Technologies that offer to diagnose Covid-19 by coughing into a smartphone are no more effective at diagnosing the disease than reporting key symptoms, according to a study by The Alan Turing Institute, the UK Health Security Agency (UKHSA) and the Royal Statistical Society (RSS).
The organisations have published three preprint papers on the study – on the vocal audio dataset, the machine learning element and audio based AI classifiers – which focused on the use of AI to detect Covid-19 by analysing smartphone recordings of the sound of people coughing.
The researchers used technology that aimed to identify a signature noise, known as a biomarker, in a person’s cough that could be linked to a positive Covid-19 PCR test result.
They found that analysing the recordings was no more effective than attempting to diagnose the virus through reporting symptoms such as a cough of loss of smell, with symptoms being present in positive cases in the general population around 65% of the time.
Researchers and policymakers had hoped that this technology could be used to identify potential Covid-19 cases, reducing the need for testing which can be costly and generate a lot of waste.
Participants who had taken a PCR test in the previous 72 hours were asked to submit a barcode they have received from the NHS. Using a mobile phone, tablet or computer, they then recorded their breathing, a cough and read a sentence.
The researchers used recordings from nearly 73,000 people who also had a PCR test. Around one third, or nearly 24,000, were positive.
The study recruited participants between June 2020 and March 2022 from both the Government’s Test and Trace programme and REACT, a population surveillance study in England that examined the prevalence of Covid-19.
Disappointment
Professor Chris Holmes, lead author and programme director for health and medical science at The Alan Turing Institute, said: “Finding new ways to quickly and easily diagnose viruses like Covid-19 is really important to stop its spread. While it’s disappointing that this technology doesn’t work for Covid-19, it may still work for other respiratory viruses in the future."
Stian Westlake, chief executive of the Royal Statistical Society, said: “Careful research isn’t just about discovering wondrous new technologies – it can also tell us when promising ideas don’t live up to expectations. This comprehensive study sheds important light on what audio analysis technology can and can’t do when it comes to diagnosing Covid.”
