Analysis of artificial intelligence-assisted cervical cytology screening combined with HPV detection in cervical cancer screening

Objective To evaluate the feasibility and application of artificial intelligence (AI) assisted cervical cytology screening combined with HPV detection in HPV detection for women in the rural areas. Methods Data were collected from free cervical cancer screening for women aged 35-64 in one county of Xuzhou from May 2019 to November 2021. According to different detection methods, the patients were divided into three groups: an artificially reading cytology screening (artificial) group, an artificial intelligence-assisted cervical cytology screening (AI) group, and an artificial intelligence-assisted cervical cytology combined with HPV screening (AI+HPV) group. The positive rates of abnormal cytology screening results (≥ASCUS) were compared between the two cytology screening methods. The positive rates of histopathological examination were compared among all three groups. The consistence between cytological results and pathological diagnosis were also analyzed. Results The AI group presented a higher rate of abnormal cytology screening than the artificial group (P<0.001). The AI group and the AI+HPV group showed a higher rate of pathological diagnosis than the artificial group (P<0.001). When ≥HSIL was used to calculate positive pathology, the positive rate of ASCUS in the AI+HPV group was higher than those in the other two groups (P<0.001). When ≥ LSIL was used to calculate positive pathology, the positive rates of ASCUS in the AI group and the AI+HPV group were higher than that in the artificial group (P<0.001). Conclusions AI-assisted cervical cytology screening can improve the detection rate of abnormal cytology and cervical precancerous lesions. AI-assisted cervical cytology combined with high-risk HPV screening reduces the referral rate of colposcopy and improves the consistence between cytological results and biopsy pathological diagnosis.