Automatic monitoring of the health and metabolic status of dairy cows

Abstract The aim of this study was to prioritise the objectives of research into new sensing systems to monitor routinely the health of dairy cows. An analysis of reasons for death and culling of cows and a survey of available sensor techniques suggested that the most suitable cow health events for routine monitoring are parturition, mastitis, and the malfunction of energy, protein, and mineral metabolism. Physical events such as parturition could be detected by image analysis and acoustic monitoring. Nutritional and metabolic conditions could be monitored by chemometric techniques to detect analytes such as acetoacetate or urea in milk and acetone in breath. Viral and bacterial proteins and hormones such as progesterone, could be detected automatically by antibody-specific biosensors provided that a suitable sample gathering system can be developed. A time sequence is proposed here to classify a hierarchy of events occurring in health or metabolic disturbance. It has been proposed that there is a primary event which should be the target for monitoring. Monitoring methods which rely on transducers to detect secondary and tertiary events, such as conductivity, milk yield change and temperature, have limited specificity and high numbers of false positives unless supported by robust models to integrate data from a number of sources. Although the technology to sense particular conditions could be developed, it is not clear whether veterinary and nutritional models are sufficiently well developed to allow decisions to be made based on the sensor data alone. Research is needed to determine the costs and reduced losses which may arise from early detection and treatment of disease. Models should include the subsequent risks for other health conditions, for example the effect of sub-clinical ketosis on the risk of mastitis or infertility.