Abstract: Objective To explore the value of artificial intelligence-assistance for cytology cervical cancer screening in liquid-based cytology. Method Totally 1000 liquid-based cytology cervical cell smears archived in Department of Pathology, Beijing Chaoyang Hospital, Capital Medical University were selected. Multiple groups of comparative experiment were designed through artificial intelligence-assisted screening system developed by iDeepWise Company and professional pathologists manual diagnosis. The dysplasia classified into atypical squamous cells of undetermined significance (ASC-US) or higher grade was regarded as the positive criteria for cervical precancerous lesions based on the positive grading criteria of TBS 2014. The differences in the results of screening methods were analyzed, and the sensitivity and specificity were calculated. Results The results of artificial intelligence-assisted screening and manual diagnosis by professional pathologists were basically consistent with the previously archived results. The sensitivity, specificity, and accuracy of artificial intelligence-assisted screening were 100.00%,90.68% and 97.80%, respectively. Conclusion The artificial intelligence-assisted screening combined with the pathologist's reading skills could effectively reduce the incidence of missed diagnosis, with high sensitivity and specificity. It also could greatly reduce the workload of pathologists.

Key words: cervical cytology, cervical cancer screening, artificial intelligence-assisted analysis