[ad_1]
Part of the challenge in controlling the coronavirus pandemic is quickly identifying and isolating infected people, which is not particularly easy when COVID-19 symptoms are not always noticeable, especially early on. Now scientists have developed a new artificial intelligence model that can detect the virus from a simple forced cough.
Evidence shows that AI can detect differences in cough that cannot be heard with the human ear, and if the detection system can be incorporated into a device such as a smartphone, the research team believes it could become a useful one. early detection tool.
The work builds on research already being done on detecting Alzheimer’s disease through coughing and speaking. Once the pandemic began to spread, the team turned their attention to COVID-19, drawing on what had already been learned about how the disease can cause very small changes in speech and the other noises we make.
“The sounds of speaking and coughing are influenced by the vocal cords and surrounding organs,” says research scientist Brian Subirana of the Massachusetts Institute of Technology (MIT).
“This means that when you speak, part of your conversation is like coughing and vice versa.”
“It also means that the things that we get easily from fluent speech, the AI can pick up just by coughing, including things like gender, mother tongue, or even the person’s emotional state. In fact, there is a feeling embedded in the way what coughs “.
The repurposed Alzheimer’s research for COVID-19 involved a neural network known as ResNet50. He trained on a thousand hours of human speech, then on a data set of words spoken in different emotional states, and then on a cough database to detect changes in respiratory and lung performance.
When the three models were combined, a noise layer was used to filter out the loudest coughs from the weakest. In around 2,500 cough recordings captured from people confirmed with COVID-19, the AI correctly identified 97.1 percent of them and 100 percent of asymptomatic cases.
It’s an impressive result, but there is still a lot of work to be done. The researchers emphasize that its main value lies in detecting the difference between healthy cough and unhealthy cough in asymptomatic people, not in the actual diagnosis of COVID-19, for which an adequate test would be required. In other words, it is an early warning system.
“The effective implementation of this group diagnostic tool could decrease the spread of the pandemic if everyone uses it before going to a classroom, a factory or a restaurant,” says Subirana.
The fact that the test is non-invasive, virtually free, and quick to apply increases its potential utility, although it is not designed to to diagnose people with COVID-19 who are already showing symptoms, it could tell you whether to isolate yourself and get properly tested when no significant signs of the virus are showing.
The researchers now want to test the engine on a more diverse data set and see if other factors are involved in achieving such an impressive detection rate. If you get to the phone app stage, there will obviously be privacy implications too, as few of us will want our devices to be constantly listening for signs of poor health.
Once we begin to move past the coronavirus pandemic, the new research could help feed back into the study of cough and Alzheimer’s screening. The data shows that neural networks only required slight adjustments to suit each condition.
“Our research uncovers a striking similarity between Alzheimer’s discrimination and COVID,” the researchers write in their published article.
“The exact same biomarkers can be used as a discrimination tool for both, suggesting that perhaps, in addition to temperature, pressure or pulse, there are some higher-level biomarkers that can sufficiently diagnose conditions in specialties than before. they were thought to be disconnected. “
The research has been published in the IEEE Open Journal of Engineering in Medicine and Biology.
