Modeling cow somatic cell count using sensor data as input to generalized additive models.

This research paper presents a study investigating if sensor data from an automatic milking rotary could be used to model cow somatic cell count (composite milk SCC: CMSCC). CMSCC is valuable for udder health monitoring and individual cow udder health surveillance could be improved by predicting CMSCC between routine samplings. Data regularly recorded in the automatic milking rotary, in one German dairy herd, were collected for analysis. The cows (Holstein-Friesian, n = 372) were milked twice daily and sampled once weekly in afternoon milkings for 8 weeks for CMSCC. From the potential independent variables, including quarter conductivity, milk flow, blood in milk, kick-offs, not milked quarters and incomplete milkings, new variables that combined quarter data were created. Past period records, i.e. lags, of up to seven days before the actual CMSCC sampling event were added in the dataset to investigate if they were of use in modeling the cell count. Univariable generalized additive models (GAM) were used to screen the data to select potential independent variables. Furthermore, several multivariable GAM were fitted in order to compare the importance of the potential independent variables and to explore how the model performance would be affected by using data from various number of days before the CMSCC sampling event. The result of the model selection showed that the best explanation of CMSCC was provided by the model incorporating all significant variables from the variable screening for the seven preceding days, including the day of the CMSCC sampling event. However, using data from only three days before the CMSCC sampling event is suggested to be sufficient to model CMSCC. Variables combining conductivity quarter data, together with quarter conductivity, are suggested to be important in describing CMSCC. We conclude that CMSCC can be modeled with a high degree of explanation using the information routinely recorded by the milking robot.

Comparison of methods for predicting cow composite somatic cell counts.

One of the most common and reliable ways of monitoring udder health and milk quality in dairy herds is by monthly cow composite somatic cell counts (CMSCC). However, such sampling can be time consuming, and more automated sampling tools entail extra costs. Machine learning methods for prediction have been widely investigated in mastitis detection research, and CMSCC is normally used as a predictor or gold standard in such models. Predicted CMSCC between samplings could supply important information and be used as an input for udder health decision-support tools. To our knowledge, methods to predict CMSCC are lacking. Our aim was to find a method to predict CMSCC by using regularly recorded quarter milk data such as milk flow or conductivity. The milk data were collected at the quarter level for 8 wk when milking 372 Holstein-Friesian cows, resulting in a data set of 30,734 records with information on 87 variables. The cows were milked in an automatic milking rotary and sampled once weekly to obtain CMSCC values. The machine learning methods chosen for evaluation were the generalized additive model (GAM), random forest, and multilayer perceptron (MLP). For each method, 4 models with different predictor variable setups were evaluated: models based on 7-d lagged or 3-d lagged records before the CMSCC sampling and additionally for each setup but removing cow number as a predictor variable (which captures indirect information regarding cows' overall level of CMSCC based on previous samplings). The methods were evaluated by a 5-fold cross validation and predictions on future data using models with the 4 different variable setups. The results indicated that GAM was the superior model, although MLP was equally good when fewer data were used. Information regarding the cows' level of previous CMSCC was shown to be important for prediction, lowering prediction error in both GAM and MLP. We conclude that the use of GAM or MLP for CMSCC prediction is promising.

Haptoglobin and serum amyloid A in milk from dairy cows with chronic sub-clinical mastitis.

New tools are needed to detect chronic sub-clinical mastitis, especially in automatic milking systems. Haptoglobin and serum amyloid A (SAA) are the two most sensitive bovine acute phase proteins, and their concentrations increase in milk from cows with clinical mastitis and in milk from cows with experimentally induced chronic sub-clinical Staphylococcus aureus mastitis. The aim of this study was to further evaluate the potential for haptoglobin and SAA in milk as indicators of chronic sub-clinical mastitis. Quarter milk samples were collected from 41 cows with a mean composite milk somatic cell count (CSCC) above 300,000 cells/mL during at least two months prior to sampling. Quarter milk samples were also taken from eleven cows with a mean CSCC below 80,000 cells/mL during at least two previous months. These samples were analysed for haptoglobin, SAA, adenosine triphosphate (ATP) activity and bacterial growth. The samples were grouped according to their ATP, haptoglobin and SAA status. ATP+ samples had ATP > 2 x 10(-10) mol/mL, Hp+ and SAA+ samples had detectable levels of haptoglobin (> or = 0.3 mg/L) and SAA (> or = 0.9 mg/L), respectively. In udder quarter samples from healthy cows, 42 out of 44 samples belonged to the ATP-Hp-SAA- group. Among cows with chronic sub-clinical mastitis, the ATP+Hp+SAA+ group contained 66 out of 164 samples while 44 samples belonged to the ATP+Hp-SAA- group. Detectable levels of haptoglobin and SAA were found in 92 and 80 samples, respectively. Growth of udder pathogens was detected in 28 samples and Staphylococcus aureus was the most common bacteria. In conclusion, haptoglobin and SAA concentrations below the detection limit were considered as good indicators of healthy udder quarters. A substantial variation in haptoglobin and SAA concentrations in milk was observed in udder quarters with chronic sub-clinical mastitis.