Combatting a Pandemic with AI

3 min read

Over the last six months, the Covid-19 pandemic has taken the world by storm, and caused so much damage to many areas of life. However, it has also proven how useful technology can be in situations like this, and how quickly it can be adapted to suit certain needs. Specifically, artificial intelligence (AI) has been instrumental in combating the pandemic with prediction models, discovering potential cures, and processing medical imaging.

Prediction Models

One of the main prediction models that has been used during the pandemic is predictive modelling to determine future hotspots, number of fatalities, and other statistics. 

However, another interesting AI model that was created during the pandemic is able to predict which newly-infected patients will develop the severe respiratory disease associated with Covid-19. The original purpose of this model was to potentially predict which patients needed hospital beds and which ones didn’t based on the severity of the illness. 

The main indicators the model used for its prediction include:

  • Changes in levels of a liver enzyme called alanine aminotransferase (ALT)
    • Higher levels of ALT are linked to the more severe forms of Covid-19.
  • Reported myalgia (muscle pain)
    • This is linked to higher inflammation in the body, which can be caused in the lungs by Covid-19.
  • Hemoglobin levels (a protein that allows blood cells to carry oxygen to tissues).
    • Higher hemoglobin levels are linked to the more severe form of Covid-19.

The model operates using decision trees to analyze the data. Decision trees are a tool that can be used to support the final decision made by the model. They include possible decisions that can be made, and outcomes based on those decisions. The image below includes a potential visualization of the decision tree used by this prediction model based on the indicators used.

This approach allows the model to predict which choices made by doctors would result in a bad outcome. The model is trained using decisions made by doctors, and the outcomes that came about based on those decisions.

With the indicators used along with others, the model was able to have 80% accuracy when predicting which patients would become more severely ill with Covid-19. To learn more about this model, click here to read the study!

Discovering a Cure

Another topic that has been on the world’s mind since the pandemic started is when the cure will be found. One of the most well-known uses of AI includes drug discovery, and AI has been used to try to discover treatments for Covid-19 as well. If you want to learn more about AI and drug discovery, check out my article about that here

The software used to discover treatments for Covid-19 was developed by a company called BenevolentAI. The algorithms used are similar to those used for a search engine. They combine data from the drug industry with portions of research papers to discover potential treatments. The parts from research papers are extracted using machine learning to determine which information may be useful. Furthermore, the algorithm is able to search from approved drugs which may be able to block the viral infection process. One of the potential treatments identified by the software is Baricitinib, used to treat rheumatoid arthritis. This drug may be able to tone down the most severe effects of Covid-19. Baricitinib is predicted to reduce the ability of the virus to infect lung cells. To learn more about this model, the company has published three research papers on this drug that can be found here, here, and here.

Medical Imaging Lung scan depicting a low risk of Covid-19. 2020.

Another crucial aspect of using AI to combat the pandemic includes using AI to process medical imaging. A great example of this includes the AI tool called qXR created by, a startup that uses AI models to detect signs of diseases from lung scans. When the pandemic began, was able to adapt qXR to monitor the progression of Covid-19 using daily lung scans from a patient. The model is able to detect the progression of the pneumonia caused by the disease, and is able to estimate a percentage of the lung that is affected by the pneumonia

The indicators used by the model include:

  • Ground glass opacities (GGO)
    • GGOs are a hazy opacity that appears on the scan that can indicate filling of the lungs.
  • Consolidation
    • Lung consolidation is when pathways in the lungs are filled with liquid instead of air.
  • Lesion localization in lung parenchyma
    • The model is able to locate lesions that can be caused by Covid-19 (round or oval growth in the lung) in under a minute. 

This technology has been used in 50 locations around the world, including sites in England, India, Italy, Mexico, and Pakistan. 

Key Takeaways

  • AI has been able to help combat the pandemic with predictive modelling, using AI for drug discovery, and processing medical imaging.
  • AI has been used to predict which newly-infected patients will develop the more severe form of Covid-19.
  • BenevolentAI created an AI model to identify potential treatments for Covid-19.
  • The AI model identified a drug called Baricitinib that may be able to tone down the severe effects of Covid-19.
  • uses AI to process images of lungs to monitor the progression of the pneumonia caused by Covid-19.

Overall, the pandemic has spotlighted the fact that technology is very necessary to difficult situations, and can be used to save lives. Leave a comment down below regarding your opinions on the use of AI for combatting Covid-19! Have you heard of any other uses of AI to combat the pandemic?

Ramandeep Saini Ramandeep Saini is a writer who covers topics in emerging tech, such as artificial intelligence. She’s served as a consultant to companies such as Walmart Canada and Wealthsimple in the past, using her expertise in tech to guide them towards their corporate goals. In her free time, she runs an art blog and enjoys volunteering with local nonprofits.

2 Replies to “Combatting a Pandemic with AI”

Leave a Reply

Your email address will not be published. Required fields are marked *