This paper presents the validation of a software tool called Cow-Gait-Analyzer (University of Bern, Switzerland) to determine gait-cycle variables in lame and non-lame dairy cows using features derived from low-cost, stand-alone 3-dimensional accelerometers (400 Hz). The Cow-Gait-Analyzer automatically extracts the relevant gait events of foot load and toe off, which characterize gait-cycle duration, stance phase, and swing phase during walking. A nonautomatic step is visual inspection of the pedograms. If the software does not automatically choose the right peaks according to pedogram definitions, peaks can be manually chosen. We validated the algorithms by comparing the accelerometer data (pedogram) with the synchronized video data, which we used as a gold standard. We carried out the measurements at the metatarsal level of paired hind limbs during walking. We included 12 non-lame cows and 5 lame cows and expressed overall differences between the Cow-Gait-Analyzer and the gold standard as relative measurement error (RME). We analyzed 34 hind limbs with a mean of 9 gait cycles. The median RME for gait-cycle duration and stance phases were 0 and 1.69%, respectively. The peaks of gait-cycle variables showed RME of 0.67 and 0.24% for foot load and toe off, respectively. The semi-automated Cow-Gait-Analyzer can accurately determine gait-cycle variables in both lame and non-lame cows, and could be used to assess gait patterns in routine clinical and research practice focusing on individual cows.
Keywords: accelerometer; algorithms; dairy cow; gait cycle.
The Authors. Published by the Federation of Animal Science Societies and Elsevier Inc. on behalf of the American Dairy Science Association®. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).