人工智能与7项肺癌自身抗体联合检测对肺结节良恶性鉴别诊断效能研究  

Study on the Effectiveness of Combined Detection of Artificial Intelligence and 7 Lung Cancer Autoantibodies in the Differential Diagnosis of Benign and Malignant Pulmonary Nodules

作　　者：刘梦[1] 任旭斌[1] 陈云凤[1] Liu Meng;Ren Xubin;Chen Yunfeng(Department of Respiratory and Critical Care Medicine,the First People's Hospital of Chengdu City,Chengdu,Sichuan 610041,China)

机构地区：[1]成都市第一人民医院呼吸与危重症医学科,四川成都610041

出　　处：《四川医学》2023年第11期1155-1159,共5页Sichuan Medical Journal

摘　　要：目的探讨人工智能(AI)与7项肺癌自身抗体(CAGE、MAGE A1、GBU4-5、GAGE7、SOX2、PGP9.5、P53)联合检测对肺结节良恶性鉴别诊断效能。方法收集2022年1月至2023年5月我院就诊的经过低剂量螺旋CT(LDCT)检测120例肺结节患者以及同期体检健康组(30例)的临床资料,通过AI结合医师诊断以及病理检查结果,将肺结节患者分为低风险结节组(45例)、良性组(28例)和肺癌组(47例)。同期完善7项肺癌自身抗体浓度检测,统计分析4组研究对象各项指标的差异。结果肺癌组中PGP9.5、SOX2、GAGE7、GBU4-5和MAGE A1表达水平明显高于其余3组(P<0.05)。在150例研究对象中,7项肺癌自身抗体联合检测共检出30例阳性,其中肺癌组阳性检出率最高,为46.8%(22/47),其次为良性组17.9%(5/28)和低风险结节组6.7%(3/45),而健康组全为阴性。将120例肺结节患者根据结节最大直径分为<1 cm,1~1.9 cm,2~3 cm,不同结节大小,AI风险概率差异有统计学意义(P<0.05),而7项肺癌自身抗体水平中仅有MAGE A1差异有统计学意义(P<0.05)。肺癌组AI风险概率(73.11±7.07)%高于良性组(68.01±7.02)%(P<0.05)。通过绘制受试者工作特征(ROC)曲线、计算曲线下面积(AUC),AI风险评估方法对肺癌的诊断AUC为0.689,敏感度为72.3%,特异度为60.7%。7项肺癌自身抗体AUC为0.864,敏感度是70.5%,特异度为81.5%。而AI与7项肺癌自身抗体二者联合检测AUC为0.935,敏感度83.6%、特异度82.1%,均优于单项检测方法。结论AI与7项肺癌自身抗体联合检测的敏感性和特异性较高,有利于早期发现恶性结节,对鉴别肺结节的良、恶性鉴别诊断具有一定的临床价值,值得推广。Objective To explore the effectiveness of artificial intelligence(AI)combined with 7 lung cancer autoantibodies(7-AABs),including CAGE、MAGE A1、GBU4-5、GAGE7、SOX2、PGP9.5、P53,in the differential diagnosis of benign and malignant pulmonary nodules.Methods The clinical data of 120 patients with pulmonary nodules detected by low-dose spiral CT(LDCT)and 30 healthy people who underwent physical examination in the same period in our hospital from January 2022 to May 2023 was collected.The pulmonary nodule patients were divided into low-risk nodule group(45 cases),benign group(28 cases)and lung cancer group(47 cases)according to the AI combined with physician diagnosis and pathological findings.At the same time,the concentration of serum 7-AABs was detected,and the differences of various indexes in the four groups were statistically analyzed.Results The expression levels of PGP9.5,SOX2,GAGE7,GBU4-5 and MAGE A1 in the lung cancer group were significantly higher than those in other three groups(P<0.05).Among the 150 subjects,a total of 30 positive cases were detected by 7-AABs,and the positive detection rate was the highest in the lung cancer group,46.8%(22/47),followed by 17.9%(5/28)in the benign group and 6.7%(3/45)in the low-risk nodule group,while all of the healthy group were negative.According to the maximum nodule diameter,120 pulmonary nodules patients were divided into<1 cm,1~1.9 cm,and 2~3 cm,and there were statistical differences in AI risk probability with different nodule sizes(P<0.05).Among the level of 7-AABs,only MAGE A1 showed statistical differences(P<0.05).The AI risk probability of lung cancer group(73.11±7.07)%was higher than that of the benign group(68.01±7.02)%(P<0.05).By drawing receiver operating characteristic(ROC)curve and calculating the area under curve(AUC),the AUC of AI risk assessment method was 0.689,the sensitivity was 72.3%and the specificity was 60.7%.The AUC of 7-AABs was 0.864,the sensitivity was 70.5%and the specificity was 81.5%.While the AUC of AI combined with 7-AABs wa

关 键 词：肺结节 人工智能 肺癌自身抗体 联合检测 

分 类 号：R734.2[医药卫生—肿瘤]