Validation of an AI-assisted screening tool for lameness detection in dairy cattle: A field-based study for preventive herd health management.

Abstract

Lameness is a major welfare and economic challenge in the dairy industry, requiring effective surveillance tools for herd-level management. While visual scoring is subjective and labor-intensive, Artificial Intelligence (AI) offers the potential for automated, objective monitoring. This study aimed to validate a field-deployable AI-assisted decision-support system as an epidemiological screening tool for the automated detection of gait abnormalities. We utilized custom-built bilateral hind-limb accelerometers on 134 cows across four diverse farm environments. The core AI algorithm employed a 1D Convolutional Neural Network (1D-CNN) to process multi-axis motion data, converting motion data into a binary classification (lame vs. non-lame) to support veterinary decision-making. Against a reference standard of expert visual locomotion scoring, the system achieved an overall accuracy of 83.6%, with a high sensitivity of 86.4% and specificity of 80.9%. The results demonstrate that this AI-driven system effectively minimizes false negatives, making it a robust tool for preventive herd health management. By automating the detection of gait asymmetry, this technology acts as a high-throughput triage tool, empowering veterinarians to prioritize at-risk animals for clinical examination and facilitating proactive herd health management.