A new AI framework , known as SHIC-XE, has been developed to detect signs of pain in horses from video analysis while providing stable, anatomically consistent explanations for its decisions. The framework was developed by an international team of researchers led by Dr. Marcelo Feighelstein, Head of the Artificial Intelligence Systems Engineering Program at Tel-Hai University 's new Cluster of Engineering and Advanced Computing.
For the first time, these AI-generated explanations can be quantitatively compared with expert assessments , representing a significant breakthrough in the field of explainable artificial intelligence and an important step toward AI systems that can be trusted in real-world clinical and healthcare environments .
The study, published in the prestigious International Journal of Computer Vision , addresses a longstanding challenge in both veterinary and human medicine: how to identify pain and distress in individuals who cannot communicate what they are experiencing.
Animals cannot express pain in words, and even experienced veterinarians and caregivers can struggle to recognize signs of suffering. Dr. Feighelstein's research aims to bridge that communication gap by using artificial intelligence to interpret facial expressions, body language, and movement patterns. Over the years, his team and collaborators have developed AI-based tools capable of recognizing pain and emotions in cats, dogs, rabbits, sheep, cattle, and now horses .
"My motivation is to build a bridge between humans and animals," said Dr. Feighelstein. "We want to give a technological voice to those who have no words, allowing caregivers and professionals to better understand their condition, their emotions, and their suffering."
A major challenge in AI-based video analysis is explainability. Existing methods often highlight different areas of an image from one video frame to the next, creating unstable and difficult-to-interpret explanations. To overcome this limitation, the researchers developed SHIC-XE, a novel framework that projects the model's attention onto a fixed three-dimensional representation of a horse's face. The result is a consistent anatomical explanation, even when the horse moves, changes its head position, or is filmed from different angles.
The system was evaluated using three independent datasets representing post-surgical pain, inflammatory and orthopedic pain, and acute mechanical pain. It achieved strong performance, with F1 scores ranging from 0.67 to 0.80, indicating high levels of accuracy and reliability in detecting pain from video across multiple clinical scenarios. Importantly, the areas identified by the AI showed statistically significant agreement with assessments made by veterinary experts using the Horse Grimace Scale, a standardized method for assessing pain in horses based on facial expressions, particularly in key facial regions such as the ears and cheek muscles.
"This is the first quantitative validation of its kind demonstrating that the model not only reaches the correct conclusion, but also focuses on the anatomically relevant regions when making that decision," said Dr. Feighelstein. “This brings us closer to opening the 'AI black box' by showing not only what the model decides, but whether the evidence it uses makes sense to experts.”
While developed for equine pain recognition, the implications extend far beyond animal welfare. The ability to consistently understand and validate what an AI system has 'seen' in video data is equally important in human healthcare, especially when patients cannot describe their condition in words. Future applications could include pain assessment in newborns, monitoring sedated and ventilated patients in intensive care units, supporting care for people with dementia, analyzing neurological movement disorders, and assisting with clinical and surgical video analysis.
"When we began studying automatic pain and emotion recognition in animals, the goal was to give them a voice," Dr. Feighelstein added. "Over time, we realized that an equally important challenge is ensuring that this voice is reliable, explaining not only that pain is present but also why the system reached that conclusion. That is what transforms technology into a true partner in care."
The study was led by Dr. Marcelo Feighelstein of Tel-Hai University of Kiryat Shmona in the Galilee, in collaboration with Prof. Anna Zamansky and the Tech4Animals Lab, Prof. Ilan Shimshoni of the University of Haifa, students Omer Bibi and Ofer Rosenbaum from the Technion, and researchers from the University of Bern, the University of Milan, the University of São Paulo, and Newcastle University.
Israel's Tel-Hai University of Kiryat Shmona in the Galilee is a rapidly growing research university addressing challenges in food security, sustainable agriculture, engineering, artificial intelligence (AI), and more. Through interdisciplinary applied research, academic excellence in a host of disciplines - including humanities, social sciences, and education - as well as close collaboration with industry and communities, Tel-Hai University of Kiryat Shmona in the Galilee is tackling some of the most pressing challenges of the 21st century, while positioning the Galilee as a global hub for innovation.
International Journal of Computer Vision
Computational simulation/modeling
Animals
SHIC-XE: Viewpoint-Invariant Explainability via Dense 2D-3D Correspondences: an Application to Equine Pain Recognition
10-Jul-2026