AI技术在宫颈癌及癌前病变细胞学筛查中的应用分析  

作　　者：蔡卫芳 刘燕 王竹音 崔万猛 刘宗磊 嵇学仙 刘莹 

机构地区：[1]浙江省德清县人民医院,313200 [2]浙江省杭州市艾迪康医学检验中心,310023 [3]云思考科技(杭州)有限公司,310000

出　　处：《浙江临床医学》2021年第6期879-880,共2页Zhejiang Clinical Medical Journal

摘　　要：目的探讨AI技术在宫颈癌及癌前病变细胞学筛查中的应用:方法收集于德清县人民医院和杭州艾迪康医学检验中心存档的共2,000例液基薄层宫颈细胞涂片样本作为研究对象,分别采用AI系统辅助、专业细胞病理医师人工阅片,比较两种阅片方法的敏感性、特异性、阳性预测值、阴性预测值以及检测所需时间。结果AI阅片与人工阅片的敏感性、阳性预测值存在显著差异(P<0.05),AI阅片具有更高的敏感性,人工阅片的阳性预测值高于AI阅片。AI阅片与人工阅片的特异性、阴性预测值比较,差异均无统计学意义(P>0.05)。人工阅片的特异性优于AI检测,但无统计学差异。AI+人工阅片的敏感性和特异性优于两种单一阅片方法(P<0.05),并且可提高工作效率结论不建议将AI辅助检测独立应用于宫颈癌及癌前病变的筛查,目前尚需要结合传统人工阅片。Objective To explore the application of AI technology in cytological screening of cervical cancer and precancerous lesions.Methods A total of 2,000 liquid-based thin-layer cervical cell smear samples archived by Deqing County People's Hospital and Hangzhou Aidikang Medical Laboratory Center were collected as the research objects,and AI system assisted and professional cytopathologists were used to read the pictures respectively,and the two readings were compared.The sensitivity,specificity,positive predictive value,negative predictive value of the method,and the time required for detection.Results The sensitivity and positive predicrive value of AI reading and manual reading are significantly different(P<0.05).AI reading has higher sensitivity,and the positive predictive value of manual reading is higher than AI reading.The specificity and negative predictive value of AI reading and manual reading were not statistically significant(P>0.05).The specificity of manual reading is better than AI detection,but there is no statistical difference.The sensitivity and specificity of AI+manual reading are better than two single reading methods(P<0.05),and it can improve work efficiency.Conclusion It is not recommended to apply AI-assisted testing to the screening of cervical cancer and precancerous lesions independently.At present,it still needs to he combined with traditional manual reading.

关 键 词：人工智能(AI) 细胞学 宫颈癌/癌前病变 筛查 

分 类 号：R73[医药卫生—肿瘤]