In Italy, they managed to detect positive cases of coronavirus from the noise of the cough

In Italy, they managed to detect positive cases of coronavirus from the noise of the cough

In Italy, they managed to detect positive cases of coronavirus from the noise of the cough

In Italy, they managed to detect positive cases of coronavirus from the noise of the cough

Detecting the positivity of the virus would be possible with analyzing the cough and making its sound waves visible. The method bears the signature of Professor Guido Sciavicco, from the Department of Mathematics and Informatics at the University of Ferrara, with data made available by the University of Cambridge.

“For some years now, scientists have been trying to discover a pattern or signal in the voice of people affected by different diseases and thus identify them or have one more element to record them. With the emergence of the COVID-19 pandemic that mainly affects the respiratory system, different softwares have been developed to measure the voice, but also to measure the cough”, Comments for Con Bienestar Iris Rodríguez (MN 64.010), a doctor specializing in laryngology and voice.

From data collected by the University of Cambridge in 2020 from several hundred cough records of asymptomatic people who were known to be positive or not COVID-19, the researchers were able to isolate features that distinguish a positive person, recreate them and also recognize them in the cough of other patients.

We made the sound visible to clearly highlight which are the frequencies and power that characterize the typical patterns of positive cough, even if it is asymptomatic -explains Sciavicco-. We then isolated these recurring characteristics and investigated and recognized them in additional cough records of the samples provided, of which the positivity was known, to validate our diagnostic system ”.

The recognition method is based on a technique called modal learning with supervision. This is how the specialist explains it: “Part of a securely labeled database and, in this case, we knew exactly if the cough we heard was from a positive patient or not. By providing this data and its label to the computer, it gives it tools to learn to distinguish positivity or negativity, exclusively analyzing the audio track of a cough, without having additional information about the patient’s medical history ”.

“There are several countries like Italy that are in this search. Israel is one of them, Canada is also, and the idea is to be able to diagnose the coronavirus infection as soon as possible, when the patient is still asymptomatic or in patients who have the disease only with a cough, “adds the doctor from the Italian Hospital of Buenos Aires and Founding Member of the Argentine Society of the Voice.

Compared to other techniques and machine learning languages, this method is not only cheaper from the point of view of computing resources, but the real strength is the ability of the computer to justify the decision made: it is no longer just the result of a long computational calculation, but a “reasoned” choice of the machine itself. The precision of this technology would allow replace or supplement with a simple microphone current non-invasive methods.

The author of the work also believes that it would be possible to think about a computer to “behave like a doctor” during a routine visit: the project team is evaluating possible implementations of this research in the form of applications for the near future. In fact, it would be enough with coughing for 15 seconds into the microphone, very similar to what we do during auscultation of the lungs.

“The objective is also that the cough can be measured in this case or remote voice, that is, the software would allow it to be sent to a population and that population would send the sample to the scientists by that means, where the sample is analyzed with a program and thus a large number of people can be measured or tested”, Concludes Rodríguez.

Leave a Reply

Your email address will not be published. Required fields are marked *