A groundbreaking study conducted by researchers from Penn State, University of Kentucky, and University of Vermont has utilized low-cost precision technology and machine learning techniques to detect bovine respiratory disease (BRD) in calves before symptoms appear. BRD is a major concern for dairy farmers, as it not only leads to antimicrobial use in calves but also contributes to a significant number of calf mortalities. Detecting the disease early is crucial for prompt treatment and minimizing long-term effects on growth and milk production. However, traditional methods of monitoring for visible signs of disease are labor-intensive, especially considering the current worker shortages in the industry.
To address this issue, the researchers proposed the use of low-cost precision technologies like wearable sensors and automatic feeders. These technologies can monitor calf behavior without the need for extensive labor and financial investments. For instance, a pedometer costs only $90 per calf with a daily maintenance cost of $0.38, while a grain feeder requires a one-time purchase of $6,500 and a daily maintenance cost of $1.62. Compared to the significant costs associated with traditional health exams, employing these technologies offers a more cost-effective solution to monitor calf health.
The study, which was published in the scientific journal IEEE Access, achieved several industry-firsts. It developed the first framework for early prediction of BRD persistency status, provided the first dataset including precision tech and manual health exams for diagnosing BRD, and featured the largest number of adopted precision technology devices. Each calf in the study wore a pedometer sensor to collect data on step count and lying time every 15 minutes. RFID ear tags were also used to track changes in appetite when approaching the automated feeder. By analyzing this data, the researchers could identify when a calf’s condition was deteriorating, allowing for early intervention.
Over a two-year period, the researchers collected data from 159 calves and employed a machine learning model paired with feature selection to maximize prediction accuracy. The system achieved an 88% accuracy rate in labeling ill and healthy calves and accurately identified 70% of calves four days before disease diagnosis. Furthermore, 80% of calves that developed chronic BRD were detected within the first five days of sickness.
The authors of the study emphasized the significance of their findings, stating that it is the first work to study BRD persistency status using cost-effective machine learning techniques and to provide such a comprehensive dataset. Melissa Cantor, the lead researcher, expressed surprise at how behavioral changes in the animals differed between those that improved with treatment and those that did not. This observation led to the concept that IoT technologies empowered with machine learning inference techniques could potentially identify sick animals earlier, offering producers more options for intervention.
In conclusion, this groundbreaking study highlights the potential of low-cost precision technology and machine learning in the early diagnosis of bovine respiratory disease. By utilizing wearable sensors and automatic feeders, dairy farmers can monitor calf health more efficiently and detect disease before symptoms appear. This approach not only reduces labor and economic requirements but also minimizes the long-term effects of the disease. Further advancements in this field could revolutionize disease management in the food and beverage industry, improving animal welfare and increasing productivity.
Source:
A Machine Learning and Optimization Framework for the Early Diagnosis of Bovine Respiratory Disease
Cantor, C. M., et al
Published: 20 June 2023
DOI: 10.1109/ACCESS.2023.3291348