LbpA.I. for Cervical Cytology is an advanced AI-driven system that automatically captures and reconstructs high-resolution, full-slide digital images from liquid-based cytology specimens. Leveraging powerful deep learning algorithms, it performs rapid and comprehensive screening, accurately identifying and localizing a wide spectrum of potentially abnormal epithelial cells, inflammatory cells, and pathogenic microorganisms. By automatically triaging negative cases, LbpA.I. significantly reduces the workload of cytopathologists while maintaining high diagnostic sensitivity.
Clinicians benefit from an intuitive interface that highlights suspicious cells for quick visual verification, streamlining the diagnostic workflow and enabling faster report turnaround. The system is trained on a robust dataset comprising over one million expertly annotated fields of view and cellular images, allowing for the high-precision classification of more than ten clinically significant cell types and microorganisms.
Tailored for the specific demands of cervical cytology, LbpA.I. enhances diagnostic accuracy, supports early detection of cervical lesions, and promotes standardization in cytological evaluation.




Training
Training Dataset: Over 2.5 million annotated fields of vi
AI Clinical Validation
Multi-center Testing: Conducted in 16 major hospitals in 2021
Real-world Sample Testing: Includes cervical cytology and ThinPrep samples
Benchmark Testing: Compared against mainstream cytological workflows and manual slide reviews
Outcome: AI accuracy in detecting LSIL: 99.2%–100%
Quantities and Ratios
NILM : ASC-US : LSIL : ASC-H : HSIL+ (SCC+HSIL+AIS) : AGC+ (ADC+AGC) : Microorganisms = 28:6:5:2:5:1:1
Number of Cases: 23,000
Robust Data & Deep Learning
ew
Cell Annotations: 6.5 million+ cells across 1.5 million+ slides
Detection Capability: Covers 13 major cervical cell types and common microorganisms
Clinical Deployment: System has analyzed over 500,000 real-world cases
Performance Metrics
Metric | Hospital-Based Slides | Cervical Screening Slides |
---|---|---|
Negative Recall (All) | 65%–80% | 96.0% |
Positive Recall (All) | 98.7% | 99.0% |
Positive Recall (LSIL+) | 99.3% | 99.6% |
Positive Recall (ASCH+) | 99.5% | 99.8% |
Key Clinical Findings
LSIL Sensitivity: >99.2%
HSIL Sensitivity: 100%
ASCH+Sensitivity: >96.7%
Specificity (Liquid-Based Cytology): 85%
Specificity (Membrane-Based Cytology): 70%





