CD161流式细胞术对活动性肺结核诊断效能的评价  

Evaluation of the diagnostic efficacy of CD161 flow cytometry in active pulmonary tuberculosis

作　　者：张雅曦 张明霞[1] 杨倩婷[1] 孟娟[1] 余志勇[1] 陈骑[1] ZHANG Yaxi;ZHANG Mingxia;YANG Qianting;MENG Juan;YU Zhiyong;CHEN Qi(Institute of Hepatology,Shenzhen Third People′s Hospital,Shenzhen,Guangdong 518112,China)

机构地区：[1]深圳市第三人民医院肝病研究所,广东深圳518112

出　　处：《国际检验医学杂志》2023年第12期1421-1424,共4页International Journal of Laboratory Medicine

基　　金：国家自然科学基金面上项目(82172286);深圳市科技计划基础研究面上项目(JCYJ20210324131603008,CJGJZD20220517141404010)。

摘　　要：目的 探讨以淋巴细胞表面分子CD161为标识的流式细胞术对活动性肺结核的诊断价值。方法 回顾性研究2020年4—12月在该院住院的活动性肺结核患者268例,其中痰涂片结果为阳性的患者170例作为涂阳组,3次或以上痰涂片结果未见结核分枝杆菌(TB)的患者98例为涂阴组。另选取该院的体检健康者138例为对照组。采用流式细胞术检测各组淋巴细胞、单核细胞和CD161阳性淋巴细胞比例。基于流式细胞术建立深度学习网络模型对活动性肺结核进行诊断,并与GeneXpert、TB-DNA、酶联免疫斑点(Elispot)试验的诊断效能进行比较。结果 各组淋巴细胞、单核细胞和CD161阳性淋巴细胞比例及OR比较,差异有统计学意义(P<0.05);GeneXpert、TB-DNA、痰培养、Elispot在涂阳患者中的检出率分别为87.65%、81.18%、60.00%、85.29%,深度学习网络模型在涂阳结核、涂阴结核患者中的检出率分别为77.05%、71.42%;深度学习网络模型鉴别涂阳组和对照组的受试者工作特征曲线下面积(AUC)为0.87(95%CI 0.84~0.91),准确率为0.82,灵敏度为0.77,特异度为0.84,鉴别涂阴组和对照组的AUC为0.79(95%CI 0.73~0.85),准确率为0.71,灵敏度为0.71,特异度为0.84。结论 以CD161为标识的流式细胞术结合深度学习网络模型可为辅助诊断活动性肺结核提供新的方向,特别是在涂阴活动性肺结核的诊断中,并对指导临床抗结核治疗用药及预测活动性肺结核复发有参考价值。Objective To explore the diagnostic value of flow cytometry with lymphocyte surface molecule CD161 in active pulmonary tuberculosis.Methods A retrospective study was conducted on totally 268 patients with active pulmonary tuberculosis who were hospitalized in this hospital from April to December 2020.Among them,170 patients with positive sputum smear results were selected as the smear positive group,and 98 patients with no Mycobacterium tuberculosis(TB)in sputum smear results for three or more times were selected as the smear negative group.Another 138 healthy people were selected as the control group.Flow cytometry was used to detect the proportion of lymphocytes,monocytes and CD161 positive lymphocytes in each group.A deep learning network model based on flow cytometry was established to diagnose active pulmonary tuberculosis,and its diagnostic efficiency was compared with GeneXpert,TB-DNA and enzyme-linked immunospot(Elispot)test.Results There were significant differences in the percentages of lymphocytes,monocytes and CD161 positive lymphocytes and OR among the groups(P<0.05).The detection rates of GeneXpert,TB-DNA,sputum culture and Elispot test in smear positive patients were 87.65%,81.18%,60.00%and 85.29%,respectively.The detection rates of the deep learning network model in smear positive tuberculosis and smear negative tuberculosis patients were 77.05%and 71.42%,respectively.The area under the receiver operating characteristic curve(AUC)of the deep learning network model for identifying subjects in the smear positive group and the control group was 0.87(95%CI 0.84-0.91),the accuracy was 0.82,the sensitivity was 0.77,and the specificity was 0.84.The AUC of the deep learning network model for identifying subjects in the smear negative group and the control group was 0.79(95%CI 0.73-0.85),the accuracy was 0.71,the sensitivity was 0.71,and the specificity was 0.84.Conclusion CD161 based flow cytometry combined with deep learning network model can provide a new direction for auxiliary diagnosis of active pulm

关 键 词：活动性肺结核 流式细胞术 CD161 深度学习网络模型 

分 类 号：R446.19[医药卫生—诊断学]