5th Jun 2020
5th Jun 2020
Curated by Endre Szvetnik
The COVID-19 pandemic has stirred up a massive effort among researchers to help tackle the crisis. Research teams are using AI techniques to identify symptoms faster, find new therapies and to inform public health policy.
How could AI help diagnose COVID-19 cases? In a recent example, Chinese researchers using data from Wuhan and Shenzhen have created an AI tool that can speed up diagnosing COVID-19 from chest CT scans. The technique uses a deep learning algorithm that is trained to identify patterns in the images of COVID-19 related pneumonia and distinguish it from other forms of pneumonia. The researchers used over four thousand hospital scans to teach their algorithm, achieving an 87% accuracy in spotting COVID-19 cases.
And what about treating cases? Another urgent task for clinicians is to identify drugs that can be used to treat COVID-19 patients. A UK company has used its machine learning tool to scan data describing the biological effects of pre-existing approved drugs so that they can be repurposed and found a medication used for rheumatoid arthritis. The drug was fast-tracked for a clinical trial to find out if it would be effective against COVID-19. In a complementary approach, Hong Kong-based researchers are using AI togenerate novel molecular structuresthat can be the basis of entirely new drugs because of their ability to bind coronavirus proteins in simulations. They have also developed an AI-based clinical trial outcome prediction tool to accelerate the validation process.
What else can AI help in against the pandemic? There are many applications, but apart from medical science, AI can also help inform policymakers. AI researchers at Lafayette College in the US are doing to help identify common responses by social media users to government decisions. The team is using natural language processing, text mining and network analysis to process over 6 million tweets in 66 languages between 22 January and 7 March at the beginning of the pandemic. With this technique, they intend to track how people’s responses to policy changed over time and also see how COVID-19-related misinformation was transmitted over Twitter. In separate studies, AI has also been deployed to predict possible geographical locations of outbreaks by analyzing text in news reports and cross-checking it with data from animal disease surveillance and airline data.
Here is the current state of science on a Sparrho pinboard. NB: The pinboard contains research papers that have not been peer-reviewed yet, meaning that they have not gone through the standard scientific validation process yet.
Do you want to stay updated with the latest Covid-19 research? Sign up to our newsletter >>