Sensors and clinical mastitis--the quest for the perfect alert

Cow key point detection in indoor housing conditions with a deep learning model

Lameness in dairy cattle is a costly and highly prevalent problem that affects all aspects of sustainable dairy production, including animal welfare. Automation of gait assessment would allow monitoring of locomotion in which the cows' walking patterns can be evaluated frequently and with limited labor. With the right interpretation algorithms, this could result in more timely detection of locomotion problems. This in turn would facilitate timely intervention and early treatment, which is crucial to reduce the effect of abnormal behavior and pain on animal welfare. Gait features of dairy cows can potentially be derived from key points that locate crucial anatomical points on a cow's body. The aim of this study is 2-fold: (1) to demonstrate automation of the detection of dairy cows' key points in a practical indoor setting with natural occlusions from gates and races, and (2) to propose the necessary steps to postprocess these key points to make them suitable for subsequent gait feature calculations. Both the automated detection of key points as well as the postprocessing of them are crucial prerequisites for camera-based automated locomotion monitoring in a real farm environment. Side-view video footage of 34 Holstein-Friesian dairy cows, captured when exiting the milking parlor, were used for model development. From these videos, 758 samples of 2 successive frames were extracted. A previously developed deep learning model called T-LEAP was trained to detect 17 key points on cows in our indoor farm environment with natural occlusions. To this end, the dataset of 758 samples was randomly split into a train (n = 22 cows; no. of samples = 388), validation (n = 7 cows; no. of samples = 108), and test dataset (n = 15 cows; no. of samples = 262). The performance of T-LEAP to automatically assign key points in our indoor situation was assessed using the average percentage of correctly detected key points using a threshold of 0.2 of the head length (PCKh0.2). The model's performance on the test set achieved a good result with PCKh0.2: 89% on all 17 key points together. Detecting key points on the back (n = 3 key points) of the cow had the poorest performance PCKh0.2: 59%. In addition to the indoor performance of the model, a more detailed study of the detection performance was conducted to formulate postprocessing steps necessary to use these key points for gait feature calculations and subsequent automated locomotion monitoring. This detailed study included the evaluation of the detection performance in multiple directions. This study revealed that the performance of the key points on a cows' back were the poorest in the horizontal direction. Based on this more in-depth study, we recommend the implementation of the outlined postprocessing techniques to address the following issues: (1) correcting camera distortion, (2) rectifying erroneous key point detection, and (3) establishing the necessary procedures for translating hoof key points into gait features.

Pathogen-specific patterns of milking traits in automatic milking systems

Early detection of intramammary infection (IMI) can improve animal health and welfare in dairy herds. The implementation of sensors and automatic milking systems (AMS) in dairy production inherently increases the amount of available data and hence also the potential for new approaches to mastitis management. To utilize the full potential of data from AMS and auxiliary sensors, a better understanding of physiological and pathological changes in milking traits associated with different udder pathogens may be imperative. This observational study aimed to investigate pathogen-specific patterns in milking traits recorded in AMS. The milking traits included; online somatic cell count (OCC), electrical conductivity (EC), milk yield (MY), and average milk flow rate (AMF). Data were collected for a study period of 2 years and included 101 492 milkings from 237 lactations in 169 cows from one farm. Measurements of OCC were recorded at cow-level and data on EC, MY, and AMF were obtained at quarter-level. In addition to the data obtained from the AMS, altogether 5756 quarter milk samples (QMS) were collected. Milk samples were obtained monthly for bacteriological culturing. We included findings of 13 known mastitis pathogens to study pathogen-specific patterns in milking traits. These patterns were compared with those in a baseline group consisting of cows that did not have any positive milk culture results throughout the lactation period. Patterns of the milking traits are described for all positive samples both across 305 d in milk (DIM), and in the 15-d period before a positive bacteriological sample. The association between a positive sample and the milking traits (ln(OCC), EC-IQR; the ratio between the quarter with the highest and the quarter with the lowest level of EC, and MY) for the 15 d before the detection of a pathogen was assessed using mixed effects linear regression models. All pathogens were associated with alterations in the level and variability of ln(OCC) relative to lactations with no positive bacteriological samples. A positive sample for Staph. aureus was associated with increased values for MY during the 15 d before a positive diagnosis. It is biologically plausible to interpret changes in OCC and EC-IQR as consequences of an intramammary infection (IMI), while higher MY in bacteriologically-positive cows is most likely linked to the increased risk of infection in high-yielding cows. In this study, the most notable changes in the traits (OCC and EC-IQR) were observed for Staph. aureus and Strep. dysgalactiae, followed by Strep. simulans, Strep. uberis, and Lactococcus lactis. Even if we did not detect significant associations between positive bacteriology and EC-IQR, visual assessment and descriptive statistics indicated that there might be differences suggesting that it could be an informative trait for detecting infection when combined with OCC and possibly other relevant traits using machine learning algorithms.

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.

Evaluation of automated monitoring calving prediction in dairy buffaloes a new tool for calving management

Abstract Buffalo is one of the leading milk-producing dairy animals. Its production and reproduction are affected due to some factors including inadequate monitoring around parturition, which cause economic losses like delayed birth process, increased risk of stillbirth, etc. The appropriate calving monitoring is essential for dairy herd management. Therefore, we designed a study its aim was, to predict the calving based on automated machine measured prepartum behaviors in buffaloes. The data were collected from n=40 pregnant buffaloes of 2nd to 5th parity, which was synchronized. The NEDAP neck and leg logger tag was attached to each buffalo at 30 days before calving and automatically collected feeding, rumination, lying, standing, no. of steps, no. of switches from standing to lying (lying bouts) and total motion activity. All behavioral data were reduced to -10 days before the calving date for statistical analysis to use mixed model procedure and ANOVA. Results showed that feeding and rumination time significantly (P<0.05) decreased from -10 to -1 days before calving indicating calving prediction. Moreover, Rumination time was at lowest (P<0.001) value at 2h before the calving such behavioral changes may be useful to predict calving in buffaloes. Similarly, lying bouts and standing time abruptly decreased (P<0.05) from -3 to -1 days before calving, while lying time abruptly increased (P<0.01) from -3 to -1 days before calving (531.57±23.65 to 665.62±18.14, respectively). No. of steps taken and total motion significantly (P<0.05) increased from -10 to -1 days before calving. Feeding time was significantly (P<0.02) lowered in 3rd parity buffaloes compared with 2nd, 4th and 5th parity buffaloes, while standing time of 5th parity buffaloes were lowered (P<0.05) as compared to 2nd to 4th parity buffalos at -1 day of prepartum. However, rumination, lying, no. of steps taken and total motion activity at -1 day of prepartum was independent (P>0.05) of parity in buffaloes. Neural network analysis for combined variables from NEDAP technology at the daily level yielded 100.0% sensitivity and 98% specificity. In conclusion NEDAP technology can be used to measured behavioral changes -10 day before calving as it can serve as a useful guide in the prediction calving date in the buffaloes.

Risk prediction model of clinical mastitis in lactating dairy cows based on machine learning algorithms

Mastitis is the most common disease among dairy cows and is known to have negative effects on both animal welfare and the profitability of dairy farms. Early detection of clinical mastitis cases is considered the best option for preventing cows from developing mastitis. In this study, we developed clinical mastitis prediction models that only required inputting common indicators from the automatic milking system. We utilized multidimensional data from the cow mastitis database of Afimilk (China) Agricultural Technology Co., Ltd. to predict mastitis in dairy cows. All data were screened for the period of 0-150 days of lactation. The data included parity, lactation day, period, mean and standard deviation of milk yield, of electrical conductivity, and of lying time, which were taken as input features. The classification of whether cows suffer from clinical mastitis was determined as output. We analyzed 426 cows with clinical mastitis and 2087 healthy cows by using four machine learning algorithms: Decision Tree, Random Forest, Back Propagation neural networks, and Support Vector Machines. In these four algorithms, the accuracy ranged from 94% to 98%, while the running times varied widely from seconds to minutes. The decision tree prediction model achieved an accuracy of 98% and the precision rate for healthy cows was 99%, while for mastitis cows it was 97%. Machine learning algorithms have played an important role in predicting cow mastitis, with the Decision Tree algorithm showing great performance and higher accuracy in our research.

Nanotechnology and E-Sensing for Food Chain Quality and Safety

Nowadays, it is well known that sensors have an enormous impact on our life, using streams of data to make life-changing decisions. Every single aspect of our day is monitored via thousands of sensors, and the benefits we can obtain are enormous. With the increasing demand for food quality, food safety has become one of the main focuses of our society. However, fresh foods are subject to spoilage due to the action of microorganisms, enzymes, and oxidation during storage. Nanotechnology can be applied in the food industry to support packaged products and extend their shelf life. Chemical composition and sensory attributes are quality markers which require innovative assessment methods, as existing ones are rather difficult to implement, labour-intensive, and expensive. E-sensing devices, such as vision systems, electronic noses, and electronic tongues, overcome many of these drawbacks. Nanotechnology holds great promise to provide benefits not just within food products but also around food products. In fact, nanotechnology introduces new chances for innovation in the food industry at immense speed. This review describes the food application fields of nanotechnologies; in particular, metal oxide sensors (MOS) will be presented.

