Apparatus and method for detecting disease in dairy animals

Abstract

Disclosed is apparatus and a method for detecting udder disease in dairy animals. An accelerometer is attached to each of a plurality of dairy animals. A processor determines a measure of the activity of the dairy animals to which the accelerometers are attached. Data is recorded by and automatically transmitted from a sensor unit secured to an animal, without the requirement for costly and time consuming chemical analysis of milk, or of visual or veterinary inspection of individual animals in a herd. The development of an udder disease in a dairy animal, such as mastitis, may be identified from a decrease in the monitored measure of activity of a dairy animal. A separate baseline measure of activity may be determined for each dairy animal and the activity of a plurality of dairy animals in one or more herds may be taken into account, in order to reduce false positives due to external effects which are not specific to a single dairy animal.

Claims

1. An Apparatus for detecting udder disease in a dairy animal, the apparatus comprising: at least one accelerometer for attachment to a leg of the dairy animal; and a processor operable to: determine a measure of activity of the dairy animal to which the accelerometer is attached based on measurements from the accelerometer, wherein the determined measure of activity is a motion index calculated based on a measured net acceleration, minus an offset for gravity, of the accelerometer, summed over a period of time; determine a measure of feeding of the dairy animal from a second sensor in a region of head or neck of the dairy animal, the measurement from the second sensor distinguishing feeding from activity associated with walking or kicking, and determine that the diary animal has the udder disease from a change in the measure of activity of the dairy animal based on the measurements from the at least one accelerometer and also the measure of feeding of the dairy animal obtained from the second sensor, and thereby detect the udder disease in the dairy animal based upon the determination.

2. The apparatus according to claim 1, comprising a health condition detection module, operable to determine an acceptable of the measure of activity or an acceptable range of the measure of activity, of the dairy animal over one or more first periods, and operable to determine that the dairy animal has or may have the udder disease from the change in the measured of activity of the dairy animal from the acceptable measure, or the acceptable range of the measure, during one or more second periods.

3. The apparatus according to claim 1, comprising a plurality of accelerometers for monitoring the measure of the activity of one or more further dairy animals.

4. The apparatus according to claim 1, comprising a central data receiving unit, operable to receive data relating to the activity of the each dairy animal, from one or more sensor units, each said sensor unit secured or securable to the dairy animal and comprising the accelerometer.

5. The apparatus according to claim 4, wherein each said sensor unit comprises a transmitter, operable to transmit the activity related data to the data receiving unit.

6. The apparatus according to claim 1, wherein the at least one accelerometer is operable to identify motion which does not involve net horizontal displacement of the dairy animal.

7. The apparatus according to claim 1, wherein the at least one accelerometer can identify motion which involves the net horizontal displacement of the dairy animal.

8. The apparatus according to claim 1, wherein the at least one accelerometer is operable to distinguish motion in more than one direction, and wherein data from the at least one accelerometer may be used to distinguish between different types of activity, wherein the different types of activity are selected from a group consisting of: walking, feeding, standing, lying and transitioning between standing and lying.

9. A method of determining a presence of an udder disease in a dairy animal, comprising: using a computer processor, determining a measure of activity of the dairy animal using at least one accelerometer attached to a leg of the dairy animal, wherein the determined measure of activity is a motion index calculated based on a measured net acceleration, minus an offset for gravity, of the accelerometer, summed over a period of time, determining a measure of feeding activity of the dairy animal from a second sensor in a region of head or neck of the dairy animal, the measurement from the second sensor distinguishing feeding from activity associated with walking or kicking, and determining, that the dairy animal has the udder disease from a change in the measure of activity of the dairy animal based on the measurements from the at least one accelerometer and also the measure of feeding of the dairy animal obtained from the second sensor, and thereby detecting the udder disease in the dairy animal based upon the determination.

10. The method according to claim 9, comprising determining an acceptable of the measure of activity, or an acceptable range of the measure of activity, of the dairy animal over one or more first periods and determining that the dairy animal has or may have an udder disease from the change in the measure of activity of the dairy animal from the acceptable of the measure of activity, or the acceptable range of the measure of activity, during one or more second periods.

11. The method according to claim 10, wherein the acceptable measure of the activity is determined by taking into account one or more of: a time of day, a time of year, or an activity which is currently being carried out, or known health conditions of the dairy animal.

