E-Paper
AI-backed smartphone app can detect Covid-19 in people using voice, claim scientists
According to the researchers, this method can be used in low-income countries where the RT-PCR tests are expensive and difficult to be carried out. The research finding was presented before the European Respiratory Society International Congress in Barcelona, Spain.
Soon, a smartphone app may accurately detect Covid-19 infection in people through their voices by using artificial intelligence (AI). The researchers have claimed that the AI model used in the research is more accurate than the rapid antigen tests, PTI reported. They also claim that the new AI model is cheap, quick and easy to use.
According to the researchers, this method can be used in low-income countries where the RT-PCR tests are expensive and difficult to be carried out. The research finding was presented before the European Respiratory Society International Congress in Barcelona, Spain.
The researchers claim that this AI model is 89 per cent accurate, whereas the accuracy of lateral flow tests varies as per the brand, PTI reported. The scientists also claim that the lateral flow tests are less accurate in detecting Covid-19 virus in people who don't show symptoms.
Wafaa Aljbawi, a researcher at Netherland-based Maastricht University told PTI that the results suggested that simple voice recordings and fine-tuned AI algorithms can ‘potentially achieve high precision in determining which patients have Covid-19 infection’.
"Such tests can be provided at no cost and are simple to interpret. Moreover, they enable remote, virtual testing and have a turnaround time of less than a minute," PTI quoted the scientist.
According to researchers, the new Covid-19 test could be used at the entry points of large gatherings and expedite rapid screening of the population. Aljbawi and her supervisors used data from the University of Cambridge's Covid-19 Sounds app which consists 893 voice samples from over 4,000 healthy and non-healthy participants, out of which 308 had tested positive.
The participants reported some basic information about demographics, medical history and then were asked to record some respiratory sounds on the app installed on the phone.
The sounds include coughing three times, breathing deeply via mouth at least three to five times and reading a short sentence on the screen thrice. A voice analysis technique called Mel-spectogram was used to identify different voice features, PTI reported. Aljbawi said to distinguish the voice of Covid-19 patients from those who were not infected, the team built different AI models and evaluated which one worked best at classifying the Covid-19 cases.
According to scientists, one model called Long-Short Term Memory outperformed the other models. The model is based on the neural networks which mimic the way human brain works and recognises the underlying relationships.
The researchers found that the overall accuracy was 89 per cent and its ability to correctly detect positive cases or sensitivity was 89 per cent. Its ability to correctly identify negative cases or specificity was 83 per cent.
