People with asymptomatic COVID-19 can spread the disease without the apparent signs of being ill. However, new research suggests that a newly developed AI with sharp algorithmic ears may be able to detect asymptomatic cases from people’s coughing sounds.
MIT’s group of researchers recently developed Artificial intelligence A model that can detect asymptomatic COVID-19 cases by listening to the subtle differences in cough between healthy and infected people. Researchers are currently testing AI in clinical trials and have already begun the process of seeking approval from the Food and Drug Administration (FDA) for use as a screening tool.
This algorithm is based on an earlier model developed by the team to detect conditions such as: pneumonia, asthma And even Alzheimer’s diseaseA memory loss condition that can cause other deterioration of the body, such as weakened vocal cords and respiratory capacity.
Relation: Coronavirus live update
Indeed, researchers have adopted a model of Alzheimer’s disease to detect COVID-19. “Both speaking and coughing sounds are affected by the vocal cords and surrounding organs,” said Brian Svilana, a research scientist and co-author at MIT’s Auto-ID Institute. Said in a statement.. “It’s easy to derive from a fluent speech. AI can easily pick up a person’s gender, native language, and even emotional state from a cough. In fact, the way you cough has emotions embedded in it. . ”
First, they created a website that allows both healthy volunteers and COVID-19 volunteers to record coughs using their mobile phones and computers. They also filled out a survey with questions about their diagnosis and the symptoms they experienced. I was asked to record a “forced cough,” such as a cough that occurs when a doctor tells me to cough while listening to my chest with a stethoscope.
According to the statement, the researchers collected more than 70,000 individual records of forced cough samples through this website. Of these, 2,660 were from patients who had COVID-19 with or without symptoms. Then train the AI model with 4,256 samples and test the model with 1,064 samples to see if it can detect the difference in cough between COVID-19 patients and healthy people. did.
They relate AI to four unique functions of COVID-19 (also used in the Alzheimer’s disease algorithm): muscle deterioration, vocal cord strength, emotions such as suspicion and frustration, respiratory organs, etc. I found that I could detect the difference in cough. Lung performance.
Coughing sound
The AI model correctly identified 98.5% of people with COVID-19 and correctly excluded COVID-19 in 94.2% of people without illness. For asymptomatic people, the model correctly identified 100% of people with COVID-19 and correctly excluded COVID-19 in 83.2% of people without illness.
These are “a fairly promising set of numbers,” and the results are “very interesting,” said Dr. Anthony Rubinski, medical director of respiratory care at NYU Langon Tissue Hospital, who did not participate in the study.
However, “whether this works well in a real environment and can be recommended for use as a screening tool needs further investigation,” Rubinsky told Live Science. In addition, he said further research is needed to ensure that AI accurately assesses cough from people of all ages and ethnicities (the author also mentions this limitation in their paper). doing).
Relation: The most promising COVID-19 vaccine candidate
If a doctor hears a forced cough in an asymptomatic COVID-19 person, they will not hear anything unusual. “It’s not something that the human ear can easily do,” Rubinsky said. Follow-up surveys are definitely needed, but if the software proves to be effective, this AI (there are linked apps if approved) will be COVID-, especially if the tool is cheap and easy. He added to use it, which may be “very helpful” in finding 19 asymptomatic cases.
AI “absolutely” helps curb its spread Pandemic Subirana emailed Live Science to help find people with asymptomatic illnesses. AI can also detect differences between people with other illnesses, such as: influenza And those who have COVID-19, he said, are much better at distinguishing between COVID-19 cases and healthy cases.
The team is currently seeking regulatory approval for apps that incorporate AI models, which could come later next month, he said. According to the paper, they are also testing AI in clinical trials in many hospitals around the world.
And they are not the only ones working on detecting COVID-19 through sound. Similar projects are underway at the University of Cambridge, Carnegie Mellon University, and the British startup Novoic. According to the BBC..
“If the pre-screening tools are always on in the background and are constantly improving, the pandemic can be a thing of the past,” the author writes in a paper. They wrote that those always-listening tools could be smart speakers or smartphones.
This study was partially supported by the pharmaceutical company Takeda Pharmaceutical Company Limited on September 30th. IEEE Open Journal of Engineering in Medicine and Biology..
Originally published in Live Science.