Review: Using animal welfare to frame discussion on dairy farm technology

The use of technology on dairy farms has increased dramatically over the last half-century. The ways that scientists describe the potential benefits and risk of technology are likely to affect if it is accepted for use on farms. The aim of our study was to identify papers that describe a linkage between technologies used on dairy farms and the welfare of dairy cattle. Our systematic review identified 380 papers, of which 28 met our inclusion criteria and were used to describe the technologies examined, the welfare-relevant measures used, and the ways in which authors framed welfare benefits and risks associated with the technologies. The large majority (27 of 28 papers) used positive frames, considering how the technology could improve welfare. Some authors carefully articulated the logic linking the specific measures to specific welfare-related outcomes (such as the use of accelerometer data to draw inferences about changes in lying times), but others made more general inferences (about health and welfare) that were not (and perhaps could not) be assessed. We conclude that much of the framing focused on animal welfare is biased toward welfare benefits and that future work should strive to address both potential benefits and harms; more balanced coverage may better inform solutions to the problems encountered by the people and animals affected by the technology. Welfare is a complex and multifaced concept, and it is unlikely that any one technology (or perhaps even a combination of technologies) can adequately capture this complexity. Thus, we encourage authors to restrict their claims to specific, welfare-relevant measures that can be assessed independently and thus validated. More general claims about welfare should be treated with skepticism.

Scientific Productions on Precision Livestock Farming: An Overview of the Evolution and Current State of Research Based on a Bibliometric Analysis

The interest in precision livestock farming (PLF)-a concept discussed for the first time in the early 2000s-has advanced considerably in recent years due to its important role in the development of sustainable livestock production systems. However, a comprehensive bibliometric analysis of the PLF literature is lacking. To address this gap, this study analyzed documents published from 2005 to 2021, aiming to understand the historical influences on technology adoption in livestock farming, identify future global trends, and examine shifts in scientific research on this topic. By using specific search terms in the Web of Science Core Collection, 886 publications were identified and analyzed using the bibliometrix R-package. The analysis revealed that the collection consisted mostly of research articles (74.6%) and reviews (10.4%). The top three core journals were the Journal of Dairy Science, Computers and Electronics in Agriculture, and Animals. Over time, the number of publications has steadily increased, with a higher growth rate in the last five years (29.0%) compared to the initial period (13.7%). Authors and institutions from multiple countries have contributed to the literature, with the USA, the Netherlands, and Italy leading in terms of publication numbers. The analysis also highlighted the growing interest in bovine production systems, emphasizing the importance of behavioral studies in PLF tool development. Automated milking systems were identified as central drivers of innovation in the PLF sector. Emerging themes for the future included "emissions" and "mitigation", indicating a focus on environmental concerns.

Can We Reliably Detect Respiratory Diseases through Precision Farming? A Systematic Review

Respiratory diseases commonly affect livestock species, negatively impacting animal's productivity and welfare. The use of precision livestock farming (PLF) applied in respiratory disease detection has been developed for several species. The aim of this systematic review was to evaluate if PLF technologies can reliably monitor clinical signs or detect cases of respiratory diseases. A technology was considered reliable if high performance was achieved (sensitivity > 90% and specificity or precision > 90%) under field conditions and using a reliable reference test. Risk of bias was assessed, and only technologies tested in studies with low risk of bias were considered reliable. From 23 studies included-swine (13), poultry (6), and bovine (4) -only three complied with our reliability criteria; however, two of these were considered to have a high risk of bias. Thus, only one swine technology fully fit our criteria. Future studies should include field tests and use previously validated reference tests to assess technology's performance. In conclusion, relying completely on PLF for monitoring respiratory diseases is still a challenge, though several technologies are promising, having high performance in field tests.

Multivariable time series classification for clinical mastitis detection and prediction in automated milking systems

In this study, we developed a machine learning framework to detect clinical mastitis (CM) at the current milking (i.e., the same milking) and predict CM at the next milking (i.e., one milking before CM occurrence) at the quarter level. Time series quarter-level milking data were extracted from an automated milking system (AMS). For both CM detection and prediction, the best classification performance was obtained from the decision tree-based ensemble models. Moreover, applying models on a data set containing data from the current milking and past 9 milkings before the current milking showed the best accuracy for detecting CM; modeling with a data set containing data from the current milking and past 7 milkings before the current milking yielded the best results for predicting CM. The models combined with oversampling methods resulted in specificity of 95 and 93% for CM detection and prediction, respectively, with the same sensitivity (82%) for both scenarios; when lowering specificity to 80 to 83%, undersampling techniques facilitated models to increase sensitivity to 95%. We propose a feasible machine learning framework to identify CM in a timely manner using imbalanced data from an AMS, which could provide useful information for farmers to manage the negative effects of CM.

Estimates of dairy herd health indicators of mastitis, ketosis, inter-calving interval, and fresh cow replacement in the Piedmont region, Italy

Test-day milk analysis has largely been used to study health and performance parameters in dairy cows. In this study, we estimated four health indicators of dairy cows using test-day data. Our purpose was to estimate (1) mastitis incidence rate, prevalence, and the probability of recovery; (2) the incidence proportion of ketosis; (3) the duration of inter-calving interval; and (4) the risk of a fresh cow being replaced, in a large cohort of dairy herds in the Piedmont region (Italy). We retrospectively analysed test day records of 261,121 lactating cows and 1315 herds during five years (2015-2020). Mastitis was defined by somatic cell count and ketosis by fat-to-protein ratio. Calving dates were used to calculate ICI and to estimate the removal of a fresh cow from the herd. Mixed-effect generalized linear models were used to adjust for unmeasured herd-level risk factors. The risk of mastitis increased by 120% with parity (Odds ratio [OR] = 2.20, confidence interval [CI]: 2.17 - 2.23), by 7% by months in milking (OR = 1.07, CI: 1.07 - 1.07), and even more if the cow was already affected during the same lactation (OR = 8.74, CI: 8.67 - 8.82). Lactose concentration on the previous test day was the best positive prognostic factor for mastitis recovery (OR = 1.12, CI: 1.08 - 1.17). Ketosis risk was the highest between 3rd and 4th lactations and itself increased the risk of having ICI longer than 440 days (OR = 1.12, CI: 1.02 - 1.22), and fresh-cow removal (OR = 1.75, CI: 1.58 - 1.93). Also, the removal of fresh cows was more likely when mastitis (OR = 1.31, CI: 1.19 - 1.45) or long ICI (OR = 1.34, CI: 1.22 - 1.48) occurred. For each health indicator, herd-level risk factors had an important role (18-56% of within-herd covariance). Our results indicate that milk analysis could be also useful for predicting mastitis, its cure rate, and ketosis. Cow-level risk factors are not enough to explain the risk of these issues. By studying a large population over a long period, this study provides an updated estimate of dairy cow health indicators in Piedmont (north-western Italy), useful for benchmarking dairy herds.

Impact of sensor data pre-processing strategies and selection of machine learning algorithm on the prediction of metritis events in dairy cattle

With all the sensor data currently generated at high frequency in dairy farms, there is potential for earlier diagnosis of postpartum diseases compared with traditional monitoring methodologies. Our objectives were 1) to compare the impact of sensor data pre-processing on classifier performance by using multiple time windows before a given metritis event, while considering other cow-level factors and farm-scheduled activities; 2) to compare the performance of random forest (RF), k-nearest neighbors (k-NN), and support vector machine (SVM) classifiers at different decision thresholds using different number of past observations (time-lags) for the detection of behavioral patterns associated with changes in metritis scores; and 3) to compare classifier performance between each one of the five behaviors registered every hour by an ear-tag 3-axis accelerometer (CowManager, Agis Autimatisering, Harmelen, Netherlands). A total of 239 metritis events were created by comparing metritis scores between two consecutive clinical evaluations from cows that were retrospectively selected from a dataset containing sensor data and health information during the first 21 days postpartum from June 2014 to May 2017. Hourly sensor data classified by the accelerometer as either ruminating, eating, not active (including both standing or lying), and two different levels of activity (active and high activity) behaviors corresponding to the 3 days before each metritis event were aggregated every 24-, 12-, 6-, and 3-hour time windows. Multiple time-lags were also used to determine the optimal number of past observations needed for optimal classification. Similarly, different decision thresholds were compared in terms of model performance. Depending on the classifier, algorithm hyperparameters were optimized using grid search (RF, k-NN, SVM) and random search (RF). All behaviors changed throughout the study period and showed distinct daily patterns. From the three algorithms, RF had the highest F₁ score followed by k-NN and SVM. Furthermore, sensor data aggregated every 6- or 12-h time windows had the best model performance at multiple time-lags. We concluded that the data from the first 3 days post-partum should be discarded when studying metritis, and either one of the five behaviors measured with CowManager could be used when predicting metritis when sensor data were aggregated every 6- or 12-hour time windows, and using time-lags corresponding to 2-3 days before a given event, depending on the time window used. This study shows how to maximize sensor data in their potential for disease prediction, enhancing the performance of algorithms used in machine learning.

