What if you could know exactly how much time each animal in a feedyard spends at the feedbunk every day, how much weight it gains, how many steps it takes or how its behavior changes from one day to the next? Or better yet, what if your computer could constantly monitor those and other behavior and performance trends, analyze countless data points and alert you as soon as an animal shows the earliest signs of sickness?
That day is approaching quickly as researchers and engineers are developing systems to remotely monitor every animal in a feedyard or dairy and detect subtle changes.
The pen rider’s job, we often hear, involves a mix of science, art, gut instincts and maybe a little magic. Skilled pen riders can spot subtle signs of sickness, but cattle feeders and veterinarians know the conventional method for spotting and pulling sick cattle is subjective and often unreliable. Numerous studies in which researchers evaluated cattle lungs at slaughter have shown that high percentages of cattle with signs of BRD were never identified or treated.
Technology allowing measurement of new indicators of health such as behavior, feed intake, water intake and daily gains could identify sick cattle significantly earlier than the visible signs upon which pen riders rely. And even if early detection remains elusive, these systems could provide more objective and reliable disease detection while reducing labor requirements.
Same goal, different paths
GrowSafeSystems®, for example, continues to test and refine its technologies for real-time monitoring of cattle behavior and health. GrowSafe, based in Alberta, Canada, is best known for systems that monitor individual feed intake, which are in wide use in genetic selection for feed efficiency by U.S. research facilities and seedstock operations. The company has extended its application of radio-frequency identification (RFID) linked to biometric measurements to monitor multiple indicators of cattle health. The GrowSafeBeef (GSB) system uses scales mounted to water troughs in feedyard pens, along with RFID. After extensive testing and evaluation, the company developed a conversion factor to determine live weights of cattle based on partial bodyweights. When an animal steps up to the water trough, its front feet are on a scale.
The system identifies the animal and transmits its ID, time spent drinking and partial bodyweight to a central computer, which converts the partial bodyweight to full bodyweight.
These frequent measurements of bodyweight allow accurate tracking of daily gains. According to GrowSafe research into standard deviations in average daily gain (ADG) estimates using a chute, measuring ADG with a chute over a three-month time period can result in errors averaging 1.2 pounds per day. In comparison, the GSB system calculated ADG with deviation of 0.4 pounds per day with a 99 percent confidence interval over just a seven-day time frame.
The water trough is equipped with an overhead spray gun designed to mark cattle. When the system identifies an animal that appears sick based on performance, water intake or other indicators, the computer can activate the sprayer, making it easy for pen riders to spot and pull suspect cattle.
The system also can be used to sort cattle into marketing groups, using color-coded spray paint to mark them based on weight and performance.
At the University of Kentucky, Craig Carter, DVM, PhD, Dipl. ACVPM, is participating in a project which takes a different approach for wireless monitoring of cattle behavior. Carter directs the university’s veterinary diagnostic laboratory. He and a team of researchers have developed a system for monitoring cattle behaviors using radio-frequency tags equipped with triple-axis accelerometers. The accelerometers continuously measure animal activity in counts per minute (CPM), reflecting the magnitude of g-forces in various directions caused by movement of the animal.
Researchers with USDA’s Agricultural Research Service at the U.S. Meat Animal Research Center in Clay Center, Neb., also have developed a system to monitor feeding behavior of feedlot cattle. Their system uses standard RFID technology designed around commercial RFID readers. In this case, multiple RFID readers track each animal’s feeding behavior and movement around the pen. The researchers currently are testing whether feeding and other behavioral data generated with the system can provide early indications of morbidity.
Dan Thomson, DVM, PhD, directs Kansas State University’s Beef Cattle Institute. Prior to joining the K-State faculty, he worked for Cactus Feeders in Texas and was involved in some early testing of systems for monitoring cattle behavior. Monitoring intake, he says, showed promise for detecting sickness early, as fluctuations in intake turn up sooner than resulting changes in daily gain. However, monitoring intake on individual animals in a feedyard setting is not economically feasible with existing technology. So, Cactus worked with GrowSafe in the late 1990s to evaluate its early systems for monitoring animal behavior at a feeding bunk in the feedyard pens. Those tests, Thomson says, indicated variation in animal-feeding behavior could identify sick cattle up to five days sooner than conventional methods.
