Authors: M.D. Rasmussen, M. Bjerring
The general conditions for hygienic milk production in the EU are currently defined in Commission Directive 89/362/EEC (1989). The directive states that the appearance of the milk shall be checked before milking and abnormal milk should be withheld from delivery. A definition of abnormal milk was proposed in deliverable D6 of the EU-project Implications of the introduction of automatic milking on dairy farms (QLK5 -2000-31006). D6 defines abnormal milk as foremilk differing in colour and homogeneity from that of normal milk. In this deliverable this definition is used to test the ability of different AMS models to detect and hence divert abnormal milk at time of milking.
Five different models of AMS were tested in six herds and sampled for 13 to
48 hours to find at least 10 cow milkings with abnormal milk and 50 cow
milkings with normal milk. Due to the short sampling periods, the CMT-score of
the foremilk was used to identify and support classification of abnormal and
normal milk. Cows and quarters with a CMT-score >3 and no clots on a 0.1 mm
filter were omitted from the calculations.
The current AMS models have systems to produce alarm lists of cows that should
be checked for abnormalities of their milk, but these systems are not intended
for automatic diversion of milk at present. This should be taken into account
when evaluating the current systems. The sensitivity of the detection for the
six herds varied from 13 to 50% when calculated for the actual milking, from
22 to 100% for the test days, and from 43 to 100% when calculated for the
previous week. Specificities for the same time periods were found to be
87-100%, 85- 100%, and 35-100%, respectively. At present, the sensitivities
and specificities are generally too low for automatic diversion of abnormal
milk, and it seems that most of the models could benefit from application of
more sophisticated algorithms.
It is proposed that calculation of sensitivity should be based on sampling of at least 20 cow milkings with abnormal milk from a total of three herds. Sampling should be carried out for about 36 hours in each herd. At least 200 cow milkings with normal milk should be used to calculate the specificity. We propose that at least 16 out of 20 cow milkings with abnormal milk should be detected, which ensures that the sensitivity is 62.5% or better for a 95% confidence interval. For calculation of specificity, we propose that at least 198 cows should test normal out of 200 truly normal cow milkings, which will give a minimum specificity of 97.6% for a 95% confidence interval. Calculation of specificity at the quarter level improves the statistical power and lowers the confidence interval.
One model of AMS was tested for its ability to detect and separate milk coloured by blood. The model separated milk with ≥6 µmol haemoglobin per L, which is the level where a red tinge to the milk can be noted. Milk mixed with blood injected into the milk stream during a short period of time in the beginning of milking was not separated. We lack data on how blood is naturally expelled into milk and to what amount. We propose that testing of sensors should be based on milk samples mixed with blood to a concentration of 0, 3, 6, and 120 µmol haemoglobin per L of milk. Milk with 120 µmol haemoglobin should be detected at the quarter level. We propose that cow composite milk with >6 µmol haemoglobin per L should be separated because at this level milk will have a red tinge.
|
|
|
|
© Animal Sciences Group -
Wageningen UR. Last update:
20-02-2008 10:03. |
 |