Scientists at the Massachusetts Institute of Technology (MIT) have developed an AI that works with Covid – 19 should be able to recognize infected people by their coughs. In tests with people demonstrably infected with the coronavirus, the detection rate was , 5 percent, in test subjects who have not yet done one Showed symptoms, it was 100 percent. The researchers now want to pour the AI ​​into a mobile app in order to be able to evaluate coughing noises with a smartphone so that infected people can be identified quickly and easily.

As from the article “Covid – Artificial Intelligence Diagnosis using only Cough Recordings “in the IEEE Open Journal of Engineering in Medicine and Biology , the method is suitable for the rapid detection of asymptomatics, i.e. infected people who have not yet been Covid – 19 – have symptoms, but can already infect people. MIT researchers found that there are differences between a common cough and a cough from Covid – 19 – infected people there – even if the cough was forced. Man is not able to hear the differences. However, the AI ​​developed by MIT can analyze the sound recording of such coughs and recognize the differences. “The nature of the sound changes when you have Covid – even if you are still asymptomatic,” says Brian Subirana, one of the scientists involved in the research project.

The number of people infected with the coronavirus is constantly increasing. This selection of articles gives an insight into the effects of the infection:

Infected people cough differently The scientists have to do this between April and May 2020 around 5320 cough records collected , according to the scientists the largest audio Covid – – cough dataset. The researchers developed a language processing framework that uses acoustic biomarker trait extractors to analyze cough records. The models based on a convolutional neural network (CNN) were trained on the basis of 4256 recordings of the test subjects and then on the remaining 1064 Cough records of the collected data set tested. In subjects who were demonstrably positive with a Covid – – test, could 98, 5 percent of the sick are correctly recognized as infected, in asymtomatic subjects it was 100 percent, according to the scientific paper of the MIT researchers.

The researchers therefore come to the conclusion that with their AI, one corresponding app and a smartphone, daily tests could be carried out quickly, easily and almost free of charge, for example when accessing schools, universities and workplaces, in order to identify infected people and prevent infection with the coronavirus.

(olb)