Artificial intelligence enables precision diagnosis of cervical cytology grades and cervical cancer

This research introduces and validates an artificial intelligence cervical cancer screening (AICCS) system for grading cervical cytology. Trained and validated on diverse datasets (10,656 participants), the AICCS uses two AI models: one for patch-level cell detection and another for whole-slide image (WSI) classification. The AICCS demonstrates high accuracy across datasets, achieving an AUC of 0.947 in prospective assessment. A randomized observational trial showed AICCS-assisted cytopathologists significantly improved specificity, accuracy, and sensitivity (13.3% increase) compared to cytopathologists alone. AICCS offers potential as a valuable tool for accurate and efficient cervical cancer screening.

Jue Wang, Yunfang Yu, Yujie Tan, Huan Wan, Nafen Zheng, Zifan He, Luhui Mao, Wei Ren, Kai Chen, Zhen Lin, Gui He, Yongjian Chen, Ruichao Chen, Hui Xu, Kai Liu, Qinyue Yao, Sha Fu, Yang Song, Qingyu Chen, Lina Zhu, Liya Wei, Jin Wang, Nengtai Ouyang, Herui Yao