Parimal G. Rajkondawar

The development of an objective lameness scoring system for dairy herds: Pilot study

Early detection of bovine lameness offers the potential for effective treatment and effectual management of hoof and leg ailments. This technical note examines statistical relationships between visually derived lameness scores and mechanically derived limb movement variables (LMVs) for sound and lame dairy cows. The visually derived lameness scores were based on observations of arched backs while cows stood and walked. The mechanically derived LMVs were captured as cows walked freely over a patent–pending force–plate system that generates signatures of ground reaction forces. A statistical model evaluated a lameness index (LI) using peak ground reaction force (PGRF). The LI allowed the 23 cows to be classified correctly according to their visually derived lameness scores, with three exceptions. This pilot study demonstrated that the outputs of our force–plate system can be used to relate lameness scores to measurable LMVs.

Enhancing the prediction accuracy of bovine lameness models through transformations of limb movement variables.

The issue of modeling bovine lameness was explored by testing the hypothesis that B-spline transformation of limb movement variables (LMV) employed in predictive models improved model accuracy. The objectives were to determine the effect of number of B-spline knots and the degree of the underlying polynomial approximation (degree of freedom) on model accuracy. Knot number used in B-spline transformation improved model accuracy by improving model specificity and to a lesser extent model sensitivity. Degree of polynomial approximation had no effect on model predictive accuracy from the data set of 261 cows. Model stability, defined as changes in predictive accuracy associated with the superimposition of perturbations (0.5 and 1.0%) in LMV on the measured data, was explored. Model specificity and to a lesser degree, sensitivity, increased with increased knot number across data set perturbations. Specificity and sensitivity increased by 43 and 11%, respectively, when knot number increased from 0 to 7 for a perturbation level of 0.5%. When the perturbation level was 1%, the corresponding increases in specificity and sensitivity were 32 and 4%, respectively. Nevertheless, different levels of LMV perturbation varied the optimal knot number associated with highest model accuracy. The optimal knot number for 0.5% perturbation was 8, whereas for 1% perturbation the optimal knot number was 7. The B-spline transformation improved specificity and sensitivity of predictive models for lameness, provided the appropriate number of knots was selected.

Diversity in the magnitude of hind limb unloading occurs with similar forms of lameness in dairy cows

The objective of the study was to evaluate the relationship of veterinary clinical assessments of lameness to probability estimates of lameness predicted from vertical kinetic measures. We hypothesized that algorithm-derived probability estimates of lameness would accurately reflect vertical measures in lame limbs even though vertical changes may not inevitably occur in all lameness. Kinetic data were collected from sound (n=179) and unilaterally lame (n=167) dairy cattle with a 1-dimensional, parallel force plate system that registered vertical ground reaction force signatures of all four limbs as cows freely exited the milking parlour. Locomotion was scored for each hind limb using a 1-5 locomotion score system (1=sound, 5=severely lame). Pain response in the interdigital space was quantified with an algometer and pain response in the claw was quantified with a hoof tester fitted with a pressure gage. Lesions were assigned severity scores (1=minimal pathology to 5=severe pathology). Lameness diminished the magnitude of peak ground reaction forces, average ground reaction forces, Fourier transformed ground reaction forces, stance times and vertical impulses in the lame limbs of unilaterally lame cows. The only effect of lameness on the opposite sound limb was increased magnitude of stance times and vertical impulses in unilaterally lame cows. Symmetry measures of the peak ground reaction forces, average ground reaction forces, Fourier transformed ground reaction forces, stance times and vertical impulses between the left and right hind limbs were also affected in unilateral lameness. Paradoxically, limbs with clinically similar lesion and locomotion scores and pain responses were associated with a broad range of load-transfer off the limb. Substantial unloading and changes in the vertical limb variables occurred in some lameness while minimal unloading and changes in vertical limb variables occurred in other lameness. Corresponding probability estimates of lameness accurately reflected changes in the vertical parameters of limbs and generated low probability estimates of lameness when minimal unloading occurred. Failure to transfer load off limbs with pain reactions, locomotion abnormalities and lesions explained much of the limited sensitivity in lameness detection with vertical limb variables.

Predictive models of lameness in dairy cows achieve high sensitivity and specificity with force measurements in three dimensions

Lameness remains a significant cause of production losses, a growing welfare concern and may be a greater economic burden than clinical mastitis . A growing need for accurate, continuous automated detection systems continues because US prevalence of lameness is 12.5% while individual herds may experience prevalences of 27.8-50.8%. To that end the first force-plate system restricted to the vertical dimension identified lame cows with 85% specificity and 52% sensitivity. These results lead to the hypothesis that addition of transverse and longitudinal dimensions could improve sensitivity of lameness detection. To address the hypothesis we upgraded the original force plate system to measure ground reaction forces (GRFs) across three directions. GRFs and locomotion scores were generated from randomly selected cows and logistic regression was used to develop a model that characterised relationships of locomotion scores to the GRFs. This preliminary study showed 76 variables across 3 dimensions produced a model with greater than 90% sensitivity, specificity, and area under the receiver operating curve (AUC). The result was a marked improvement on the 52% sensitivity, and 85% specificity previously observed with the 1 dimensional model or the 45% sensitivities reported with visual observations. Validation of model accuracy continues with the goal to finalise accurate automated methods of lameness detection.