Innovations in Cattle Farming: Application of Innovative Technologies and Sensors in the Diagnosis of Diseases

Precision livestock farming has a crucial function as farming grows in significance. It will help farmers make better decisions, alter their roles and perspectives as farmers and managers, and allow for the tracking and monitoring of product quality and animal welfare as mandated by the government and industry. Farmers can improve productivity, sustainability, and animal care by gaining a deeper understanding of their farm systems as a result of the increased use of data generated by smart farming equipment. Automation and robots in agriculture have the potential to play a significant role in helping society fulfill its future demands for food supply. These technologies have already enabled significant cost reductions in production, as well as reductions in the amount of intensive manual labor, improvements in product quality, and enhancements in environmental management. Wearable sensors can monitor eating, rumination, rumen pH, rumen temperature, body temperature, laying behavior, animal activity, and animal position or placement. Detachable or imprinted biosensors that are adaptable and enable remote data transfer might be highly important in this quickly growing industry. There are already multiple gadgets to evaluate illnesses such as ketosis or mastitis in cattle. The objective evaluation of sensor methods and systems employed on the farm is one of the difficulties presented by the implementation of modern technologies on dairy farms. The availability of sensors and high-precision technology for real-time monitoring of cattle raises the question of how to objectively evaluate the contribution of these technologies to the long-term viability of farms (productivity, health monitoring, welfare evaluation, and environmental effects). This review focuses on biosensing technologies that have the potential to change early illness diagnosis, management, and operations for livestock.

The costs of chronic mastitis: A simulation study of an automatic milking system farm

Mastitis is a production disease in dairy farming that causes economic losses. Especially chronic mastitis (i.e., mastitis cases continuing longer than 28 days) can substantially affect the risk of transmission of intramammary infections (IMI) and total milk production losses. Insights into the impact of chronic mastitis on production and farm economics are needed to guide chronic mastitis decision-making. We aimed to estimate the costs of chronic mastitis with a Monte Carlo simulation model in which the costs of chronic mastitis were estimated as part of the total mastitis costs. The model simulated milk yields, IMI dynamics, somatic cell count (SCC), and pregnancy status on an average Dutch dairy farm with 100 cow places over 9 years. The model was parameterized using information from the literature and actual sensor data from automatic milking system (AMS) farms. The daily subclinical milk production losses were modeled using a generalized additive model and sensor data. Transmission of IMI was modeled as well. The model results indicated median total costs of mastitis of € 230 per generic IMI case (i.e., a weighted average of all pathogens). The most substantial cost factors were the extra mastitis cases due to transmission, culling, and milk production losses. Other significant costs originated from dry cow treatments and diverted milk. The model also indicated median total costs due to chronic mastitis of € 118 (51 % of the total mastitis costs). The share of chronic mastitis relative to the total mastitis costs was substantial. Transmission of contagious bacteria had the largest share among the chronic mastitis costs (51 % of the costs of chronic cases). The large share of chronic mastitis costs in the total mastitis costs indicates the economic importance of these mastitis cases. The results of the study point to the need for future research to focus on chronic mastitis and reducing its presence on the AMS dairy farm.

Sensitivity and Specificity for the Detection of Clinical Mastitis by Automatic Milking Systems in Bavarian Dairy Herds

In automatic milking systems (AMSs), the detection of clinical mastitis (CM) and the subsequent separation of abnormal milk should be reliably performed by commercial AMSs. Therefore, the objectives of this cross-sectional study were (1) to determine the sensitivity (SN) and specificity (SP) of CM detection of AMS by the four most common manufacturers in Bavarian dairy farms, and (2) to identify routinely collected cow data (AMS and monthly test day data of the regional Dairy Herd Improvement Association (DHIA)) that could improve the SN and SP of clinical mastitis detection. Bavarian dairy farms with AMS from the manufacturers DeLaval, GEA Farm Technologies, Lely, and Lemmer-Fullwood were recruited with the aim of sampling at least 40 cows with clinical mastitis per AMS manufacturer in addition to clinically healthy ones. During a single farm visit, cow-level milking information was first electronically extracted from each AMS and then all lactating cows examined for their udder health status in the barn. Clinical mastitis was defined as at least the presence of visibly abnormal milk. In addition, available DHIA test results from the previous six months were collected. None of the manufacturers provided a definition for clinical mastitis (i.e., visually abnormal milk), therefore, the SN and SP of AMS warning lists for udder health were assessed for each manufacturer individually, based on the clinical evaluation results. Generalized linear mixed models (GLMMs) with herd as random effect were used to determine the potential influence of routinely recorded parameters on SN and SP. A total of 7411 cows on 114 farms were assessed; of these, 7096 cows could be matched to AMS data and were included in the analysis. The prevalence of clinical mastitis was 3.4% (239 cows). When considering the 95% confidence interval (95% CI), all but one manufacturer achieved the minimum SN limit of >80%: DeLaval (SN: 61.4% (95% CI: 49.0%−72.8%)), GEA (75.9% (62.4%−86.5%)), Lely (78.2% (67.4%−86.8%)), and Lemmer-Fullwood (67.6% (50.2%−82.0%)). However, none of the evaluated AMSs achieved the minimum SP limit of 99%: DeLaval (SP: 89.3% (95% CI: 87.7%−90.7%)), GEA (79.2% (77.1%−81.2%)), Lely (86.2% (84.6%−87.7%)), and Lemmer-Fullwood (92.2% (90.8%−93.5%)). All AMS manufacturers’ robots showed an association of SP with cow classification based on somatic cell count (SCC) measurement from the last two DHIA test results: cows that were above the threshold of 100,000 cells/mL for subclinical mastitis on both test days had lower chances of being classified as healthy by the AMS compared to cows that were below the threshold. In conclusion, the detection of clinical mastitis cases was satisfactory across AMS manufacturers. However, the low SP will lead to unnecessarily discarded milk and increased workload to assess potentially false-positive mastitis cases. Based on the results of our study, farmers must evaluate all available data (test day data, AMS data, and daily assessment of their cows in the barn) to make decisions about individual cows and to ultimately ensure animal welfare, food quality, and the economic viability of their farm.

Practical challenges and potential approaches to predicting low-incidence diseases on farm using individual cow data: A clinical mastitis example

Clinical mastitis (CM) incidence is considerable in terms of cows affected per year, but cases are much less common in terms of detections per cow per milking. From a modeling perspective, where predictions are made every time any cow is milked, low CM incidence per cow day makes training, evaluating, and applying CM prediction models a challenge. The objective of this study was to build models for predicting CM incidence using time-series sensor data and choose models that maximize net return based on a cost matrix. Data collected from 2 university dairy farms, the University of Florida and Virginia Polytechnic Institute and State University, were used to gather representative data, including 110,156 milkings and 333 CM cases. Variables used in the models were milk yield, protein, lactose, fat, electrical conductivity, days in milk, lactation number, and activity as the number of steps, lying time, lying bouts, and lying bout duration. Models that predicted either likelihood of CM caused by gram-negative (GN) or gram-positive (GP) bacteria on each day were derived using extreme gradient boosting with weighting favoring true-positive cases, logistic responses, and log-loss errors. Model accuracies were determined using data randomly held out from the training set on each run. All variables considered were in terms of change (slope) over previous days, including the day CM was visually detected. The GN models had a median sensitivity (Se) of 52.6% and specificity (Sp) of 99.8%, whereas the GP models had a median Se of 37.5% and Sp of 99.9% when tested on the held-out data. In our models optimized to reduce cost from predictions, the Se was much less than Sp, suggesting that CM models might benefit from greater model weighting placed on Sp. Results also highlight the importance of positive predictive value (true positive cases per predicted positive case) along with Sp and Se, as models built on sparse data tend to predict too many false-positive cases. The calculated partial net return of our GN and GP models were -$0.15 and -$0.10 per cow per lactation, respectively, whereas International Organization for Standardization (ISO) standard models with Se of 80% and Sp of 99% would return -$1.32 per cow per lactation. Models chosen that minimized the cost to the farmer differed markedly from models that met ISO guidelines, showing asymmetry in targets between Sp and Se when the disease incidence rate is low. Because of the unique challenges that low-incidence diseases like CM present, we recommend that future CM predictive models consider the economic and practical implications in addition to the traditional model evaluation metrics.

