"More accurate than lateral flow testing" according to the Covid app that analyses your speech for viruses.
Researchers have created a tool that can recognize Covid in your voice.
It is said that the AI-powered technology is more user-friendly and accurate than lateral flow testing.
Users of the myCOPD app are instructed to cough, take a deep breath, and then read a brief text three times.
In testing, it quickly and accurately recognized positive instances 83% of the time and negative cases 89% of the time.
These encouraging results show that basic speech recordings and well-tuned AI algorithms might possibly reach high precision in identifying whether patients have Covid-19 infection, according to researcher Wafaa Aljbawi from the University of Maastricht in the Netherlands.
Such tests are easy to understand and may be given for free. Additionally, they support virtual, remote testing and have a turnaround time of under a minute.
They might be employed, for instance, at the entrances to major meetings to provide quick crowd screening.
She stated lateral flow testing missed 44 out of 100 positive instances, whereas the app would miss just 11. However, compared to one in 100 for lateral flow, the app may incorrectly diagnose 17 out of every 100 non-infected individuals.
"These results demonstrate a considerable increase in the accuracy of diagnosing Covid-19 compared to cutting-edge procedures like the lateral flow test," Ms. Aljbawi continued.
Although the lateral flow test's specificity rate is greater at 99.5%, its sensitivity is just 56%. This is significant because it shows that, compared to our test, the lateral flow test frequently misclassifies infected individuals as Covid-19 negative.
In other words, "the lateral flow test would miss 44 out of 100 cases whereas the AI LSTM model may miss 11 out of 100 patients who would go on to spread the virus."
The upper respiratory system and vocal chords are frequently impacted by covid, altering a person's voice.
The programme makes use of data gathered by Cambridge University from 4,352 people, including 308 Covid patients, who provided 893 audio samples.
Mel Spectrogram, a voice analysis method, was employed in the study. The researchers created an artificial intelligence system based on neural networks, which simulate how the human brain functions, to identify Covid patient voices from non-sufferers.