Walking Acceleration Patterns as a Method for Lameness Detection N. Chapinal 1 , M. Pastell 2 , L. Hänninen 3 L Munksgaard 4 , J. Rushen 5 , A.M. de Passillé 5 1 Animal Welfare Program, Canada. 2 Department of Agrotechnology, University of Helsinki, Finland. 3 Department of Production Animal Medine, University of Helsinki, Finland. 4 Faculty of Agricultural Sciences, Aarhus University, Denmark. 5 Agriculture and Agri-Food Canada. Several methods of automatically recording the behavior of cattle have great potential for detecting lameness (Rajkondawar et al., 2002; Pastell et al., 2006; Rushen et al., 2007). Commercially available 3-dimensional accelerometers attached to the legs can measure time spent lying down, which is usually longer for lame cows (Chapinal et al., 2009). Accelerometers can also detect alterations in walking gait that are indicative of lameness (Pastell et al., in press). The aim of this study was to validate 3-dimensional accelerometers placed on the back and on each leg of cows as a method of lameness detection (Experiment 1), and to assess the effect of flooring surfaces on walking accelerations patterns in dairy cattle (Experiment 2 and 3). Materials and Methods Lactating Holstein cows housed in sand bedded free-stalls in groups of 12 to 36 cows were fitted with 5 tri-axial accelerometers (HOBO Pendant G Acceleration Data Logger, Onset Computer Corporation, Pocasset, MA), attached to each leg and to the back. The accelerometers can measure a range of ± 3 g, and were set to sample at 33.3 Hz. Cows were gait scored using the system described by Flower and Weary (2006), which assessed 6 different gait components and assigns a continuous overall gait score on a 1-to-100 analog scale. In Experiment 1, 10 cows were selected to include some lame and some non-lame cows. Cows were walked down a 13 m long by 1.3 m wide non-grooved concrete passageway. In Experiment 2, 30 cows were walked down a 13 m x 1.3 m passageway with a surface of either non-grooved concrete or rubber. The rubber flooring consisted of a bottom layer of revulcanized rubber mats (1.9 cm thick each; Animat, Saint-Élie d’Orford, Quebec) and an upper layer of anti-slippery rubber (0.6 cm thick each, #125 2-ply; Cobelt Canada Inc., Saint-Laurent, QC, Canada). Cows were walked from 6 to 8 times, alternating the walking surface. In Experiment 3, 15 cows were walked down a 27.4 m X 1.2 m passageway with a surface of either grass pasture or concrete. Cows were walked from 8 to 10 times, alternating the walking surface. The measurements provided forward, lateral horizontal acceleration and vertical acceleration relative to the cows’ legs. We calculated the variance of acceleration for each axis and the sum vector of acceleration for each leg. We analyzed the symmetry of gait by calculating the symmetry of acceleration between the front pair and the hind pair of legs, dividing the variance of the leg with lesser variance by the variance of the leg with greater variance. The differences in acceleration features for different treatments in each trial were tested with a linear mixed model with cow included as the random effect. R 2.90 (R Development Core Team, 2008) with nlme package was used in the statistical analysis. Results In Experiment 1, we found overall differences between sound and lame animals in the patterns of acceleration when walking. There was a negative linear relationship between the overall gait The First North American Conference on Precision Dairy Management 2010