قالب وردپرس درنا توس
Home / Health / The MIT Group cough detector identifies 97% of COVID-19 cases even in asymptomatic people

The MIT Group cough detector identifies 97% of COVID-19 cases even in asymptomatic people



Part of the challenge in controlling the coronavirus pandemic is identifying and isolating infected people quickly – especially not so easily when COVID-19 symptoms are not always noticeable, particularly especially soon. Now scientists have developed a new model of artificial intelligence that can detect a virus from a simple forced cough.

Evidence suggests that AI can detect differences in coughing sounds that the human ear cannot hear, and if the detection system could be integrated into a device like a smartphone, the team thinks it can be a useful early screening tool.

Construction based on research has been discovered Alzheimer’s disease through coughing and speaking. As the pandemic began to spread, the team turned their attention to COVID-1

9, exploring what had been learned about how illness could cause very small changes to speech and other noises. that we create.

Research scientist Brian Subirana from Massachusetts Institute of Technology (MIT) said: “Sounds when speaking and coughing are both affected by the larynx and surrounding organs.

“This means that when you talk, your speaking part is like coughing, and vice versa.”

“That also means things we easily gain from our ability to speak fluently, AI can recognize simply from coughing spells, including things like gender, mother tongue or even state of mind. The person’s emotional attitude. Actually emotion is tied to the way you cough. “

Research in Alzheimer’s disease was rebranding for COVID-19 involving a neural network called ResNet50. It is trained per thousand hours of human talk, then on a dataset of words spoken in different emotional states, and then on a cough database for detection. changes in lung and respiratory activity.

When three models are combined, a noise layer is used to filter out the stronger cough noise from the weaker models. Out of about 2,500 recorded cough records of people confirmed to have COVID-19, AI correctly identified 97.1% of them – and 100% of cases were asymptomatic.

It’s an impressive result, but much more remains to be done. Researchers emphasize that its main value lies in detecting the difference between a healthy cough and a non-healing one in asymptomatic people – not in the actual diagnosis of COVID-19, a Appropriate testing will be required. In other words, it’s an early warning system.

“The effective implementation of this group diagnostic tool can reduce the spread of the pandemic if people use it before going to the classroom, factory or restaurant,” Subirana said.

The fact that the test is non-invasive, virtually free to run and applied quickly adds to its potential usefulness – while it’s not designed for diagnose People with COVID-19 already have symptoms that can tell you if you should be isolated and properly tested when there are no major signs of the virus.

The researchers now want to test the engine on a more diverse set of data and see if there are other factors involved in achieving such an impressively high rate of detection. If it does get into the phone application phase, there will obviously also be privacy implications, as few of us want our devices to constantly hear signs of poor health.

Once we start to push back the pandemic coronavirus behind, new research could help bring back the study of cough and the detection of Alzheimer’s disease. The data shows that neural networks only require a little bit of adjustment to fit each condition.

“Our study found a striking similarity between Alzheimer’s disease and COVID,” the researchers wrote in their publication.

“Identical markers can be used as a distinguishing tool for both, suggesting that perhaps, in addition to temperature, pressure, or circuit, there are some higher level biomarkers there are.” fully diagnose conditions between specialists that were thought to be almost disconnected. “

Research has been published above IEEE Open Journal of Engineering in Medicine and Biology.


Source link