Agreement between subjective gait assessment and markerless video gait-analysis in endurance horses.
Authors: de Chiara, Montano, De Matteis, Guidi, Buono, Auletta, Del Prete, Pasolini
Journal: Equine veterinary journal
Summary
# Editorial Summary: AI Motion Tracking in Endurance Horse Gait Assessment Subjective lameness evaluation by FEI-accredited officials remains the standard for determining fitness-to-compete in endurance horses, but introduces inconsistency and potential bias into competition veterinary decisions. De Chiara and colleagues investigated whether markerless artificial intelligence motion-tracking (AI-MTS) software—capable of quantifying head and pelvis vertical asymmetry from smartphone video—could provide objective corroboration for these subjective assessments. Across 188 gait examinations conducted at three endurance competitions, overall agreement between the AI system and official veterinary opinion was fair (κ = 0.26), though performance varied significantly by severity category: the technology demonstrated substantial agreement for detecting severe asymmetry (κ = 0.75) but struggled with mild cases (κ = 0.13), whilst showing only fair agreement for symmetrical gaits (κ = 0.25). The most practical implication is that AI-MTS could serve as a valuable decision-support tool for officials making marginal fitness calls, particularly when horses present with obvious lameness, though the current system cannot reliably distinguish subclinical or mild gait deviations—exactly where veterinary judgment is often most challenged and where consistent, objective measurement would provide greatest value.
Read the full abstract on PubMed
Practical Takeaways
- •AI motion tracking systems can confidently assist with identifying severe lameness/gait asymmetry at endurance competitions, improving objectivity and consistency of 'fitness-to-compete' decisions
- •Mild lameness remains difficult to detect with current markerless AI technology; subjective OEV assessment remains essential for borderline cases
- •Smartphone video combined with AI analysis offers a practical, portable tool for veterinarians to document and verify gait evaluations at competitions without specialized equipment
Key Findings
- •Overall fair agreement (κ = 0.26) between AI markerless motion tracking and subjective OEV lameness assessment across all gait categories
- •Substantial agreement (κ = 0.75) achieved for severe asymmetry detection, suggesting AI-MTS reliably identifies clinically significant lameness
- •No agreement (κ = 0.13) for mild asymmetry category, indicating this remains challenging for objective motion analysis systems
- •188 gait examinations performed during three endurance competitions with smartphone-based video recording and post-hoc AI analysis