Smart diagnostics: Leveraging artificial intelligence for accurate at-home COVID-19 testing

Area of Science:

• Medical Diagnostics
• Artificial Intelligence in Healthcare
• Infectious Disease Management

Background:

• Reverse transcriptase polymerase chain reaction (RT-PCR) is the gold standard for COVID-19 diagnosis but is slow and expensive.
• Rapid antigen tests offer a faster, more accessible alternative but can suffer from reduced diagnostic performance due to subjective visual interpretation.
• Improving the accuracy of at-home COVID-19 testing is crucial for timely diagnosis and public health management.

Purpose of the Study:

• To evaluate the performance of artificial intelligence (AI) in interpreting at-home COVID-19 lateral flow antigen tests.
• To compare AI-assisted interpretation against traditional visual interpretation of antigen tests.
• To determine if AI can enhance the diagnostic accuracy of rapid antigen tests for COVID-19.

Main Methods:

• A study involving 99 participants from COVID-19 treatment trials who performed at-home antigen tests.
• Results from lateral flow antigen tests were interpreted independently by an AI model and through visual review.
• Performance metrics (accuracy, sensitivity, specificity) were compared against RT-PCR as the reference standard.

Main Results:

• AI-assisted interpretation achieved 96.97% accuracy, 80% sensitivity, and 100% specificity compared to RT-PCR.
• Visual interpretation yielded 68.82% accuracy, 78.57% sensitivity, and 67.09% specificity.
• A statistically significant improvement in diagnostic accuracy was observed with AI-assisted interpretation (p < 0.0001).

Conclusions:

• AI-based interpretation significantly enhances the accuracy and specificity of at-home COVID-19 antigen tests.
• AI offers a more reliable method for interpreting rapid antigen test results than visual review.
• The integration of AI tools holds promise for improving the overall reliability and effectiveness of point-of-care COVID-19 diagnostics.