Animal Welfare Program, University of British Columbia, Vancouver, Canada, V6T 1Z4.
J Dairy Sci. 2012 Nov;95(11):6523-8. doi: 10.3168/jds.2012-5742. Epub 2012 Sep 7.
Weight shifting between legs and steps taken when cows stand may be a useful tool to assess cow comfort and lameness. Weight shifting is assessed by measuring the distribution of weight applied to each leg when standing on a weighing platform, whereas frequency of steps is traditionally measured with live observation or video recording. The objectives of this study were to validate an automated method to count steps from weight distribution measurements (experiment 1) and to assess the accuracy of the frequency of steps in detecting lameness (experiment 2). In experiment 1, 6 nonlame multiparous cows stood on a weighing platform covered with either concrete or rubber (1h/cow per surface) while stepping behavior was video recorded. Receiver operating characteristic curves were constructed, using the steps observed in the video recordings as the gold standard, to calculate the optimal threshold (based on the sum of sensitivity and specificity) of the weight applied to a leg to define a step. Optimal thresholds were similar between surfaces. The optimal thresholds, when pooling the 2 surfaces, were 127 and 98 kg for the front and rear pair for legs, respectively, with a specificity and sensitivity ≥0.96. Thresholds were used to construct an algorithm to count steps. In experiment 2, 57 cows (26 of them considered lame according to their gait score) stood for 15 min on the weighing platform. Frequency of steps taken with the front and rear pair of legs was calculated from the weight distribution measurements using the algorithm calculated in experiment 1. Lame cows took more steps per minute with the rear legs than did nonlame cows (1.6 vs. 1.0 steps/min; SE of the difference=0.2). As previously shown for weight shifting, the frequency of steps taken with the rear legs was a good predictor of lameness (area under the curve of the receiver operating characteristic curve=0.67; 95% confidence interval=0.52, 0.81). A positive relationship was observed between the frequency of steps and weight shifting (measured as SD of the weight applied over time to the legs) in both the front (R(2)=0.35) and rear (R(2)=0.49) legs, yet the slopes differed from 1 and the intercepts differed from 0, indicating that the 2 measures were related but not the same. In conclusion, weighing platforms can accurately calculate the frequency of steps automatically, and this measure shows promise as a tool to assess lameness.
奶牛在站立时腿部的负重转移和迈出的步伐可能是评估奶牛舒适度和跛行的有用工具。腿部负重转移通过测量奶牛站在称重平台上时每条腿承受的重量分布来评估，而步伐频率传统上通过现场观察或视频记录来测量。本研究的目的是验证一种从称重分布测量自动计算步伐数量的方法（实验 1），并评估检测跛行的步伐频率的准确性（实验 2）。在实验 1 中，6 头非跛行经产奶牛站在覆盖有混凝土或橡胶的称重平台上（每头奶牛在每种表面上站立 1 小时），同时用视频记录其行走行为。使用视频记录中的观察到的步伐作为金标准，构建受试者工作特征曲线，计算将腿部承受的重量定义为一步的最佳阈值（基于灵敏度和特异性的总和）。两种表面的最佳阈值相似。两种表面的最佳阈值分别为 127 和 98 千克，用于定义前腿和后腿的一步，特异性和敏感性≥0.96。使用这些阈值构建了一种算法来计算步伐数量。在实验 2 中，57 头奶牛（其中 26 头根据其步态评分被认为跛行）在称重平台上站立 15 分钟。使用实验 1 中计算的算法，从称重分布测量中计算前腿和后腿的步伐频率。与非跛行奶牛相比，后腿跛行奶牛每分钟迈出的步伐更多（后腿为 1.6 步/分钟，而非后腿为 1.0 步/分钟；差异的标准误=0.2）。与腿部负重转移一样，后腿的步伐频率是跛行的良好预测指标（受试者工作特征曲线的曲线下面积=0.67；95%置信区间=0.52，0.81）。在前后腿（前腿 R(2)=0.35，后腿 R(2)=0.49）中，都观察到步伐频率与腿部负重转移（以时间加权的腿部承重的标准差衡量）之间呈正相关，但斜率不等于 1，截距也不等于 0，这表明这两个指标相关但并不相同。总之，称重平台可以自动准确地计算步伐频率，该指标有望成为评估跛行的一种工具。
