On-farm detection of claw lesions in dairy cows based on acoustic analyses and machine learning

Claw lesions are a serious problem on dairy farms, affecting both the health and welfare of cow. Automated detection lameness with practical, on-farm application would support early treatment lame cows, potentially reducing number severity claw lesions. Therefore, in this study, method was proposed for based acoustic analysis cow's gait. A panel constructed to measure impact sound animals walking over it. The recorded edited, 640 files from 64 cows were analyzed. classification animal-lameness status performed using machine-learning process random forest algorithm. gold standard 2-point scale hoof-trimming results (healthy vs. affected), 38 properties used as influencing factors. prediction model classifying cow built This validated by comparing reference output concerning sound. Altering likelihood settings changing cutoff value predict improved model. At at 0.4, decreased false-negative rate generated, false-positive only increased slightly. obtained sensitivity 0.81 specificity 0.97. With procedure, Cohen's Kappa 0.80 showed good agreement between diagnoses hoof-trimming. In summary, enabled study shows that can be detected machine learning hoofs cows.