Woman Coughing with Mask On

New AI Technology Can Detect Covid-19 From The Sound Of Your Cough 

One of the biggest concerns that’s been raised surrounding the Covid-19 pandemic regards asymptomatic individuals who unknowingly spread the virus. Now, thanks to newly developed AI technology, healthcare officials may be able to detect asymptomatic cases simply by the sounds of their coughs. 

The technology came from a group of researchers at the Massachusetts Institute of Technology (MIT), who claim that the model can detect asymptomatic cases by listening to subtle differences between a healthy cough and a Covid-19 cough. As of right now the AI technology is still in its clinical trial period, however, the group has already started the process of seeking approval from the US Food and Drug Administration (FDA).

Embed from Getty Images

The algorithm is based on previous AI models that were able to detect lung conditions such as pneumonia and asthma, and it’s even been used in some screenings for Alzheimer’s disease, as that can cause some major strain or degradation on the vocal cords and respiratory system of certain patients. In fact, the specific Alzheimer’s model is the main one that was used to develop the Covid-19 detecting technology. 

One of the studies co-authors, Brian Subirana, who’s also a research scientist in MIT’s Auto-ID Laboratory recently explained in a statement how this technology is able to analyze our vocal cords to make such concrete conclusions. 

Things we easily derive from fluent speech, AI can pick up simply from coughs, including things like the person’s gender, mother tongue or even emotional state. There’s in fact sentiment embedded in how you cough.”

The process began with the group creating a website where volunteers were able to record their coughs using their cell phones or computers. Volunteers were both healthy and infected with Covid-19 so the technology could really hear every little difference. Participants would then fill out survey questionnaires regarding their diagnosis and any symptoms they may be experiencing. 

Embed from Getty Images

People were also asked to record forced coughs, like the one you let out when your doctor tells you to cough so they can listen to your lungs. Researchers have so far gathered over 70,000 individual recordings of forced coughs. Of those recordings 2,660 were from patients with Covid-19, with or without symptoms. After they compiled the recordings they used 4,256 of the samples to train their AI models and 1,064 of the samples to test the model to see whether or not it could detect the difference in a regular cough and a Covid-19 cough. 

They then found that the AI model could tell the differences based on four determining factors: muscular degradation, vocal cord strength, respiratory function and lung performance. So far, the model has been able to correctly identify 98.5% of people with Covid-19 and correctly rule out Covid-19 in 94.2% of people who were healthy and participated in the study. 

In order for this technology to be approved and distributed throughout the nation as a valid form of testing, further research needs to be done. This would mean expanding the sample size of volunteers to include coughs from people of all ages and ethnicities. If the software does prove effective, this technology will likely take the form of a downloadable app on smart devices that would make testing very useful for the remainder of this pandemic. 

The team of researchers are seeking regulatory approval for the app which may come as early as next month, but for now researchers are continuing to perfect the technology so it can be used to its fullest potential and really help curve the spread of Covid-19.