12. The method according to claim 9, comprising monitoring the measure of the activity of one or more further dairy animals to thereby distinguish the change in the measure of activity of the dairy animal from a change in the measure of activity of the one or more further dairy animals.

13. The method according to claim 12, wherein the acceptable measure of activity or the acceptable range of the measure of activity, of the dairy animal may be determined taking into account the measure of activity of one or more of the one or more further dairy animals.

14. The method according to claim 9, wherein one or more further motion sensors is or are secured to the each dairy animal.

15. The method according to claim 9, comprising thereby detecting mastitis using the determined measure of activity of the dairy animal to which the accelerometer is attached.

16. The method according to claim 9, comprising a estimating somatic cell count.

17. A non-transitory computer readable medium encoded with a computer software program which, when executed on a computer processor, causes the computer processor to: determine a measure of activity of the dairy animal using an accelerometer attached to a leg of the dairy animal, wherein the determined measure of activity is a motion index calculated based on a measured net acceleration, minus an offset for gravity, of the accelerometer, summed over a period of time, determine a measure of feeding activity of the dairy animal from a second sensor in a region of head or neck of the dairy animal, the measurement from the second sensor distinguishing feeding from activity associated with walking or kicking, and determine that the dairy animal has an udder disease from a change in the measure of activity of the dairy animal based on the measurements from the at least one accelerometer and also the measure of feeding activity of the dairy animal obtained from the second sensor, and thereby detect the udder disease in the dairy animal based upon the determination.

Description

DESCRIPTION OF THE DRAWINGS

(1) An example embodiment of the present invention will now be illustrated with reference to the following Figures in which:

(2) FIG. 1 is a schematic diagram of apparatus for detecting udder diseases in cattle;

(3) FIG. 2 shows measured activity levels of animals before and during an observation period;

(4) FIG. 3 shows four sub-plots of the residuals of data acquired from three dairy animals during the observation period; and

(5) FIG. 4 shows data acquired during the observation period fitted to Formula A.

DETAILED DESCRIPTION OF AN EXAMPLE EMBODIMENT

(6) An example of apparatus according to the invention is shown schematically in FIG. 1. A cow 1 is provided with a sensor unit 3, strapped to a hind leg. The sensor unit comprises a low power, low radio frequency transceiver, and a high power, high radio frequency transmitter, a three axis accelerometer (functioning as a motion sensor), a motion data storage unit and a battery.

(7) The sensor unit is operable to store motion data measured by the accelerometer in use and to transmit it to a central receiver 9. In an example embodiment, the sensor unit is operable between a dormant mode, wherein the transceiver, accelerometer and motion data storage unit is functional, and a higher power active mode, wherein the accelerometer, motion data storage unit and the transmitter are functional. When brought into the proximity of a tablet reader 5 (typically positioned at an entrance to a milking parlour) the sensor unit communicates with the tablet reader via low power, low radio frequency signals 7 to transmit the stored data. The sensor unit is thus caused to switch from a low power dormant mode to a high power active mode, in which the transmitter transmits stored motion data acquired when the sensor unit was dormant) to a central receiver 9 by high power, high radio frequency signals 10 and, optionally transmits (continuously or periodically) motion data for a further period of time, after which the sensor reverts to a dormant mode. Optionally, further communication with the tablet reader (for example when exiting the milking parlour) causes the sensor unit to revert to a dormant mode. The duration for which the sensor unit is in an active mode is thus minimized and battery life thereby optimised.

(8) The receiver is in wired communication with a server 11 (function as a data receiving unit) which communicates wirelessly (or, in some embodiments, over a wired connection) with a laptop computer 13 running data processing software (thus functioning as a processor).

(9) The server is also optionally in wireless communication (or in communication over the internet) with further computers 15 located remotely (typically each in communication with sensors secured to cows of different herds).

(10) The server monitors motion data from each of a plurality of cows and calculates a motion index for each cow (wherein the motion index is a measure of activity). The motion index is the sum of the measured net acceleration (minus an offset for gravity) to which a sensor unit attached to a respective cow is subjected, summed over a period of time, and is therefore representative of the kinetic energy transferred to the sensor unit by the cow to which it is attached.

(11) In some embodiments, the data measured by the accelerometer is transmitted to the central receiver and forwarded to the server in raw form. Thus, the server calculates the motion index. In alternative embodiments, the sensor unit comprises a processor which receives the motion data and generates activity related data, derived from the raw motion data, to be transmitted to the receiver. Thus, in these embodiments, the sensor unit calculates the motion index.

