VirtualVet: intelligence platform for early identification of physiological and environmental disorders in beef cattle

Abstract

In 2018, in Brazil, expenses on animal health related to beef cattle ranged around R$ 2 billion. This value was considered underestimated since it does not account for financial losses related to animal morbidity and mortality. Late detection of low-performing animals due to poor management and health issues negatively impact animal welfare and increases the risk of spreading contagious diseases among animals. Therefore, an intelligence platform that allows early identification of animal issues will enable a rapid management intervention and the treatment of diseases at early stages. The early animal intervention will have a positive impact on its performance, health, profitability, and animal welfare. It can be said that we are in the fourth industrial revolution, characterized by a set of technologies that allow the fusion of physical, digital and biological worlds. Following this trend, the livestock farming industry has been evolving in search for sensors integration, which collect biological data from animals with an intelligent network of processes associated to predictive algorithms. The use of these sensors in livestock allows to accomplish the first step: database structure, which allocates daily individual animal observations for the development of predictive algorithms. The main objective of this project is to collect data for the development of predictive algorithms, which will be based on the study of the circadian rhythm of behavioral variables recorded from beef cattle in feedlot. Our challenge is to develop algorithms that allow a quick and accurate detection of animal issues for early intervention to minimize negative impacts on animal welfare and feedlot performance. The final approach will be an intelligence platform (VirtualVet) that allows identification, in real time, of environmental (weather, management errors and other related external factors) and physiological disorders in beef cattle. Animals monitored by the BeefTrader technology (FAPESP 2015/07855-7) will have their data collected through automatic scales and intelligent cameras. Behavioral variables will also be recorded, while collecting live animal weight. Weather data will be recorded daily (temperature, humidity, and precipitation). Based on the daily data recorded, low performing and sick animals will be identified and flagged for individual clinical evaluation, including blood and laboratory tests. Environmental effects will also be evaluated as explanatory variables. The DFC (Degree of Functional Coupling) parameter that represents the strength of the circadian rhythm (0 to 100%) of behavioral variables will be calculated. For each animal, the daily DFC response curve will be outlined. Machine learning techniques, such as random regression models and neural networks, will be used to measure variation between individuals and detect deviations. Our main expected results are: 1) to create a database with individualized animal data collected daily, which allows characterizing health, management and weather issues that impact animal well-being, performance and, consequently, affecting feedlot profit; 2) to obtain individual average DFC parameter as an animal welfare score metric; 3) to focus on the circadian rhythm of behavioral variables of beef cattle for the development of algorithms with the potential to identify early physiological and environmental disorders. These results will be used in the development of the VirtualVet intelligence platform and will be presented at an international technical-scientific event. (AU)