A visual precision-livestock technology has outperformed expert humans in identifying behavioural signs of illness in pigs.
In a study led by researchers at Kansas State University, NUtrack, which uses camera technology and software based on machine learning, had a consistently-higher rate of diagnostic accuracy than a veterinarian and trained technician when faced with pigs showing sickness symptoms.
The 192 weaned pigs in the study were either challenged with lipopolysaccharide, a bacterial component which causes ‘sickness behaviours’, or given a sham treatment of saline. Pens of 16 pigs either all consisted of one treatment, or a mix of the two.
Technology provides longer-term consistency
The human participants and the technology were tasked with identifying which pigs had the lipopolysaccharide treatment, based on behavioural signs such as food and water intake, resting, as well as social behaviours.
The human experts correctly identified over 70% of challenged pigs and over 85% of non-challenged pigs on days 0 and 1 of the study, though this rate began to drop at day 2, and was lower when tested against pigs from the mixed pens.
In contrast, NUtrack identified over 79% of challenged pigs and over 94% of unchallenged pigs, up to day 3 of the work, while its performance was unaffected by what pens pigs were from.
Crucially, the human observers provided both more false positives and negatives than the precision technology over the course of the study.
“In commercial swine operations, false-positive pigs may be treated with antibiotics, which can impact the rate of antibiotic-resistant pathogens,” the scientists wrote in the journal animals.
“A more likely challenge in production is false negative pigs. False negative pigs may go undetected, even by experienced technicians. These pigs can potentially serve as vectors, especially as they are commingled into the finishing phase.”
Identifying social behaviours
The work demonstrated that NUtrack was better at observing more subtle behavioural signs in the pigs, they explained, such as animals pivoting away from ear-biting attempts by others in their pen.
However, such technologies should be seen as complementary to human input, they added. Human caretakers checking on pigs’ health can incite behaviours that are then picked up by machine learning algorithms, for example.
“The potential impact of this research may improve the labourer’s ability to treat animals at the individual level rather than the group level,” the team concluded.
