A new trial found that an AI app-based reader for lateral flow devices reduced false Covid negatives in assisted test sites.
Researchers show that machine learning-based apps could be beneficial for diagnostic testing at home.
An artificial intelligence app to read COVID-19 lateral flow tests helped to reduce false results in a new trial out today.
Published in Cell Reports Medicine, a team of researchers from the University of Birmingham, Durham University and Oxford University tested whether a machine learning algorithm could improve the accuracy of results from antigen lateral flow devices for COVID-19.
The LFD AI Consortium team worked at UK Health Security Agency assisted test centres and with health care workers conducting self-testing to trial the AI app. More than 100,000 images were submitted as part of the study, and the team found that the algorithm was able to increase the sensitivity of results, determining between a true positive and false negative, from 92% to 97.6% accuracy.
Professor Andrew Beggs, Professor of Cancer Genetics & Surgery at the University of Birmingham and lead author of the study said:
“The widespread use of antigen lateral flow devices was a significant moment not just during the pandemic, but has also introduced diagnostic testing to many more people in society. One of the drawbacks with LFD testing for Covid, pregnancy and any other future use is the ‘faint line’ question – where we can’t quite tell if it’s a positive or not.
One of the drawbacks with LFD testing for Covid, pregnancy and any other future use is the ‘faint line’ question – where we can’t quite tell if it’s a positive or not.
“The study looked at the feasibility of using machine learning to take the guesswork out of the faint line tests, and we’re pleased to see that the app saw an increase in sensitivity of the tests, reducing the numbers of false negatives. The promise of this type of technology could be used in lots of applications, both to reduce uncertainty about test results and provide a crucial support for visually impaired people.”
Professor Camila Caiado, Professor of Statistics at Durham University and chief statistician on the project, said:
“The increase in sensitivity and overall accuracy is significant and it shows the potential of this app by reducing the number of false negatives and future infections. Crucially, the method can also be easily adapted to the evaluation of other digital readers for lateral flow type devices.”
For media enquiries please contact Tim Mayo, Press Office, University of Birmingham, tel: +44 (0)7920 405 040: email: t.mayo@bham.ac.uk
The University of Birmingham is ranked amongst the world’s top 100 institutions. Its work brings people from across the world to Birmingham, including researchers, teachers and more than 6,500 international students from over 150 countries.
The University of Birmingham is a member of Birmingham Health Partners (BHP), a strategic alliance which transcends organisational boundaries to rapidly translate healthcare research findings into new diagnostics, drugs and devices for patients. Birmingham Health Partners is a strategic alliance between five organisations who collaborate to bring healthcare innovations through to clinical application:
Full citation: The LFD AI Consortium, Caiado, C.C.S., Branigan, M., Lewis-Borman, P., Patel, N., Fowler, T., Dijkstra, A., Chudzik, P., Yousefi, P., Javer, A., Van Meurs, B., Tarassenko, L., Irving, B., Whalley, C., Lal, N., Robbins, H., Leung, E., Lee, L., Banathy, R., Beggs, A.D, Machine learning for determining lateral flow device results for testing of SARS-CoV-2 infection in asymptomatic populations, Cell Reports Medicine (2022), DOI: https://doi.org/10.1016/j.xcrm.2022.100784.
Professor of Cancer Genetics & Surgery
Staff profile for Andrew Beggs, Professor of Cancer Genetics & Surgery, Institute of Cancer and Genomic Sciences, College of Medical and Dental Sciences.
