Covid app that detects virus in your voice ‘more accurate than lateral flow test’

An app that can detect the coronavirus in your voice has been developed in a major scientific breakthrough.

Scientists say AI-driven technology is easier to use and more accurate than lateral flow tests.

The mobile app took less than a minute to flag positive cases and gave accurate results 89% of the time and negative cases 83% of the time.

In contrast, the accuracy of the lateral flow test varies by brand, and nasopharyngeal swabs are not very good at detecting infected people who are asymptomatic.

The new app could be used to quickly screen for the vulnerability before people attend large events such as concerts and major sporting events.

It can also be deployed in poorer countries where gold standard PCR tests are very expensive and often difficult to distribute.

The coronavirus typically affects the upper respiratory tract and vocal cords, causing changes in a person’s voice, Dutch researchers said.

User requests to record breath sounds

The team decided to investigate whether it was possible to detect the novel virus in people’s voices.

In developing it, they used data from the University of Cambridge’s crowdsourced Covid-19 Sounds app, which contains 893 audio samples from 4,352 participants, 308 of whom tested positive for the virus.

The app was installed on users’ phones, and participants reported some basic details about demographics, medical history and smoking status.

They were then asked to record some breath sounds, including coughing 3 times, taking deep breaths through their mouths 3 to 5 times, and reading short sentences aloud 3 times on a screen.

The researchers used a speech analysis technique called mel-spectrogram analysis, which can identify different speech characteristics such as loudness, power, and changes over time.

To differentiate the voices of Covid-19 patients from those who did not, the team built different AI models and assessed which model was best for classifying positive cases.

One model called long short-term memory (LSTM) outperformed the others.

It is based on a neural network that mimics the way the human brain works and identifies potential relationships in data.

It works with sequences, which makes it suitable for modeling signals collected over time, such as from speech, because of its ability to store data in memory.

Tests are available for free

Wafaa Aljbawi, researcher at Maastricht University, said: “These promising results show that simple recordings and fine-tuned AI algorithms can achieve high accuracy in determining which patients are infected with Covid-19.

“Such tests are available for free and are easy to interpret. Additionally, they support remote virtual testing and have a turnaround time of less than a minute.

“For example, they can be used as entry points for large gatherings, enabling rapid screening of crowds.

“These results demonstrate a significant improvement in the accuracy of diagnosing Covid-19 compared to state-of-the-art tests, such as the lateral flow test.

“The lateral flow test has a sensitivity of only 56 percent, but a higher specificity of 99.5 percent.

“This is important because it shows that lateral flow tests classify infected people as negative for Covid-19 more frequently than our tests.

“In other words, with the AI ​​LSTM model, we might miss 11 out of 100 cases that continue to spread the infection, while the lateral flow test would miss 44 out of 100 cases.

“The high specificity of the lateral flow test means that only 1 in 10 people will be falsely told they are Covid-19 positive when in fact they are not infected, whereas the LSTM test will incorrectly diagnose every 100 people who are not infected. 17 of those were positive.

“However, since the test is actually free, people can be invited to have a PCR test if the LSTM test shows they are positive.”

Further research is required before the app can be used

The team said further studies with more participants are needed before the app starts appearing on people’s phones.

Since the start of the project, 53,449 audio samples from 36,116 participants have been collected, which can be used to improve and validate the accuracy of the model.

The team is also doing more analysis to understand which parameters in speech are affecting the AI ​​model.

The findings will be presented at the European Respiratory Society International Congress in Barcelona, ​​Spain.

Leave a Comment

Your email address will not be published.