Validation of a handheld smartphone markerless gait-analysis tool using an estimated groundline in horses.
Authors: Key Karsten, Kirkegaard Jakob, Berg Katja, Andresen Kristian Ringkjær, Skov Hansen Sabrina
Journal: Equine veterinary journal
Summary
# Validation of Smartphone-Based Gait Analysis in Horses Equine professionals now have access to a potentially game-changing tool: the RealHorse® app uses smartphone computer vision to analyse gait without requiring expensive multicamera motion-capture systems, but its accuracy needed rigorous validation against established gold-standard technology. Researchers compared vertical displacement measurements from the smartphone algorithm against a Qualisys three-dimensional optical system across 59 horses trotting in straight lines and 24 horses lunged on circles, measuring motion at four anatomical landmarks (eye, withers, back and croup) to assess stride-to-stride and overall trial-level accuracy. On straight lines, the app achieved impressive trial-level precision of 1.1–1.4 mm mean absolute error across all landmarks; circling increased this slightly to 1.8–3.3 mm, with the croup showing notably higher variability during curved work. The back proved the most consistently reliable measurement point regardless of movement pattern, whilst croup measurements during lunging warrant caution due to larger errors. These results suggest the RealHorse® system is sufficiently accurate for practical gait assessment, offering farriers, veterinarians and rehabilitation specialists an accessible alternative to laboratory-based motion capture for identifying subtle asymmetries and monitoring treatment responses, though users should recognise its limitations during circular work and recognise the distinction between stride-level versus overall trial accuracy when interpreting results.
Read the full abstract on PubMed
Practical Takeaways
- •RealHorse offers a practical, accessible alternative to expensive motion capture systems for routine gait assessment, with accuracy sufficient for clinical decision-making when used at trial level rather than individual stride analysis
- •The system performs reliably for straight-line trotting evaluations but users should be aware of reduced accuracy during circular work, particularly for croup vertical displacement measurements
- •Back and croup measurements are most reliable; eye and withers show greater variability, so clinical interpretation should account for keypoint-specific accuracy limitations
Key Findings
- •RealHorse smartphone-based markerless gait analysis achieved mean absolute error of 3.8 mm on straight line and 5.5 mm on circle at stride level when measuring vertical displacement
- •Trial-level analysis showed superior accuracy with errors of 1.1-1.4 mm on straight line and 1.8-3.3 mm on circle across eye, withers and croup keypoints
- •Back keypoint consistently demonstrated lowest errors across both stride and trial levels
- •Croup Mindiff measurements during circling showed the largest errors and highest values, indicating potential limitation of the system during curved locomotion