(12) The motion index associated with a cow is monitored over an extended period of time. The time averaged value of the motion index is periodically compared with a threshold value, functioning as acceptable motion. If the time average value of the motion index drops below the threshold for an extended period of time, an alert is generated by the server that the cow in question may be suffering from a health condition, such as mastitis.

(13) When determining the threshold value, the server takes into account corresponding values of the motion index calculated in respect of other cows in the herd. Thus, external factors which affect the entire herd can be taken into account, reducing false alarms. Values of the motion index calculated in respect of cows in other herds can also be taken into account to further improve the reliability of alerts.

(14) It may be that only the motion index calculated during certain activities, e.g. milking, or at certain times of the day is taken into account.

(15) Use of the apparatus was demonstrated in a number of experiments. In a first experiment, four lactating Holstein-Friesian cows were each fitted with sensor units comprising a three axis accelerometer on both of the rear legs. The sensor units were IceTag3D devices available from IceRobotics Limited of South Queensferry, United Kingdom.

(16) The cows were separated from herd on day one and were given three days to settle into their new pen. The animals were milked each morning and each afternoon (at approximately 7:00-7:30 am and 15:30-16:00 pm, respectively) on each day of the experimental period.

(17) They were infected with mastitis, by injecting Streptococcus uberis into two udder quarters, after scheduled afternoon milking on day four. The cows were observed for four further days before being treated with antibiotics after the scheduled morning milking of day eight.

(18) The specific time period over which the effect of the illness was considered was from midday of day four to midday on day eight. The observational period therefore included eight milkings.

(19) One of the cows was later diagnosed with a pre-existing Staphylococcus aureus infection, and went on to develop acute mastitis on day five. Data concerning this animal was therefore excluded from the data analysis. A tag on a second animal developed a fault during the observational period and the corresponding data was also disregarded.

(20) Somatic cell count and bacteriological level in milk samples from each udder quarter of three cows were taken at the eight milkings, during the observational period, and are shown in Table 1. Bacteriology data are presented in colony forming units/ml, somatic cell count data are present in cells/μl.

(21) TABLE-US-00001 TABLE 1 Results of milk analysis (average figures for all four quarters of each animal) Dairy Dairy Dairy animal 7159 animal 7189 animal 7387 Somatic Somatic Somatic Bacteri- cell Bacteri- cell Bacteri- cell Day/milking ology count ology count ology count 4 pm 0 41.25 0 63 0 18.25 5 am 0 16.25 85 42.75 5 13.75 pm 0 35.5 325 58.5 0 21.25 6 am 0 13 25050 42.25 30 10.5 pm 0 44.5 2540 674.75 5455 56.25 7 am 0 19.5 100 456.5 20 604 pm 0 35 0 346 0 1144.25 8 am 0 22 0 254.75 0 995

(22) Data shown in Table 1 indicate that that only two dairy animals out of the three dairy animals of interest were successfully infected (the milk samples from Dairy animal 7159 showing no significant levels of bacteria).

(23) For cows that have not previously suffered from mastitis, a somatic cell count above 100 cells/μl from any udder quarter is considered to be indicative of the disease. The data therefore indicate that the two remaining cows developed mastitis in the second half of the observational period, but that none of the three cows under consideration developed anything beyond sub-clinical mastitis (i.e. low level mastitis, which is typically not identified by visual inspection of the udders).

(24) Activity related data was recorded during the observation period (and transmitted by each receiving unit during milking) for a period of 255 minutes before and after transmission was triggered by the reader, immediately prior to milking, is shown in FIG. 2. Data corresponding to each milking period are presented for each animal as a summation motion index (related to total expended by each animal). In alternative embodiments, data is summarised over different periods. For example, in some cases, milking is conducted more, or less frequently, or data may be summarised in units of a calendar day.

(25) Initial elevated activity levels following the relocation of the animals tapers prior to the observational period (beginning at A).

(26) Data during the observational period were fitted to a linear model of Formula A:
y.sub.ijk=μ+DAIRY ANIMAL.sub.i+β log.sub.10(SCC.sub.ij)+e.sub.ijk  (Formula A)
Where:
y is the summarised motion index (MI) (an activity measure), i refers to each of the dairy animals, j=1, 2, 3, . . . , 8 denotes the different milkings during the observational period and k=1, 2 and indicates the leg (left or right) on which the measurement was made on each animal.

