Breakthrough in AI-based nematode identification

Put a spade in the ground and you are likely to find dozens, if not hundreds, of nematode species. These thread-like worms, measuring between 0.2 and 3 millimetres, occur almost everywhere in the soil - including farmland. While some nematode species contribute positively to soil health, others are far less welcome, explains researcher Leendert Molendijk. “When harmful nematodes are present in the soil, such as stem nematodes or root-knot nematodes, this can mean that crops like ornamental flower bulbs, onions or seed potatoes can no longer be exported. Crops may also become deformed, making them difficult or impossible to sell.” Harmful nematodes are reported to cause dozens of billions of euros in damages yearly to growers all over the world. Researchers estimate that 10% of the world’s agricultural output is affected.

Within Integrated Nematode Management (INM) - managing harmful nematode species while minimizing the use of nematicides - accurate identification is crucial, says Molendijk. “You do not want to target harmless or beneficial nematodes. Measures also vary in effectiveness from one species to another. Some species decline after rotation with certain crops or green manures, but others thrive on many different crops. Timing of cultivation can also help reducing certain species. Sometimes, a drastic measure is needed, such as inundation, where a field is temporarily flooded.”

“Identifying nematode species requires a great deal of expertise”, adds researcher Pella Brinkman. “The differences between species are often extremely small. It really comes down to details: the shape of the stylet knobs, the length of a transparent section at the tail tip, or the number of head rings. In many cases, you can only determine the species once the nematode has reached adulthood.” Identification is usually carried out manually in specialized laboratories using microscopes, but sometimes additional molecular analysis is required. Worldwide, only a few dozen labs have the right equipment and expertise, including WUR. “It is an expensive and highly specialized field,” Brinkman notes.

One of the most difficult nematodes to identify is Meloidogyne chitwoodi, also known as the Columbia root-knot nematode. Molendijk: “It takes a lot of knowledge and careful identification to be certain that you are dealing with this species. It is particularly similar to Meloidogyne fallax, also called the False Columbia root-knot nematode.” When the agritech company Veridi Technologies approached WUR with the idea of developing an AI-based identification system, Meloidogyne chitwoodi was an obvious place to start. “If it works for such a difficult species, it should also work for species that are easier to distinguish,” Molendijk says.

Brinkman explains the role WUR played in supplying input for the AI system. “We provided nematodes from our cultures and identified them accurately — the annotation. This is essential for training the AI model on large numbers of images. We also carried out validation checks to assess reliability.” The researchers used field samples known to contain high numbers of root-knot nematodes. When the system made mistakes, the team helped analysing possible causes and advised on what morphological features the AI should focus on to improve the identification accuracy.

Tests showed that Veridi’s AI-powered microscope (Nemascope^TM)  achieved an accuracy of 96 percent for Meloidogyne chitwoodi identification. The results demonstrate that AI could become a valuable tool for nematode identification. Molendijk: “Since the 1990s, we have been working on technical ways to make identification more efficient. Achieving this with AI is a major milestone. If we can apply it to other nematode species, it could have a significant impact worldwide.” This is particularly relevant in regions where taxonomic expertise and advisory services are scarce. Providing farmers all over the world with access to an affordable identification system could improve understanding of soil health, leading to better yields, and less reliance on nematicides.

Within the context of a European Innovation Council grant, Veridi Technologies and WUR are building on this initial success to perform additional trials and research to expand the Nemascope’s capabilities to non-parasitic, free-living nematodes. Beyond only being parasites, nematodes are indeed excellent indicators of soil biodiversity. This subject is gaining importance due to the newly adopted European Soil Monitoring Law.