By Mads Sejer Nielsen, AAU Communication and Public Affairs
Photo: Colourbox

Recurring or prolonged discomfort, such as pain and stress in animals, fundamentally compromises animal welfare. Since animals cannot verbally communicate their experiences, these issues are often overlooked or misinterpreted. A new project at Aalborg University aims to play a part in changing that, and the initial results are promising.

"Society is placing increasing demands on us to treat animals ethically justifiable, but in order to do that, we need to better understand how animals are actually feeling. That’s why we are currently developing AI tools that can automatically and efficiently detect subtle facial expressions in horses. In the long term, this will allow us to recognize signs of stress and pain in many animals at once", says Rikke Gade, Associate Professor at Aalborg University and leader of the project.

Previous research has shown that horses are capable of communicating both stress and pain, however as the horse is a prey animal, they tend to hide these signs as much as possible. 

Facial expressions, being more subtle than body language, are not as easily concealed, but detecting them requires expert analysis which for the time being must study video recordings in slow motion to identify the brief facial movements that indicate various forms of discomfort. This process is both time-consuming and resource-intensive.

The goal is therefore to develop a method for automatically detecting and classifying combinations of facial muscle movements that signal pain or stress and the researchers already have access to large datasets in which each individual movement is described.

“We are collaborating with professors in veterinary science from the University of Copenhagen and the Swedish University of Agricultural Sciences, who have studied the horse’s behavior and expressions for many years. Through these partnerships, we have access to invaluable data that now enables us to develop modern AI methods to detect these expressions,” explains Rikke Gade.

The researchers aim to develop Computer Vision algorithms that use artificial intelligence to identify and analyze the most distinct differences between neutral expressions, pain, and stress. For example, they are examining the frequency and types of eye blinks, as well as specific ear movements.

Based on these indicators, the system will automatically classify when a horse is showing signs of discomfort.

Researchers have begun working with video data from controlled experimental studies and will later expand the algorithm to handle footage of horses moving freely in a stall.

The project is expected to lay the foundation for an automated method to detect signs of pain and stress, not only in horses, but also in other farm animals such as cows, sheep, and pigs, which exhibit similar facial expressions.

The project is supported by the Independent Research Fund Denmark with a grant of  2,879,662 kr.

Translated by Rikke Bajlum Nielsen, AAU Communication and Public Affairs