周围型小细胞肺癌的CT形态学特征及预测模型的建立  

Morphological Features on Computed Tomography and Predicting Model of Peripheral Small-cell Lung Cancer

作　　者：张磊[1] 沈文荣 蒋峰[1] 尹娜[1] ZHANG Lei;SHEN Wenrong;JIANG Feng(CT/MRI Room,Jiangsu Cancer Hospital,Nanjing,Jiangsu Province 210009,P.R.China)

机构地区：[1]南京医科大学附属肿瘤医院影像科,210009

出　　处：《临床放射学杂志》2021年第1期60-65,共6页Journal of Clinical Radiology

基　　金：江苏省卫健委课题(编号:CXTDA2017002);江苏省卫健委保健局课题(编号:BJ19032)。

摘　　要：目的探讨3 cm以下的周围型小细胞肺癌(SCLC)的临床表现与CT形态学特征并建立预测模型。方法搜集本院2011年5月至2019年9月CT发现肺部周围型结节,经病理证实为SCLC 43例,共43个病灶,另搜集77例周围型非小细胞肺癌(NSCLC)患者,共77个病灶作为对照组,所有病灶直径均≤3 cm,分析其临床表现及CT征象,行单因素及多因素Logistic回归明确SCLC的主要影响因素及回归系数,并以回归系数为基础计算风险评分预测模型(risk score),最后用受试者工作特征(ROC)曲线评判该模型的预测效能。结果单因素分析显示两组在性别、年龄、有无吸烟史、血液神经元特异性烯醇化酶(NSE)、有无合并慢性阻塞性肺疾病(COPD)、直径、切迹、磨玻璃、毛刺、支气管充气征、血管纠集、胸膜牵拉间差异有统计学意义(P<0.05),且切迹征在两者之间的CT表现有差异,前者较深且边缘圆钝;形状虽无统计学差异(P>0.05),但SCLC独特的"腊肠样"及"鹿角样"外形有别于少部分NSCLC斑块状不规则外形,这对诊断SCLC有一定的价值;空泡、空洞方面差异无统计学意义(P>0.05)。多因素Logistic回归显示有吸烟史(OR=37.793;P<0.05)、血液NSE阳性(OR=12.841;P<0.05)、切迹(OR=6.909;P<0.05)、无毛刺(OR=7.419;P<0.05)、无血管纠集(OR=8.884;P<0.05)、无胸膜牵拉(OR=22.293;P<0.05)为主要影响因素,通过使用风险评分模型risk score计算各主要影响因素所得风险总分并绘制ROC,曲线下的面积(AUC)为0.967(P<0.05),当风险总分切点选择为1.4125时,该模型的灵敏度为88.4%,特异度为98.7%,约登指数(准确率)为87.1%。结论3 cm以下周围型SCLC多见于高龄、男性、重度吸烟患者,血液NSE常升高,常合并COPD,CT上切迹征更常见,"腊肠样"或"鹿角状"外形是其特征;基于吸烟史、血液NSE水平、切迹、无毛刺、无血管纠集、无胸膜牵拉主要影响因素的回归系数而建立的risk score预测模型对SCLC具有较好的�Objective To investigate the clinical manifestations and CT morphological characteristics of peripheral small cell lung cancer(SCLC)below 3 cm and establish a prediction model.Methods From May 2011 to September 2019,CT Chest performed in our hospital found peripheral lung nodules.43 cases were pathologically confirmed to be SCLC,a total of 43 lesions,and 77 patients with peripheral non-small cell lung cancer(NSCLC)were collected.A total of 77 lesions were used as the control group.All lesions were≤3 cm in diameter.The clinical manifestations and CT signs were analyzed.Single-factor and multi-factor Logistic regression was performed to determine the main influencing factors and regression coefficients of SCLC,and the risk was calculated based on the regression coefficients.Score the prediction model(risk score),and finally use the receiver operating characteristic(ROC)curve to evaluate the predictive performance of the model.Results Univariate analysis showed that the difference between the two groups was statistically significant in sex,age,smoking history,NSE,chronic obstructive pulmonary disease(COPD),diameter,notching,ground glass,burr,air bronchogram,vascular convergence and pleural traction(P<0.05).Notching was different between the two groups,the former was deeper and blunter.Although there was no significant statistical difference in shape(P>0.05),the botuliform and antler-like shape of SCLC was different from irregular shapes of a few part of NSCLC,which is usefull for diagnosis of SCLC;The difference in vacuole sign and cavity was not statistically significant(P>0.05).Multivariate logistic regression showed that presence of smoking history(OR=37.793;P<0.05),NSE(OR=12.841;P<0.05),and notching(OR=6.909;P<0.05),absence of burr(OR=7.419;P<0.05),vascular convergence(OR=8.884;P<0.05),pleural traction(OR=22.293;P<0.05)were the main influencing factors.The area under ROC curve(AUC)based on all risk score calculated from risk score model was 0.967(95%CI,0.933-1.0).It revealed that risk score 1.4125 maybe an opt

关 键 词：小细胞肺癌 影响因素 切迹 risk score模型 

分 类 号：R734.2[医药卫生—肿瘤]