Detecting coronavirus from asymptomatic is no easy task. Luckily, researchers from the Massachusetts Institute of Technology (MIT) have found a way to detect them through cellphone-recorded coughs.
A recent paper published in the IEEE Journal of Engineering in Medicine and Biology reports an artificial intelligence (AI) model that lets people voluntarily submit their recorded coughs via desktop and mobile.
MIT News reported that MIT's research scientists Brian Subirana, Jordi Laguarta, and Ferran Hueto are the paper's co-authors.
Subrina hopes that the effective implementation of such a tool could eradicate and flatten the spread of the virus if used in public places, such as classrooms, factories, and restaurants.
How Does It Work?
A user simply coughs into their device, and then the model gives accurate feedback to find out whether they're asymptomatic to Covid-19.
To test the model, researchers have dug over 70,000 cough samples, and spoken words, 2,500 of them were from those who had confirmed with Covid-19.
MIT News reveals that it accurately identified almost 99% of coughs from infected people. The AI reaches 100% perfection on asymptomatic samples - those who show no symptoms but tested positive.
However, the team is still gearing up to launch a user-friendly and FDA-approved app as of this writing. Whether it's on the desktop or mobile, the app aims for free, convenient, and mass-scale use.
Supported by Takeda Pharmaceutical Company Limited, researchers have been working in the same fashion to detect asthma and pneumonia lung infection. Vocal cord strength, sentiment, respiratory performance, and muscular degradation are what the AI sees.
Subirana stresses a 'striking similarity' in voice recognition between Alzheimer's and Covid-19 patients. The technology similarly picks up gender information, mother tongue, or even their emotional state. The tool is empowered to discern asymptomatic coughs from healthy coughs. Several hospitals have partnered with the team to provide a much larger and more diverse sample to train its accuracy.
"So we thought, why don't we try these Alzheimer's biomarkers (to see if they're relevant) for Covid," said Subirana.
Read the full report here.
Coronavirus in Numbers
Since the earliest symptom recorded back on December 1, 2019, SARS-CoV-2 has killed over 1,188,141 lives, while 33,058,492 others are lucky enough to recover.
Up to this writing (30/10), worldwide's total number has reached over 45,464,208 cases. The USA is leading with 9 million patients, followed by India and Brazil with, respectively, 8 million and 5 million.
A vaccine could be a long way to go, so in the meantime, researchers try to find a way to somehow flatten the curve by detecting the undetected. This not-so-new AI could be a light in the dark.