Scientists train AI to detect pain—in goats
by University of FloridaThis article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:
fact-checked
peer-reviewed publication
trusted source
proofread
The patient arrived with a bladder stone, grimacing in pain and moping about. He wouldn't even chew his cud. The patient, you see, was a goat. And while treated for his bladder stone—a common ailment in the small ruminants—he was also contributing to new research that aims to accurately measure pain not only in goats, but other domestic animals as well and even, one day, in people.
"If we solve the problem with animals, we can also solve the problem for children and other non-verbal patients," said Ludovica Chiavaccini, D.M.V., D.E.S., M.S., a clinical associate professor of anesthesiology at the University of Florida's College of Veterinary Medicine.
Chiavaccini and her colleagues filmed the faces of goats that were in pain and those that were comfortable. Then they fed the data into an artificial intelligence-based model that learned to distinguish goats in pain by their faces alone.
The researchers published their findings Nov. 7 in the journal Scientific Reports.
The system, trained and tested on 40 goats so far, was anywhere from 62 to 80% accurate at identifying pained faces, depending on how the scientists tested the model. With data on more goats and other animal species, these kinds of AI models may help clinicians treat pain effectively in patients they can't speak with.
"It's not just an animal-welfare issue," Chiavaccini said. "We also know animals that are in pain don't gain weight and are less productive. Farmers are becoming more and more aware of the need to control acute and chronic pain in animals."
Implementing AI-powered pain scales at veterinary clinics will require more research, but would help solve a longstanding problem in animal care. Pain assessment in animals has historically been both difficult and subjective. Traditionally, veterinarians had to rely on decades of experience to make judgment calls.
Researchers have developed standardized pain scales for different species in recent years to reduce subjectivity, but the quality of those measures varies wildly.
When Chiavaccini and her team started the study—inspired by a graduate student's love of goats—there wasn't any pain scale available for goats at all. Today, a single pain score for goats exists. But it is only validated for male goats undergoing castration, demonstrating the need for a more generalizable system, Chiavaccini said.
More information: Ludovica Chiavaccini et al, Automated acute pain prediction in domestic goats using deep learning-based models on video-recordings, Scientific Reports (2024). DOI: 10.1038/s41598-024-78494-0
Journal information: Scientific Reports
Provided by University of Florida