Suitability of milk lactate dehydrogenase and serum albumin for pathogen-specific mastitis detection in automatic milking systems

In response to intramammary infection (IMI), blood-derived leukocytes are transferred into milk, which can be measured as an increase of somatic cell count (SCC). Additionally, pathogen-dependent IgG increases in milk following infection. The IgG transfer into milk is associated with the opening of the blood-milk barrier, which is much more pronounced during gram-negative than gram-positive IMI. Thus, milk IgG concentration may help to predict the pathogen type causing IMI. Likewise, lactate dehydrogenase (LDH) and serum albumin (SA) cross the blood-milk barrier with IgG if its integrity is reduced. Because exact IgG analysis is complicated and difficult to automate, LDH activity and SA concentration aid as markers to predict the IgG transfer into milk in automatic milking systems (AMS). This study was conducted to test the hypothesis that LDH and SA in milk correlate with the IgG transfer, and in combination with SCC these factors allow the differentiation between gram-positive and gram-negative IMI or even more precisely the infection-causing pathogen. Further, the expression of these parameters in foremilk before (BME) and after (AME) milk ejection was tested. In the AMS, quarter milk samples (n = 686) from 48 Holstein-Friesian cows were collected manually BME and AME, followed by an aseptic sample for bacteriological culture. Mixed models were used to (1) predict the concentration of IgG transmitted from blood into milk based on LDH and SA; (2) use principal component analysis to evaluate joint patterns of SCC (cells/mL), IgG (mg/mL), LDH (U/L), and SA (mg/mL) and use the principal component scores to compare gram-positive, gram-negative, and control IMI types and BME versus AME samples; and (3) predict gram-positive and gram-negative IMI by inclusion of combined SCC-LDH and SCC-SA as predictors in the model. Overall, the SA and LDH had similar ability to predict IgG transmission from blood into milk. Comparing the areas under the curve (AUC) of the receiver operator characteristic curves, the SCC-LDH versus SCC-SA had lower gram-positive (AUC = 0.984 vs. 0.986) but similar gram-negative (AUC = 0.995 vs. 0.998) IMI prediction ability. The SCC, IgG, LDH, and SA were greater in gram-negative than in gram-positive IMI (BME and AME) in early lactation. All measured factors had higher values in milk samples taken BME than AME. In conclusion, LDH and SA could be used as replacement markers to indicate the presence of IgG transfer from blood into milk; in combination with SCC, both SA and LDH are suitable for differentiating IMI type, and BME is better for mastitis detection in AMS.

A Systematic Review of Automatic Health Monitoring in Calves: Glimpsing the Future From Current Practice

Infectious diseases, particularly bovine respiratory disease (BRD) and neonatal calf diarrhea (NCD), are prevalent in calves. Efficient health-monitoring tools to identify such diseases on time are lacking. Common practice (i.e., health checks) often identifies sick calves at a late stage of disease or not at all. Sensor technology enables the automatic and continuous monitoring of calf physiology or behavior, potentially offering timely and precise detection of sick calves. A systematic overview of automated disease detection in calves is still lacking. The objectives of this literature review were hence: to investigate previously applied sensor validation methods used in the context of calf health, to identify sensors used on calves, the parameters these sensors monitor, and the statistical tools applied to identify diseases, to explore potential research gaps and to point to future research opportunities. To achieve these objectives, systematic literature searches were conducted. We defined four stages in the development of health-monitoring systems: (1) sensor technique, (2) data interpretation, (3) information integration, and (4) decision support. Fifty-four articles were included (stage one: 26; stage two: 19; stage three: 9; and stage four: 0). Common parameters that assess the performance of these systems are sensitivity, specificity, accuracy, precision, and negative predictive value. Gold standards that typically assess these parameters include manual measurement and manual health-assessment protocols. At stage one, automatic feeding stations, accelerometers, infrared thermography cameras, microphones, and 3-D cameras are accurate in screening behavior and physiology in calves. At stage two, changes in feeding behaviors, lying, activity, or body temperature corresponded to changes in health status, and point to health issues earlier than manual health checks. At stage three, accelerometers, thermometers, and automatic feeding stations have been integrated into one system that was shown to be able to successfully detect diseases in calves, including BRD and NCD. We discuss these findings, look into potentials at stage four, and touch upon the topic of resilience, whereby health-monitoring system might be used to detect low resilience (i.e., prone to disease but clinically healthy calves), promoting further improvements in calf health and welfare.

Assessment of Sensitivity and Profitability of an Intravaginal Sensor for Remote Calving Prediction in Dairy Cattle

One critical point of dairy farm management is calving and neonatal first care. Timely calving assistance is associated with the reduction of calf mortality and postpartum uterine disease, and with improved fertility in dairy cattle. This study aimed to evaluate the performance and profitability of an intravaginal sensor for the prediction of stage II of labor in dairy farms, thus allowing proper calving assistance. Seventy-three late-gestating Italian Holstein cows were submitted to the insertion of an intravaginal device, equipped with light and temperature sensors, connected with a Central Unit for the commutation of a radio-signal into a cell phone alert. The remote calving alarm correctly identified the beginning of the expulsive phase of labor in 86.3% of the monitored cows. The mean interval from alarm to complete expulsion of the fetus was 71.56 ± 52.98 min, with a greater range in cows with dystocia (p = 0.012). The sensor worked correctly in both cold and warm weather conditions, and during day- or night-time. The intravaginal probe was well tolerated, as any cow showed lesions to the vaginal mucosa after calving. Using sex-sorted semen in heifers and beef bull semen in cows at their last lactation, the economic estimation performed through PrecisionTree™ software led to an income improvement of 119 € and 123 €/monitored delivery in primiparous and pluriparous cows, respectively. Remote calving alarm devices are key components of "precision farming" management and proven to improve animal welfare, to reduce calf losses and to increase farm incomes.

MasPA: A Machine Learning Application to Predict Risk of Mastitis in Cattle from AMS Sensor Data

Mastitis is a common disease that prevails in cattle owing mainly to environmental pathogens; they are also the most expensive disease for cattle in dairy farms. Several prevention and treatment methods are available, although most of these options are quite expensive, especially for small farms. In this study, we utilized a dataset of 6600 cattle along with several of their sensory parameters (collected via inexpensive sensors) and their prevalence to mastitis. Supervised machine learning approaches were deployed to determine the most effective parameters that could be utilized to predict the risk of mastitis in cattle. To achieve this goal, 26 classification models were built, among which the best performing model (the highest accuracy in the shortest time) was selected. Hyper parameter tuning and K-fold cross validation were applied to further boost the top model’s performance, while at the same time avoiding bias and overfitting of the model. The model was then utilized to build a GUI application that could be used online as a web application. The application can predict the risk of mastitis in cattle from the inhale and exhale limits of their udder and their temperature with an accuracy of 98.1% and sensitivity and specificity of 99.4% and 98.8%, respectively. The full potential of this application can be utilized via the standalone version, which can be easily integrated into an automatic milking system to detect the risk of mastitis in real time.

Precision Technologies to Address Dairy Cattle Welfare: Focus on Lameness, Mastitis and Body Condition

Simple Summary

The welfare of farm animals is a growing concern in the EU and across the world. In milk production, there is a strong need to assess the welfare of dairy cows. One of the most sound assessment initiatives has been practiced using protocols developed by the Welfare Quality project. These protocols mainly support the assessment of cow welfare with animal-based indicators. However, evaluating these indicators is time-consuming and expensive, so using precision livestock farming (PLF) solutions is a way forward and is becoming a reality in the dairy industry. This review presents advances in PLF solutions, particularly in the last five years, and for assessing the animal-based indicators of lameness, mastitis, and body condition in dairy cattle farming.

Abstract

Specific animal-based indicators that can be used to predict animal welfare have been the core of protocols for assessing the welfare of farm animals, such as those produced by the Welfare Quality project. At the same time, the contribution of technological tools for the accurate and real-time assessment of farm animal welfare is also evident. The solutions based on technological tools fit into the precision livestock farming (PLF) concept, which has improved productivity, economic sustainability, and animal welfare in dairy farms. PLF has been adopted recently; nevertheless, the need for technological support on farms is getting more and more attention and has translated into significant scientific contributions in various fields of the dairy industry, but with an emphasis on the health and welfare of the cows. This review aims to present the recent advances of PLF in dairy cow welfare, particularly in the assessment of lameness, mastitis, and body condition, which are among the most relevant animal-based indications for the welfare of cows. Finally, a discussion is presented on the possibility of integrating the information obtained by PLF into a welfare assessment framework.

Novel ways to use sensor data to improve mastitis management

Current sensor systems are used to detect cows with clinical mastitis. Although, the systems perform well enough to not negatively affect the adoption of automatic milking systems, the performance is far from perfect. An important advantage of sensor systems is the availability of multiple measurements per day. By clearly defining the need for detection of subclinical mastitis (SCM) and clinical mastitis (CM) from the farmers' management perspective, detection and management of SCM and CM may be improved. Sensor systems may also be used for other aspects of mastitis management. In this paper we have defined 4 mastitis situations that could be managed with the support of sensor systems. Because of differences in the associated management and the epidemiology of these specific mastitis situations, the required demands for performance of the sensor systems do differ. The 4 defined mastitis situations with the requirements of performance are the following: (1) Cows with severe CM needing immediate attention. Sensor systems should have a very high sensitivity (>95% and preferably close to 100%) and specificity (>99%) within a narrow time window (maximum 12 h) to ensure that close to all cows with true cases of severe CM are detected quickly. Although never studied, it is expected that because of the effects of severe CM, such a high detection performance is feasible. (2) Cows with mastitis that do not need immediate attention. Although these cows have a risk of progressing into severe CM or chronic mastitis, they should get the chance to cure spontaneously under close monitoring. Sensor alerts should have a reasonable sensitivity (>80%) and a high specificity (>99.5%). The time window may be around 7 d. (3) Cows needing attention at drying off. For selective dry cow treatment, the absence or presence of an intramammary infection at dry-off needs to be known. To avoid both false-positive and false-negative alerts, sensitivity and specificity can be equally high (>95%). (4) Herd-level udder health. By combining sensor readings from all cows in the herd, novel herd-level key performance indicators can be developed to monitor udder health status and development over time and raise alerts at significant deviances from predefined thresholds; sensitivity should be reasonably high, >80%, and because of the costs for further analysis of false-positive alerts, the specificity should be >99%. The development and validation of sensor-based algorithms specifically for these 4 mastitis situations will encourage situation-specific farmer interventions and operational udder health management.

