Is Markerless More or Less? Comparing a Smartphone Computer Vision Method for Equine Lameness Assessment to Multi-Camera Motion Capture.
Authors: Lawin Felix Järemo, Byström Anna, Roepstorff Christoffer, Rhodin Marie, Almlöf Mattias, Silva Mudith, Andersen Pia Haubro, Kjellström Hedvig, Hernlund Elin
Journal: Animals : an open access journal from MDPI
Summary
# Editorial Summary: Smartphone Computer Vision for Equine Lameness Detection Researchers from Swedish institutions have evaluated a smartphone-based artificial intelligence system against the gold-standard multi-camera motion capture technology to assess its reliability for detecting lameness in horses. Twenty-five horses were recorded simultaneously using both a single smartphone camera (running markerless motion analysis software) and a 13-camera optical system with reflective markers whilst trotting on a 30 m runway, with vertical head and pelvis displacement analysed across 655 and 404 matched strides respectively. Mean differences between the two systems were minimal at 2.2 mm for both head and pelvis measurements, with within-trial variability broadly comparable between methods (3.1–28.1 mm for multi-camera versus 3.6–26.2 mm for smartphone), suggesting the smartphone approach captures motion data with clinically acceptable accuracy. The smartphone system's ability to detect asymmetries equivalent to a multi-camera setup—whilst requiring no markers, minimal equipment, and substantially less technical expertise—makes it a genuinely accessible tool for practitioners seeking to objectively monitor gait changes and lameness progression over time in working and competition horses. For farriers, veterinarians and physiotherapists, this technology offers the potential to track subtle movement asymmetries during rehabilitation or training programmes without the logistical demands and expense of traditional motion capture laboratories.
Read the full abstract on PubMed
Practical Takeaways
- •Smartphone gait analysis can now provide objective lameness detection comparable to expensive multi-camera systems, enabling routine monitoring at the barn or racetrack without specialized equipment
- •The 2.2 mm mean difference between systems is small enough for clinical practice; use repeated assessments over time rather than single snapshots to maximize sensitivity for detecting lameness progression or improvement
- •This tool offers an accessible way to document subtle asymmetries early and track response to treatment, potentially improving outcome documentation for insurance, sales, and client communication
Key Findings
- •Smartphone single-camera markerless motion capture achieved mean differences of 2.2 mm for head and 2.2 mm for pelvis compared to multi-camera optical system
- •Within-trial standard deviations ranged 3.1-28.1 mm for multi-camera and 3.6-26.2 mm for smartphone across 655 head and 404 pelvis stride segments
- •Smartphone application successfully detected vertical displacement asymmetries of clinically relevant magnitude using artificial neural networks for body segment tracking
- •Good agreement between markerless smartphone method and gold-standard multi-camera system demonstrates feasibility for objective lameness assessment in field conditions