SHIC-XE Makes Equine Pain Detection Explainable

Horses often suppress visible signs of pain when people are nearby, making assessments based on facial expressions and posture difficult and subjective. Conventional AI heat maps can also shift as the animal or camera moves, obscuring which anatomical features drive a diagnosis. SHIC-XE addresses that black-box problem by projecting a model’s attention from two-dimensional video onto a fixed three-dimensional horse-face model, giving veterinarians spatially consistent explanations and offering a potential template for monitoring pain in patients who cannot communicate.

Led by Marcelo Feighelstein at Tel-Hai University, the international project also involved the University of Haifa, the Technion–Israel Institute of Technology and researchers in Europe and Brazil. The paper was published in the International Journal of Computer Vision on July 10, 2026, and highlighted by Tel-Hai on Aug. 18. Video-level F1 scores were 0.80 for orthopedic pain, 0.67 for post-surgical pain and 0.70 for lip-twitch detection. The team sees possible applications in newborn care and in monitoring sedated or ventilated intensive-care patients.