AI approach outperformed human experts in identifying cervical precancer
A deep learning algorithm can use cervical images to identify precancerous changes more accurately than human experts and conventional cytology.
Researchers trained a deep‑learning algorithm on over 60,000 digitized cervical images from a Costa Rican population study. The algorithm, named automated visual evaluation, achieved an AUC of 0.91 for detecting precancerous lesions, outperforming both expert reviewers (AUC 0.69) and conventional cytology (AUC 0.71).
The method provides a scalable, low‑cost, and minimally trained screening tool that could greatly improve cervical cancer detection in low‑resource settings, potentially reducing the disease burden and death rates worldwide.
Evidence level: Állatkísérletes. Állatmodellben vizsgálták.
Related signals
- A multinational cross-sectional study on human papillomavirus and cervical cancer knowledge, vaccination attitudes, and risk factors in the Middle East
- Genomic features of cervical cancer
- Screening for Cervical Cancer
- Cervical cancer: Epidemiology, risk factors and screening
- Targeted treatments for cervical cancer: a review