Some in healthcare are turning to AI to battle COVID-19

Published On May 28, 2020

While COVID-19 has presented complex, multi-faceted challenges to humanity and the healthcare industry, it has also created a unique opportunity for healthcare to apply AI in creative, problem-solving ways. Here are some examples:

• Pooling COVID-19 data for AI systems: According to an article in World Economic Forum, there are several projects working to pool data so that AI systems can sift through that data and help identify patterns:

  • The COVID-19 Open Research Dataset provides the text of over 24,000 research papers

  • The COVID-net open-access neural network is working to develop a system to identify the virus in lung scans

  • There is also an initiative asking individuals to donate their anonymized data including location data to help track who has the virus and who has interacted with them

• Vocal signatures and COVID-19: According to an article in Business Insider, a team of researchers from Harvard and MIT are using machine learning to comes through voice recording from COVID-19 patients as well as health people to try to identify specific vocal signatures that could indicate someone has the virus.

  Analyzing peoples’ speech, coughing and breathing patterns isn’t new, according to the article. In fact, the study of cough sounds has been around for decades.

  This research is still in early stages but the goal would be to develop an AI tool that could tell people whether they have COVID-19 based on an audio recording

• Using AI to discover a treatment for COVID-19: According to the National Science Foundation, the race is on to find a vaccine, drug or combination of treatments for COVID-19. In order to speed up this process, scientists are combining AI with physics-based drug docking and molecular dynamics simulations to identify the most promising molecules. One such project is using the most powerful supercomputers to run millions of simulations in order to train a machine learning system to identify the factors that might make a molecule a good candidate and then exploring the most promising results.

• Scanning radiology images to detect COVID-19: According to an article in the MIT Technology Review, early research has shown that the most severe cases of COVID-19 displayed distinct lung abnormalities in radiology images. One application that is already using a combination of deep learning models to detect common types of lung abnormalities is working to rework its application to detect the specific lung abnormalities COVID-19 produces in the lungs.

  A preliminary study of this tool analyzed the images of 11,000 patients and found the application to distinguish between COVID-19 patients and non-COVID-19 patients with 95% accuracy.

Keys to success: Data quality and human intervention

While these innovative applications of AI are inspiring, it’s important not to lose sight of the critical role data quality and human intelligence play in realizing AI’s true potential during this pandemic. The quality of the data is key. In order for AI to do its job accurately, it needs unbiased, trustworthy data that algorithms and logic it can read and learn from.

Furthermore, human intelligence also plays a crucial role in the success of AI. Data scientists and domain experts who can think outside of the box, and apply and connect insights from an expanded set of experiences and relational situations, are all important in ensuring AI is successful.

Ultimately, in order to realize the full capabilities of AI, data quality, scope and structure will need to be continually evaluated and optimized. Furthermore, it will also be important to look for opportunities to teach AI tools how to read, analyze and connect data differently.