Preliminary Study on a Data-driven Prediction Method for the Early Detection of Coccidiosis in Intensive Poultry Systems

Coccidiosis in poultry is still one of the main enteric diseases that can influence the performance of animals raised under intensive production system. Unfortunately, these enteric pathologies are not preventable, and diagnosis is available only when the disease is full-blown. The available diagnostic methods such as oocyst count and lesion scoring are extremely time-consuming and costly. For this reason, the use of a prediction method, that works in real-time, could provide valuable and rapid information for farmers with clear and suitable alerts in their daily routine. The quick reaction to any change in the health, well-being or productive states is the fundamental element for reduction of drugs usage, prevention and control of coccidiosis, and improvement of animal welfare. The main object of this research was to assess the possible relationship between the air quality data and the number of oocysts hosted in three different broiler houses. Prototypes have been developed and tested in three different experimental poultry farms based on collected data of the Volatile Organic Compounds (VOCs - organic chemicals and detectable to the human nose and to an electronic device such as electronic nose) emitted from broilers during the entire life cycle of the animals. A data-driven machine learning algorithm was built to relate VOCs data to the number of oocysts during time. For each broiler production cycle, the results showed that variations in the VOCs were related to the change of oocysts number, and specific critical VOCs values were associated to abnormal levels of oocysts count at an early stage of the disease. In conclusion, the results of this study support the feasibility of building an automatic data-driven machine learning algorithm for early warning of coccidiosis for intensive broiler production.