Using artificial intelligence (AI), a mobile phone app may identify COVID in people’s voices with “possibly high precision,” according to experts.
Users will be needed to provide their medical history, smoking status, and demographic information, as well as record certain respiratory sounds, such as coughing and reading a short text.
An AI model reportedly has an accuracy of 89% and is inexpensive to employ, allowing it to be implemented in low-income nations where PCR testing is more expensive.
According to scientists, the results may be supplied in less than a minute and are a “substantial improvement” over the accuracy of lateral flow testing.
Infection typically affects the upper respiratory tract and vocal cords, therefore researchers decided to use an AI model to detect COVID by analyzing vocal alterations.
Wafaa Aljbawi, a researcher at the Maastricht University Institute of Data Science in the Netherlands, stated: “These encouraging results show that simple speech recordings and refined AI algorithms have the potential to achieve high precision in identifying COVID-19-infected patients.
“These tests can be provided for free and are easy to understand. In addition, they allow for remote virtual testing and have a turnaround time of less than one minute.
“They may be utilized, for instance, at the entry points to major meetings, allowing for speedy population screening.”
The COVID19 Sounds app of the University of Cambridge was utilized to collect data. This was comprised of 893 audio samples from 4,352 healthy and unhealthy individuals.
The user must provide information regarding their medical history, smoking status, and demographics, as well as record some respiratory noises, such as a cough, and read a short text.
Mel-spectrogram is a speech analysis technique that discovered several voice characteristics to “decompose the many qualities of the participants’ voices.”
Ms. Aljbawi added: “These results demonstrate a considerable improvement in the diagnostic accuracy of COVID-19 when compared to cutting-edge procedures such as the lateral flow test.
“The sensitivity of the lateral flow test is approximately 56%, while its specificity is 99.5%. This is significant because it indicates that the lateral flow test incorrectly classifies infected individuals as negative for COVID-19 more frequently than our test.
In other words, the AI LSTM model may miss 11 out of 100 cases that propagated the virus, whereas the lateral flow test would miss 44 out of 100 cases.
The AI model is also being applied to an app that predicts chronic obstructive pulmonary disease exacerbations.
Monday, the research is scheduled to be presented at the International Congress of the European Respiratory Society in Barcelona.
