AI is for the birds: How machine learning can help predict and manage avian flu outbreaks

The active and ongoing global spread of avian influenza virus has impacted more than 14 million birds in Canada and 160 million in the USA.

This recent outbreak has resulted in major economic losses, and a rise in egg prices in the past few years. This trend can cause disruptions in poultry supply chain and significant increases in the price of other poultry products.

A virus like avian influenza is carried by birds, but it can “jump” species and infect livestock such as dairy or sheep or even pets like dogs and cats.

Read more: Bird flu detected in Colorado dairy cattle - a vet explains the risks of the highly pathogenic avian influenza virus

And most, if not all, human pandemic influenza viruses have had an avian origin in the past few decades. Experts warn it is only a matter of time before we face another pandemic threat.

The good news is, we are better prepared than ever to meet that challenge. Not just because we have vaccines or treatments, although those are critical. But because we have something that can change the game entirely: artificial intelligence (AI).

Vast amounts of information

AI can offer much in the way of advance pandemic information and planning. Remember the early days of COVID-19? What if we had more time to prepare? What if health officials had known weeks earlier where the virus was spreading, which neighbourhoods were most at risk, and what we needed to do to stop it?

AI can analyze vast amounts of information, from wildlife health reports, geographical data, satellite images to social media trends, online content, farm data and even weather patterns to answer some questions about how, when and why pandemics happen. It spots patterns, anomalies and relationships humans cannot see in real-time.

AI can alert monitors to where an avian influenza outbreak might occur before a region is impacted, how severe an outbreak might be and what type of intervention may be most effective. AI can help responders and governments act quickly, precisely and efficiently.

Predicting outbreaks

At the University of Guelph, my research team and I are working on AI solutions to help track and predict the avian influenza outbreaks. Our research — which is currently under review — has used AI to filter out misinformation about avian influenza from social media platforms and Reddit, as well as Google search data, and other online sources.

This helps us understand public discussion about avian influenza. We have also combined these online activities with other data sources to monitor avian influenza online mentions and trends — we’ve found that AI can use this information to predict if an outbreak might occur in a specific area.

With the availability of online and social media data, an outbreak surge can be predicted up to four weeks in advance in specific regions.

Our research team has also created and tested decision support tools that use different types of information from wild bird reports, satellite images, climate change data and farm information. These tools help predict avian influenza outbreaks and how serious they might be in a certain area; through testing, we achieved an accuracy of 85 per cent.

We’re currently in the process of building a Canadian tool to predict where bird flu might emerge, helping farmers and public health officials get ahead of outbreaks — this could mean the difference between a contained outbreak and a global crisis.

More than a public health issue

Avian influenza spreads through the food chain, wildlife and global trade. An outbreak in poultry can devastate agriculture and threaten our food security. Worse, it can jump to human populations with little warning.

This issue is not just a public health issue. It is also an economic and social concern. But if we harness AI properly, we can give ourselves a better chance at combating these threats. We can predict where the next outbreak might come from and take action before it spreads.

Read more: Soaring U.S. egg prices and millions of dead chickens signal the deep problems and risks in modern poultry production

Using AI to predict avian flu outbreaks and spread can be applied to other situations, including other illnesses and the weather and environmental conditions that could contribute to disease spread.

AI-based decision tools can also include augmented reality that enables the testing of thousands of hypothetical scenarios related to avian influenza. These include how outbreaks might spread, what the impacts of different intervention strategies could be, how changes in the economy and environment might occur, and how the supply chain could be impacted.

We have the technology in our labs. But to make it work, we need strong partnerships between government, universities, farmers, industry and communities. We need to make sure that we generate high quality data, use the data ethically in a privacy-preserving manner, develop the AI tool responsibly and apply it fairly to ensure that no one is left behind.

Rozita Dara receives funding from Ontario Ministry of Agriculture, Food and Agribusiness Alliance Tier I, funding and the University of Guelph's Food from Thought.