Author:

Abstract:

Lameness is an important health problem causing severe welfare deteriorations and economic losses up to € 53 per cow per year in dairy cattle. Timely detection and treatment can help to minimize economic losses and preserve cow welfare. Current visual detection methods are labor intensive and subjective, and require training to allow detection of subtle changes in a cow’s gait. As a result, the problem is underestimated in practice, and lameness is often detected in a late stage when losses have already run high. Due to further intensification in the dairy sector, less time will be available to spend on monitoring individual animals in the future, implying that more objective and less time consuming methods are desirable. Automatic lameness detection systems can be a solution, and may enable early detection and treatment, but no truly cost-efficient systems are currently available on the market. The development and market introduction of existing prototypes is being held up by the unknown economic value and maximum investment cost of such systems, and the lack of knowledge on the potential adoption rate and farmers’ preferences concerning lameness detection performance and system cost. Promising results have been reported, but prototypes are often costly and some are difficult to implement in existing dairy barns, whereas their detection performance still seems insufficient for use in practice. Therefore, in this PhD research, user-centric design criteria for further development of existing prototypes into market-ready lameness detection systems were derived. It was investigated which factors influence economic value, and how this economic value can be quantified for specific farms and systems. Farmers’ preferences for the detection performance and cost of an automatic lameness detection system were investigated using a choice-experiment. Simultaneously, the effect of providing extra information on lameness and its consequences to the farmer was investigated. The gathered information was used to define how system developers could use this information to further develop existing prototypes, and to get an idea on the current adoption potential of automatic lameness detection in the Flemish dairy sector. An attempt was made to implement the derived design criteria in a walkover pressure mat by lowering the system cost and spatial requirements to increase the easy with which the system can be implemented in practice. In addition, new automatically measured gait variables that describe cow gait were derived from this sensor and used in new, improved individual lameness detection algorithms. Analysis of the economic value indicated several knowledge gaps that impede accurate economic value calculations. Especially the effect of early detection and treatment on the economic losses caused by lameness and the unknown system lifespan were important unknown drivers for economic value. System-specific and farm-specific information was incorporated to account for the fact that system cost and detection performance as well as the herd size can influence the economic value of a lameness detection system substantially. Automatic lameness detection systems proved capable of generating positive economic values, but the assumptions made to estimate the economic value should be kept in mind. The choice experiment led to the conclusion that dairy farmers prefer systems that miss few lame cows with a low number of false alarms for a low cost. Systems capable to indicate which leg is lame were preferred over systems that did not have this feature. Flemish dairy farmers were willing to pay more for a system with better detection performance. In general, visual detection was still preferred over automatic detection, except for those farmers who already have experience with automatic estrus detection systems. It was concluded that the detection performance should be sufficiently high for farmers to consider investing in an automatic lameness detection system. Providing extra information on lameness influenced farmers’ preferences positively, implying that sensitizing actions can improve the future adoption rate of automatic lameness detection systems. Also, the adoption can be supported by making systems cheaper, and by improving their detection performance. It was concluded that the Gaitwise sensor can be shortened from 4.88 to 3.28 meter to decrease the system cost and increase the ease of implementation in existing dairy barns. The sensor resolution can be lowered without affecting the lameness detection performance to reduce system cost further, leading to an estimated total cost reduction of 83 %. New variables describing how cows distribute their weight in time, and how within-stance times change as a result of lameness were derived. The new variables indeed differed between non-lame and lame cows, implying that they can be interesting to use in lameness detection algorithms. Finally, a new monitoring setup was built, and daily automatic measurements were executed with a walkover pressure mat (Gaitwise) to allow for the development of new detection algorithms with higher detection performance. The influence of environmental factors that affect cow gait, such as darkness and slipperiness of the walking surface, was reduced as much as possible. In a first step, a detection model based on group thresholds was developed, resulting in a still insufficient detection performance with a sensitivity of 36.9 % and specificity of 86.9 %. Cows were often distracted during measurements, implying that gait patterns of non-lame and lame cows could not easily be differentiated. In a second step, cow gait was monitored individually, but due to many missing and failed measurements resulting in a measurement success rate of 27.6 %, it was not possible to develop well-working individual detection algorithms. Nevertheless, suggestions were formulated to improve sensor implementation and data collection in the future to allow for better individual monitoring. Suggestions included keeping the number of obstacles and distractions as low as possible, and motivating cows to walk at a sufficiently high pace. Future research could use the presented results to support further development and adoption of automatic lameness detection systems in practice. Drivers for economic value should be investigated further to allow for more accurate estimations of the economic value, which can subsequently be used to define development goals. The economic value can be increased by lowering system cost and improving detection performance, and by integration of the used technologies with other health monitoring systems. However, future research should also use the presented results to investigate which preventive and other lameness-reducing measures should be incorporated in good lameness management, and whether further development is still feasible for all existing automatic lameness detection system prototypes.