This AI sends you a voice sample
Researcher Wafaa Aljbawi and a team of researchers from Maastricht University have developed an artificial intelligence that can detect a corona infection. All she needs is a voice recording and a completed questionnaire.
We have been dealing with the corona virus for more than two and a half years and there is no end in sight. Of course, there are positive developments in terms of dealing with the pandemic.
However, it is clear that the corona virus will remain an integral part of our lives for a while. It is therefore all the more important that science and we as a society find better and better solutions to deal with the virus.
Contents
Voice message instead of quick test: New AI can reliably detect corona infection
A research group from the University of Maastricht in the Netherlands has now reached an important milestone along this path. The team around the student Wafaa Aljbawi presented a new technology at a conference in Barcelona, Spain presented.
Specifically, the group has developed an artificial intelligence (AI) that can very reliably detect a corona infection. However, no nose or throat swab is necessary for the analysis.
Instead, the technology, dubbed “Mel-spectrogram,” just requires a voice message and some medical information.
This is how the AI works, which can change the way we deal with Covid-19 in the long term
The technical analysis is based on some acoustic recordings. The patients have to read a short sentence three times in the corresponding app, cough three times and take five deep breaths.
In the second step, the subjects have to fill out a short questionnaire. In this, the researchers or the AI inquire about previous illnesses and (unhealthy) habits such as smoking status.
Based on this information and the color of the voice, the app now recognizes whether a person has contracted the corona virus.
89 percent success rate in positive corona patients
The results of artificial intelligence are particularly convincing. In 89 percent of the cases, the Corona AI correctly identified positive cases. Conversely, the technology recognized 83 percent of healthy patients as negative.
For comparison: With ordinary antigen rapid tests, the rate of positive cases is only around 56 percent. On the other hand, 99.5 percent of healthy people can be detected with a rapid test. The safety is therefore significantly higher in the exclusion principle.
In total, the research team worked with 893 audio files from 4,352 healthy and sick people, 308 of whom tested positive for the coronavirus.
Corona AI: Inexpensive alternative to the complex test structure
If the results of the Corona AI from the University of Maastricht are as good on a larger scale, an interesting alternative to the conventional rapid test could potentially come onto the market.
With an app and a short voice recording, the hurdles are very low for the patients. In addition, no complex local test infrastructure is required, which is particularly advantageous in structurally weak regions of the world.
Accordingly, it remains to be hoped that in the future we will really be able to detect a corona infection via an app in less than a minute. It would definitely be great progress.
Also interesting:
