DEVICE AND METHOD FOR IN-OVO DETERMINATION OF THE SEX OF A FERTILISED BIRD EGG

Abstract

A device for the in-ovo determination of the sex of a fertilized bird egg, having a light source for emitting excitation radiation for exciting fluorescence in a region inside the bird egg, a spectroscopic apparatus for analysing, in a temporally and/or spectrally-resolved manner, fluorescence radiation emitted from the region inside the bird egg, an evaluation unit for determining the sex from the data determined by means of the spectro-scopic apparatus and a measuring head for jointly transmitting the excitation radiation into the bird egg and receiving the fluorescence radiation from the bird egg. The measuring head has an optical fibre system with a head end for transmitting the excitation radiation and receiving the fluorescence radiation.

Claims

1. Device for in-ovo determination of the sex of a fertilized bird egg, comprising: a light source for emitting excitation radiation for exciting fluorescence in a region inside the fertilized bird egg; a spectroscopic apparatus for analysing fluorescence radiation emitted from the region inside of the fertilized bird egg in a temporally and/or spectral-resolved manner; an evaluation unit for determining the sex from the data determined by means of the spectroscopic apparatus; and a measuring head for jointly emitting the excitation radiation into the fertilized bird egg and receiving the fluorescence radiation from the fertilized bird egg; wherein the measuring head comprises a light guide system with a head end for emitting the excitation radiation and receiving the fluorescence radiation.

2. Device according to claim 1, wherein the light guide system is a Y-shaped light guide system having two light guide strands which are brought together on the side of the head end.

3. Device according to claim 2, wherein the light guide strands are formed as light guide bundles, the individual light guides of which are intertwined at the side of the head end.

4. Device according to claim 1, wherein the spectroscopic apparatus is a device for analysing fluorescence radiation in a temporally-resolved or temporally and spectrally-resolved manner.

5. Device according to claim 1, wherein the evaluation unit is an evaluation unit based on machine learning.

6. Device according to claim 1, wherein the evaluation unit is an evaluation unit based on feature engineering.

7. Measuring head for a device according to claim 1, which is designed for jointly emitting excitation radiation into the fertilized bird egg and receiving fluorescence radiation from the fertilized bird egg, wherein the measuring head comprises a light guide system with a head end for emitting the excitation radiation and receiving the fluorescence radiation.

8. Method for in-ovo determination of the sex of a fertilized bird egg, comprising the steps of: emitting excitation radiation for exciting fluorescence in a region inside the fertilized bird egg by means of a light source; analysing fluorescence radiation emitted from the region inside the fertilized bird egg by means of a spectroscopic apparatus in a temporally and/or spectrally-resolved manner; and determining the sex from the data obtained by the spectroscopic apparatus, wherein by means of a measuring head the excitation radiation is emitted into the fertilized bird egg and jointly therewith the fluorescence radiation is received from the fertilized bird egg, wherein the measuring head comprises a light guide system and the emission of the excitation radiation into the fertilized bird egg and the reception of the fluorescence radiation from the fertilized bird egg is effected via a head end of the light guide system which is brought close to the fertilized bird egg.

9. Method according to claim 8, wherein by means of the measuring head the excitation radiation is emitted into the fertilized bird egg and jointly therewith the fluorescence radiation is received from the fertilized bird egg through a single hole in an eggshell of the fertilized bird egg.

10. Method according to claim 9, wherein the preceding step of: forming the hole in the eggshell without opening an egg skin underlying the eggshell.

11. Method according to claim 8, wherein the region inside the fertilized bird egg is a germinal disc region and/or a bloodstream region and/or a region of embryonic structures.

Description

[0037] In the following, the invention is explained with reference to the accompanying drawings based on exemplary embodiments, wherein the features shown below may each individually or in combination represent an aspect of the invention. In the drawings:

[0038] FIG. 1 is a structure with a bird egg and a device for in-ovo determination of the sex of said bird egg according to a preferred embodiment of the invention;

[0039] FIG. 2 is a representation of the corresponding spectra representing values a.sub.jk of (a.sub.j1, . . . , an) corresponding to the wavelength λ=500.89 nm for a male or female chicken egg; and

[0040] FIG. 3 is the representation of a classifier (black line) for the first scenario and Experiment 2 (6-day, through the egg skin).

