一种基于人工智能的细胞病理诊断系统在肺癌诊断中的初步应用  

Preliminary application of an artificial intelligence-based cytopathological diagnostic system in diagnosis of lung cancer

作　　者：王娟[1] 杨震[1] 朱强[1] 贾艳红[1] 孙小蓉 陈良安[1] 梁志欣[1] WANG Juan;YANG Zhen;ZHU Qiang;JIA Yanhong;SUN Xiaorong;CHEN Liang'an;LIANG Zhixin(Department of Respiratory and Critical Care Medicine,the First Medical Center,Chinese PLA General Hospital,Beijing 100853,China;Wuhan Landing Medical high-tech CO.Ltd,Wuhan 430000,Hubei Province,China)

机构地区：[1]解放军总医院第一医学中心呼吸与危重症医学科,北京100853 [2]武汉兰丁医学高科技有限公司,湖北武汉430000

出　　处：《解放军医学院学报》2020年第9期897-900,共4页Academic Journal of Chinese PLA Medical School

摘　　要：目的评价一种基于人工智能的细胞病理诊断系统在肺癌诊断中的应用价值。方法选取解放军总医院第一医学中心2019年3-5月临床拟诊肺癌患者101例,收集肺组织活检标本94例、胸腔积液标本6例及腹水标本1例,男性79例,女性22例,平均年龄(57±10.55)岁。将所有标本涂片、染色,分别采用人工智能细胞病理诊断系统及快速现场评估(rapid on-site evaluation,ROSE)方法进行结果判读,判读结果与后续的金标准病理检查结果进行比较。结果最终对纳入的101例合格标本进行统计学分析,结果显示人工智能细胞病理诊断与金标准病理诊断的符合率为66.3%,敏感度为67%,特异性为60%(优势性检验χ^2=18.380,P=0.000),提示人工智能细胞病理学诊断对标本属性判断的正确率不高。ROSE诊断符合率为64.4%,敏感度为63.7%,特异性为70%(优势性检验χ^2=15.630,P=0.000),提示ROSE诊断对标本属性判断的正确率也不高。人工智能细胞病理诊断与ROSE方法诊断性能相近,但与金标准病理方法有较大差异。结论人工智能细胞病理诊断系统可以达到ROSE同等的诊断效果,可以有效减轻病理学检查医务人员工作负荷,提高工作效率,在肺癌诊断领域有一定的推广应用价值。Objective To explore the value of an artificial intelligence-based cytopathological diagnostic system in the diagnosis of lung cancer.Methods A total of 101 specimens from patients with pathologically diagnosed as lung cancer in Chinese PLA General Hospital from March to May in 2019 were collected,including 94 lung biopsy specimens,6 pleural fluid specimens and 1 ascites specimen.Of the 101 cases,there were 79 males and 22 females with an average age of 57±10.5 years old.All specimens were smeared and stained,then the artificial intelligence cytopathological diagnostic system and rapid on-site evaluation were applied.The results were compared with the gold standard of pathological diagnosis.Results Totally 101 eligible samples were analyzed.The results showed that the consistency rate between the artificial pathological diagnosis and the gold standard of pathological diagnosis was 66.3%,with sensitivity of 67% and specificity of 60% (χ^2=18.380,P=0.000).The consistency rate of ROSE group and the pathological diagnosis was 64.4%,with sensitivity of 63.7% and specificity of 70% (χ^2=15.630,P=0.000).The consisitency between ROSE and the artificial intelligence cytopathological diagnosis was high.Conclusion The artificial intelligence cytopathological diagnostic system can achieve the equal diagnostic performance of ROSE,which can effectively reduce the workload of medical staff for pathological examination and improve the work efficiency,which has a certain application value in the field of lung cancer diagnosis.

关 键 词：人工智能 细胞病理 快速现场评估 肺癌 活检标本 

分 类 号：R730.43[医药卫生—肿瘤]