Artificial intelligence can help build pollen mosaics of current and ancient flora

An emerging system that combines rapid imaging with artificial intelligence could help scientists build a comprehensive picture of current and historical environmental change by quickly and accurately analyzing pollen. body.

Pollen grains of different plant species are unique and identifiable based on their shape. Analyzing which pollen grains are captured in samples such as sediment cores from lakes helps scientists understand which plants thrived at any given time in history, potentially dating back ranging from thousands to millions of years.

Until now, scientists have manually counted pollen types in sediments or from air samples using light microscopes – a specialized and time-consuming job.

Now, scientists at the University of Exeter and Swansea University are combining cutting-edge technologies including imaging flow cytometry and artificial intelligence to build a system capable of identifying and sort pollen at a much faster rate.

As well as building a more complete picture of past flora, the team hopes this technology can one day be applied to more accurately read pollen in today’s environments, which This may help people with hay fever relieve symptoms. The paper is titled “Automatically inferred pollen classification in environmental samples via imaging flow cytometry and exploratory deep learning”, and is published in the journal New botanist.

Dr Ann Power, of the University of Exeter, said: “Pollen is an important environmental indicator and piecing together different types of pollen in the atmosphere, both today and in the past, can help us We build a picture of biodiversity.” and climate change.”

“However, identifying plant pollen under a microscope is extremely laborious and not always possible. The system we are developing will significantly cut down on time and improve This means we can build a richer picture of pollen in the environment much faster, showing how climate, human activity and biodiversity have changed over time or better understand the allergens in the air we breathe.”

The team used the system to automatically analyze a slice of a 5,500-year-old lake sediment core, quickly classifying more than a thousand pollen grains. Previously, it would have taken an expert up to eight hours to count and classify this task – a task the new system completed in less than an hour.

The new system uses imaging flow cytometry – a technology commonly used to study cells in medical research – to quickly capture images of pollen. Then, a unique type of artificial intelligence was developed based on deep learning technology to identify different types of pollen in environmental samples. This can create these differences even if the sample is not perfect.

Dr Claire Barnes, from Swansea University, said: “To date, AI systems being developed for pollen classification will learn from and test on the same pollen library—meaning each sample are perfect and belong to species the network has seen before. These systems cannot recognize pollen from the environment that has encountered some collisions along the way, nor can they classify pollen that is not in the training library.”

“Incorporating a unique version of deep learning into our system means the artificial intelligence is smarter and takes a more flexible approach to learning. It can process high quality images. quality and can use common species characteristics to predict the plant family to which the pollen belongs even if the system has never seen it before during training.”

In the coming years, the team hopes to perfect and launch the new system, and use it to learn more about grass pollen, a particular irritant for people with hay fever. . Dr Power said: “Some types of grass pollen are more allergenic than others. If we can better understand which types of pollen are common at particular times, that will lead to improvements.” improvements in pollen forecasting, which can help people with hay fever make plans to reduce exposure.”