A mobile app detects covid-19 cases from people’s voices

The system, which will be presented at the European Respiratory Society International Congress in Barcelona this Monday, will also cheaper than antigen testsThis means that these tests can be used in low-income countries where they are expensive or difficult to obtain.

According to Wafaa Aljbawi, a researcher at the Institute for Data Science at Maastricht University (Netherlands), this artificial intelligence model 89% accuracya percentage that varies depending on the brand in tests.

“Our results are promising and show that audio recordings and fine-tuned AI algorithms can be used. very accurate in identifying which patients have covid-19 infection”he assures.

Fast and virtual response

“These tests are free and easy to interpret. remote virtual tests and response times less than a minute so they can be used, for example, at entry points of large gatherings to provide rapid identification in the population.”

Covid infection usually affects the upper respiratory tract and vocal cords, causing changes in a person’s voice.

From there, Aljbawi, Sami Simons, a pulmonologist at Maastricht University Medical Center, and Visara Urovi of the Data Science Institute explored whether it was possible to use artificial intelligence to analyze sounds and detect infections.

For this, they used the open application “Covid-19 Sounds” created by Cambridge University. examining symptoms coronavirusA database containing 893 audio samples from 4,352 healthy and unhealthy participants, of which 308 tested positive for covid-19.

The application is installed on the user’s mobile phone, the participants basic information and data about your medical history and habits such as smokingThey are then asked to record their breathing, coughing, and voice.

Using Mel’s sound analysis technique called spectrogram analysis, which describes different characteristics of the voice such as loudness, power, and variation, they were able to isolate the different features of the participants’ voices.

Next, to distinguish the voices of Covid-19 patients from healthy ones, the scientists created different AI models and examined which one worked best to classify cases.

89% accuracy

The “Long Short Term Memory” (LSTM) model, based on neural networks that mimic the way the human brain works and recognize key relationships in data, 89 percent accuracy It detects positive cases and negative cases with 83 percent accuracy.

this study results will be confirmed in a larger study With 53,449 audio samples from 36,116 participants.

In a second study, Henry Glyde of the University of Bristol showed that AI (via an app called “myCOPD”) can predict exacerbations (severe exacerbations) in patients with chronic obstructive pulmonary disease (COPD).

“MyCOPD” is an interactive app developed by patients and doctors, made available by the UK’s National Health Service since 2016 and currently helping more than 15,000 COPD patients manage their disease.

Researchers gathered 45,636 records of 183 patients Between August 2017 and December 2021 (45,007 records of stable disease and 629 exacerbations) and used this data to train AI models.

“The most recent AI model we developed has a sensitivity of 32% and a specificity of 95%. This means that the model is very good at telling patients that they will not experience an exacerbation, which can help them avoid exacerbations. Glyde concludes by saying “unnecessary treatment.”