Thomson says the GrowSafe system potentially offers benefits at the beginning and end of the feeding period. During the first 30 days after arrival — typically the period of highest risk for respiratory disease — the system can facilitate treatment decisions. During the final 30 days, weight and gain data could help managers sort and market cattle at their optimal economic endpoints.
Sorting through the data
As often is the case, developers of these systems have found that collecting data can be relatively easy compared with accurately interpreting and applying those data. “We’re engineers and computer geeks, not veterinarians,” says GrowSafe CEO Alison Sunstrum. So her company has worked closely with veterinarians and the Noble Foundation in Oklahoma to test correlations between animal behaviors and health, and to develop an algorithm to identify sick animals based on behavior data both in the feedlot and on pasture.
Carter also says data management and analysis are the most challenging steps in developing the University of Kentucky system, which collects millions of data points on each animal every month. The team spent considerable time developing a novel mathematical algorithm to identify behavioral changes that could indicate early signs of morbidity. A patent application has been filed on the algorithm.
In initial trials in 2010 and 2011, the group used their electronic system to monitor activity levels on groups of about 100 animals in feedyard pens at the University of Kentucky Woodford Farm. At the same time, farm staff monitored the pens as they normally would, pulling and treating cattle that appeared sick and returning them to their home pens after a typical stay in a hospital pen. At the end of each test, the researchers compared the lists of animals the pen riders identified as sick with those flagged as morbid based on analysis of activity data. They found, Carter says, a very close correlation between the two systems, and in some cases, the electronic system identified sick cattle several days before caretakers observed physical signs of morbidity.
The researchers created graphs showing average activity, measured in CPM, for cattle identified as sick and for healthy cattle. The CPM line for healthy cattle consistently ran well above that for sick cattle prior to treatment. Following treatment, the lines converged, indicating activity level was a reliable and accurate indicator of health status.
Carter concludes that activity data, subjected to appropriate analysis, appears to generate nearly the same quality of information on the health status of cattle as human caretakers. The team is hopeful that further trials will demonstrate the system can exceed performance of human caretakers.
Carter believes remote monitoring of cattle health is the wave of the future but says commercial development for the Kentucky system will require several more years of testing. The ear tags are inexpensive, with a signal range of more than a mile, allowing data collection in a central location such as a feedyard office. The veterinarian could then access the information from a remote location. The team now is planning a large-scale test in a commercial feedyard setting to validate the results from earlier trials.
GrowSafe also has focused on software in recent trials. Beginning in 2010, the company carried out GSB trials in two commercial feedyards, intended to determine if multiple measurements acquired continuously could be built into a computerized algorithm that profiles and identifies disease occurrence in advance of visual assessment.
Sunstrum says the trial intended to compare the GSB system’s ability to detect sickness with that of feed yard cowboys using traditional methods. There is a general lack of agreement as to what constitutes a sick animal, at least in the early stages, she notes. Even in the scientific literature, a common assumption is that if an animal is diagnosed by visual observation and treated, it is considered sick.
In the GrowSafe trials, the feedyard veterinarian instructed health crews on protocols, and crew members made decisions on health interventions based on those protocols. Researchers monitored feed intake, water intake and bodyweight in real time with GSB and GrowSafe Feed Intake technology. The researchers collected carcass data, lung scores and liver scores on the cattle at slaughter.
Measurements collected during the trial exceeded 1.9 billion data points. Sunstrum says developing a real-time algorithm to analyze that volume of data required novel data acquisition, processing, analytics and visualization techniques, and incorporation of real-time feedyard data into software processing routines.
In this trial with high-risk calves, morbidity was above industry averages at 69 percent, and mortalities were within the industry average at 3.4 percent. Of 748 high-risk cattle that were enrolled in the trial, feedyard cowboys pulled 525 based on visual signs of sickness. There were 988 health interventions in total, including repeated treatments.
However, only 8 percent of the animals pulled and treated showed any reduction of water intake, feed intake and weight loss prior to treatment, and 129 animals that did not perform were not identified by the feedyard crew. GrowSafe data indicated that these animals were not drinking, feeding or gaining. The researchers found limited correlation between visual symptoms and subsequent treatment decisions, temperature profiles and performance.
Using the data collected during this trial, GrowSafe developed its algorithm to identify non-performing cattle based on water intake, normalized growth events and performance parameter changes.
The researchers concluded the value of the algorithm may not be in early identification, as thought, but in the consistency of disease identification.