Lameness models that use spline transformation to enhance their performance: Theory and field testing

Lameness is an important welfare and costly production problem for the dairy industry. Losses due to lameness are caused by diminished milk yields, loss of reproductive efficiency and increased involuntary culling. These costs could be reduced with early detection and proper treatment of the ailment. StepMetrix, an automatic, early lameness detection system for dairy cattle has been developed for this purpose. The automatic lameness detection of this system is ideal for the increasing farm sizes, the demand for more efficient production and the growth of labor costs in a modern dairy environment.;The aim of this thesis was to improve the performance of the lameness prediction model currently used in the StepMetrix system by using B-Spline transformations of limb movement variables. Seven lameness prediction models (including the StepMetrix lameness prediction model) were derived and tested using various databases. The issue of modeling bovine lameness was explored by testing the hypothesis that B-spline transformation of limb movement variables employed in predictive models improved model accuracy. The objectives were to determine the effect of the number of B-Spline knots and the degree of the underlying polynomial approximation (degree of freedom) in B-Spline transformation on model accuracy.;The testing results showed that the lameness prediction models using B-Spline transformed limb movement variables did not outperform the StepMetrix lameness prediction model. However, by applying the StepMetrix lameness prediction model and lameness prediction models using B-Spline transformed limb movement variables in series, model performance was improved in sensitivity (from 61% to 65.9%) as well as in specificity (from 64.3% to 70%). Overall, this thesis demonstrated that models using B-Spline transformation can be used in combination with the StepMetrix lameness prediction model to improve the model performance of the StepMetrix system.