RS researchers are looking to help U.S. beekeepers better detect and treat their honey bee populations. Honey bees are important pollinators in agriculture. With an estimated 50–80% of crops pollinated by honey bees, they generate approximately $20 billion annually in market value in the United States alone. However, pathogens and the Varroa destructor parasitic mite threaten their health, costing the industry over $500 million each year.
Commercial beekeepers face challenges in protecting their hives from bacterial, viral, and fungal diseases. Accurate diagnosis of brood disease, especially distinguishing European Foulbrood (EFB), a bacterial disease, from viral infections with symptoms similar to EFB, remains challenging and time consuming. Misdiagnosis often results in inappropriate antibiotic use, leading to increased antimicrobial resistant bacteria and significant disruption of the honey bees’ gut microbiome.
In a published study, researchers at the ARS Carl Hayden Bee Research Center in Tucson, AZ, developed an image-based artificial intelligence (AI) diagnostic tool to rapidly identify and differentiate between bacterial and viral infections in honey bee brood. Over many years, the researchers collaborated with apiary inspectors and university extension agents to generate a dataset of 2,759 honey bee larvae images gathered from multiple apiaries. Each image was paired with definitive molecular evidence of viral or bacterial infection.
Using image analysis, the proof-of-concept models achieved 73–88% accuracy in identifying the diseases.
“This study represents a significant step toward addressing the challenge of brood disease diagnosis, which has traditionally relied on visual inspection by experienced apiarists with variable accuracy,” said ARS Research Microbiologist Duan Copeland.
“AI models can provide rapid, objective disease diagnostics which can reduce the unnecessary use of antibiotics and preserve the integrity of the gut microbiome, a significant contributor to colony health in commercial beekeeping operations.”
According to Copeland, the goal is to develop a phone-app diagnostic tool whereby beekeepers can upload photos and receive accurate diagnoses in seconds. The findings show promise, but more research and development are required to produce field-ready diagnostic tools. Future research will focus on including all major bee pathogens and extending landscape representation throughout the U.S.
The research was funded by the National Institute of Food and Agriculture (NIFA)’s Agriculture and Food Research Initiative grant and conducted in collaboration with Apiary Inspectors of America, Michigan State University, and the ARS Bee Research Laboratory. – by Jessica Ryan, ARS Office of Communications