A collaborative study led by the Chinese Academy of Medical Sciences' Peking Union Medical College, involving numerous esteemed medical institutions across China, has been published in the prestigious journal Nature Communications. The research evaluates the clinical application of a deep learning-assisted liquid-based cytology (LBC) diagnostic model, particularly focusing on its effectiveness in real-world scenarios.

The AI model, developed with core algorithm design and deployment by Guangzhou Anbiping Medical Science & Technology Co., Ltd., was trained on 17,397 full-field LBC slides and validated through a three-phase trial involving 10,826 real cases from multiple centers nationwide. The study demonstrated that the AI system significantly enhances diagnostic efficiency, aids in triage decisions, and strengthens screening capabilities, especially in primary healthcare settings.

Key findings include:

• Improved Diagnostic Consistency: In a multi-reader, multi-case study involving      550 cases, AI assistance notably increased diagnostic agreement among      pathologists and reduced average slide review time from 218 seconds to 30      seconds.
• Enhanced Sensitivity in Community Screening: In community-based screenings      (n=3,001), the AI model achieved a sensitivity of 87.8% for detecting      high-grade lesions (CIN2+), comparable to experienced cytopathologists,      and identified cases that were previously missed.
• Optimized Hospital Screening Efficiency: In opportunistic hospital screenings      (n=1,472), AI support elevated the sensitivity of junior cytopathologists      from 65.7% to 85.7% and specificity from 73.7% to 84.0%, while reducing      unnecessary colposcopy referrals from 28.2% to 19.3%.

The AI system's compatibility with various liquid-based preparation methods and scanning devices ensures its stable operation across different medical institutions and hardware environments.

This advancement signifies a pivotal step toward equitable healthcare by addressing the shortage of skilled pathologists in primary care settings and standardizing diagnostic outputs. Guangzhou Anbiping, as China's first publicly listed company specializing in pathological diagnosis, continues to innovate in tumor screening and diagnosis, with platforms encompassing LBC, immunohistochemistry (IHC), fluorescence in situ hybridization (FISH), and digital pathology. The company is committed to further clinical integration of this AI system, aiming to provide robust technical support for early detection and treatment of gynecological diseases.