Seeing beyond positivity: computer vision approach to decoding BinaxNOW COVID-19 tests and forecasting negative...

Area of Science:

• Biomedical Diagnostics
• Medical Imaging Analysis
• Public Health Technology

Background:

• COVID-19 spurred widespread use of lateral flow assays (LFA) for self-diagnosis.
• Limitations of LFAs include subjective interpretation, qualitative results, and lower sensitivity/specificity than PCR.
• Public health decisions increasingly relied on lay interpretation of LFA results.

Purpose of the Study:

• Develop an objective image analysis algorithm for COVID-19 LFA Rapid Antigen Tests (RAT).
• Enable accurate viral load estimation using RAT image analysis.
• Predict the duration of infectiousness post-positive RAT by determining when users test negative.

Main Methods:

• Developed an image analysis algorithm for interpreting RAT results.
• Validated the algorithm's ability to estimate viral load against PCR.
• Assessed the algorithm's predictive accuracy for negative test results over time.

Main Results:

• Viral load estimated within 1.44 Cycle Threshold (CT) of PCR testing.
• Predicted negative test results at least 5 days post-symptom onset with 99% sensitivity, 73% specificity, and 94% accuracy.
• Integrated the algorithm into a user-friendly mobile application.

Conclusions:

• Objective image analysis algorithms for LFAs significantly improve diagnostic accuracy.
• Smartphone-integrated solutions offer a scalable advancement for LFA RAT technologies.
• This technology enhances the reliability of self-administered diagnostic tests for various diseases.