- One recent study found that an artificial intelligence model can tell the difference between a regular cough and a cough caused by COVID-19 – even coughs caused by an asymptomatic carrier.
- AI has identified 98.5% of coughs from people infected with coronavirus, with no symptoms or other cases.
- The researchers hope to incorporate this model into one app, as well as smart speakers and cell phones, as a real-time screening tool.
- Visit the Business Insider homepage for more stories.
At least one in five people with coronavirus will have no symptoms and may unknowingly pass the virus on to others. People who do not feel sick and have not been informed of the exposure cannot know that they should get tested.
But researchers at the Massachusetts Institute of Technology may have found a way to identify these silent coronavirus carriers without testing.
A study published in September describes an artificial intelligence model that can differentiate coughing spells of people infected with coronavirus and healthy people. It can even tell from the involuntary coughs, which force whether the person is healthy or the carrier is asymptomatic, based on subtle vocal variations that cannot be distinguished by the human ear. .
The MIT model accurately found that 98.5% of people in the study were infected with coronavirus based on their coughs, including asymptomatic carriers. In fact, the program failed to misdiagnose a cough from an asymptomatic carrier.
“We think this shows how sound changes when you have COVID, even when you don̵7;t have symptoms,” said Brian Subirana, a scientist and co-author of the study.
Subirana and his colleagues think their AI can be integrated into smart speakers and cell phones.
“A pandemic could be a thing of the past if pre-screening tools are always enabled in the background and constantly improving,” they wrote.
‘Emotions are tied to the way you cough’
Before the pandemic, researchers trained AI models to detect other diseases like Alzheimer’s and pneumonia based on people’s coughs. This could happen because the way we speak and cough can reflect the strength of the vocal cords and the organs around them, according to Subirana. For example, patients with Alzheimer’s disease tend to have weaker vocal cords due to neuromuscular deterioration.
“Things we easily get from our ability to speak fluently, AI can recognize simply from coughing spells, including things like gender, mother tongue or even emotional states,” he said. emotions. There are actually emotions attached to the way you cough.
Therefore, Subirana’s team applied previous models to patients with COVID-19.
They collected 70,000 sound samples of coughs from both healthy and infected people, and asked the second group to report any coronavirus symptoms or deficiency. More than 2,600 recordings were submitted by people who tested positive for COVID-19.
The researchers then showed their artificial intelligence model about 4,250 recordings, including those from the infected people. AI was able to build a picture of sick and healthy coughs. It identifies specific patterns of vocal cords strength, lung activity, emotions and muscle deterioration in leopard dogs only in patients with COVID-19.
When the model was ready, Subirana’s team then listened for more than 1,000 coughs. It identifies 100% of all coughs belonging to asymptomatic coronavirus carriers.
A tool used ‘before going to a classroom, factory or restaurant’
Subirana’s team is working to incorporate the AI model into one free application, and hopes to get permission from the Food and Drug Administration to use the model as a COVID-19 pre-screening tool. .
Theoretically, a person could cough into their phone then immediately find out if they could be an asymptomatic carrier of the coronavirus. However, they still need tests to confirm the diagnosis.
This kind of pre-screening tool “can reduce the spread of a pandemic if people use it before going to a classroom, factory or restaurant,” Subirana said.
However, the AI cannot determine if your cough is caused by another illness, specifically the flu or a cold. It is calibrated only to detect COVID-19. It also cannot tell if you have a fever or sore throat caused by COVID-19.
According to Anthony Lubinsky, director of respiratory care at Langone Tisch Hospital, New York University, the results are very encouraging.
But he told Live Science that “whether this works well enough in a real-world context to recommend its use as a screening tool will need more research.”
The study authors said they collaborated with hospitals in the US, Mexico and Italy to collect additional cough recordings to improve and test their models.