In the near future, artificial intelligence will be able to identify patients with COVID-19 by coughing A unique development in this direction at the Open Innovations forum in Skolkovo was presented by Russian scientists. "Acoustery" technology, capable of detecting a sick person by coughing in the near future, may appear in airports, carsharing cars, buses, trains, train stations, shops, pharmacies and other places with a large crowd of people, tvbrics.com with reference to
The principle of the system is simple: to understand whether a person has a disease, he just needs to cough into a special terminal - and the system will give a conclusion “here and now”. According to the developers, the miracle system can be used individually, simply by installing a special application on a smartphone or smart watch.
The technology is based on the analysis of the visual representation of sound - a spectrogram. Cough in various diseases has characteristic differences. However, it can be difficult to determine by ear, and even the most experienced doctor can be mistaken. But if you interpret the sound as a picture, it will be much more informative and more accurate. Each type of cough, captured graphically, has its own unique pattern, including the pattern of a cough with coronavirus.
In order to literally compose a complete "picture" of cough, the creators of the development of "Acoustery" for six months, with the help of doctors from different clinics and medical universities in Russia, Belarus and Kazakhstan, collected thousands of sound samples of the cough of people with COVID-19. Then the developers "taught" the system to detect exactly the "coronavirus" cough.
“We've been in machine learning technology for over five years. Before the pandemic, our developments performed well in the automotive industry - they were used to carry out acoustic diagnostics of problems in cars and motorcycles. And now, thanks to cooperation with doctors, we have been able to create a technology that will help to recognize a "covid" cough ", - said Dmitry Mikhailov, head of the Acoustery project at Meditex.
According to the authors of the project, after the successful completion of clinical trials, the system of preliminary diagnostics created by them will be able to be used in medical institutions - it is planned to develop a version of the product for diagnosing patients, as well as monitoring the effectiveness of treatment and rehabilitation. With its help, it will be possible to observe the dynamics of the patient's disease and the effectiveness of the treatment.
Also, Russian scientists intend to extend this experience of preliminary diagnostics using a neural network to other diseases. The algorithms at the heart of the system will be able to detect not only the cough characteristic of COVID-19, but also asthma, whooping cough and other lung diseases.