Association between early postpartum rumination time and peak milk yield in dairy cows

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.

Use of rumination time in health risk assessment of prepartum dairy cows

The objectives of this observational cohort study were to evaluate the associations of rumination time (RT) in the last week of pregnancy with transition cow metabolism, inflammation, health, and subsequent milk production, reproduction, and culling. Pregnant nulliparous (n = 199) and parous (n = 337) cows were enrolled 21 d before the expected calving. RT and physical activity were monitored automatically by sensors from d -21 to 15 relative to calving. Blood samples were collected on d -14, -5, 4, 8, and 12 ± 1 relative to calving. Diagnoses of clinical health problems were performed by researchers from calving to 15 d in milk (DIM). In classification 1, cows were ranked based on average daily RT in the last week of pregnancy and classified into terciles as short RT (SRT), moderate RT (MRT), or long RT (LRT) for association analyses. In classification 2, RT deviation from the parity average was used in a receiver operating characteristic curve to identify the best threshold to predict postpartum clinical disease. Cows were then classified as above the threshold (AT) or below the threshold (BT). Compared with cows with LRT, cows with SRT had greater serum concentrations of NEFA (0.47 vs 0.40 ± 0.01 mmol/L), BHB (0.58 vs 0.52 ± 0.01 mmol/L), and haptoglobin (0.22 vs 0.18 ± 0.008 g/L) throughout the transition period, and reduced concentrations of glucose, cholesterol, albumin, and magnesium in a time-dependent manner. Parous cows with SRT had higher odds of postpartum clinical disease (adjusted odds ratio [AOR]: 3.7; 95% confidence interval [CI]: 2.1-6.4), lower odds of pregnancy by 210 DIM (AOR: 0.34; CI: 0.15-0.75), and lower milk production (46.9 vs 48.6 ± 0.5 kg/d) than parous cows with LRT. Deviation in prepartum RT had good predictive value for clinical disease in parous cows (area under the curve [AUC]: 0.65; CI: 0.60-0.71) but not in nulliparous (AUC: 0.51; CI: 0.42-0.59). Separation of parous cows according to the identified threshold (≤-53 min from the parity average) resulted in differences in serum concentrations of NEFA (AT = 0.31 ± 0.006, BT = 0.38 ± 0.014 mmol/L), BHB (AT = 0.49 ± 0.008, BT = 0.53 ± 0.015 mmol/L), and globulin (AT = 32.3 ± 0.3, BT = 34.8 ± 0.5 g/L) throughout the transition period, as well as in serum cholesterol, urea, magnesium, albumin, and haptoglobin in a time-dependent manner. BT parous cows had higher odds of clinical disease (AOR: 3.7; CI: 2.1-6.4), reduced hazard of pregnancy (AHR: 0.64, CI: 0.47-0.89), greater hazard of culling (AHR: 2.1, CI: 1.2-3.6), and lower milk production (46.3 ± 0.7 vs 48.5 ± 0.3 kg/d). External validation using data from 153 parous cows from a different herd and the established threshold in RT deviation (≤-53 min) resulted in similar predictive value, with the odds of postpartum disease 2.4 times greater in BT than AT (37.5 vs 20.1%). In conclusion, RT in the wk preceding calving was a reasonable predictor of postpartum health and future milk production, reproduction, and culling in parous cows but not in nulliparous cows.

Predictive modelling of deviation from expected milk yield in transition cows on automatic milking systems

The transition period is a pivotal time in the production cycle of the dairy cow. It is estimated that between 30% and 50% of all cows experience metabolic or infectious disease during this time. One of the most common and economically consequential effects of disease during the transition period is a reduction in early lactation milk production. This has led to the utilisation of deviation from expected milk yield in early lactation as a proxy measure for transition health. However, to date, this analysis has been used exclusively for the retrospective assessment of transition cow health. Statistical models capable of predicting deviations from expected milk yield may allow producers to proactively manage animals predicted to suffer negative deviations in early lactation milk production. The objective of this retrospective cohort study was first, to explore the accuracy with which cow-level production and behaviour data collected on automatic milking systems (AMS) from 1-3 days in milk (DIM) can predict deviation from expected 30-day cumulative milk yield in multiparous cows. And second, to assess the accuracy with which predicted yield deviations can classify cows into groups which may facilitate improved transition management. Production, rumination, and physical activity data from 31 commercial AMS were accessed. A 3-step analytical procedure was then conducted. In Step 1, expected cumulative yield for 1-30 DIM for each individual cow-lactation was calculated using a mixed effect linear model. In Step 2, 30-Day Yield Deviation (YD) was calculated as the difference between observed and expected cumulative yield. Lactations were then assigned to one of three groups based on their YD, RED Group (0% YD). In Step 3, yield, rumination, and physical activity data from days 1-3 in lactation were used to predict YD using machine learning models. Following external validation, YD was predicted across the test data set with a mean absolute error of 9%. Categorisation of animals suffering large negative deviations (RED group) was achieved with a specificity of 99%, sensitivity of 35%, and balanced accuracy of 67%. Our results suggest that milk yield, rumination and physical activity patterns expressed by dairy cows from 1-3 DIM have utility in the prediction of deviation from expected 30-day cumulative yield. However, these predictions currently lack the sensitivity required to classify cows reliably and completely into groups which may facilitate improved transition cow management.

