If you hear AI (artificial intelligence) brought up in conversation, chances are it will end up with someone wondering when the AI uprising will pose some kind of existential threat to humans.
Even Elon Musk has spoken out about AI and labelled it as one of the biggest threats facing the human race while speaking to Joe Rogan a few weeks back.
However, it seems like researchers in the UK are utilising AI models to create an application that will be able to diagnose COVID-19.
Innovation during a pandemic
While the focus at the moment around the globe is on vaccine rollouts, it’s clear from our experience during the last year that new and innovative testing methods are required when dealing with a pandemic.
Much of the initial failings in controlling COVID-19’s spread stemmed from a lack of tests which hampered the public health response, especially in poorer nations.
DeepCough3D is the AI model that is being tested as an innovative way to diagnose COVID-19. The tool analyses audio samples of people coughing from hospitals in Mexico and Spain.
The tool analysed 8 000 samples, with about 2 000 of the patients being COVID-19 positive. Impressively during the experiment, the tool was able to identify patients COVID-19 status with 98% accuracy accurately.
AI in healthcare
While AI is already used to assist in a number of healthcare settings, it seems clear that there is still a lot of opportunities to leverage AI to find real-world solutions that previously would not have been possible.
DeepCough3D isn’t the only tool of it’s kind, though. Analysing cough sounds has been investigated as a means of diagnosing COVID-19 for some time now.
MIT researchers recently developed an algorithm that claimed a similar 98% success rate in diagnosing COVID-19. However, Essex University researchers are backing their claims with a robust body of data that they feel sets their tool apart.
In addition to diagnosing infection, researchers feel that the data will eventually be able to predict and diagnose COVID-19 severity levels. This kind of information could be invaluable to healthcare facilities to allocate staff and resources in high demand.
The researchers’ next step is to expand their research into interventional studies with an eye to wide-scale adoption and certification of the tool for use in real-world settings.