The significance of computer-aided analysis of liquid-based cytology in cervical cancer screening applications
YANG Wei-kang
ZHANG Yan
YI Li-mo
WU Qing-jun
LIU Yan-qun
GUO Jun
WANG Man
LING Sheng-rong
PENG Ru-jiao
SUO Ting-jiao
XIANG Ting
Abstract：Objective To evaluate the potential of computer-aided analysis(CAA)technology for liquid-based thin-layer cytology in improving the accuracy and efficiency of cervical cancer screening.Methods Using artifi-cial intelligence deep learning models,this study analyzed 220,000 images of liquid-based thin-layer cytology samples.The diagnostic outcomes and efficiency of the analysis methods were statistically evaluated.Cervical cytology diagnoses fol-lowed the 2014 Bethesda System(TBS),with positive results classified as ASC-US and above.Results The accuracy and efficiency of three different cervical cancer screening methods were compared.The sensitivity,specificity and accuracy of pathologists were 96.80％,93.61％ and accuracy 95.21％,respectively;those of AI-assisted analysis were 98.60％,79.61％ and 89.11％,respectively;and those of AI analysis system-assisted pathologist review were 100.00％,95.74％ and 97.87％,respectively.The results indicate that compared to traditional manual microscopic re-view,computer-aided analysis significantly improves the accuracy of pathologists'reviews.Regarding review times,three pathologists required approximately 11 000 hours to review all 220 000 samples manually,while AI-only analysis took 3 663 hours.AI-assisted review required approximately 5 500 hours,doubling the efficiency compared to manual review.Conclusion Computer-aided analysis of liquid-based thin-layer cytology holds significant value in cervical cancer screening.It effectively enhances screening accuracy and efficiency,alleviating pathologists'workloads.By combining CAA technology with clinical experience,it can serve as an important auxiliary tool for cervical cancer screening.
Keywords：cervical cancer screeningcomputer-aided analysisdeep learningliquid-based thin-layer cy-tologydiagnostic accuracy
Publication Date：2025-04-15
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：7( 495-501 )