The Prediction of Clinical Mastitis in Dairy Cows Based on Milk Yield, Rumination Time, and Milk Electrical Conductivity Using Machine Learning Algorithms

In commercial dairy farms, mastitis is associated with increased antimicrobial use and associated resistance, which may affect milk production. This study aimed to develop sensor-based prediction models for naturally occurring clinical bovine mastitis using nine machine learning algorithms with data from 447 mastitic and 2146 healthy cows obtained from five commercial farms in Northeast China. The variables were related to daily activity, rumination time, and daily milk yield of cows, as well as milk electrical conductivity. Both Z-standardized and non-standardized datasets pertaining to four specific stages of lactation were used to train and test prediction models. For all four subgroups, the Z-standardized dataset yielded better results than those of the non-standardized one, with the multilayer artificial neural net algorithm showing the best performance. Variables of importance had a similar rank in this algorithm, indicating the consistency of these variables as predictors for bovine mastitis in commercial farms with similar automatic systems. Moreover, the peak milk yield (PMY) of mastitic cows was significantly higher than that of healthy cows (p < 0.005), indicating that high-yielding cattle are more prone to mastitis. Our results show that machine learning algorithms are effective tools for predicting mastitis in dairy cows for immediate intervention and management in commercial farms.

Prediction of persistency for day 305 of lactation at the moment of the insemination decision

When deciding on the voluntary waiting period of an individual cow, it might be useful to have insight into the persistency for the remainder of that lactation at the moment of the insemination decision, especially for farmers who consider persistency in their reproduction management. Currently, breeding values for persistency are calculated for dairy cows but, to our knowledge, prediction models to accurately predict persistency at different moments of insemination are lacking. This study aimed to predict lactation persistency for DIM 305 at different insemination moments (DIM 50, 75, 100, and 125). Available cow and herd level data from 2005 to 2022 were collected for a total of 20,508 cows from 85 herds located in the Netherlands and Belgium. Lactation curve characteristics were estimated for every daily record using the data up to and including that day. Persistency was defined as the number of days it takes for the milk production to decrease by half during the declining stage of lactation, and calculated from the estimated lactation curve characteristic 'decay'. Four linear regression models for each of the selected insemination moment were built separately to predict decay at DIM 305 (decay-305). Independent variables included the lactation curve characteristics at the selected insemination moment, daily milk yield, age, calving season, parity group and other herd variables. The average decay-305 of primiparous cows was lower than that of multiparous cows (1.55 *10-3 vs. 2.41*10-3, equivalent to a persistency of 447 vs. 288 days, respectively). Results showed that our models had limitations in accurately predicting persistency, although predictions improved slightly at later insemination moments, with R2 values ranging between 0.27 and 0.41. It can thus be concluded that, based only on cow and herd milk production information, accurate prediction of persistency for DIM 305 is not feasible.

Association between change in body weight during early lactation and milk production in automatic milking system herds

The objective of this observational study was to investigate the association between percent body weight (BW) change in early lactation and the 90-d cumulative milk yield of dairy cows in automatic milking system (AMS) herds. Retrospective daily cow data were collected from the Lely T4C (Lely Industries, Maassluis, the Netherlands) software on 34 farms. Cows were categorized by parity into parity 1 (P1), parity 2 (P2), or parity 3 and greater (P3+). The BW change over the first 21 d of lactation was calculated as the percentage difference between the cow's average BW across d 20 through 22 and the average BW across d 2 through 4 (initial BW) postpartum. The 90-d cumulative milk yield was the outcome variable in a mixed linear regression model, with BW change, parity, their interaction, and season of calving as explanatory variables. Farm and cow nested within farm (n = 4,695) were random effects in the model. On average, cows in all 3 parity groups lost BW during the first 21 d in milk. The 21-d BW change had a negative quadratic relationship with 90-d cumulative milk yield for all parity groups; P1, P2, and P3+ cows with a 21-d BW change of -7.42%, -5.02%, and -4.52%, respectively, were more productive over 90 d in milk (P1 = 3,123 ± 52.6 kg, P2 = 4,271 ± 52.8 kg, and P3+ = 4,548 ± 52.2 kg). The findings of this study highlight the benefits of monitoring BW change in early lactation and may contribute to future research aimed to develop or improve predictive models for milk production in herds using AMS.

