Israeli-led team develops explainable AI to improve pain detection in horses

Researchers have created SHIC-XE, an innovative AI system that not only detects pain in horses but also offers an interpretable visual explanation, promising to improve animal welfare and clinical decision-making.

A new AI system developed by an Israeli-led team could help vets read equine pain more clearly.
Its chief innovation is not diagnosis alone, but an explanation of what the software is seeing.
That may matter as much as the result.

Researchers led by Dr Marcelo Feighelstein at Tel-Hai University in Kiryat Shmona have built SHIC-XE, a video-based framework designed to detect pain in horses and map the signs to stable anatomical regions. According to the team’s paper in the International Journal of Computer Vision, the system was tested on three datasets covering post-surgical, inflammatory, orthopaedic and acute mechanical pain.

The model produced F1 scores between 0.67 and 0.80. More importantly for clinicians, it translated its attention into a standardised three-dimensional horse model, avoiding the shifting visual heatmaps that can make video explanations hard to trust.

The researchers found some agreement with veterinary assessments, particularly around the ears and cheek muscles, though the eye region diverged. That gap underlines the wider point: explainable AI is meant to support, not replace, professional judgement.

The work, which involved institutions in Israel, Brazil, Italy, Switzerland and the UK, points to a broader use for interpretable pain-recognition tools in animal welfare and, potentially, human medicine.