Related Experiment Video
Updated: Sep 8, 2025
05:33
Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
Published on: July 11, 2025
221
Deep Learning-Assisted System Improves Practical Effects in Cervical Cytopathology Diagnosis: A Comparative Study of
Zichen Ye1, Peiyu Zhang2, Ronggan Wei3
1School of Population Medicine and Public Health, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China.
Summary
Deep learning (DL) reading modes significantly enhance cervical cancer screening accuracy and efficiency for cytopathologists. The concurrent mode offers the best efficiency, while second and triage modes excel in sensitivity and specificity, respectively.
Area of Science:
- Medical Imaging
- Computational Pathology
- Artificial Intelligence in Medicine
Background:
- Deep learning (DL) shows promise in improving cytopathology diagnostics.
- Limited data exists on the comparative effectiveness and user experience of different DL-assisted reading modes.
Purpose of the Study:
- To evaluate four distinct DL-assisted reading modes (unassisted, concurrent, second, triage) in cervical cytopathology.
- To assess their impact on diagnostic performance, efficiency, and cytopathologist preferences.
Main Methods:
- A randomized, controlled, four-way crossover study involving 108 cytopathologists reading 1620 cervical slides.
- Utilized four reading modes: unassisted, concurrent, second, and triage.
- Included a questionnaire survey on adoption, confidence, and preferences.
Main Results:
- All DL-assisted modes improved sensitivity and specificity compared to unassisted reading.
- Concurrent mode significantly reduced reading time (53s/slide), while triage mode also showed efficiency gains (130s/slide).
- Over 90% of cytopathologists found DL modes useful; concurrent mode was preferred, but triage mode scored highest.
Conclusions:
- Second mode enhances sensitivity, triage mode improves specificity, and concurrent mode maximizes efficiency.
- Findings offer guidance for selecting optimal DL-assisted modes in cytopathology practice.
- DL integration can enhance diagnostic confidence and performance in cervical screening.
