Round Corner
Department of Computer and Information Science


Prediction of osteochondrosis and skeletal health in pigs using computer tomography (CT) images and machine learning

Norsvin SA have since 2008 been using CT to improve the body composition, health and meat quality of pigs. Each year 3500 pigs get CT-scanned. With the new test station opening in Canada as part of Topigs Norsvin global test system for purebred boars, the number of tested animals increase to almost 10.000 tested animals annually.

One of the parameters being measured from the CT images is osteochondrosis. Today, this parameter is scored manually by investigating the skeleton from the CT images in 8 different locations, using a score between 0 and 4 (4 is the worst). This measurement is subjective based on the smoothness and homogenicity of the bone surface of condyles in humerus and femur. 0 indicate a smooth an even surface, while 4 indicates a fragmented and damaged surface. We would like to train a model or network on the manual annotation done by trained operators and try to predict the level of osteochondrosis in condyles in humerus and femur joints (and possibly add some more joints in pelvis and scapula) in breeding pigs.

Expected outcome of the task is to find out if
machine learning could be used to predict if pigs are likely to get osteochondrosis. The outcome of these results could then be implemented in Norsvins breeding program and help select animals with low or no risk of developing lameness due to osteochondrisis The solution should be evaluated according to established research metnods


John Krogstie John Krogstie
214 IT-bygget
735 93677 
NTNU logo