Assessment on effect of liquid-based cytology screening based on AI technique in diagnosis for precancerous lesions of cervix
10.3969/j.issn.1672-8270.2025.09.007
- VernacularTitle:基于人工智能技术的液基细胞学筛查在宫颈癌前病变诊断中的效果评估
- Author:
Hui JIANG
1
;
Xin WEI
1
;
Jie CHEN
1
Author Information
1. 中国医学科学院北京协和医院病理科 北京 100730
- Publication Type:Journal Article
- Keywords:
Artificial intelligence(AI);
Liquid-based cytology screening;
Precancerous lesions of cervix;
Effectiveness assessment
- From:
China Medical Equipment
2025;22(9):32-35
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To investigate application effect of liquid-based cytology screening based on artificial intelligence(AI)technique in diagnosing precancerous lesions of cervix,so as to provide basis for early diagnosis and treatment about cervical cancer.Methods:A total of 2 406 cervical liquid-based cytology specimens that were collected from the Health Medical Physical Examination Center of Peking Union Medical College Hospital during January and February 2024.AI-powered pathology platforms were employed to perform pathological diagnosis on submitted tissue samples,and the AI-assisted diagnostic results were compared with microscopic diagnosis results by senior cytopathologists.The pathological images of liquid-based cell sheet were automatically screened through deep convolutional neural network(DCNN),and the results of senior cytopathologists were used as"gold standard"to judge all diagnostic results.The accuracy of AI was analyzed in rapidly identifying precancerous lesions or early-stage cervical cancer.Results:Among 2 406 pathological diagnoses,2 054 specimens were negative and 352 positive.Of the 352 positive cases,123 cases Atypical squamous cells cannot exclude high-grade squamous intraepithelial lesions(ASCH)66 cases showed atypical squamous cells(ASCUS),83 cases showed low-grade squamous intraepithelial neoplasia(LSIL),9 cases showed high-grade squamous intraepithelial neoplasia(HSIL),70 cases showed infectious lesions(including bacterial vaginosis,Candida albicans,trichomoniasis,and actinomycosis),and 1 case showed atypical glandular cells(AGC).The AI-assisted imaging system achieved an overall diagnostic positive accuracy rate of 95.20%(335/352)and 90.76%,and the sensitivities of that in diagnosing ASCUS and LSIL were respectively 36.72%and 98.00%,all of which were higher than the diagnostic results of physicians with junior and intermediate professional titles,and the differences were significant(x2=11.81,21.97,P<0.05).The time of AI system in reading images was 49 s,while those of pathologists with junior and intermediate professional titles were respectively 125 s and 80 s.Conclusion:The AI system has higher sensitivity in liquid-based cytology screening for cervix,and its diagnostic efficiency has a certain of advantages.The application of AI-assisted diagnosis has substantially improved the accuracy and reliability of diagnosing precancerous lesions of cervix.
