Detecting COVID-19 just by analyzing the way you cough

An MIT team developed an AI algorithm detecting COVID-19 just by analyzing the sound of your cough. With an overall accuracy of 98% and detecting 100% of asymptomatic patients, this app. could change the way we test for COVID-19.

MIT Covid-19 Coughing AI algorithm

Medical Chatbots are used extensively to check for COVID-19. Their accuracy is still highly debatable. Jordi Laguarta et al., from the Massachusetts Institute of Technology, Auto-ID laboratory, just published an article in the Journal of Engineering in Medicine and Biology. If confirmed, their Artificial Intelligence-based app. could lead to a revolution in the way Covid-19 is detected.

The second wave of COVID-19 outbreak is especially difficult to contain partly due to the lack of efficient and convenient viral or serology tests. Testing an entire population on a daily is also practically and financially unrealistic. New methodologies are needed to reduce the dramatic economic and human costs of the second wave of infection we are currently facing.

Real-time Artificial Intelligence detection algorithms may not be the ultimate answer but could prioritize standard testing access, especially in asymptomatic patients. By analyzing the cough’s sound, the diagnostic algorithm developed by MIT researchers can detect COVID-19 with 98.5% accuracy and 94.2% specificity. The asymptomatic detection rate was 100% and 88% accuracy on all subjects.

To achieve such impressive results, the model was trained on tens of thousands of volunteers who uploaded samples of their coughs, as well as spoken words on a dedicated website.

MIT’s Auto-ID laboratory, headed by Brian Subirana, is not new to using coughs and Artificial Intelligence to detect diseases. They already analyzed diseases such as pneumonia, asthma, or Alzheimer’s disease using such methodologies. The common point between COVID-19 and these conditions is that they will induce temporary or permanent neuromuscular weaknesses of the muscles surrounding the vocal cords.

Interestingly even asymptomatic Covid-19 carriers experienced a modification of the sound they produce when coughing. A human ear can not detect the differences in the sound produced. The Deep-learning algorithm can somehow detect a shift in the sound produced and achieve great results.

The test is non-invasive, delivers results in real-time, has zero variable cost, can be used to monitor the patients longitudinally, and is accessible to anyone. It carries a lot of potential for quick screening in Schools, workplaces, or even restaurants.

MIT’s’ researchers next steps are to develop an FDA approved easy-to-use app for iOS and Android that will allow convenient self-testing. The model is currently being improved in collaboration with Mount Sinai and White Planes Hospitals in the US, Catalan Health Institute in Catalonia, Hospitales Civiles de Guadalajara in Mexico, and Ospedale Luigi Sacco in Italy.

This A.I. powered model has the ability to open new doors in the way Covid-19 is being tested. By detecting Covid-19 early, especially in asymptomatic patients, this model could be a game-changer for limiting the virus’s spread. We are impatiently waiting to test the app. to know more about the potential of this breakthrough technology.

Was this article helpful?