Dutch researchers report they could detect a COVID-19 infection with better sensitivity than antigen tests by a mobile app that uses AI to analyze voice recordings.
At the International Congress of the European Respiratory Society, Dutch researcher Wafaa Aljbawi from Maastricht University reported that using the University of Cambridge’s COVID-19 Sounds app, a COVID-19 infection was detected with a higher sensitivity (84%) but lower specificity (83 %) than by using antigen tests (sensitivity 56.2%, specificity 99.5%). The goal of the researchers is to make a test for fast COVID-19 diagnosis in countries with limited medical resources available.
The app was cheaper and easier to use than antigen test, said Aljbawi but it was not sure that the deep learning algorithm used to train the app can distinguish between COVID-19 and other respiratory diseases.
Further analysis has given that a sensitivity of up to 89% is possible. However, the accuracy of antigen tests varies greatly depending on the brand and from whether COVID-19 patients showed symptoms.
Infection with COVID-19 leads to changes in the patient’s voice. Thus, the research group used the University of Cambridge’s COVID-19 Sounds app, which contains 893 voice samples from 4,352 healthy and non-healthy participants – 308 of whom had tested positive for COVID-19.
.