The Early Prediction of Common Disorders in Dairy Cows Monitored by Automatic Systems with Machine Learning Algorithms

We use multidimensional data from automated monitoring systems and milking systems to predict disorders of dairy cows by employing eight machine learning algorithms. The data included the season, days in milking, parity, age at the time of disorders, milk yield (kg/day), activity (unitless), six variables related to rumination time, and two variables related to the electrical conductivity of milk. We analyze 131 sick cows and 149 healthy cows with identical lactation days and parity; all data are collected on the same day, which corresponds to the diagnosis day for disordered cows. For disordered cows, each variable, except the ratio of rumination time from daytime to nighttime, displays a decreasing/increasing trend from d-7 or d-3 to d0 and/or d-1, with the d0, d-1, or d-2 values reaching the minimum or maximum. The test data sensitivity for three algorithms exceeded 80%, and the accuracies of the eight algorithms ranged from 65.08% to 84.21%. The area under the curve (AUC) of the three algorithms was >80%. Overall, Rpart best predicts the disorders with an accuracy, precision, and AUC of 81.58%, 92.86%, and 0.908, respectively. The machine learning algorithms may be an appropriate and powerful decision support and monitoring tool to detect herds with common health disorders.

Antimicrobial Selection for the Treatment of Clinical Mastitis and the Efficacy of Penicillin Treatment Protocols in Large Estonian Dairy Herds

Clinical mastitis (CM) is the most common microbial disease treated in dairy cows. We analyzed the antimicrobial usage in cows with CM (n = 11,420) in large dairy herds (n = 43) in Estonia. CM treatment data were collected during a 12-month study period. The antimicrobial usage was observed during the 21 days from the initiation of treatment, and the incidence of antimicrobial-treated CM was calculated for each study herd. The effect of intramammary (IMM), systemic, and combined (systemic and IMM) penicillin treatment of CM on the post-treatment somatic cell count (SCC) was analyzed using the treatment records of 2222 cows from 24 herds with a mixed multivariable linear regression model. The median incidence of antimicrobial-treated CM was 35.8 per 100 cow-years. Procaine benzylpenicillin and marbofloxacin were used in 6103 (35.5%, 95% CI 34.8-36.2) and 2839 (16.5%, 95% CI 16.0-17.1) CM treatments, respectively. Post-treatment SCC was higher after IMM penicillin therapy compared to systemic or combination therapy. Treatment of CM usually included first-choice antimicrobials, but different antimicrobial combinations were also widely used. The effect of procaine benzylpenicillin to post-treatment SCC was dependent on the administration route, cow parity, and days in milk. Further studies should evaluate the factors affecting veterinarians' choice of antimicrobial used in the treatment of CM.

Using rumination time to manage health and reproduction in dairy cattle: a review

Early detection of disease is the key to successful management of the dairy cattle which leads to timely treatment and prevention of costs associated with prolonged treatment and reduced milk yield. Electronic systems that allow for monitoring of physiological parameters like rumination, are now commercially available. This review paper discusses different aspects of rumination time that could be used to monitor the health and reproduction of dairy cattle. This review paper explored different areas where rumination time could be utilized in monitoring dairy cattle at calving, during the estrus period, during heat stressed conditions, and to detect diseases and transition cow disorders. In conclusion, rumination time could be used as an indicator of the health status in dairy cattle.

Unraveling the Relationship between Milk Yield and Quality at the Test Day with Rumination Time Recorded by a PLF Technology

Simple Summary

Precision livestock farming, by real time monitoring of dairy cows, has the potential to generate a huge amount of data to be used for farm management purposes, as well as in breeding programs. Daily rumination time (RT) recorded by commercial systems is promising in this context because it may be related to individual milk yield and composition. However, it is necessary to assess the ability of sensor data to be used in a predictive model, but also to evaluate and standardize the correct phenotypes, and how they are related to individual variability rather than from other sources. RT data and milk test day (TD) records collected from 691 cows, monitored for thirteen months, were analyzed for the already mentioned goals and to better characterize the effect of high-, medium- and low-level daily RT on milk yield and composition. Our results showed that “animal” in a farm major contributed to the RT total variability, confirming a possible use in breeding program. The higher RT class reported the best productive performance for milk and each solid yield, in spite of a small reduction in their contents, and appears to be related to a higher degree of saturation in the fatty acid profile.

Abstract

The study aimed to estimate the components of rumination time (RT) variability recorded by a neck collar sensor and the relationship between RT and milk composition. Milk test day (TD) and RT data were collected from 691 cows in three farms. Daily RT data of each animal were averaged for 3, 7, and 10 days preceding the TD date (RT_D). Variance component analysis of RT_D, considering the effects of farm, cow, parity, TD date, and lactation phase, showed that a farm, followed by a cow, had major contributions to the total variability. The RT₁₀ variable best performed on TD milk yield and quality records across models by a multi-model inference approach and was adopted to study its relationship with milk traits, by linear mixed models, through a 3-level stratification: low (LRT₁₀ ≤ 8 h/day), medium (8 h/day < MRT₁₀ ≤ 9 h/day), and high (HRT₁₀ > 9 h/day) RT. Cows with HRT₁₀ had greater milk, fat, protein, casein, and lactose daily yield, and lower fat, protein, casein contents, and fat to protein ratio compared to MRT₁₀ and LRT₁₀. Higher percentages of saturated fatty acid and lower unsaturated and monounsaturated fatty acid were found in HRT₁₀, with respect to LRT₁₀ and MRT₁₀ observations.