Validation of AI-assisted ThinPrep® Pap test screening using the Genius^TM Digital Diagnostics System

Area of Science:

• Gynecologic Pathology
• Digital Pathology
• Artificial Intelligence in Medicine

Background:

• Whole-slide imaging and AI offer potential advancements in Pap test screening.
• Limited studies exist on validating AI-based digital systems for clinical Pap test screening.
• This study addresses the need for validation of AI systems in cervical cytology.

Purpose of the Study:

• To validate the Genius™ Digital Diagnostics System for Pap test screening.
• To compare the performance of AI digital evaluation against traditional manual light microscopy.
• To assess diagnostic accuracy and efficiency of AI-assisted Pap smear analysis.

Main Methods:

• Prospective assessment of 319 ThinPrep® Pap test cases.
• Evaluation by six cytologists and three cytopathologists using both light microscopy and digital methods.
• Comparison of results against the original ground truth Pap test diagnosis.

Main Results:

• Significantly higher concordance with ground truth diagnosis using digital vs. manual review across exact (62.1% vs 55.8%) and condensed (76.8% vs 71.5%) Bethesda categories.
• Improved concordance for clinical management categories with digital (71.5%) vs. manual (65.2%) review.
• Shorter evaluation time for digital review (3.2 min) compared to manual review (5.9 min).

Conclusions:

• AI-based digital Pap test evaluation demonstrates improved diagnostic accuracy over light microscopy.
• The AI system significantly reduces screening time per case.
• Participants reported a positive user experience with the AI digital system.