Invited review: Toward a common language in data-driven mastitis detection research

Sensor technologies for mastitis detection have resulted in the collection and availability of a large amount of data. As a result, scientific publications reporting mastitis detection research have become less driven by approaches based on biological assumptions and more by data-driven modeling. Most of these approaches try to predict mastitis events from (combinations of) raw sensor data to which a wide variety of methods are applied originating from machine learning and classical statistical approaches. However, an even wider variety in terminologies is used by researchers for methods that are similar in nature. This makes it difficult for readers from other disciplines to understand the specific methods that are used and how these differ from each other. The aim of this paper was to provide a framework (filtering, transformation, and classification) for describing the different methods applied in sensor data-based clinical mastitis detection research and use this framework to review and categorize the approaches and underlying methods described in the scientific literature on mastitis detection. We identified 40 scientific publications between 1992 and 2020 that applied methods to detect clinical mastitis from sensor data. Based on these publications, we developed and used the framework and categorized these scientific publications into the 2 data processing techniques of filtering and transformation. These data processing techniques make raw data more amendable to be used for the third step in our framework, that of classification, which is used to distinguish between healthy and nonhealthy (mastitis) cows. Most publications (n = 34) used filtering or transformation, or a combination of these 2, for data processing before classification, whereas the remaining publications (n = 6) classified the observations directly from raw data. Concerning classification, applying a simple threshold was the most used method (n = 19 publications). Our work identified that within approaches several different methods and terminologies for similar methods were used. Not all publications provided a clear description of the method used, and therefore it seemed that different methods were used between publications, whereas in fact just a different terminology was used, or the other way around. This paper is intended to serve as a reference for people from various research disciplines who need to collaborate and communicate efficiently about the topic of sensor-based mastitis detection and the methods used in this context. The framework used in this paper can support future research to correctly classify approaches and methods, which can improve the understanding of scientific publication. We encourage future research on sensor-based animal disease detection, including that of mastitis detection, to use a more coherent terminology for methods, and clearly state which technique (e.g., filtering) and approach (e.g., moving average) are used. This paper, therefore, can serve as a starting point and further stimulates the interdisciplinary cooperation in sensor-based mastitis research.

Detecting and predicting changes in milk homogeneity using data from automatic milking systems

To ensure milk quality and detect cows with signs of mastitis, visual inspection of milk by prestripping quarters before milking is recommended in many countries. An objective method to find milk changed in homogeneity (i.e., with clots) is to use commercially available inline filters to inspect the milk. Due to the required manual labor, this method is not applicable in automatic milking systems (AMS). We investigated the possibility of detecting and predicting changes in milk homogeneity using data generated by AMS. In total, 21,335 quarter-level milk inspections were performed on 5,424 milkings of 624 unique cows on 4 farms by applying visual inspection of inline filters that assembled clots from the separate quarters during milking. Images of the filters with clots were scored for density, resulting in 892 observations with signs of clots for analysis (77% traces or mild cases, 15% moderate cases, and 8% heavy cases). The quarter density scores were combined into 1 score indicating the presence of clots during a single cow milking and into 2 scores summarizing the density scores in cow milkings during a 30-h sampling period. Data generated from the AMS, such as milk yield, milk flow, conductivity, and online somatic cell counts, were used as input to 4 multilayer perceptron models to detect or predict single milkings with clots and to detect milking periods with clots. All models resulted in high specificity (98-100%), showing that the models correctly classified cow milkings or cow milking periods with no clots observed. The ability to successfully classify cow milkings or cow periods with observed clots had a low sensitivity. The highest sensitivity (26%) was obtained by the model that detected clots in a single milking. The prevalence of clots in the data was low (2.4%), which was reflected in the results. The positive predictive value depends on the prevalence and was relatively high, with the highest positive predictive value (72%) reached in the model that detected clots during the 30-h sampling periods. The misclassification rate for cow milkings that included higher-density scores was lower, indicating that the models that detected or predicted clots in a single milking could better distinguish the heavier cases of clots. Using data from AMS to detect and predict changes in milk homogeneity seems to be possible, although the prediction performance for the definitions of clots used in this study was poor.

Technical note: Evaluation of a commercial on-farm milk leukocyte differential tester to identify subclinical mastitis cases in dairy cows

The objective of this study was to validate the precision and accuracy of a milk leukocyte differential tester to identify subclinical mastitis cases in dairy cows. Milk samples from individual quarters (n = 320) of 80 Holstein cows were aseptically collected and analyzed in this study. Each sample was divided into 2 replicate samples after mixing. One replicate was analyzed for somatic cell count (SCC) using the current gold standard of flow cytometry immediately after milking. The second sample was evaluated using the on-farm milk leukocyte differential tester directly after milking, where total leukocyte count (TLC; cells/mL) was obtained. The SCC and TLC were used to calculate somatic cell score (SCS) and TLC score [TLS = log₂ (TLC/100,000) + 3]. Two subclinical mastitis thresholds were set: >200,000 (low) and >400,000 (high) cells/mL. First, precision was determined between the 2 methods. Total leukocyte count and calculated TLS from the milk leukocyte differential device were compared with the gold standard using correlation and regression coefficient of determination analyses. Correlation coefficients (r) were 0.97 for TLC and SCC and 0.90 for TLS and SCS. The coefficient of determination for regression (R²) was 0.94 for TLC and SCC and 0.80 for TLS and SCS. Slopes of regression for scores and measures were 0.36 [95% confidence interval (CI): 0.35-0.37] and 0.69 (CI: 0.65-0.73), respectively; both were significantly different from 1. Sensitivity, specificity, and diagnostic accuracy were calculated for correct diagnosis of the 2 SCC thresholds using the gold standard as reference. The sensitivity of the on-farm test was 58% (95% CI: 44 to 71%) and 73% (95% CI: 56 to 86%) for the low and high thresholds, respectively. The specificities for the on-farm test were 100% (95% CI: 99 to 100%) and 100% (95% CI: 98 to 100%) for the low and high thresholds, respectively. Subclinical diagnosis accuracies were 93% (95% CI: 89 to 95%) and 96% (95% CI: 92 to 98%) for the low and high thresholds, respectively. The on-farm milk leukocyte differential tester was precise but not overall accurate for total cell counts; it had high specificity and accuracy for diagnosis compared with a standard diagnostic tool. These results suggest that the tested system is a promising technology to detect subclinical mastitis on-farm.

Challenges and Tendencies of Automatic Milking Systems (AMS): A 20-Years Systematic Review of Literature and Patents

Over the last two decades, the dairy industry has adopted the use of Automatic Milking Systems (AMS). AMS have the potential to increase the effectiveness of the milking process and sustain animal welfare. This study assessed the state of the art of research activities on AMS through a systematic review of scientific and industrial research. The papers and patents of the last 20 years (2000-2019) were analysed to assess the research tendencies. The words appearing in title, abstract and keywords of a total of 802 documents were processed with the text mining tool. Four clusters were identified (Components, Technology, Process and Animal). For each cluster, the words frequency analysis enabled us to identify the research tendencies and gaps. The results showed that focuses of the scientific and industrial research areas complementary, with scientific papers mainly dealing with topics related to animal and process, and patents giving priority to technology and components. Both scientific and industrial research converged on some crucial objectives, such as animal welfare, process sustainability and technological development. Despite the increasing interest in animal welfare, this review highlighted that further progress is needed to meet the consumers' demand. Moreover, milk yield is still regarded as more valuable compared to milk quality. Therefore, additional effort is necessary on the latter. At the process level, some gaps have been found related to cleaning operations, necessary to improve milk quality and animal health. The use of farm data and their incorporation on herd decision support systems (DSS) appeared optimal. The results presented in this review may be used as an overall assessment useful to address future research.

