Lameness scoring system for dairy cows using force plates and artificial intelligence

Abstract

Lameness scoring is a routine procedure in dairy industry to screen the herds for new cases of lameness. Subjective lameness scoring, which is the most popular lameness detection and screening method in dairy herds, has several limitations. They include low intra-observer and inter-observer agreement and the discrete nature of the scores which limits its usage in monitoring the lameness. The aim of this study is to develop an automated lameness scoring system comparable with conventional subjective lameness scoring by means of artificial neural networks. The system is composed of four balanced force plates installed in a hoof-trimming box. A group of 105 dairy cows was used for the study. Twenty-three features extracted from ground reaction force (GRF) data were used in a computer training process which was performed on 60 per cent of the data. The remaining 40 per cent of the data were used to test the trained system. Repeatability of the lameness scoring system was determined by GRF samples from 25 cows, captured at two different times from the same animals. The mean sd was 0.31 and the mean coefficient of variation was 14.55 per cent, which represents a high repeatability in comparison with subjective vision-based scoring methods. Although the highest sensitivity and specificity values were seen in locomotion score groups 1 and 4, the automatic lameness system was both sensitive and specific in all groups. The sensitivity and specificity were higher than 72 per cent in locomotion score groups 1 to 4, and it was 100 per cent specific and 50 per cent sensitive for group 5.

Lameness detection challenges in automated milking systems addressed with partial least squares discriminant analysis

“…Multivariate methods (e.g., principal component analysis) have also been used for analysis in the animal science field (Bro et al, 2002;Dumas et al, 2005;Miekley et al, 2013), often focusing on data reduction (Sloth et al, 2003;Gorzecka et al, 2011). Based on automated data collection in dairy herds, pattern recognition has been the objective in several studies using principal component analysis, neural networks, or classification trees (Nielen et al, 1995;Klaas et al, 2004;Cavero et al, 2008;Ghotoorlar et al, 2012;Piwczy ski et al, 2013).…”

Evaluation of a gait scoring system for cattle by using cluster analysis and Krippendorff's α reliability

“…For instance, Pastell et al 17 based their study on an accelerometer system on a preselection of cows using an LS. Also, the use of a four‐force plate‐balanced system as a technique to detect lameness presupposed the classification of the cows depending on their previously fixed LS 18 . Van Hartem et al 19 defined a visual five‐point LS as reference for automatic lameness detection based on consecutive 3D‐video recording.…”

Space-use patterns highlight behavioural differences linked to lameness, parity, and days in milk in barn-housed dairy cows

“…Increasing demand for animal products and intensification of farming practices in general, means that there is a need for automated behavioural monitoring systems that can act as an ‘early warning’ to detect and predict the health status of managed animals, including dairy cows suffering from lameness and other diseases [7,8,26,83]. Automated lameness detection technology systems have been developed based on the identification of an abnormality of gait or posture [83], using force plate technology [7,84] or kinematics [85]. Meanwhile, automated monitoring of feeding behaviour in cattle has relied on electronic feed troughs [27,30].…”