[0041] FIG. 1 shows a schematic representation of a measurement setup with several bird eggs 10 and a device 12 for optical in-ovo determination of the sex of one of these bird eggs 10. The device 12 comprises a light source 14 for emitting excitation radiation in the form of light pulses for exciting fluorescence in a region 16 inside the bird egg 10, a spectroscopic apparatus 18 for analysing fluorescence radiation emitted from the region inside the bird egg 10 in a temporally and/or spectrally-resolved manner, a computer-based evaluation unit 20 for determining the sex from the data determined by the spectroscopic apparatus 18 and a measuring head 22 for jointly emitting the excitation radiation into the bird egg 10 and receiving the fluorescence radiation from the bird egg 10 through a single hole 24 in the eggshell 26 of the bird egg 10. The measuring head 22 comprises a light guide system 28 having a head end 30 for emitting the excitation radiation and receiving the fluorescence radiation. This head end 30 is also the head end of the measuring head 22 and is held against the hole 24 in the eggshell 26 for measuring.

[0042] The light guide system 28 of the measuring head 22 is Y-shaped and comprises two light guide strands 32, 34, one light guide strand 32 for the excitation radiation and one light guide strand 34 for the fluorescence radiation. The two light guide strands 32, 34 are brought together on the side of the head end 30. The light guide strands 32, 34 are each formed as light guide bundles 36, 38 whose individual light guides (not shown in detail) are intertwined on the side of the head end 30.

[0043] In the example shown, the light source 14 comprises a laser 40 for generating laser pulses and a frequency doubler device 42 connected downstream of the laser 40. Thus, the light source 14 is a pulsed laser light source.

[0044] In the example shown, the spectroscopic apparatus 18 comprises a spectrometer (a spectrograph) 44, a multichannel detector 46 configured as an ICCD camera, and a shutter 48 interconnected between the spectrometer 44 and the multichannel detector 46. A coupling between the laser 40 on the one hand and the multichannel detector 46 and the shutter 48 on the other hand is implemented via a so-called Q-switch 50 and a pulse generator 52 on the detector side. A computer serves as the evaluation unit 20.

[0045] In the example, the laser is an ND:YAG laser with a wavelength of 1064 nm. By means of frequency doubling, a wavelength of 266 nm is achieved for the excitation radiation. The laser-induced fluorescence signals are recorded by the multi-channel detector ICCD camera 46. The fiber bundle of the light guide system 28 not only guides the laser beam onto the sample, i.e. the inside of the bird egg 10, but also guides the generated fluorescence light to the detector 46. During the measurement, the complete spectrum is recorded at any time. This allows ZLIF measurements to be made directly on surfaces. The emission range is between 300 and 600 nm.

[0046] The crucial region inside the bird egg 10 where fluorescence radiation is excited is a germinal disc region, a bloodstream region and/or an region of embryonic structures.

[0047] An example of in-ovo determination of the sex of 3- to 6-day-old incubated chicken eggs by temporally-resolved fluorescence spectroscopy (ZLIF) by use of a device 12 is discussed below.

[0048] In all examinations, the orientation of the region 16 of the embryonic structures (in short, of the embryo) to the measurement head 22 is crucial. The position of the embryo can be determined by candling. This example relates to the examinations of 3- or 6-day-old incubated eggs. By use of the measuring system, chicken eggs can be examined in three ways:

[0049] 1. measurement through the eggshell 26;

[0050] 2. measurement without eggshell 26 through the intact egg skin;

[0051] 3. measurement directly on the embryo.

[0052] For the examinations of 3- or 6-day-old incubated eggs 10, the chicken eggs were aligned and then measured from above through the eggshell 26. In alternative two, the eggshell 26 was carefully removed so that a measurement could be made through the intact egg skin. In these two types of studies, the alignment of the embryo cannot be ensured without error. For this reason, in the third alternative, the egg skin was removed and a measurement was repeated directly on the embryo and on the blood veins.

[0053] The ZLIF measurements can be taken as a function of equidistantly increasing times τ.sub.1, . . . , T.sub.n for a fixed n as series of values or equivalently as signal series or vectors a=(a.sub.1, . . . , a.sub.n)∈.sup.n. Since at each time τ.sub.j, j=1, . . . , n the spectrum is recorded, it is possible to define the value series as (mathematical) matrices A∈.sup.mxn:

[00001]$A=\left(\begin{array}{cccc}{a}_{11}& a_{12}& .Math.& a_{1n}\\ a_{21}& a_{22}& .Math.& a_{2n}\\ .Math.& .Math.& .Math.& .Math.\\ a_{m1}& a_{m2}& .Math.& a_{\mathrm{mn}}\end{array}\right)$

wherein the columns (a.sub.1j, . . . , a.sub.mj).sup.T∈.sup.mj the above spectra at time τ.sub.j. Thus, for each individual bird egg, the result of the ZLIF measurement corresponds to a matrix of associated measurement values. Depending on the type of the ZLIF measurement as well as the sex of 3- or 6-day-old incubated eggs, such matrices are assigned to one of the following classes:

[0054] {Sex (m/f), Day (3/6), type (without eggshell, through int. egg skin, embryo)}.

[0055] Here, for error-free assignment of the sexes, the samples were analysed from the embryo in the Gene Laboratory by quantitative real-time polymerase chain reaction (qPCR) and corresponding primer sets.

[0056] Machine learning (ML) methods can be used to learn a system which allows a direct determination of sex from ZLIF measurements. These are the classical WIL methods as well as the own NIL relevant scientific publications of the inventors involved in this patent application.

[0057] Classes of different sexes of yet same incubation periods and same types are of interest for learning the system. This is not a limitation, since usually both the incubation periods and the type of measurement are known. The reasoning leads to the binary classification problem for the two classes: [0058] C.sub.X (for female) and C.sub.Y (for male).

[0059] For a set of chicken eggs of the same incubation periods, ZLIF measurements of the same type are carried out. Furthermore, the chicken eggs are assigned by qPCR to one of the classes C.sub.X or C.sub.Y. Thus we obtain objects with A.sub.X∈C.sub.X or A.sub.Y∈C.sub.Y with the structures corresponding to matrices A described above.

[0060] For learning the system for direct sex determination, the following feature extraction from object A is carried out. Here, it is assumed that the values of individual rows of A, i.e., (a.sub.j1, . . . , a.sub.jn), j=1, . . . , m are described by statistical distributions. Physically, the rows mean the temporally-resolved measurements for certain fixed wavelengths. FIG. 2 represents the example of a row (a.sub.j1, . . . , a.sub.jn) for a fixed j corresponding to the wavelength=500.89 nm for the measurement on a male (m) or female (w) chicken egg. It can be seen from the plot that by using different forms of the measurement series, the classification is possible in principle.

[0061] Central moments of a.sub.j:=(a.sub.j1, . . . , a.sub.jn), j=1, . . . , m describe statistical properties of a.sub.j. The first three moments are mean value μ, standard deviation δ and skewness S:

[00002]$\mu \left(a_j\right)=\frac{1}{n}{.Math.}_{k=1}^n{a}_{jk},\sigma \left(a_j\right)=\sqrt{\frac{1}{n}{.Math.}_{k=1}^n{\left(a_{jk}-\mu \left(a_j\right)\right)}^2},$$S\left(a_j\right)=\frac{1}{n}{.Math.}_{k=1}^n{\left(\frac{a_{jk}-\mu \left(a_j\right)}{\sigma }\right)}^3.$

[0062] For each object, m features (a.sub.j) and m features S(a.sub.j), j=1, . . . , m are extracted respectively. Thus, the feature vector f∈.sup.2.Math.m of each object is.

f=(σ(a.sub.1), . . . ,σ(a.sub.m),S(a.sub.1), . . . ,S(a.sub.m)).

[0063] To satisfy the accuracy of the classification of C.sub.X and C.sub.Y a feature selection is carried out. Here, by use of “Fisher's Linear Discriminant Analysis” (LDA), the features with weak separating properties are eliminated, since they contribute too little to the classification and may worsen results. The features with strong separating properties are used to construct a classification hyperplane by use of LDA.

[0064] Several datasets have been generated to train and validate the system for direct sex determination. For this purpose, a quantity M.sub.1 of 12+13 chicken eggs were taken and measurements were carried out. In total there were about 30 chicken eggs, nevertheless the sex by means of qPCR could not be determined for all of them. At a later time, the next ZLIF measurement took place on another quantity M.sub.2 of 15+10 chicken eggs. Here there were initially 40 chicken eggs, but as in the first test, the sex could not be determined for all. These two tests should show that the described method is suitable for direct sex determination. At least two measurements are necessary to validate the results.

