A newly published study conducted at the University of Georgia showed that new artificial intelligence (AI)-based tools could help poultry producers detect Salmonella more quickly using smartphone images and bird activity patterns.
The project was funded by the USPOULTRY Foundation to advance innovative solutions for poultry health and food safety.
The project was led by principal investigator Dr Guoming Li from University of Georgia Department of Poultry Science and Institute for Artificial Intelligence.
Why Salmonella
Salmonella is one of the leading causes of foodborne illness in the US, causing symptoms such as diarrhea, fever and stomach cramps.
Poultry products are recognized as one of the sources of human Salmonella infections. Detecting Salmonella quickly on poultry farms is critical for improving food safety, protecting flock health and reducing the risk of contamination entering the food supply.
Currently, the standard method for detecting Salmonella on farms involves collecting litter or fecal samples and sending them to a laboratory for microbiological testing. While accurate, this process is labor intensive, expensive and can take several days to produce results.
Two primary objectives
A research team led by Dr Li explored whether AI and computer vision technologies could provide a faster, more practical approach to monitoring Salmonella in poultry flocks.
The project had two primary objectives:
- First, the team developed and evaluated an AI system capable of predicting the presence of Salmonella using smartphone images of poultry feces.
- Second, researchers investigated whether changes in bird activity and daily movement patterns, known as the Broiler Activity Index, could serve as early indicators of Salmonella infection.
Research findings
The AI model performed very well when it was trained and tested using fecal images collected from the same geographic region, achieving greater than 93% accuracy.
However, when the model was applied to images collected from different regions, accuracy dropped significantly to between 40% and 60%.
These findings indicate that environmental and regional differences influence how well AI models perform and that additional data from diverse locations will be needed before a broadly applicable commercial system can be developed.
Researchers also found that healthy, uninfected chicks exhibited normal daily activity cycles, known as circadian rhythms, while Salmonella-infected birds showed measurable differences in their activity patterns.
This suggests that continuous monitoring of bird behavior using computer vision could serve as a digital biomarker for identifying flocks that may be infected with Salmonella.
As part of the project, the research team made the software code and smartphone application publicly available through GitHub, allowing industry and technology developers to build upon the platform.
Overall, the findings demonstrate the potential for AI and computer vision to provide rapid, low-cost tools for monitoring Salmonella on poultry farms, offering producers new opportunities to strengthen flock health management and biosecurity.
Although additional work is needed to improve performance across different geographic regions, the technology could eventually reduce reliance on laboratory testing, provide earlier detection of disease, improve flock management, strengthen on-farm biosecurity and support the development of automated precision poultry farming systems.