KIRYAT SHMONA, Israel – Israeli-led researchers have developed an artificial intelligence framework that could give veterinarians a new way to recognize something their patients cannot explain in words: where and when they hurt.
The system, known as SHIC-XE, analyzes video of horses for signs associated with pain and, crucially, attempts to explain what the artificial intelligence is seeing rather than simply delivering a yes-or-no assessment.
Dr. Marcelo Feighelstein, head of the Artificial Intelligence Systems Engineering Program at Tel-Hai University of Kiryat Shmona in the Galilee, led the international research team.
Researchers from the Technion and the University of Haifa worked alongside scientists from São Paulo State University in Brazil, the University of Milan, the University of Bern, and Newcastle University. The peer-reviewed research was published in the International Journal of Computer Vision.
The technology addresses a longstanding challenge in veterinary medicine.
Animals cannot verbally tell a veterinarian how much pain they are experiencing or precisely where it hurts. Horses present an additional difficulty because they can suppress or conceal outward indications of pain, particularly when humans are nearby.
That can make assessment dependent on sometimes subtle changes in facial expression, posture, and behavior.
SHIC-XE attempts to make those signals easier to identify while giving veterinary professionals a way to examine the reasoning behind the computer’s conclusion.
Making AI Explain Itself
The innovation is not simply that artificial intelligence can recognize patterns associated with equine pain.
Researchers have already been investigating automated animal pain recognition.
The central advance behind SHIC-XE concerns explainability.
Conventional AI visualization methods can highlight the portions of an image that influenced an algorithm’s decision. But video creates a difficult problem. As an animal moves its head or the camera changes position, those highlighted areas can jump from frame to frame, producing explanations that are unstable and difficult for a veterinarian to interpret.
SHIC-XE takes a different approach.
The framework maps the model’s attention onto a standardized three-dimensional representation of the horse. That allows the system to associate its findings with consistent anatomical areas even as the horse moves or is filmed from different angles.
Instead of asking a veterinarian simply to trust an algorithm that declares an animal to be in pain, the system can help show which anatomical regions influenced the assessment.
That distinction is important in medicine, where an apparently accurate result is not necessarily enough for clinicians to trust an automated system.
The researchers also compared the AI’s areas of attention with pain indicators used by veterinary experts. They found meaningful agreement in some anatomical regions, including the ears and cheek muscles, while the eye region showed divergence between the AI and expert assessments.
That finding underscores the purpose of explainable AI. The objective is not merely to produce an impressive computer-generated answer, but to let experts examine and challenge how the system reached it.
Tested Across Different Types Of Pain
The researchers evaluated the system using three independent datasets involving different clinical pain conditions.
Those included post-surgical pain, inflammatory and orthopedic pain, and acute mechanical pain. Reported F1 scores ranged from 0.67 to 0.80 across the datasets.
The results point toward the potential for AI-assisted pain assessment while also showing that the technology remains a research framework rather than an infallible replacement for veterinary judgment.
For Feighelstein, the objective reaches beyond computer science.
“My motivation is to build a bridge between humans and animals,” he said in a Tel-Hai University release.
The implications could eventually extend beyond horses.
A system capable of identifying and explaining nonverbal indications of pain could potentially contribute to animal welfare across other species. Researchers also see broader possibilities for explainable pain-recognition technology in human healthcare, particularly involving patients who have difficulty communicating their condition reliably.
Considerable work remains between a promising research framework and routine clinical use. Artificial intelligence must perform reliably across different animals, environments, camera conditions, and medical circumstances before veterinarians can depend on it in everyday practice.
But SHIC-XE represents a distinctly Israeli-led contribution to one of AI’s most consequential emerging fields: developing systems that do not merely make decisions, but help human experts understand why they made them.
For veterinarians, that could eventually mean another set of eyes that can recognize suffering.
For the horse, it could mean something even more fundamental.
A way to be heard without saying a word.
TheJ.ca receives no government money. Independent journalism depends on readers who believe these stories must be told. Consider supporting our efforts.
Stand Proud. Stand Strong. Stand Zionist.