Progression of different udder inflammation indicators and their episode length after onset of inflammation using automatic milking system sensor data

In automatic milking systems (AMS), sensors can measure cow behavior and milk composition at every milking. The aim of this observational study of previously collected data was to gain insight into the differences in dynamics of udder inflammation indicators between cows that recover and those that do not recover after detection of an initial inflammation. Milk diversion (milk separated from the bulk tank and thus indicating farmer intervention), conductivity, and somatic cell count (SCC) data from 4 wk before the initial inflammation to 12 wk after the initial inflammation were used to analyze 2,584 cases of udder inflammation. An udder inflammation case was defined as an initial observation of SCC ≥200,000 cells/mL as well as 1 additional SCC measurement >200,000 cells/mL within 10 d after the initial case, among other requirements. The data originated from 15 AMS herds in 6 countries. Four subsets of cows were created based on whether milk was diverted after the initial inflammation and whether the udder inflammation cases recovered, using a 10-d rolling average SCC threshold of 200,000 cells/mL and checking whether this rolling mean was below the threshold within 90 d after the initial inflammation as the indication of recovery. This formed the following subsets of cow lactations: milk diverted-recovered, milk diverted-not recovered, no milk diverted-not recovered, no milk diverted-recovered. Thresholds of 100,000 SCC/mL and 300,000 SCC/mL for the definition of case and recovery were also applied in a sensitivity analysis but with no substantial difference in results. Linear mixed models were used for each subset to study the variation in SCC (natural logarithm of SCC divided by 1,000) and σ-conductivity (natural logarithm of standard deviation of quarter conductivities). When observing the fraction of cows with SCC <200,000 cells/mL in the recovery subsets, most cows recovered within 20 d after the initial inflammation. In the recovery subsets, both σ-conductivity and SCC stabilized, mostly within 3 to 4 wk after the initial inflammation. σ-Conductivity stabilized above the pre-onset level in all subsets and did not show a clear increase in the no-milk-diverted subgroups, whereas SCC stabilized closer to the pre-onset level. Overall, this study indicated a cutoff point between nonchronic and chronic changes in indicators 3 to 4 wk after the initial inflammation for SCC and σ-conductivity.

A survey on sensor systems used in Italian dairy farms and comparison between performances of similar herds equipped or not equipped with sensors

Sensor systems (SS) were developed over the last few decades to help dairy farmers manage their herds. Such systems can provide both data and alerts to several productive, behavioral, and physiological indicators on individual cows. Currently, there is still a lack of knowledge on both the proportion of dairy farms that invested in SS and type of SS installed. Additionally, it is still unclear whether the performances of herds equipped with SS differ from those of similar herds managed without any technological aid. Therefore, the aims of this study were (1) to provide an insight into SS spread among Italian dairy farms and (2) to analyze the performances of similar herds equipped or not equipped with SS. To reach the former goal, a large survey was carried out on 964 dairy farms in the northeast of Italy. Farmers were interviewed by the technicians of the regional breeders association to collect information on the type of SS installed on farms and the main parameters recorded. Overall, 42% of the surveyed farms had at least 1 SS, and most of them (72%) reared more than 50 cows. Sensors for measuring individual cow milk yield were the most prevalent type installed (39% of the surveyed farms), whereas only 15% of farms had SS for estrus detection. More sophisticated parameters, such as rumination, were automatically monitored in less than 5% of the farms. To reach the latter goal of the study, a subset of 100 Holstein dairy farms with similar characteristics was selected: half of them were equipped with SS for monitoring at least individual milk yield and estrus, and the other half were managed without any SS. Average herd productive and reproductive data from official test days over 3 yr were analyzed. The outcomes of the comparison showed that farms with SS had higher mature-equivalent milk production. Further clustering analysis of the same 100 farms partitioned them into 3 clusters based on herd productive and reproductive data. Results of the Chi-squared test showed that the proportion of farms equipped with SS was greater in the cluster with the best performance (e.g., higher milk yield and shorter calving interval). However, the presence of a few farms equipped with SS in the least productive cluster for the same parameters pointed out that although the installation of SS may support farmers in time- and labor-saving or in data recording, it is not a guarantee of better herd performance.

Point-of-care tests for bovine clinical mastitis: what do we have and what do we need?

Mastitis, inflammation of the bovine mammary gland, is generally caused by intramammary infection with bacteria, and antimicrobials have long been a corner stone of mastitis control. As societal concern about antimicrobial use in animal agriculture grows, there is pressure to reduce antimicrobial use in dairy farming. Point-of-care tests for on-farm use are increasingly available as tools to support this. In this Research Reflection, we consider available culture-dependent and culture-independent tests in the context of ASSURED criteria for low-resource settings, including convenience criteria, scientific criteria and societal criteria that can be used to evaluate test performance. As tests become more sophisticated and sensitive, we may be generating more data than we need. Special attention is given to the relationship between test outcomes and treatment decisions, including issues of diagnostic refinement, antimicrobial susceptibility testing, and detection of viable organisms. In addition, we explore the role of technology, big data and people in improved performance and uptake of point-of-care tests, recognising that societal barriers may limit uptake of available or future tests. Finally, we propose that the 3Rs of reduction, refinement and replacement, which have been used in an animal welfare context for many years, could be applied to antimicrobial use for mastitis control on dairy farms.

Using Sensor Data to Detect Lameness and Mastitis Treatment Events in Dairy Cows: A Comparison of Classification Models

The aim of this study was to develop classification models for mastitis and lameness treatments in Holstein dairy cows as the target variables based on continuous data from herd management software with modern machine learning methods. Data was collected over a period of 40 months from a total of 167 different cows with daily individual sensor information containing milking parameters, pedometer activity, feed and water intake, and body weight (in the form of differently aggregated data) as well as the entered treatment data. To identify the most important predictors for mastitis and lameness treatments, respectively, Random Forest feature importance, Pearson’s correlation and sequential forward feature selection were applied. With the selected predictors, various machine learning models such as Logistic Regression (LR), Support Vector Machine (SVM), K-nearest neighbors (KNN), Gaussian Naïve Bayes (GNB), Extra Trees Classifier (ET) and different ensemble methods such as Random Forest (RF) were trained. Their performance was compared using the receiver operator characteristic (ROC) area-under-curve (AUC), as well as sensitivity, block sensitivity and specificity. In addition, sampling methods were compared: Over- and undersampling as compensation for the expected unbalanced training data had a high impact on the ratio of sensitivity and specificity in the classification of the test data, but with regard to AUC, random oversampling and SMOTE (Synthetic Minority Over-sampling) even showed significantly lower values than with non-sampled data. The best model, ET, obtained a mean AUC of 0.79 for mastitis and 0.71 for lameness, respectively, based on testing data from practical conditions and is recommended by us for this type of data, but GNB, LR and RF were only marginally worse, and random oversampling and SMOTE even showed significantly lower values than without sampling. We recommend the use of these models as a benchmark for similar self-learning classification tasks. The classification models presented here retain their interpretability with the ability to present feature importances to the farmer in contrast to the “black box” models of Deep Learning methods.

Economic value of information from an alert system on physiological imbalance in fresh cows

Physiological imbalance is an abnormal physiological condition that cannot be directly observed but is assumed to precede subclinical and clinical diseases in the beginning of lactation. Alert systems to detect the physiological imbalance in a cow using Fourier transform mid-infrared spectroscopy in milk have been developed. The objective of this study was to estimate the value of information provided from such system with different indicator accuracies, herd prevalence and prices. A decision tree was created to model the probabilities of detection and associated costs of test outcome, intervention and occurrence of disease. We assumed that the negative effect of physiological imbalance was the development of subclinical ketosis and that this negative effect was prevented by drenching the cows with propylene glycol for 5 days. We simulated the economic impact of subclinical ketosis mediated through physiological imbalance to be $194 per case. The results showed that if the alert system was highly accurate (Se = 0.99/Sp = 0.99), and the prevalence of physiological imbalance was 30 %, the value of information provided from the system is $19 per cow-year. In case the prevalence is 5 % or 50 %, the value of information is $3 and $13, respectively. These estimates for the value do not cover the capital costs and operational costs of the alert system. This study furthermore clearly demonstrated that in order to estimate the value of information correctly, it is important to consider that drenching all cows and not drenching any of the cows are the two relevant alternative options in the absence of the alert system. In conclusion, the decision tree and sensitivity analysis developed in this study show that final economic results are highly variable to the prevalence of physiological imbalance and highest at an intermediate prevalence. Other relevant factors are the costs associated with drenching and the cost associated with treating false positives and not treating false negatives. In addition, this study highlights the benefits of simulation to pinpoint where additional information is needed to further quantify the economic value and required accuracy of an indication-based intervention system.

