AI can diagnose asymptomatic covid-19 patients using cellphone-recorded coughs

OIP (1)

An artificial intelligence (AI) model has been developed by scientists that they claim can differentiate asymptomatic Covid-19 patients from healthy people via their cough recordings and display the results on a smartphone app.

The team identified an AI model that differentiates asymptomatic individuals from healthy individuals via forced-cough recordings in a paper published in the IEEE Journal of Engineering in Medicine and Biology.They said that these recordings were submitted voluntarily by individuals via web browsers and computers such as cellphones and laptops.

The researchers educated the model on tens of thousands of cough samples, as well as words spoken. When they fed new cough recordings to the model, 98.5 percent of coughs from individuals who were reported to have Covid-19 were correctly classified.
This included 100% of asymptomatic coughs, those who according to the researchers, indicated that they had no symptoms but had tested positive for the virus.

The team is working on integrating the model into a user-friendly app that could potentially be a free, simple non-invasive pre-screening tool to identify individuals who are likely to be asymptomatic of Covid-19 if accepted and implemented on a large scale. A user may log in on a regular basis, cough into their phone, and receive information immediately about whether they could be contaminated and should therefore confirm with a formal test.

“If everyone uses it before going to a classroom, a factory, or a restaurant, effective implementation of this group diagnostic tool could decrease the spread of the pandemic,” said Brian Subirana, a research scientist at the Auto-ID Laboratory of MIT.

Similarly, to study forced-cough recordings, the MIT team developed AI models to see if they could detect symptoms of Alzheimer’s, a condition associated with not only memory loss, but also neuromuscular deterioration such as impaired vocal cords.

In April, the team set out to collect as many recordings of coughs as they could, including those from Covid-19 patients. They established a website where people can record a series of coughs, through a cellphone or other web-enabled device.

Participants also fill out a survey of symptoms they are experiencing, whether or not they have Covid-19, and whether they were diagnosed through an official test, by a doctor’s assessment of their symptoms, or if they self-diagnosed. They also can note their gender, geographical location, and native language.

Till date, more than 70,000 recordings have been collected by the researchers, each containing several coughs, amounting to some 200,000 audio samples of forced cough, which is the largest dataset of research cough.
In order to balance the dataset, the team used 2,500 Covid-associated recordings, along with 2,500 more recordings that they randomly picked from the array.
4,000 of these samples were used by them to train the AI model. The remaining 1,000 recordings were then fed into the model to see if it could distinguish coughs reliably from patients with Covid and healthy people.

The researchers were able to pick up patterns that are unique to Covid-19 in the four biomarkers—vocal cord intensity, sentiment, lung and respiratory efficiency, and muscular degradation.4,000 of these samples were used by them to train the AI model. The remaining 1,000 recordings were then fed into the model to see if it could distinguish coughs reliably from patients with Covid and healthy people.

The model identified 98.5 per cent of coughs from people confirmed with Covid-19, and of those, it accurately detected all of the asymptomatic coughs.

.

3 Likes