.77,95%CI:0.68~0.85,准确度为70.0%,灵敏度为0.67,特异度为0.76,阳性预测值为79.3%,阴性预测值为61.8%;验证组AUC=0.77,95%CI:0.63~0.91,准确度为71.1%,灵敏度为0.79,特异度为0.62,阳性预测值为70.4%,阴性预测值为72.2%。Rad-score、结节分型、吸烟史均为独立预测因子。结论通过基于胸部CT Rad-score、图像语义特征及临床特征资料建立的的联合模型所得到的列线图,对预测NSCLC EGFR基因突变具有一定价值。
Objective To develop a nomogram based on chest CT imaging to predict epidermal growth factor receptor(EGFR)gene mutations in non-small cell lung cancer(NSCLC).Methods The clinical data of 153 NSCLC patients admitted from January 2019 to August 2020 were retrospectively analyzed,including 90 cases of EGFR mutation and 63 cases of wild EGFR.The patients were randomly divided into the training group(n=108)and the validation group(n=45).The radiomic features were extracted with the GE Analysis Kit(AK),and the Rad-score was calculated.Meanwhile,the radiomic feature model was established.A combined model was established by incorporating the Rad-scores,the semantic features of the image and the clinical data of the patients,using logistic regression analysis,then the nomogram was constructed and validated.Receiver operating characteristic curve(ROC)was used to evaluate the predictive efficacy of the radiomic feature model,clinical-semantic feature model and combined model for EGFR gene mutation in NSCLC.Results The combined model had a good performance effect for identification of EGFR mutations in NSCLC.In the training cohort the area under the ROC curve(AUC)was 0.77(95%CI:0.68-0.85),the accuracy,sensitivity,specificity,positive predictive value and negative predictive value were 70.0%,0.67,0.76,79.3%and 61.8%,respectively.In the validation cohort,the AUC was 0.77(95%CI:0.63-0.91),the accuracy,sensitivity,specificity,positive predictive value and negative predictive value were 71.1%,0.79,0.62,70.4%and 72.2%,respectively.Rad-scores,nodule classification and smoking history were independent predictive factors.Conclusion A predictive nomo-gram has been developed with the combination of Rad-scores based on chest CT imaging,semantic feature of the image and clinical feature,which has a certain predictive value of EGFR gene mutations in NSCLC patients.
作者
甄涛
王罗羽
沈起钧
ZHEN Tao;WANG Luoyu;SHEN Qijun(Department of Radiology,Hangzhou First People's Hospital,Hangzhou 310006,China)
出处
《浙江医学》
CAS
2021年第19期2078-2083,2127,共7页
Zhejiang Medical Journal
基金
浙江省自然科学基金项目(LSY19H180009)
浙江省卫生科技计划项目(2021KY240)
浙江大学临床科学研究基金(YYJJ2019Z06)。
关键词
非小细胞肺癌
影像组学
列线图
表皮生长因子受体
Non-small cell lung cancer
Algorithm
Nomogram
Epidermal growth factor receptor