Imagine being at the grocery store, picking up a package of strawberries, then giving them a quick scan that will immediately tell you exactly the types and quantities of pesticide residues that are on the fruit before you buy them and bring them home to your family. This may sound like something out of a futuristic science fiction film, but a team of interdisciplinary researchers at the University of Missouri have used the power of artificial intelligence to make the technology that would enable this scene a reality.
The team, which is led by Mengshi Lin, professor of food science, and includes Jianlin Cheng, Curators’ Distinguished Professor of electrical engineering and computer science; John Snyder, assistant adjunct professor of statistics, and two Ph.D. students, Akshata Hegde and Mehdi Hajikhani; began with an interest in leveraging artificial intelligence to improve food safety.
“AI is currently experiencing a boom akin to the gold rush era,” Lin said. “There is a surge of exciting things happening in that space, and this is a new approach to blend nanotechnology and AI to boost food safety. We are combining food science, chemistry, and computer science to create this ecosystem.”
Lin further explained that exposure to toxic pesticides can cause a variety of health problems ranging from short-term acute conditions to long-term chronic illnesses. For this reason, fruits and vegetables sold to consumers in the US undergo a thorough inspection process to ensure consumer safety.
Prior to the development of this new technology, it would take hours to sample fruit or vegetables to determine the pesticides present, but with the team’s new AI-powered technology, it is possible to process a large number of samples within minutes.
Lin points out that the speed and accuracy of his technology are potentially very powerful tools for the food industry, government food inspectors, and the academic community. With this tool, a single scan only takes a fraction of a second, and it can tell you exactly what pesticides are present and their quantities with 98% accuracy.
The team used surface-enhanced Raman spectroscopy (SERS) technique, coupled with nanotechnology, to analyze various pesticides on the surface of fruits and vegetables, which, in layman’s terms, means they used a special laser light to observe scattered signals emitted by pesticide molecules. Then, they used that information to build a state-of-the-art AI transformer model that can quickly extract critical information and identify the pesticides.
“This research is highly original,” Cheng said. “It is the first time a transformer AI model similar to large language models like ChatGPT has been used in this way.” Cheng believes the technology developed in this study has the potential to be applied to a wide array of other fields.
“There is so much potential with this to capture all kinds of physical processes,” Snyder said. “There is potential to identify the presence of almost anything.”
The research was funded by a $649,483 USDA National Institute of Food and Agriculture (NIFA) grant.