[ad_1]
How artificial intelligence should help with coronary diagnosis. Scientists from the Massachusetts Institute of Technology presented groundbreaking research in the shadow of the US election.
Icon image. Photo: pixabay / regionalHeute.de
06.11.2020
Boston. As the US election hit the headlines, researchers from the Massachusetts Institute of Technology (MIT) unveiled a groundbreaking research result. Their AI-based program should be able to recognize corona-infected people based on their cough, even if people who cough otherwise don’t have any symptoms. According to the researchers, this should only work with audio recordings.
Read also: Just a cold or Corona? A new study explains the differentiating factors
The accuracy of the process is phenomenal. The researchers trained the program with tens of thousands of audio recordings of people coughing. As a result, the software reliably identified 98.5% of those who had already tested positive for the coronavirus. Healthy subjects even correctly recognized their intelligence at 100%. The audio recordings came from volunteers broadcasting their coughs via an internet platform.
Coughing can look so different:
Podcast: Marvin König / Sound Samples courtesy of the researchers / MIT
Changes in the body even without symptoms
People infected with Corona without any symptoms of the disease have caused virologists headaches since the start of the pandemic. You can pass the virus to others without getting sick and thus represent an almost unpredictable variable in the fight against the spread of the infection.But even with otherwise asymptomatic carriers of a corona infection, the research results of MIT scientists occur after the changes. As a result, even a forced cough will sound different from that of an uninfected person. The difference is not audible to the human ear. However, artificial intelligence appears to be able to detect these differences.
The research team is currently working on integrating the AI model into a user-friendly smartphone app. If this is approved by the authorities, according to the researchers, artificial intelligence in smartphones could represent an effective and free tool to help determine a coronavirus infection. Conventional PCR tests shouldn’t be replaced by this, the app should help you a lot more in deciding if a PCR test is really needed.
“Effective implementation of a diagnostic tool like this could significantly reduce the spread of the pandemic if everyone used it before going to class, factory or restaurant,” said co-author Brian Subirana, a researcher at MIT.
Alzheimer’s and Corona have similarities
However, the researchers aren’t really breaking new ground with their work – even before the outbreak of the pandemic, training algorithms were being worked on to detect pneumonia and asthma, even using smartphone-recorded coughs. Similarly, a team from MIT has developed a method that can be used to identify possible signs of Alzheimer’s disease, including through coughing. Symptoms of this disease involve not only neurological failures, but also a change in the vocal cords due to muscle weakness. The model performed well in this area. Based on these research findings, Subirana wondered early in the pandemic whether the AI model could not reliably identify a COVID-19 disease as well. “The sounds of spoken words and coughing are both affected by the vocal cords and the organs that surround them. That is, when someone speaks, part of the speech sounds like coughing and vice versa. AI can do things like gender only by cough of a person a person who can recognize their mother tongue or even their emotional state “, explains the scientist and notes:” There is actually a lot of sensation in the cough. Surprisingly, however, the researchers concluded that the neurological symptoms of Alzheimer’s and COVID-19 disease are strikingly similar.
To the home page
.
[ad_2]
Source link
