Related Experiment Video
Updated: Sep 10, 2025
08:20
Author Spotlight: AI-Driven Trypanosome Species Detection from Microscopic Images
Published on: October 27, 2023
1.7K
Image-based honey bee larval viral and bacterial diagnosis using machine learning
Duan C Copeland1, Brendon M Mott2, Oliver L Kortenkamp2,3
1USDA-ARS Carl Hayden Bee Research Center, 2000 E. Allen Rd, Tucson, AZ, 85719, USA. duan.copeland@usda.gov.
Scientific Reports
|August 21, 2025
Summary
An AI tool shows promise for diagnosing honey bee brood diseases, distinguishing European Foulbrood (EFB) from viral infections. This could reduce unnecessary antibiotic use and protect bee health.
Area of Science:
- Apiculture
- Veterinary Entomology
- Artificial Intelligence in Agriculture
Background:
- Honey bees are crucial pollinators, but hive losses due to brood diseases pose significant threats to agriculture and ecosystems.
- Accurate diagnosis of brood diseases, like European Foulbrood (EFB) and viral infections, is challenging for beekeepers, often leading to improper antibiotic use.
- Misdiagnosis exacerbates antibiotic resistance and harms beneficial bee gut microbiota, increasing colony vulnerability.
Purpose of the Study:
- To explore the feasibility of an image-based Artificial Intelligence (AI) diagnostic tool for honey bee brood diseases.
- To differentiate between European Foulbrood (EFB) and common viral infections affecting honey bee larvae using machine learning.
Main Methods:
- A dataset of 2,759 honey bee larvae images was collected from Michigan apiaries, molecularly verified for EFB and viral pathogens (ABPV, DWVA, DWVB).
- Image datasets were augmented, and deep convolutional neural networks (ResNet-50v2, ResNet-101v2, InceptionResNet-v2) were fine-tuned using transfer learning.
- Models were trained to discriminate between EFB and viral infections and tested on an independent dataset from Illinois.
Main Results:
- Proof-of-concept AI models achieved 73-88% accuracy on training/validation sets for distinguishing EFB and viral infections.
- On an independent dataset, models showed higher accuracy for EFB (72-88%) compared to viral infections (28-68%).
- The study highlights the potential of AI but also identifies limitations, particularly with diverse viral pathogens.
Conclusions:
- Image-based AI diagnostic tools offer a promising approach to reduce unnecessary antibiotic treatments in beekeeping.
- Accurate diagnosis supports the maintenance of microbiome integrity crucial for honey bee colony health.
- Further development requires expanded datasets including more pathogens and diverse geographic data for field readiness.
