Researchers from the University of Nottingham found that machine learning could enhance and improve a vet’s ability to accurately diagnose herd mastitis. Mastitis is a costly disease that affects dairy cattle. The disease costs around 170 million pounds per year in the U.K.
The team wanted to create an automated diagnostic support tool that can diagnose the herd level mastitis origin and be an early step in the Agriculture and Horticulture Development Board mastitis control plan.
The first step in controlling mastitis is identifying where mastitis pathogens originate. Vets have to figure out if the bacteria is coming from the cow’s environment or is it contagiously spread through the milking parlor? This requires time and special veterinary training to answer these questions.
Machine learning algorithms approach diagnostic problems in a similar manner to a student doctor or vet. The algorithm learns from data and applies what it learned to new patients.
Mastitis data from 1,000 herds were gathered over several three-month periods. Machine learning algorithms were used to classify herd mastitis origin and compared the results to expert diagnosis from a specialized vet.
The algorithm had a classification accuracy of 98 percent for environmental versus contagious mastitis. It also achieved 78 percent accuracy in classification of lactation versus dry period environmental mastitis compared to the expert’s diagnosis.
A paper on this research was published in Scientific Reports.