Automatic Measurement of Walking Speed to determine Mobility Score of Dairy Cows
In 2012 a study was conducted by Mottram et al. to assess the potential for automatically measuring mobility scores for cows by deriving this from their walking speed. The paper was ultimately not published but the study still proves useful for the general study of cow behaviour.
Mobility scores range from 0 (no impairment of mobility, perfect locomotion) to 3 (severe impairment of mobility) with 1 (imperfect mobility) and 2 (impaired mobility) as the intermediate scores. Walking speed was recorded using time of passage (TOP) between two automatic radio frequency identification (RFID) scanning points.
The typical approach to screening lameness involves watching every cow take at least three purposeful, unimpeded strides and a turn. Cows are then allocated a score based on the observation of a set of five key behaviours associated with lameness: namely even weight bearing on all four limbs, even rhythm of walking, stride length, posture and speed of locomotion. DairyCo have published guidance for farmers looking to score their herds for mobility (Anon. 2011a; Anon. 2011b) which has become the industry standard approach in the UK. The recommendation is that farmers should screen their herds regularly, ideally scoring at least monthly. However, the quality of this screening remains questionable given there are variations in the environment in which cows are scored, a clear view of each cow is not always achieved when groups are being moved, there are difficulties with identifying cows, intervals between scoring is longer than ideal for managing lameness and the intra- and inter-observer repeatability are a concern.
The hypothesis to be tested was that the time of passage would be longer in cows with mobility scores of 2 and 3 which are indicators of cows likely to benefit from lameness treatment. A commercial dairy farm was recruited for this trial and in order to measure walking speed two portal antennas (ATL Agricultural Technology Limited, UK) were placed five metres apart in the parlour exit race. When a cow was within ± 200 mm of either antenna her identity was transmitted to a computer which logged her time passing under each antenna, from which the time of passage was computed. The herd was scored for mobility as cows exited the sorting gate located adjacent to the RFID antennae. All scoring was done by one veterinary surgeon (IB) trained by NJB. Video clips were used for training, to minimise observer drift and to ensure scoring was consistent with a panel of international experts.
To account for variations in TOP due to cows queuing, TOP data were collected over five days centred over a mobility scoring session on the middle day. It was important to match these two sets of data in this way as mobility scores can change over time with the onset and recovery of lameness. Any cows with only one TOP recording were removed from the data set. The 5-day period means of the minimum and maximum TOPs for each cow were calculated, and minimum TOP compared between lame (mobility scores of 2-3) and not lame cows (mobility scores of 0-1).
Minimum TOP was considered the preferred way to discriminate between lame and sound cows as sound cows may walk slowly due to curiosity or due to cows in front of them impeding progress: however, lame cows would not be capable of achieving the equivalent high speeds through the passage regardless of circumstances. T-tests were used to investigate if there was a significant difference between these two groups.
TOPs over time were not solely correlated to mobility score. Some animals walk faster or slower and this may be related to age, temperament or other factors not explored in this study. Cows were motivated to walk through the shedding gate as feed was present beyond the monitoring area and this was considered important for reducing variation due to anything other than appetite (a behaviour which would apply to all lactating dairy cows) and lameness. Consequently, it would appear there is no fixed value that could be used to distinguish between cows with different mobility scores at an individual cow level. Furthermore, congestion in cow flow as cows passed through the shedding gate and social interactions add variation to TOPs. Therefore, the minimum TOP value over a period of five days was used to assess mobility. By comparing a cow to her baseline data built up over time it may be possible to improve the correlation between mobility score and minimum time of passage using rolling means from the preceding days. Further work is needed to investigate whether mean minimum TOP over five days is optimal and assess whether this approach applies across different types of farm and breed of cow.
The conclusion of this study is that mobility scores can be compared reliably with minimum time of passage measured between pairs of portal antennae that can detect an individual animal within a consistent and limited range.