Technological interventions and advances in the diagnosis of intramammary infections in animals with emphasis on bovine population-a review

Mastitis, an inflammation of the udder, is a challenging problem in dairy animals accounting for high economic losses. Disease complexity, degree of economic losses and increasing importance of the dairy industries along with public health concerns envisages devising appropriate diagnostics of mastitis, which can offer rapid, accurate and confirmatory diagnosis. The various diagnostic tests of mastitis have been divided into general or phenotypic and specific or genotypic tests. General or phenotypic tests are those that identify general alterations, which are not specific to any pathogen. Genotypic tests are specific, hence confirmatory for diagnosis of mastitis and include specific culture, polymerase chain reaction (PCR) and its various versions (e.g. qRT-PCR), loop-mediated isothermal amplification, lateral flow assays, nucleotide sequencing, matrix-assisted laser desorption ionization time-of-flight mass spectrometry, and other molecular diagnostic methods. However, for highly specific and confirmatory diagnosis, pure cultures still provide raw materials for more sophisticated diagnostic technological interventions like PCR and nucleotide sequencing. Diagnostic ability of like infra-red thermography (IRT) has been shown to be similar to California mastitis test and also differentiates clinical mastitis from subclinical mastitis cases. As such, IRT can become a convenient and portable diagnostic tool. Of note, magnetic nanoparticles-based colorimetric biosensor assay was developed by using for instance proteolytic activity of plasmin or anti-S. aureus antibody. Last but not least, microRNAs have been suggested to be potential biomarkers for diagnosing bovine mastitis. This review summarizes the various diagnostic tests available for detection of mastitis including diagnosis through general and specific technological interventions and advances.

Prediction of quarter level subclinical mastitis by combining in-line and on-animal sensor data

We investigated the potential for automatic detection of subclinical mastitis (SCM) in pasture-based automatic milking systems. The objective of the study was to determine the ability of electrical conductivity (EC), together with relative changes in daily activity (activity) and daily rumination (rumination) recorded using heat and rumination–long-distance tags, to predict quarter-level SCM. Activity (arbitrary unit/day) and rumination (min/day) data were determined across 21 days using heat and rumination–long-distance tags for 170 cows. Cows were allocated into the following three groups: SCM (n = 32, EC ≥ 7.5 millisiemens/cm (mS/cm) in one or more quarters and a positive bacteriological culture in the corresponding quarter(s)); true-negative (TN, n = 9, EC ≥ 7.5 mS/cm and a negative culture in all four quarters); and apparently healthy (n = 129, no culture test and EC &lt; 7.5 mS/cm). Group mean differences in activity and rumination were compared using Welch’s t-tests. Logistic mixed models were used to predict SCM by EC, activity and rumination changes before mastitis detection, including parity information between SCM and TN groups. Cow- and quarter-specific information were included as random effects, followed by model assessment by producing receiver operating-characteristic curve and area under the curve (AUC) value. In total, 287 quarters were used in the prediction model, including 143 quarters with a positive culture (Gram-positive; n = 131, Gram-negative; n = 6, mixed; n = 6) and 144 quarters with a negative culture. On average, SCM group had 4.65% greater (P &lt; 0.01) activity and 9.89% greater (P &lt; 0.001) rumination than did the TN group and 11.70% greater (P &lt; 0.001) activity than did the apparently healthy group. A combined model with terms for EC, activity changes, rumination changes prior to detect SCM and parity had a better SCM prediction (AUC = 0.92) ability than did any of them separately (all AUC &lt; 0.8). Hence, we conclude that EC in combination with activity and rumination information can improve the accuracy of prediction of quarter-level SCM.

Identifying gram-negative and gram-positive clinical mastitis using daily milk component and behavioral sensor data

Opportunities exist for automated animal health monitoring and early detection of diseases such as mastitis with greater on-farm adoption of precision technologies. Our objective was to evaluate time series changes in individual milk component or behavioral variables for all clinical mastitis (CM) cases (ACM), for CM caused by gram-negative (GN) or gram-positive (GP) pathogens, or CM cases in which no pathogen was isolated (NPI). We developed algorithms using a combination of milk and activity parameters for predicting each of these infection types. Milk and activity data were collated for the 14 d preceding a CM event (n = 170) and for controls (n = 166) matched for breed, parity, and days in milk. Explanatory variables in the univariate and multiple regression models were the slope change in milk (milk yield, conductivity, somatic cell count, lactose percentage, protein percentage, and fat percentage) and activity parameters (steps, lying time, lying bout duration, and number of lying bouts) over 7 d. Slopes were estimated using linear regression between d -7 and -5, d -7 and -4, d -7 and -3, d -7 and -2, and d -7 and -1 relative to CM detection for all parameters. Univariate analyses determined significant slope ranges for explanatory variables against the 4 responses: ACM, GN, GP, and NPI. Next, all slope ranges were offered into the multivariate models for the same 4 responses using 3 baselines: d -10, -7, and -3 relative to CM detection. In the univariate analysis, no explanatory variables were significant indicators of ACM, whereas at least 1 parameter was significant for each of GN, GP, and NPI models. Superior sensitivity (Se) and specificity (Sp) estimates were observed for the best GP (Se = 82%, Sp = 87%) and NPI (Se = 80%, Sp = 94%) multiple regression models compared with the best ACM (Se = 73%, Sp = 75%) and GN (Se = 71%, Sp = 74%) models. Sensitivity for the GN model was greater at the baseline closest to the day of CM detection (d -3), whereas the opposite was observed for the GP and NPI model as Se was maximized at the d -10 baseline. Based on this screening of relationships, milk and activity sensor data could be used in CM detection systems.

Symposium review: Real-time continuous decision making using big data on dairy farms

We are developing a real-time, data-integrated, data-driven, continuous decision-making engine, The Dairy Brain, by applying precision farming, big data analytics, and the Internet of Things. This is a transdisciplinary research and extension project that engages multidisciplinary scientists, dairy farmers, and industry professionals. Dairy farms have embraced large and diverse technological innovations such as sensors and robotic systems, and procured vast amounts of constant data streams, but they have not been able to integrate all this information effectively to improve whole-farm decision making. Consequently, the effects of all this new smart dairy farming are not being fully realized. It is imperative to develop a system that can collect, integrate, manage, and analyze on- and off-farm data in real time for practical and relevant actions. We are using the state-of-the-art database management system from the University of Wisconsin-Madison Center for High Throughput Computing to develop our Agricultural Data Hub that connects and analyzes cow and herd data on a permanent basis. This involves cleaning and normalizing the data as well as allowing data retrieval on demand. We illustrate our Dairy Brain concept with 3 practical applications: (1) nutritional grouping that provides a more accurate diet to lactating cows by automatically allocating cows to pens according to their nutritional requirements aggregating and analyzing data streams from management, feed, Dairy Herd Improvement (DHI), and milking parlor records; (2) early risk detection of clinical mastitis (CM) that identifies first-lactation cows under risk of developing CM by analyzing integrated data from genetic, management, and DHI records; and (3) predicting CM onset that recognizes cows at higher risk of contracting CM, by continuously integrating and analyzing data from management and the milking parlor. We demonstrate with these applications that it is possible to develop integrated continuous decision-support tools that could potentially reduce diet costs by $99/cow per yr and that it is possible to provide a new dimension for monitoring health events by identifying cows at higher risk of CM and by detecting 90% of CM cases a few milkings before disease onset. We are securely advancing toward our overarching goal of developing our Dairy Brain. This is an ongoing innovative project that is anticipated to transform how dairy farms operate.

On-farm use of disease alerts generated by precision dairy technology

Wearable precision dairy monitoring (PDM) technologies currently used to detect estrus may provide insight into disease detection. However, the incorporation of PDM into farm management and its perceived usefulness for dairy producers have not been explored. As the targeted end users of these products, information is needed on how producers use generated disease alerts as well as barriers to adoption and usefulness. The objective of this research was to assess the perceived usefulness producers attributed to alerts from a daily generated alert list designed to identify sick or injured cows or cows that experienced a major management change. Data from 1,171 cows on 4 commercial farms in Kentucky were collected from October 2015 to October 2016. Each cow was equipped with 2 PDM technologies: a leg tag (measuring activity in steps/d and lying time in h/d) and a neck collar (measuring eating time in h/d). Alerts were generated based on an individual cow's decrease of ≥30% in activity, lying, and eating time compared with each cow's 10-d moving mean. Producers sorted alerts into 3 overall categories: (1) the cow alert was perceived to be true and the cow was visually checked, (2) the cow alert was perceived to be true, but the cow was not visually checked, and (3) the cow alert behavior change was doubted by the producer and the cow was not visually checked. Further subdivisions were also provided to explain the choice for an overall category. Over the year, 24,012 cow alerts were generated (eating time, n = 9,543; lying time, n = 9,777; activity, n = 1,590; or a combination of behaviors, n = 3,102). Only 8% of the alerts were doubted by the producer. Although 55% of alerts were perceived to be true, producers visually assessed cows based on only 21% of the alerts with a large variation between farms (2 to 45% of the alerts visually assessed). Producers were more likely to completely ignore alerts over time. Producers were more likely to perceive cow alerts to be true and visually check cows when ≤20 alerts occurred per day, cows were fresh or in early lactation, alerts occurred during the work week, or when cow alerts were for eating time, activity, or a combination of multiple behaviors. Behavioral disease alerts must be improved and correspond to an actionable change for producers to use them. Incorporating herd management software, creating and managing alerts by lactation stage, and focusing on behaviors producers already find useful could improve future alert utilization.

