Application Analysis of Artificial Intelligence Technology in Female Cervical Cancer Screening
Objective To explore the effectiveness and accuracy of an automated cytology assisted diagnosis system based on artificial intelligence technology in cervical cancer screening.Methods Relying on the artificial intelligence cervical cancer diagnosis platform independently developed and designed by the Yangtze River Delta Information Intelligence Research Institute,the technology was applied to the cervical cancer screening project of 64,788 women of appropriate age in Wuhu City.The positive samples of"abnormal cervical epithelial cells"detected by the intelligent auxiliary diagnosis platform and 10%negative samples were randomly selected,and the manual review results were used as the standard to calculate the coincidence rate,sensitivity,specificity and other indicators in the classification of AI reading in two-classification.The accuracy of the automated cytology diagnostic platform for cervical cancer screening was analyzed based on the histopathological results.Results A total of 64,788 women were screened for cervical cancer using human-computer interaction mode,and the consistency rate of artificial intelligence diagnosis and manual screening was 98.81%.Under the strict implementation of the secondary audit process of primary examination and review,the positive detection rate was stable at 6%-7%,and the daily issue report increased from the original manual reading of 100 copies/day to the maximum remote diagnosis number of 1,000 copies/day.Based on the manual review results,the sensitivity and specificity of AI interpretation were 97.08%and 98.94%,while the positive predictive value and negative predictive value were 87.56%and 99.77%,respectively.AI achieved high sensitivity and specificity value in two-classification.Conclusion The binary classification results of the artificial intelligence cervical cancer assisted diagnosis system in cervical cytology detection have high consistency and specificity with manual film reading,which can reduce the workload of cytopathologists,with high efficiency and accuracy.
NETL
NSTL
万方数据