(27) The model assumes that the activity of a specific dairy animal around a given milking is equal to the sum of:

(28) 1) an overall, time-invariant mean activity level μ

(29) 2) a dairy animal specific term DAIRY ANIMAL (thus, the model takes into account the natural

(30) variation in activity between individual dairy animals)

(31) 3) a term proportional to the logarithm of the somatic cell count (SCC) of the milk

(32) 4) a residual term e

(33) Statistical analysis based on this model one indicates that SCC has a significant effect on the activity level of an animal, as shown in FIG. 4. In particular, an elevated SCC (indicating mastitis) correlates with lower activity levels.

(34) The analysis assumes that the residuals are independent, normally distributed and have constant variance. The last two of these assumptions may be verified using residual plots, which are shown in FIG. 3.

(35) The scatter plot of the residuals versus the predicted values (top-right subplot) shows an approximately even scattering around zero, suggestive of a constant variance.

(36) The bottom sub-plots are normal and half-normal plots, both of which follow a straight line, supporting the assumption of normality.

(37) FIG. 4 shows data acquired during the observation period fitted to Formula A. Straight lines are plotted through the data of each cow, fitted to a scatter plot of Motion Index versus log.sub.10(SCC).

(38) The decrease in activity with high somatic cell count can be clearly seen from the downward slope of the lines.

(39) The step count of each animal (determined from a further analysis of accelerometer data) was also determined, as was a “standing percentage” (percentage of time each animal was standing still during each period).

(40) Using the same type of mathematical model as set out in Formula A, but taking either the standing percentage or the step count as the activity related data, instead of the motion index, no statistically significant evidence of a corresponding correlation between SCC and these alternative measures of activity was shown.

(41) We have also found that the sampling duration either side of each milking, over which the Motion index summed, does not give additional results either. A reduced effect of the variable DAIRY ANIMAL.sub.i is observed in the model, which may be indicative that shorter sampling durations would be more appropriate following initial installation of the apparatus.

(42) Experimental results show a strong link between mastitis and a fall in activity levels, enabling a correlation between even sub-clinical mastitis to be identified, which may only be detected currently through expensive and time consuming chemical analysis.

(43) Far greater reductions in activity were seen in subsequent trials of animals having acute mastitis, detectable through activity as measured by motion index and, in addition, but also the step count and (in some cases) the standing percentage.

(44) The post relocation data prior to the sampling period also demonstrate that the importance of taking the behaviour of other individuals in a herd (or, a herd average) into account.

(45) In subsequent experimental trials, correlations have been observed between a decrease in activity and both clinical milk score and clinical udder score.

(46) In a further trial, data were acquired for animals in a herd over a period of several months and data concerning ten cases of mastitis (diagnosed conventionally) occurring in seven of the animals was retrospectively analysed in order to test whether the accelerometer data could be processed by the methods of the invention to identify the subsequently observed cases of mastitis.

(47) The parameters of motion index, step count and lying % were calculated from the data for each animal, for a period of one month before and after each diagnosed case of mastitis. (The motion index is a first measure of activity and step count and lying % are each second measures, being measures of activity types).

(48) Daily values were generated for each of: a 4-day TREND (a rolling value resulting from a linear fit to the previous 4 days of activity data) a *DIFF value (a daily value calculated from the difference in activity between an average or summation value from the preceding 24 hours as compared to the preceding 6 days)

(49) A range of coarse acceptability criteria were then applied to the values.

(50) Unlike the results shown in FIG. 4, where herd wide behaviour was taken into account by discounting post relocation elevation in the activity of all animals, herd wide behaviour and pre-existing health conditions were not taken into account.

(51) It was observed that all ten instances of mastitis corresponded to periods for which the *DIFF value exceed twice the *std value, however using this simple test, a number of false positives (which may be indicative of herd-wide variations in activity or other health conditions) were also observed, i.e. periods for which the *DIFF value exceeded twice the *std value which did not correspond to an observed case of mastitis. Similarly, it was observed that a TREND value of −5 or less was also found to be indicative of a possible mastitis case.

(52) Thus, it was demonstrated that the data could be used to identify that an animal may be suffering from mastitis.

(53) Further variations and modifications may be made within the scope of the invention herein disclosed.