Neural-network-assisted analysis and microscopic rescreening in presumed negative cervical cytologic smears. A...

Area of Science:

• Medical technology
• Cytopathology
• Artificial intelligence in healthcare

Background:

• Cervical cancer screening relies on accurate cytological analysis.
• Conventional rescreening of negative smears is a quality control measure.
• False negative results in cervical smears can lead to delayed diagnosis.

Purpose of the Study:

• To compare the effectiveness of neural network-assisted (NNA) review using the PAPNET Testing System versus conventional unassisted rescreening.
• To evaluate the detection yield of abnormalities in cervical smears initially diagnosed as negative.

Main Methods:

• A multicenter clinical trial involving over 10,000 cervical smears.
• Comparison of false negative detection yields between NNA review and conventional microscopic rescreening on a subset of negative smears.
• False negative detection yield defined as the percentage of rescreened negatives reclassified as abnormal.

Main Results:

• NNA review detected a statistically significantly greater proportion of false negative smears (6.2%) compared to conventional rescreening (0.6%).
• This improvement was observed even when controlling for time intervals, post-Clinical Laboratory Improvement Act data, and focusing solely on squamous intraepithelial lesions.

Conclusions:

• Cytologists using NNA review identified a significantly higher rate of previously undetected cervical abnormalities.
• Neural network-assisted review enhances the accuracy of cervical smear analysis, improving the detection of potentially serious conditions.