Interactive neural-network-assisted screening. A clinical assessment

Area of Science:

• Medical technology
• Artificial intelligence in healthcare
• Gynecologic oncology

Background:

• Cervical cancer screening relies on accurate detection of epithelial abnormalities.
• Current screening methods have limitations in sensitivity.
• Interactive, neural network-assisted (INNA) systems offer potential improvements.

Purpose of the Study:

• To evaluate the clinical utility and effectiveness of INNA systems in cervical screening.
• To compile and analyze data from existing clinical studies on INNA screening.
• To provide quantitative metrics for INNA system performance.

Main Methods:

• Systematic review and meta-analysis of published and unpublished clinical studies.
• Development of a taxonomy to classify study results.
• Data pooling based on study design and effectiveness metrics.

Main Results:

• INNA sensitivity estimates range from 89% to 100%.
• Relative yield metrics indicate 8% to 49% improvement over unassisted screening.
• INNA screening showed 18-40% increase in abnormality yield in primary screening modes.

Conclusions:

• A substantial evidence base supports the use of INNA screening.
• INNA screening demonstrates superior sensitivity for cervical epithelial abnormalities compared to unassisted methods.
• INNA systems can be effectively used in both substitutive and augmentative roles.