Short communication: Prediction of hyperketonemia in dairy cows in early lactation using on-farm cow data and net energy intake by partial least square discriminant analysis

Evaluating the performance of herd-specific Long Short-Term Memory models to identify automated health alerts associated with a ketosis diagnosis in early lactation cows

The growing use of automated systems in the dairy industry generates a vast amount of cow-level data daily, creating opportunities for using these data to support real-time decision-making. Currently, various commercial systems offer built-in alert algorithms to identify cows requiring attention. To our knowledge, no work has been done to compare the use of models accounting for herd-level variability on their predictive ability against automated systems. Long Short-Term Memory (LSTM) models are machine learning models capable of learning temporal patterns and making predictions based on time series data. The objective of our study was to evaluate the ability of LSTM models to identify a health alert associated with a ketosis diagnosis (HAK) using deviations of daily milk yield, milk FPR, number of successful milkings, rumination time, and activity index from the herd median by parity and DIM, considering various time series lengths and numbers of d before HAK. Additionally, we aimed to use Explainable Artificial Intelligence method to understand the relationships between input variables and model outputs. Data on daily milk yield, milk fat-to-protein ratio (FPR), number of successful milkings, rumination time, activity, and health events during 0 to 21 d in milk (DIM) were retrospectively obtained from a commercial Holstein dairy farm in northern Indiana from February 2020 to January 2023. A total of 1,743 cows were included in the analysis (non-HAK = 1,550; HAK = 193). Variables were transformed based on deviations from the herd median by parity and DIM. Six LSTM models were developed to identify HAK 1, 2, and 3 d before farm diagnosis using historic cow-level data with varying time series lengths. Model performance was assessed using repeated stratified 10-fold cross-validation for 20 repeats. The Shapley additive explanations framework (SHAP) was used for model explanation. Model accuracy was 83, 74, and 70%, balanced error rate was 17 to 18, 26 to 28, and 34%, sensitivity was 81 to 83, 71 to 74, and 62%, specificity was 83, 74, and 71%, positive predictive value was 38, 25 to 27, and 21%, negative predictive value was 97 to 98, 95 to 96, and 94%, and area under the curve was 0.89 to 0.90, 0.80 to 0.81, and 0.72 for models identifying HAK 1, 2, and 3 d before diagnosis, respectively. Performance declined as the time interval between identification and farm diagnosis increased, and extending the time series length did not improve model performance. Model explanation revealed that cows with lower milk yield, number of successful milkings, rumination time, and activity, and higher milk FPR compared with herdmates of the same parity and DIM were more likely to be classified as HAK. Our results demonstrate the potential of LSTM models in identifying HAK using deviations of daily milk production variables, rumination time, and activity index from the herd median by parity and DIM. Future studies are needed to evaluate the performance of health alerts using LSTM models controlling for herd-specific metrics against commercial built-in algorithms in multiple farms and for other disorders.

The Use of Multilayer Perceptron Artificial Neural Networks to Detect Dairy Cows at Risk of Ketosis

Simple Summary

Ketosis is a serious metabolic disease in high-yield dairy cows, that affects productive herds throughout the world. Subclinical ketosis is one of the most dominant metabolic disorders in dairy herds during early lactation, so early detection and prevention are important for both economic and animal welfare reasons. Neural networks, which offer a high degree of accuracy in predicting various phenomena and processes where there is no clear causal correlation or there are no rules that allow the establishment of a logical cause-and-effect relationship, can be used to address problems related to prediction, classification, or control. A Multi-Layer perceptron (MLP) is a feedforward artificial neural network model that takes input data for a set of proper output. This study investigated the performance of four algorithms used to train MLP networks. The experimental results demonstrate that the MLP network model improved the accuracy of process recognition of subclinical ketosis in dairy cows. The received artificial model’s results were saved in the predictive model markup language (PMML) and can be used to describe the learning set, the algorithm used in the data mining application and related information.

