Machine Learning-Assisted Gait Analysis for Lameness Detection in Heterogeneous Dog Populations.

Machine learning gait analysis reliably detects clinical lameness but requires refinement for subclinical cases.

This proof-of-concept study evaluated machine learning algorithms for detecting lameness in dogs using pressure-sensitive treadmill gait analysis. A total of 173 dogs were assessed during walking (119 normal, 54 injured) and 129 during trotting (98 normal, 31 injured). Classification models were trained on 85% of data using F1 scoring methodology and validated on the remaining 15%. The walking model achieved perfect sensitivity for clinical lameness detection but failed to identify subclinical cases. The trotting model detected one of two subclinical lameness cases. Point biserial correlation analysis revealed distinct parameter patterns suggesting future models may localize lameness. Results demonstrate machine learning's potential to enhance canine gait analysis for detecting subtle abnormalities and determining lameness location and type, thereby improving clinical decision-making and directing further diagnostic investigations. The findings support continued development of these methods as adjunctive diagnostic tools for musculoskeletal and neurological disorders in dogs.

This summary was distilled by AI and may occasionally misinterpret data. Confirm critical details with the primary literature before clinical application.