Quickly and effectively discovering who is infected with covid-19, even when asymptomatic, is a priority. Now, MIT could have found a key
Ending covid-19 remains the main mission of thousands of scientists around the planet. During the last months, numerous investigations have been carried out with the aim of trying to understand how a virus spreads relatively easily and, especially, when the person has not yet shown symptoms – or even when they never show them. Finding a remedy that allows the improvement of a patient or a vaccine that can eradicate the virus are the main objectives, although not the only ones: the latest tool developed by MIT can help a return to normality as soon as possible.
The Massachusetts Institute of Technology (MIT) is developing a series of neural networks to help health authorities control the pandemic. The latest development they have carried out is really impressive and seems almost like science fiction, but the reality is that its high degree of effectiveness makes it a fundamental element: artificial intelligence that can detect if a person has covid-19 by person’s coughing.
The main problem with the Coronavirus is, in reality, its ability to spread when the patient is not yet aware that they have contracted it. In fact, one of the main mysteries around it is not only why it is contagious when someone is presymptomatic, but especially why there are asymptomatic people who do not suffer from any physical problem but who are the main vector of contagion. How could someone who does not have any symptoms find out that they are actually sick and can infect more people? That is the main problem for the spread of covid-19 that now this tool can help stop.
Actually, its operation is very simple. First, it has a neural network that is responsible for measuring the sounds associated with the strength of the vocal cords; then another neural network is in charge of detecting signals related to the subject’s state of mind, being able to interpret joy, anger, sadness… Finally, another third network is in charge of interpreting very subtle changes in the performance of our lungs. The three models are combined through an algorithm and this determines if there is any type of muscle degradation. If so, the person is infected with covid-19.
It might sound too futuristic, but the truth is that the tests they have carried out have obtained really impressive results. About 70,000 recordings of different people, with different coughs, have been used, equivalent to a total of 200,000 audio samples. After using this AI, the researchers were able to recognize COVID-19 sensitivity of 98.5% with a specificity of 94.2% (AUC: 0.97); and, what may be even more important, they managed to discover 100% of asymptomatic people.
This MIT study, developed by Jordi Laguarta, Ferran Hueto and Brian Subirana, and published in the ‘IEEE Journal of Engineering a Medicine and Biology’, confirms the great results of this impressive tool, although the researchers emphasize that the use of this guide should never substitute for a PCR test or an antigen test. It must be an auxiliary element that helps medical services to determine if a person has contracted Covid-19 or if, on the contrary, they are healthy.
“The effective implementation of this group diagnostic tool could diminish the spread of the pandemic if everyone uses it before going to a classroom, a factory, or a restaurant,” Subirana said in a statement. Although it may be imperceptible to the human ear, our voice can hold many secrets, even if we are sick with covid-19. A tool that can help stop the spread of the great pandemic of the 21st century.