Range of Projects

Computer-aided automated machine learning

This latest international project investigates the possibility of a new approach of screening for pain in horses. The main aim of this interdisciplinary project is to develop a fully automated facial decoding system for recognition of pain in horses via video films.

Our collaborators are leading computer vision science researchers from the Machine Learning Lab at the University of California in the USA. These computer vision researchers have already developed machine learned technology for automated decoding of human expressions of emotions, including pain.

Introducing a veterinary medicine perspective to their computer vision work, we will produce a set of videos containing horses with and without pain. These videos are then used to “train” a machine learning system to discriminate between horses in mild to moderate pain, and horses without pain. This “training” is done by running the image detection algorithm Faster R-CNN to examine frames from each video, based on the paper Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks by Shaoqing Ren, Kaiming He, Ross Girshick, Jian Sun, presented at the Conference on Neural Information Processing Systems (NIPS), 2015.

Such an automated machine learning project may also open the road for automated recognition of pain in a number of other relevant species, including cattle, swine and rodents.

Lameness evaluation through optical symmetry system

Lameness in horses is a clinical sign characterized by an aberration from a normal gait pattern due to locomotor dysfunction or structural pathology caused by pain, mechanic dysfunction or both.

Our team – including researchers at Utrecht University - is using a brand new algorithm for the university in-house kinematic objective motion system to detect possible low-grade lameness at walk. Using this objective lameness evaluation, we wish to explore whether this system can help produce an additional parameter in low-grade pain.

As there is no golden standard for uncovering the true pain state of the horse, there could be added power in attempting to describe pain states using differing but complimentary pain modalities, including a potential low-grade degree of lameness.

Supervising pain related student thesis work

A number of student projects related to horses in pain are currently being conducted by veterinary students as well as veterinary tech students at the Swedish University of Agricultural Sciences in Uppsala. They include projects looking at facial expressions of sedated horses, documenting possible pain faces of horses during lameness evaluation flexion tests, and investigating differences of pain behaviors when there are humans present vs absent at the stall.