Unweaving tangled mortality and antibiotic consumption data to detect disease outbreaks - Peaks, growths, and foresight in swine production

As our capacity to collect and store health data is increasing, a new challenge of transforming data into meaningful information for disease monitoring and surveillance has arisen. The aim of this study was to explore the potential of using livestock mortality and antibiotic consumption data as a proxy for detecting disease outbreaks at herd level. Changes in the monthly records of mortality and antibiotic consumption were monitored in Danish swine herds that became positive for porcine reproductive and respiratory syndrome (PRRS) and porcine pleuropneumonia. Laboratory serological results were used to identify herds that changed from a negative to a positive status for the diseases. A dynamic linear model with a linear growth component was used to model the data. Alarms about state changes were raised based on forecast errors, changes in the growth component, and the values of the retrospectively smoothed values of the growth component. In all cases, the alarms were defined based on credible intervals and assessed prior and after herds got a positive disease status. The number of herds with alarms based on mortality increased by 3% in the 3 months prior to laboratory confirmation of PRRS-positive herds (Se = 0.47). A 22% rise in the number of weaner herds with alarms based on the consumption of antibiotics for respiratory diseases was found 1 month prior to these herds becoming PRRS-positive (Se = 0.22). For porcine pleuropneumonia-positive herds, a 10% increase in antibiotic consumption for respiratory diseases in sow herds was seen 1 month prior to a positive result (Se = 0.5). Monitoring changes in mortality data and antibiotic consumption showed changes at herd level prior to and in the same month as confirmation from diagnostic tests. These results also show a potential value for using these data streams as part of surveillance strategies.

Factors associated with milking-to-milking variability in somatic cell counts from healthy cows in an automatic milking system

Fully automated on-line analysis equipment is available for analysis of somatic cell count (SCC) at every milking in automatic milking systems. In addition to results from on-line cell counters (OCC), an array of additional cow-level and quarter-level factors considered important for udder health are recorded in these systems. However, the amount of variability in SCC that can be explained by available data is unknown, and so is the proportion of the variability that may be due to physiological or normal variability. Our aim was to increase our knowledge on OCC as an indicator for disturbances in udder health by assessing the variability in OCC in cows free from clinical mastitis. The first objective was to evaluate how much of the variability in OCC could be explained by different potential sources of variability, including intramammary infection (IMI) status (assessed by bacterial culture of quarter milk samples). The second objective was to evaluate the repeatability of the OCC sensor used in our study and the agreement between OCC values and SCC measured in a dairy herd improvement (DHI) laboratory. A longitudinal study was performed in the research herd of the Norwegian University of Life Sciences from January 5th 2016 to May 22nd 2017. Data from 62,471 milkings from 173 lactations in 129 cows were analyzed. We used ln-transformed OCC values (in 1000 cells/ml) as the outcome (lnOCC) in linear mixed models, with random intercepts at cow-level and lactation-level within cow. We were able to explain 15.0% of the variability in lnOCC with the following fixed effects: lactation stage, parity, milk yield, OCC in residual milk from the previous milking, inter-quarter difference between the highest and lowest conductivity, season, IMI status, and genetic lineage. When including the random intercepts, the degree of explanation was 55.2%. The individual variables explained only a small part of the total variability in lnOCC. We concluded that physiological or normal variability is probably responsible for a large part of the overall variability in OCC in cows without clinical mastitis. This is important to consider when using OCC data for research purposes or in decision-support tools. Sensor repeatability was evaluated by analyzing milk from the same sample multiple times. The coefficient of variation was 0.11 at an OCC level relevant for detection of subclinical mastitis. The agreement study showed a concordance correlation coefficient of 0.82 when comparing results from the OCC with results from a DHI laboratory.

Validation of a novel milk progesterone-based tool to monitor luteolysis in dairy cows: Timing of the alerts and robustness against missing values

Automated monitoring of fertility in dairy cows using milk progesterone is based on the accurate and timely identification of luteolysis. In this way, well-adapted insemination advice can be provided to the farmer to further optimize fertility management. To properly evaluate and compare the performance of new and existing data-processing algorithms, a test data set of progesterone time-series that fully covers the desired variability in progesterone profiles is needed. Further, the data should be measured with a high frequency to allow rapid onset events, such as luteolysis, to be precisely determined. Collecting this type of data would require a lot of time, effort, and budget. In the absence of such data, an alternative was developed using simulated progesterone profiles for multiple cows and lactations, in which the different fertility statuses were represented. To these, relevant variability in terms of cycle characteristics and measurement error was added, resulting in a large cost-efficient data set of well-controlled but highly variable and farm-representative profiles. Besides the progesterone profiles, information on (the timing of) luteolysis was extracted from the modeling approach and used as a reference for the evaluation and comparison of the algorithms. In this study, 2 progesterone monitoring tools were compared: a multiprocess Kalman filter combined with a fixed threshold on the smoothed progesterone values to detect luteolysis, and a progesterone monitoring algorithm using synergistic control, PMASC, which uses a mathematical model based on the luteal dynamics and a statistical control chart to detect luteolysis. The timing of the alerts and the robustness against missing values of both algorithms were investigated using 2 different sampling schemes: one sample per cow every 8 h versus 1 sample per day. The alerts for luteolysis of the PMASC algorithm were on average 20 h earlier compared with the ones of the multiprocess Kalman filter, and their timing was less sensitive to missing values. This was shown by the fact that, when 1 sample per day was used, the Kalman filter gave its alerts on average 24 h later, and the variability in timing of the alerts compared with simulated luteolysis increased with 22%. Accordingly, we postulate that implementation of the PMASC system could improve the consistency of luteolysis detection on farm and lower the analysis costs compared with the current state of the art.

Biosensors for On-Farm Diagnosis of Mastitis

Bovine mastitis is an inflammation of the mammary gland caused by a multitude of pathogens with devastating consequences for the dairy industry. Global annual losses are estimated to be around €30 bn and are caused by significant milk losses, poor milk quality, culling of chronically infected animals, and occasional deaths. Moreover, mastitis management routinely implies the administration of antibiotics to treat and prevent the disease which poses serious risks regarding the emergence of antibiotic resistance. Conventional diagnostic methods based on somatic cell counts (SCC) and plate-culture techniques are accurate in identifying the disease, the respective infectious agents and antibiotic resistant phenotypes. However, pressure exists to develop less lengthy approaches, capable of providing on-site information concerning the infection, and in this way, guide, and hasten the most adequate treatment. Biosensors are analytical tools that convert the presence of biological compounds into an electric signal. Benefitting from high signal-to-noise ratios and fast response times, when properly tuned, they can detect the presence of specific cells and cell markers with high sensitivity. In combination with microfluidics, they provide the means for development of automated and portable diagnostic devices. Still, while biosensors are growing at a fast pace in human diagnostics, applications for the veterinary market, and specifically, for the diagnosis of mastitis remain limited. This review highlights current approaches for mastitis diagnosis and describes the latest outcomes in biosensors and lab-on-chip devices with the potential to become real alternatives to standard practices. Focus is given to those technologies that, in a near future, will enable for an on-farm diagnosis of mastitis.

The detection of intramammary infections using online somatic cell counts

Timely and accurate identification of cows with intramammary infections is essential for optimal udder health management. Various sensor systems have been developed to provide udder health information that can be used as a decision support tool for the farmer. Among these sensors, the DeLaval Online Cell Counter (DeLaval, Tumba, Sweden) provides somatic cell counts from every milking at cow level. Our aim was to describe and evaluate diagnostic sensor properties of these online cell counts (OCC) for detecting an intramammary infection, defined as an episode of subclinical mastitis or a new case of clinical mastitis. The predictive abilities of a single OCC value, rolling averages of OCC values, and an elevated mastitis risk (EMR) variable were compared for their accuracy in identifying cows with episodes of subclinical mastitis or new cases of clinical mastitis. Detection of subclinical mastitis episodes by OCC was performed in 2 separate groups of different mastitis pathogens, Pat 1 and Pat 2, categorized by their known ability to increase somatic cell count. The data for this study were obtained in a field trial conducted in the dairy herd of the Norwegian University of Life Sciences. Altogether, 173 cows were sampled at least once during a 17-mo study period. The total number of quarter milk cultures was 5,330. The most common Pat 1 pathogens were Staphylococcus epidermidis, Staphylococcus aureus, and Streptococcus dysgalactiae. The most common Pat 2 pathogens were Corynebacterium bovis, Staphylococcus chromogenes, and Staphylococcus haemolyticus. The OCC were successfully recorded from 82,182 of 96,542 milkings during the study period. For episodes of subclinical mastitis the rolling 7-d average OCC and the EMR approach performed better than a single OCC value for detection of Pat 1 subclinical mastitis episodes. The EMR approach outperformed the OCC approaches for detection of Pat 2 subclinical mastitis episodes. For the 2 pathogen groups, the sensitivity of detection of subclinical mastitis episodes was 69% (Pat 1) and 31% (Pat 2), respectively, at a predefined specificity of 80% (EMR). All 3 approaches were equally good at detecting new cases of clinical mastitis, with an optimum sensitivity of 80% and specificity of 90% (single OCC value).