Intercenter Study of Interobserver Variability With Optical and Digital Imaging in Cervicovaginal Cytology....

Area of Science:

• Cytopathology
• Medical Diagnostics
• Artificial Intelligence in Medicine

Background:

• Accreditation of cytology laboratories mandates participation in diagnostic intercomparison programs.
• Cervico vaginal cytology (CVC) intercomparison programs reveal significant diagnostic discrepancies.
• Investigating artificial intelligence (IA) systems to enhance diagnostic concordance in CVC is crucial.

Purpose of the Study:

• To evaluate the diagnostic concordance of IA systems compared to traditional light microscopy in CVC.
• To assess if IA systems improve diagnostic agreement in simulated intercomparison programs.

Main Methods:

• A simulated intercomparison program involved four observers analyzing 30 CVC cases.
• Cases included negative diagnoses, ASCUS, LSIL, and HSIL, examined via light optic microscopy (OM) and two IA systems (Genius G1 and G2).

Main Results:

• Diagnostic agreement between OM and IA systems was comparable (82% for G1, 78% for G2).
• Both IA systems demonstrated a higher affinity for diagnosing glandular lesions not identified by OM.
• Kappa values indicated similar agreement across all three diagnostic methods.

Conclusions:

• Acceptable diagnostic concordance was achieved with both OM and IA systems.
• IA systems show a notable tendency to identify glandular lesions.
• Potential differences between IA devices may necessitate calibration adjustments.