[0065] In the first test with the quantity M.sub.1, 12 chicken eggs (6 m+6 w, 3-day-old embryos) as well as 13 chicken eggs (6 m+7 w, 6-day-old first without eggshell through the intact egg skin, then at embryos) were measured with ZLIF. In the second test with the quantity M.sub.2 15 chicken eggs (3 m+12 w, 3-day-old embryos) as well as 18 chicken eggs (14 m+4 w, 6-day-old first without eggshell through the intact egg skin, then on embryos) were measured. On each chicken egg the ZLIF measurements have been repeated 15 times. Thus, each chicken egg provides 15 objects for one of the classes. Table 1 summarizes the two tests with three experiments each.

TABLE-US-00001 TABLE 1 Two tests with three experiments each for learning and validating the system. The term #obj. means number of objects. # Obj. m/f # Obj. m/f (test 1) (test 2) Days Type 1 90/90   45/180 3 Embryo 2 90/105 210/60 6 Through egg skin 3 90/105 210/60 6 Embryo

[0066] For the selected temporal resolution, n is 20, the spectrum is equidistantly recorded starting with λ≈200.08 nm to λ=596.75 nm and m=1024. The feature vectors of each object thus correspond to:

f=(σ(a.sub.1), . . . ,σ(a.sub.1024),S(a.sub.1), . . . ,S(a.sub.1024))

and have the lengths 2.Math.m=2048.

[0067] Two scenarios have been tested for learning and validating the system. In the first case, the classes C.sub.X and C.sub.Y consist of the objects from both tests M.sub.1 and M.sub.2 and for teaching and testing the system, no distinction is made between tests. Thus, in C.sub.Y in experiment 1 there are 135 objects, in experiment 2 there are 300 and in experiment 3 there are also 300 objects. In C.sub.X in experiment 1 there are 270 objects, in experiment 2 there are 165 and in experiment 3 there are 165 objects.

[0068] Then, the classes C.sub.X and C.sub.Y have each been randomly subdivided into training objects and test objects. The classifier is designed for the training objects and the accuracy of the classifier is tested on the test objects. The 5×2 cross-validation method is used to determine the accuracy of the classifier. This is an established method for determining the accuracy of a classifier. The idea of the 5×2 cross-validation is to take an equal number of training objects and test objects, so that one can switch the training set to the test set and vice versa. The results of the 5×2 cross-validation provide a reliable mean value for the accuracy. Table 2 lists the results.

TABLE-US-00002 TABLE 2 Results of 5 × 2 cross-validation for objects of both tests Experiment Accuracy (Training) Accuracy (test) 1 3-day-old, embryo 88.89% 84.69% 2 6-day-old, through egg skin 96.46% 92.27% 3 6-day-old, embryo 96.13% 90.51%

[0069] FIG. 3 shows the representation of a corresponding classifier (black line) for the first scenario and experiment 2 (6-day-old, through egg skin).

[0070] In the second scenario, the tests distinguish between teaching and testing. In the second test, there are more objects than in the first. It is expected here that the more objects are used for training, the more stable the system is. For this reason, the classifier is trained on the objects from experiment 2. and tested on the objects from experiment 1. The results can be found in Table 3 below.

TABLE-US-00003 TABLE 2 Results for accuracy during training on objects from M.sub.2 and testing on M.sub.1. Experiment Accuracy (Training) Accuracy (test) 1 3-day-old, embryo 90.67% 76.67% 2 6-day-old, through egg skin 95.56% 89.23% 3 6-day-old, embryo 95.56% 87.69%

[0071] Conclusion: The experiments carried out show that a direct sex determination is possible in incubated chicken eggs using ZLIF. The incubation periods are between 3 and 6 days. A reliable accuracy between 76.67% and 92.27% can be achieved. This method has a great potential to improve the accuracy. The potentials are based on the optimization of the ZLIF experiment as well as on the optimization of the feature extraction, feature selection and the design of the classifier.

REFERENCE SYMBOLS

[0072] 10 bird egg [0073] 12 device for in-ovo determination of sex [0074] 14 light source [0075] 16 region (inside bird egg) [0076] 18 spectroscopic apparatus [0077] 20 evaluation unit [0078] 22 measuring head [0079] 24 hole (in eggshell) [0080] 26 eggshell [0081] 28 light guide system [0082] 30 head end [0083] 32 light guide strand [0084] 34 light guide strand [0085] 36 light guide bundle [0086] 38 light guide bundle [0087] 40 laser [0088] 42 frequency doubler device [0089] 44 spectrometer [0090] 46 multichannel detector [0091] 48 shutter [0092] 50 Q-Switch [0093] 52 pulse generator