Bite-o-Mat: a device for the early detection of tail biting in pigs

Tail biting in pigs impairs animal welfare and causes economic losses. Detecting early indicators of tail biting would allow farmers to implement intervention measures and minimize negative consequences. The Bite-o-Mat, a device to automatically detect enrichment manipulation behaviour, was used to assess the suitability of this behaviour as an early indicator of tail biting. The Bite-o-Mat was installed in two rearing and two fattening pens on four farms (16 pens in total, 18–40 undocked pigs per pen). Based on daily inspections of each pig’s tail, days were classified as days with (= TL) or without tail lesions (= control, C). Control days within the 14-day period prior to tail lesions were designated as pre-tail lesion days and included in the TL class. A multi-layer perceptron classifier was used to classify TL and C days based on Bite-o-Mat data. (accuracy = 0.92, sensitivity = 1.00, specificity = 0.71, mean class accuracy = 0.84). Subsequently a time series classifier was trained to predict the class of the following day based on the Bite-o-Mat data of the preceding seven days (accuracy = 0.96, sensitivity = 0.97, specificity = 0.95, mean class accuracy = 0.98). As pre-tail lesion days were included in the TL class, the predicted tail lesion could occur any time from the next day up to 14 day later. Altogether, these results show the potential of the Bite-o-Mat for the early detection of tail lesions due to tail biting in pigs during multiple production periods.