It is a development of the Catalan researcher Brian Subirana and the Massachusetts Institute of Technology (MIT).
The coronavirus pandemic brought with it several applications that perform self-diagnostics, as in the case of Caring, but none so far capable of analyzing a person’s cough to determine the diagnosis ofe asymptomatic with Covid-19.
The Artificial Intelligence (AI) model developed by the team of Catalan researcher Brian Subirana, director of the Auto-ID Lab at MIT, started training in April with forceful coughing sounds sent directly by thousands of volunteers through the Internet, in which they had to specify their symptoms, their language and if they were finally diagnosed with coronavirus.
Using this data as training, AI is able to differentiate with 98.5 percent accuracy which coughs belong to healthy people and which ones are from asymptomatic Covid patients, as the MIT reported in a statement.
The researchers are implementing their tool in a mobile phone app that people can record their cough and check if they are likely to be asymptomatic positives, recommending in that case requesting conventional tests.
“The effective implementation of this diagnostic tool could decrease the spread of the pandemic if everyone uses it before going to class, the factory or a restaurant, “said Subirana, who had Jordi Laguarta and Ferran Hueto as co-authors.
The idea of using an AI to analyze the cough of potential asymptomatic patients comes from the team’s previous research, which used voice recordings to detect Alzheimer’s symptoms and that determined that the state of the vocal cords, the attitude of the speaker, respiration and muscle breakdown are biomarkers of degenerative disease.
The researchers used these markers to analyze 200,000 recordings of imposed coughs of 70,000 volunteers, of which 2,500 were infected with Covid-19, many of them without symptoms.
Analysis of the databases revealed “a striking similarity” between the results with asymptomatic Covid patients and those with Alzheimer’s symptoms.
“The way we produce sounds changes when we have Covid, even without symptoms,” Subirana concluded.
Previously, a group of researchers from the Federal Polytechnic School of Lausanne (Switzerland) revealed in April the development of an application that could also determine whether or not a person suffers from the disease after recording and analyzing your cough.
The app in question, called “coughvid” (a play on words between Covid and “cough”, which means “cough” in English) can be used in its version to internet browsers. Stakeholders, including Covid-19 patients, can record their coughs to help create a “database” that will use Artificial Intelligence (AI) to make diagnoses.
The application sought “to be an alternative to conventional tests,” said Atienza, who explained that the results will be known immediately and will have a reliability index of 70 percent.
At the moment, “coughvid” has no confirmed release date.