Say goodbye to throat-piercing? AI mobile phone program detects COVID-19 through sound, with an accuracy rate of 89%

I don’t know what everyone thinks about doing nucleic acid. Anyway, the throat of Digestella has become cocooned.

However, for the overall situation of epidemic prevention and to know and confirm your own health status, it is inevitable to confirm that you are not infected with the new coronavirus through nucleic acid.

Wait... Is there any other way to detect whether I have been tricked?

It’s best not to go out.

Don’t tell me, such technology is really possible.

A study presented at the European Respiratory Society International Congress in Barcelona, ​​Spain, on September 8th showed that a mobile application uses artificial intelligence to You can tell whether you are infected with COVID-19 by your voice.

Currently, the accuracy of this model has reached 89%.

Does this mean that in the future, uploading your own voice at home can replace nucleic acid testing? It feels wonderful to think about it...

Telling whether you are positive by sound is better than rapid antigen testing

New coronavirus pneumonia can affect the upper respiratory tract and vocal cords, causing changes in a person’s voice.

On this basis, Ms. Wafaa Aljbawi from the Institute of Data Science at Maastricht University and her supervisor, Maastricht University Dr. Sami Simon, a pulmonologist at the medical center, and Dr. Visara Urovi, also from the Data Science Institute, decided to investigate the possibility of using artificial intelligence to analyze voices to detect COVID-19 positivity.

They used data from the University of Cambridge’s COVID-19 Sounds app, which contains 893 audio samples from 4,352 healthy and non-healthy participants, 308 of whom had COVID-19. Tested positive.

The app is installed on the user's phone, and participants report some basic information such as demographics, medical history, and smoking status, and then are asked to record some breathing sounds, including coughing three times, using Take three to five deep breaths through your mouth and read short sentences on the screen three times.

The researchers used a speech analysis technology called "Mel-spectrogram", which can identify different speech characteristics, such as loudness and frequency. and changes over time.

"This way we can break down the many properties of the participant's voice," Ms. Aljbawi said. "To distinguish the voices of COVID-19 positive patients and negative normal people, we built different artificial intelligence models and evaluated which model was best suited to classify these cases."

They found LSTM model outperforms other models. LSTM is based on neural networks, which simulate the way the human brain works and identify underlying relationships in data. It excels at timing analysis, which makes it suitable for modeling signals collected over time, such as sound.

In the end, the overall accuracy of this model was 89%, and the ability to correctly detect positive cases (true positives) The rate is 89%, and the ability to correctly identify negative cases (true negatives) is 83%.

“These results show a significant improvement in the accuracy of diagnosing COVID-19 disease compared with state-of-the-art testing methods such as lateral flow testing,” said Aljbawi. The accuracy of the method (rapid antigen test) is only 56%, but the specificity is higher at 99.5%. This is important because it means that the rapid antigen test incorrectly classifies an infected person as negative more often than our test More common. In other words, using the AI ​​LSTM model, we might miss 11/100 cases that would go on to spread the infection, whereas a rapid antigen test would miss 44/100 cases."

Should we do nucleic acid or nucleic acid testing?

The reason why it is compared with rapid antigen testing is because many countries no longer provide free nucleic acid virus testing. ——That is, large-scale nucleic acid testing conducted in China.

Nucleic acid virus detection is a direct detection of collected viral nucleic acids. It has the characteristics of strong specificity and high sensitivity. It is the main method for detecting new coronaviruses.

Rapid antigen testing is relatively simple and can be done by yourself. It mainly distinguishes by detecting the antigen of the virus. It can be used as one of the main basis for the diagnosis of new coronavirus. However, The accuracy is relatively low.

Many countries currently distribute rapid antigen self-test kits, and only those who test positive will require further testing.

Of course, because this kind of self-test package is relatively difficult to operate, the accuracy is not very high. Therefore, if it can be judged as positive through the sound, it can not only save resources, but also obtain relatively accurate results. , is indeed a good thing.

So, although AI performs relatively poorly in terms of false positives, with 17% of people being misdiagnosed as positive, it can be used as a preliminary screening method to allow voice detection to show positive results. If it is positive, go to the next step for testing.

Moreover, this technology is targeted more at low-income countries where nucleic acid testing is expensive and/or difficult to distribute.

As for us, we should do nucleic acid or nucleic acid...

In addition, the researchers said that their results still require a large amount of data To verify, 53,449 audio samples collected from 36,116 participants since the start of the project can be used to improve and validate the accuracy of the model. They are also conducting further analysis to understand which parameters in speech are affecting the AI ​​model.

Related reports:

https://www.news-medical.net/news/20220905/AI-model-detects-COVID-19-infection-in -peoplee28099s-voices.aspx

The above is the detailed content of Say goodbye to throat-piercing? AI mobile phone program detects COVID-19 through sound, with an accuracy rate of 89%. For more information, please follow other related articles on the PHP Chinese website!