The tool was able to detect nearly 99% of COVID-19 infections using thousands of cough recordings and 100% of asymptomatic cases, per MIT.
One of the myriad challenges associated with mitigating the spread of COVID-19 involves the high number of asymptomatic patients who are infected but do not demonstrate noticeable symptoms. While these individuals may not show physical symptoms of the disease, they are still capable of transmitting COVID-19 to others. To assist with this dire conundrum, MIT researchers have developed an algorithm that could help identify asymptomatic COVID-19 cases using a smartphone and a recording of their cough.
SEE: TechRepublic Premium editorial calendar: IT policies, checklists, toolkits, and research for download (TechRepublic Premium)
“The effective implementation of this group diagnostic tool could diminish the spread of the pandemic if everyone uses it before going to a classroom, a factory, or a restaurant,” said co-author Brian Subirana, a research scientist in MIT’s Auto-ID Laboratory, in an MIT report.
On Thursday, MIT released a report detailing a recent study published in the IEEE Journal of Engineering in Medicine and Biology. The report illustrates the research team’s previous research efforts, the model training process as well as a potential app to help mitigate the spread of COVID-19.
The MIT research teams’ efforts to use AI models to detect signs of disease in cough recordings began before the onset of the coronavirus pandemic. Initially, the researchers were using AI models to analyze recordings of forced coughs to detect Alzheimer’s diseases using AI models.
The team eventually developed an AI framework and the results demonstrated that “together, vocal cord strength, sentiment, lung and respiratory performance, and muscular degradation were effective biomarkers for diagnosing the disease.”
At the onset of the coronavirus pandemic, Subirana “wondered” about the possibility of using the team’s Alzeheimer’s detection framework to help diagnose COVID-19 cases, due to evidence of similar neurological symptoms experienced by COVID-19 patients, according to MIT.
The research team trained the COVID-19 detection algorithm using a combination of spoken words and “tens of thousands” of cough recordings. When the model was fed recordings of new coughs, the algorithm was able to identify nearly 98.5% of coughs of confirmed COVID-19 infections. This includes 100% of coughs associated with asymptomatic COVID-19 cases. (The asymptomatic cases had not experienced symptoms, but did test positive for COVID-19, per MIT.)
SEE: Artificial intelligence is predicting coronavirus outbreaks before they start (TechRepublic)
The researchers are attempting to incorporate the algorithm into an easy-to-use app. Pending FDA-approval and subsequent adoption at-scale, this app “could potentially be a free, convenient, noninvasive prescreening tool to identify people who are likely to be asymptomatic for Covid-19,” per MIT.