人工智能辅助诊断系统预测肺腺癌PD-L1表达水平的应用研究  

An Applied Study of Artificial Intelligence-Assisted Diagnostic System for Predicting PD-L1 Expression Level in Lung Adenocarcinoma

作　　者：佘李岚 黄妹兰 薛蕴菁[1] SHE Lilan;HUANG Meilan;XUE Yunjing(Department of Radiology,Fujian Medical University Union Hospital,Fuzhou Fujian 350001,China)

机构地区：[1]福建医科大学附属协和医院放射科,福建福州350001

出　　处：《中国卫生标准管理》2023年第23期106-109,共4页China Health Standard Management

基　　金：福建省卫生健康科技计划项目资助(2022QNA023)。

摘　　要：目的探讨人工智能辅助诊断系统对肺腺癌细胞程序性死亡配体1(programmed cell death 1 ligand 1,PD-L1)表达水平的预测研究。方法回顾性分析2021年1月—2022年4月福建医科大学附属协和医院胸外科收治的106例肺腺癌患者(PD-L1阳性56例,PD-L1阴性50例)的临床资料。所有患者均于术前或穿刺前2周内行肺部CT平扫,并将CT肺窗影像导入肺结节人工智能(artificial intelligence,AI)辅助诊断系统,自动生成长径、体积、CT值最大值、CT最小值、CT中位数、CT平均值、CT标准差、实性成分占比,紧凑度、球形度、CT值的峰度和偏度、熵等量化参数。分析临床指标和AI参数在两组间的差异性。结果两组年龄、性别、吸烟史、表皮生长因子受体(epidermal growth factor receptor,EGFR)状态比较,差异有统计学意义(P<0.05)。相较于PD-L1阴性组,PD-L1阳性组多为年龄较大且有吸烟史的男性患者,具有更高的EGFR突变率。两组肿块的最大径、体积、CT最小值、CT平均值、CT中位数、实性成分占比、紧凑度、CT值的峰度和偏度比较,差异有统计学意义(P<0.05);相较于PD-L1阴性组,PD-L1阳性组肿块尺寸更大,密度更高,含有更多的实性成分,且形态较不规则。结论人工智能辅助诊断系统对肺腺癌PD-L1表达的预测具有较高的应用价值。Objective To explore the prediction study of programmed death ligand 1(PD-L1)expression level in lung adenocarcinoma cells by artificial intelligence-assisted diagnosis system.Methods The clinical data of 106 patients with lung adenocarcinoma(56 cases of PD-L1 positive and 50 cases of PD-L1 negative)admitted to Fujian Medical University Union Hospital from January 2021 to April 2022 were retrospectively analyzed.All patients underwent CT plain scanning of the lungs before surgery or within 2 weeks before puncture,and the CT lung window images were imported into the artificial intelligence(AI)artificial intelligence assisted diagnostic system for lung nodules,which automatically generated quantitative parameters such as longitudinal diameter,volume,maximum CT value,minimum CT value,median CT value,mean CT value,standard deviation of CT,percentage of solid component,compactness,sphericity,kurtosis and skewness of CT values and entropy.Clinical indicators and AI parameters were analyzed for differences between the two groups.Results There were significant differences in age,sex,smoking history and epidermal growth factor receptor(EGFR)status between the two groups(P<0.05).Compared with PD-L1 negative group,PD-L1 positive group was mostly older male patients with smoking history,and had higher EGFR mutation rate.There were significant differences in the maximum diameter,volume,minimum CT value,average CT value,median CT value,proportion of solid components,compactness,kurtosis and skewness of CT value between the two groups(P<0.05).Compared with PD-L1 negative group,PD-L1 positive group had larger size,higher density,more solid components and more irregular shape.Conclusion The artificial intelligence-assisted diagnosis system has high application value for the prediction of PD-L1 expression in lung adenocarcinoma.

关 键 词：细胞程序性死亡配体1 人工智能辅助诊断 肺腺癌 计算机体层摄影 免疫检查点 表皮生长因子受体 

分 类 号：R734[医药卫生—肿瘤]