Abstract

Subclinical ketosis is one of the most dominant metabolic disorders in dairy herds during lactation. Cows suffering from ketosis experience elevated ketone body levels in blood and milk, including β-hydroxybutyric acid (BHB), acetone (ACE) and acetoacetic acid. Ketosis causes serious financial losses to dairy cattle breeders and milk producers due to the costs of diagnosis and management as well as animal welfare reasons. Recent years have seen a growing interest in the use of artificial neural networks (ANNs) in various fields of science. ANNs offer a modeling method that enables the mapping of highly complex functional relationships. The purpose of this study was to determine the relationship between milk composition and blood BHB levels associated with subclinical ketosis in dairy cows, using feedforward multilayer perceptron (MLP) artificial neural networks. The results were verified based on the estimated sensitivity and specificity of selected network models, an optimum cut-off point was identified for the receiver operating characteristic (ROC) curve and the area under the ROC curve (AUC). The study demonstrated that BHB, ACE and lactose (LAC) levels, as well as the fat-to-protein ratio in milk, were important input variables in the network training process. For the identification of cows at risk of subclinical ketosis, variables such as BHB and ACE levels in milk were of particular relevance, with a sensitivity and specificity of 0.84 and 0.61, respectively. It was found that the back propagation algorithm offers opportunities to integrate artificial intelligence and dairy cattle welfare within a computerized decision support tool.

The Whole and the Parts—A New Perspective on Production Diseases and Economic Sustainability in Dairy Farming

The levels of production diseases (PD) and the cow replacement rate are high in dairy farming. They indicate excessive production demands on the cow and a poor state of animal welfare. This is the subject of increasing public debate. The purpose of this study was to assess the effect of production diseases on the economic sustainability of dairy farms. The contributions of individual culled cows to the farm’s economic performance were calculated, based on milk recording and accounting data from 32 farms in Germany. Cows were identified as ‘profit cows’ when they reached their individual ‘break-even point’. Data from milk recordings (yield and indicators for PD) were used to cluster farms by means of a principal component and a cluster analysis. The analysis revealed five clusters of farms. The average proportion of profit cows was 57.5%, 55.6%, 44.1%, 29.4% and 19.5%. Clusters characterized by a high proportion of cows with metabolic problems and high culling and mortality rates had lower proportions of profit cows, somewhat irrespective of the average milk-yield per cow. Changing the perception of PD from considering it as collateral damage to a threat to the farms’ economic viability might foster change processes to reduce production diseases.

Monitoring and Improving the Metabolic Health of Dairy Cows during the Transition Period

Simple Summary

The transition from late gestation to early lactation is a challenging period for dairy cows. A successful transition period depends on metabolic adaptation to the new physiological state in early lactation and proper management in order to support the cow’s requirements. This review paper will discuss various aspects of routine and consistent approaches to collect and analyze herd records, to detect unintended disruptions in performance. In addition, we discuss how to incorporate methods to assess health, production, nutrition, and welfare information to monitor cows during the transition period. Lastly, we discuss management strategies that can be implemented to improve the metabolic health and performance of transition dairy cows.

Abstract

The peripartum period of a dairy cow is characterized by several physiological and behavioral changes in response to a rapid increase in nutrient demands, to support the final stages of fetal growth and the production of colostrum and milk. Traditionally, the transition period is defined as the period 3 weeks before and 3 weeks after parturition. However, several researchers have argued that the transition period begins at the time of dry-off (~60–50 days prior to calving) and extends beyond the first month post-calving in high producing dairy cows. Independent of the definition used, adequate adaptation to the physiological demands of this period is paramount for a successful lactation. Nonetheless, not all cows are successful in transitioning from late gestation to early lactation, leading to approximately one third of dairy cows having at least one clinical disease (metabolic and/or infectious) and more than half of the cows having at least one subclinical case of disease within the first 90 days of lactation. Thus, monitoring dairy cows during this period is essential to detect early disease signs, diagnose clinical and subclinical diseases, and initiate targeted health management to avoid health and production impairment. In this review, we discuss different strategies to monitor dairy cows to detected unintended disruptions in performance and management strategies that can be implemented to improve the metabolic health and performance of dairy cows during the transition period.