Hyperspectral offal inspection turns slaughter data into herd-health feedback

In Australia, an applied project on bovine offal shows how the abattoir can become a more useful source of animal-health data. Charles Sturt University and P&P Optica, supported by MLA and Tey’s Australia, are developing a system that combines hyperspectral imaging and machine learning to identify abnormalities in lungs, livers and kidneys. Its intended role is not to replace official inspection, but to complement it with an objective, repeatable assessment.

The model was trained with thousands of images and differentiates healthy tissue from six findings: liver abscesses, liver fluke, kidney cysts, pneumonia, pleurisy and nephritis. Hyperspectral imaging adds information beyond what is visible to the human eye. In pre-commercial trials under realistic processing conditions, the team reports an 88% accuracy rate for the six defined conditions.

The technical significance lies in closing the information loop. Inspection determines whether tissues are suitable and safe for use; the system can also generate comparable data on health patterns that can be returned to producers. Lesions detected on the line could therefore become signals for reviewing management, feeding, health programmes or lot-level trends.