AI-assisted Digital Video Analysis Reveals Changes in Gait Among Three-Day Event Horses During Competition.

This study uses artificial intelligence (AI) to conduct a digital video analysis to assess changes in horse gait during performance competitions. The AI model was trained to identify key points in the horse’s movement, and this data was then used to quantify their gait characteristics, providing a more efficient and objective method for this kind of research.

Objective of The Research

• The research aims to offer an economical, precise, repetitive, and objective technique for the field quantification of equine locomotion. The researchers aimed to overcome the constraints of traditional qualitative assessments, which are not suitable for extensive biomechanical research studies.

Methodology

• The researchers utilized a regular digital video camera to capture high-resolution videos of horses in the trot during mandatory inspections for international-level competitions.
• They examined a total of 194 horses from five different competition venues. These inspections were recorded both pre-competition and post-cross-country inspections to serve as a model for gait change after exertion.
• They used DeepLabCut, an artificial intelligence-based strategy, to label twenty-six keypoints on each frame from the videos. This formed the basis for the derived quantitative gait parameters.
• After the DeepLabCut model was trained, it managed to label the 388 videos within minutes—an accomplishment that would have taken months of human effort to complete.

Results

• The Generalized Linear Mixed Model (GLMM) examined seven gait parameters: they found significant changes in several parameters, including duty factor, speed, and forelimb swing range, following the completion of the cross-country phase.

Conclusion

• Despite certain limitations, the study concluded that video analysis through artificial intelligence can quantify several gait parameters efficiently. It indicated that AI tools can be of significant use in future biomechanical research studies without requiring specialized equipment. Doing so promises to improve our understanding of athletic horse locomotion and possibly aids in furthering the welfare of performance horses.