基于能谱CT参数及临床特征构建非小细胞肺癌PD-L1表达的列线图预测模型  

Construction of a nomogram prediction model for PD-L1 expression in non-small cell lung cancer using spectral CT parameters and clinical features

作　　者：朱凯博 邓靓娜 王海升 刘建强[1] 雒攀 周俊林[1] ZHU Kaibo;DENG Liangna;WANG Haisheng;LIU Jianqiang;LUO Pan;ZHOU Junlin(Department of Radiology,Lanzhou University Second Hospital/the Second Clinical Medical School,Lanzhou University/Key Laboratory of Medical Imaging of Gansu Province/Gansu International Scientific and Technological Cooperation Base of Medical Imaging Artificial Intelligence,Lanzhou 730030,China)

机构地区：[1]兰州大学第二医院放射科/兰州大学第二临床医学院/甘肃省医学影像重点实验室/医学影像人工智能甘肃省国际科技合作基地,甘肃兰州730030

出　　处：《中国医学物理学杂志》2025年第4期443-449,共7页Chinese Journal of Medical Physics

基　　金：国家自然科学基金(82371914)。

摘　　要：目的:基于临床信息、常规CT征象及能谱CT参数构建列线图模型术前预测非小细胞肺癌(NSCLC)程序性死亡配体1(PD-L1)的表达水平。方法:回顾性分析经病理证实为NSCLC的52例患者,术前均行能谱CT检查。依据PD-L1表达水平分为阳性组和阴性组,收集两组患者的临床信息、常规CT征象及能谱CT定量参数。临床信息包括性别、年龄、Ki-67、肿瘤标志物。常规CT征象包括肿瘤密度、边界、钙化征、毛刺征、分叶征、胸膜凹陷征、空洞征等。测量两组病灶在动脉期和静脉期的能谱参数,包括有效原子序数(Eff-Z)、碘浓度(IC)、水浓度(WC),并计算归一化碘浓度(NIC)。比较两组间的差异,采用多因素Logistic回归来筛选独立预测因子并构建预测模型,采用受试者工作特征曲线、校准曲线和决策曲线评估列线图模型的预测效能及准确性。结果:临床信息中,两组在性别方面的差异有统计学意义(P<0.05)。在动脉期和静脉期,PD-L1阳性组的能谱定量参数IC、NIC及Eff-Z均大于PD-L1阴性组,差异有统计学意义(P<0.05)。多因素Logistic回归分析显示性别(P=0.024)、静脉期Eff-Z(P=0.002)及静脉期IC(P=0.003)为PD-L1表达的独立预测因素。基于以上独立预测因子构建的列线图预测模型曲线下面积为0.80,敏感度为88.00%,特异度为59.00%。校准曲线表明模型预测值和实际值具有较高的一致性。决策曲线显示高风险阈值界于0.10~0.83时,模型可以获得最大净收益。结论:基于能谱CT定量参数和临床信息的列线图模型在术前预测NSCLC的PD-L1表达水平中有一定的价值。Objective To investigate the preoperative prediction of the expression level of programmed cell death ligand 1(PD-L1)in non-small cell lung cancer(NSCLC)by a nomogram model constructed with clinical data,conventional CT signs and spectral CT parameters.Methods A retrospective analysis was conducted on 52 patients with pathologically confirmed NSCLC and undergoing preoperative spectral CT examination.The patients were categorized into positive and negative groups according to PD-L1 expression level,and their clinical data,conventional CT signs and spectral CT parameters were collected.Specifically,clinical data included gender,age,Ki-67 and tumor markers;conventional CT signs included tumor density,margins,calcification,spiculation,lobulation,pleural indentation and cavitation;and spectral CT parameters measured in the arterial and venous phases included effective atomic number(Eff-Z),iodine concentration(IC),water concentration(WC)and normalized iodine concentration(NIC).Intergroup differences were analyzed,and multivariate Logistic regression was used to identify independent predictors and establish the prediction model which was evaluated for prediction performance and accuracy using receiver operating characteristic(ROC)curves,calibration curve and decision curve analyses.Results For clinical data,only the difference in gender between two groups had statistical significance(P<0.05).The spectral CT parameters(IC,NIC and Eff-Z)in the arterial and venous phases of PD-L1 positive group were all greater than those of PD-L1 negative group,with statistically significant differences(P<0.05).Multivariate Logistic regression analysis identified gender(P=0.024),venous-phase Eff-Z(P=0.002),and venous-phase IC(P=0.003)as independent predictive factors for PD-L1 expression.The nomogram prediction model constructed with these independent predictors had an area under curve of 0.80,a sensitivity of 88.00%,and a specificity of 59.00%.The calibration curve showed that the predicted values had a high consistency with the actual v

关 键 词：非小细胞肺癌 列线图 能谱CT 程序性死亡配体1 

分 类 号：R318[医药卫生—生物医学工程]