Application of AI-assisted diagnostic system in cervical liquid-based thin-layer cytology
MA Yi1, 2, HE Qiuxiang3, MA Jun3, HU Leyin3, CHEN Chen4, LI Jianmin3.
1.School of Information and Engineering, the First Clinical Medical College, Wenzhou Medical University, Wenzhou 325035, China;2.Department of Pathology, Sanmen People’s Hospital, Taizhou 317100, China; 3.Department of Pathology, the First Affiliated Hospital of Wenzhou Medical University, Wenzhou 325015, China; 4.Renji College, Wenzhou Medical University, Wenzhou 325035, China
MA Yi,HE Qiuxiang,MA Jun, et al. Application of AI-assisted diagnostic system in cervical liquid-based thin-layer cytology[J]. JOURNAL OF WEZHOU MEDICAL UNIVERSITY, 2024, 54(10): 806-810,816.
Abstract：Objective: To verify with clinical data the application value of AI-assisted diagnostic system in cervical liquid-based thin-layer cytology examinations and to explore the feasibility of in-depth development of AI technology in the field of pathological medicine. Methods: A retrospective analysis was conducted on 1 500 cases of cervical liquid-based thin-layer cytology smears collected from February to November 2023 at the First Affiliated Hospital of Wenzhou Medical University. The dataset comprised 843 membrane preparations and 657 sedimentary preparations. The sensitivity, specificity, and accuracy of cytopathologists alone, the AI-assisted diagnostic system alone, and cytopathologists utilizing the AI-assisted diagnostic system were evaluated, alongside a comparison of reading time. Results: Cytopathologists utilizing the AI-assisted diagnostic system demonstrated a sensitivity of 92.83%, specificity of 98.56%, and accuracy of 97.33%, compared with the AI-assisted diagnostic systems alone (with a sensitivity of 81.93%, specificity of 92.71%, and accuracy of 90.40%) and the cytopathologists alone (with a sensitivity of 82.24%, specificity of 97.63%, and accuracy of 94.33%), showing significant difference (P<0.05). Compared with the AI-assisted diagnostic system alone and the cytopathologists alone, cytopathologists utilizing the AI-assisted diagnostic system demonstrated significant improvement in sensitivity, specificity, and higher accuracy in both membrane-based and sedimentation-based slide preparation methods,with statistical significance (P<0.05). In addition, the average time for AI-assisted physicians to read images was 58 s/image,which was significantly lower than 70 s/image for cytopathologists alone. Conclusion: Cytopathologists leveraging AI-assisted diagnostic system benefit from heightened sensitivity, specificity, and accuracy. The adoption of AIassisted diagnostic system enhances screening efficiency and alleviates the workload of cytopathologists.
