A Model for Detection of Individual Cow Mastitis Based on an Indicator Measured in Milk

Abstract

A dynamic deterministic biological model was developed that generates, for a given cow on a given day, a value for her risk of having mastitis. The model combines real-time information from a mastitis indicator measured in milk with additional factors that are other known risk factors of mastitis but that are not reflected in the indicator. l-Lactate dehydrogenase (LDH), an enzyme whose activity is increased because of mastitis, is used as an example of a mastitis indicator. The additional factors incorporated in the model are days from calving, breed, parity, milk yield, udder characteristics, other disease records, electrical conductivity, and herd characteristics. The model is designed to run each time a new LDH value is recorded and can run in the absence of the additional factors. Electrical conductivity measurements and disease records, where available, also trigger the model to run. As an input, milk LDH activity values (μmol/min per L) are multiplied by milk yield (L) to produce the amount of LDH (μmol/min) and are then smoothed using an extended Kalman filter before being processed by the biological model. The output comprises a risk of acute mastitis and a relative degree of chronic mastitis. The model also produces a days-to-next sample value that allows sampling frequency to be either increased or reduced depending on the risk of mastitis. The days-to-next sample value was designed to make the best use of opportunities afforded by automated, inline sampling technology. The model functionality was investigated using simulated data, and real-farm data of naturally occurring mastitis were then used to validate the model. The results demonstrated that the model is robust to sampling frequency and random noise in the LDH measurements. It was able to detect mastitis reasonably well: Using a threshold mastitis risk of 0.7, sensitivity for detecting clinical mastitis was 82%. Specificity, that is, the ability to avoid misclassifying healthy observations as mastitis, was 99%.