AI技术建立宫颈鳞状上皮病变组织学切片模型及应用价值分析

Establishment of Histological Section Model of Cervical Squamous Epithelial Lesions by AI Technology and Analysis of its Application Value

目的探讨人工智能(AI)技术在建立宫颈鳞状上皮病变组织学切片模型中的应用。方法收集2016年1月-2020年12月在杭州市妇产科医院手术的113例宫颈鳞状上皮病变患者的组织学切片,采用麦克奥迪数字切片扫描与应用系统对切片进行数字扫描,扫描完成后导入浙江赛尔微因公司的人工智能图像分析系统CellVigen v11.0进行标注训练、验证及测试。再用AI对切片进行判读,将判读结果与病理结果进行比对分析。结果通过训练集形成数据库及验证建立的AI模型与病理医生诊断符合率高,癌前病变、早期癌及浸润癌诊断符合率分别为86.70%、92.00%、92.00%。对AI诊断与病理医师的诊断结果比较分析,差异有统计学意义(P<0.05)。结论AI能够实现对宫颈癌前病变、早期癌及浸润癌的有效判读诊断,与诊断医师具有较高的符合率,在辅助诊断中具备一定的应用价值。

Objective To explore the application of artificial intelligence(AI)technology in establishing a histological slice model of cervical squamous epithelial lesions.Methods The histological sections of 113 patients with cervical squamous epithelial lesions who underwent surgery in Hangzhou Women’s Hospital from January 2016 to December 2020 were collected.The sections were digitally scanned using the Motic Digital Slice Scanning and Application System.After the scanning was completed,it was imported into the artificial intelligence image analysis system CellVigen v11.0 of Zhejiang CellVigene Company for labeling training,verification and testing.AI was used to interpret the sections,and the interpretation results were compared with the pathological results.Results The AI model established by training set formation database and verification had a high diagnostic coincidence rate with pathologists,the diagnostic coincidence rates of precancerous lesions,early cancer and invasive cancer were 86.70%,92.00%and 92.00%,respectively,and there was significant difference between the diagnosis of AI and that of pathologists(P<0.05).Conclusion AI can achieve effective interpretation and diagnosis of cervical precancerous lesions,early cancer and invasive cancer.It has a high coincidence rate with the diagnostician and has certain application value in the auxiliary diagnosis.