Smoothing centre-of-mass tracking to quantify body oscillations and ambulation in broilers

Abstract

Lameness is a key welfare issue in broilers, and the development of non-invasive, automated methods to identify impaired mobility is important for broiler welfare assessment. Current methods for assessing lameness in broilers can be invasive and often require broilers to walk, which may induce stress or pain in lame birds. The aim of this study was to develop a computational approach to quantify broiler ambulation by tracking broilers in an open field and exploring correlations between smoothed movement metrics and bodyweight. At 28 days of age, male broilers (n = 46) were video-recorded during a 180 s open field test. The centre-of-mass was tracked using EthoVision® XT17 at 15 Hz. Zigzag trajectories were detected in 22 broilers and smoothed in RStudio using a 40-point moving average (~2.7 s window). The Euclidean distance (EuD) was calculated between the raw zigzag and smoothed trajectories, and steps were classified based on the variation in EuD over time using a > 2 cm noise threshold. Computed steps were compared with observer counts from the video. Smoothing zigzag trajectories reduced the estimated distance by 57.04% and speed by 56.14%. Computed step counts were 23.49% lower than observer counted steps. There was a strong positive association between raw and smoothed distance, speed, and observer-counted vs computed steps. Bodyweight did not correlate significantly with movement metrics. Smoothing may reduce the influence of lateral body oscillations on estimated distance and speed, although the computational step-detection method requires further refinement to reduce underprediction.