BIOMETRIC IDENTIFICATION OF LABORATORY ANIMALS
Field of the Invention
The invention relates generally to a biometric identification of individuals and, in particular, to an ear image capturing apparatus, an image processing system and methods for identification of laboratory animals.
Background
In medical science and the pharmaceutical industry there is often a need for testing the reaction of drugs or treatments in living beings, for example test objects in the form of laboratory animals such as mice, rats and guinea pigs used for experimental purposes. In such cases it is common to take samples, inject substances or otherwise treat animals at scheduled intervals in order to allow observation of gradual responses in the test object. Such experiments can proceed from days up to 1-2 months. To maintain their natural behaviour, multiple animals coexist unconstrained in the same cage and each individual animal is taken from the cage only at events like weighing, sampling or treatment. After handling, the animal can be put back in the same cage or for various reasons in another cage.
It is of great importance in such handling to identify each individual animal because different animals can react differently on one and the same medical treatment. It is also of importance to use an inexpensive, last and simple method which does not hurt the animal or otherwise introduce pain to the animal. Preferably the method should be a non- invasive method.
Description of the related Art
A few identification systems which have been used in test objects like laboratory animals are known in the art. In U.S. Patent No. 4,399,821 "Free moving animal physiological monitoring and identification system and method" is described an identification of an individual animal based on the use of an electronic implant, which can hold animal identification data and can communicate with external interrogation devices. In U.S. Patent No. 4,262,632 "Electronic livestock identification system" is disclosed an electronic identification for livestock, particularly adapted to tagging ruminating animals, such as cattle and sheep. Both identification systems require a medical surgery of the animal. Further an implant of this type is too expensive to be used on mice, rats or guinea pigs which generally are used in medical testing.
Labelling with ear notch has also been used. This identification has also many drawbacks. One is that the notch area can be infected which sometimes results in unreadable codes, another is that the number of codes are limited.
Biometrics defined broadly is the scientific discipline of observing and measuring relevant attributes of living individuals to identify active properties or unique characteristics. Biometrics can look for patterns of change by measuring attributes over time or look for consistency by measuring attributes of identity or unique differentiation. When looking for patterns of change, biometric technology can be considered a tool for research, diagnosis, or even medical monitoring. When looking for consistency, biometrics becomes a useful vehicle for security. The most well-known example is identification of individuals using fingerprint or hand geometry but systems based on voice or iris recognition are also known in the art. Such techniques are described in, for example, "High Confidence Visual Recognition of Persons by a Test of Statistical Independence", J. G. Daugman (1993), IEEE Trans and "Pattern Analysis and Machine Intelligence", 15(11), pp. 1148-1161.
In U.S. Patent No. 6,229,905 "Animal identification based on irial granule analysis" is disclosed an animal identification which is designed to determine outlines of a pupil and irial granule of an eye. Similar systems are disclosed in U.S. Patent No. 6,320,973, U.S. Patent No. 5,572,596, U.S. Patent No. 6,757,409 and in WO01/35321. The inventions relates generally to a system designed to identify animals such as horse or cattle and not laboratory animals.
In British Patent No. 2,156,127 there is described a method and apparatus for the identification of individuals by means of scans of subcutaneous patterns of blood vessels using a radiation detector. One difficulty encountered with such system is the poor signal-to- noise ratio when endeavouring to detect blood vessel locations, for example, in the back of the hand. This may be due to hairs, or the non-planar nature of the locations or uneven illumination. The patent does not disclosed an animal ear image processing system and a method for identifying of a laboratory animal.
In U.S. Patent No. 5,787,185 "Biometric identification of individuals by use of subcutaneous vein patterns" captured images of vein pattern at predetermined regions are used for biometric identification of individuals and, in particular, to methods and systems for detecting the locations of subcutaneous blood vessels and for encoding such locations for storage on identity cards. It finds particular application in the verification of identity in transactions involving such cards' identity. The method is not for identification of laboratory animals.
SUMMARY OF THE INVENTION
The invention relates to a biometric identification of individuals and, in particular, to an animal ear image capturing apparatus, an image processing system and a method for biometric identification of an individual laboratory animal used in medical testing. The invention relates more specifically to an ear sensor for capturing an image of visible blood vessels in the ear of a laboratory animal and to electronic means for converting said captured image to a format which is searchable in a database of stored images. The invention relates further to a method for identification of an individual laboratory animal using a captured image and data from said database of collected images.
In a first aspect the invention provides an apparatus for capturing an image of an animal's ear comprising: a light source for illuminating the ear; an image sensor for capturing an image of the ear produced by light transmitted through or reflected by the ear; a communication interface for connecting the apparatus to a computer; a control unit for controlling the light source, the image sensor and the communication interlace.
In a second aspect the invention provides an animal ear image processing system designed to process an image of an ear and identify an animal, comprising: an ear sensor comprising a light source and optionally a fixture for holding the ear of the animal and an image sensor for capturing a contrasted image of the blood vessels of the animal's ear; electronic means with software and a first algorithm which identifies critical blood vessel patterns in the captured contrasted image and convert it to a searchable parameterized image representation; software with a second algorithm which can compare and recognize a parameterized image representation from a collection of images in a database; a central database with identification data of each individual animal optionally with the experimental data.
In a third aspect the invention provides an identification method for identifying individual laboratory animals comprising; capturing an image of the blood vessels of an ear of the laboratory animal;
converting said image to a parameterized searchable image representation; identifying the animal by comparing images from a collection in a database.
In a fourth aspect the invention provides software and an algorithm which can retrieve and identify unique structures of the blood vessel pattern of an animal's ear, the algorithm featuring the elements: extraction of the structures of the blood vessel pattern from a captured image; positioning and orientation of branching points of the blood vessel pattern in the image; determination of the total length of line segments at a certain orientation for line segments of a certain orientation; decomposition of the line segments in different lengths.
The advantages of the invention will be more readily apparent from the following detailed description.
BRIEF DESCRIPTION OF THE DRAWINGS
The invention will be described in detail below with reference to the attached drawings in which:
Figure 1 is a photograph of an ear which is fixed between two transparent plates.
Figure 2 is a photograph of an ear which is not fixed.
Figure 3 is a schematic view of an embodiment of an apparatus according to the invention.
Figure 4 is a schematic view of a part of the apparatus with an inserted ear. Figure 5 is a schematic view of a part of the apparatus with a clamped inserted ear.
Figures 6A, 6B, 6C are schematic views of a light pattern impinging on an ear, an equidistant dot pattern observed in the direction of the light pattern and a non-equidistant dot pattern observed in a direction at an angle to the light pattern, respectively.
Figures 7A, 7B, 7C are a red channel image, a green channel image and an image in which the green channel is subtracted from the red channel.
Figure 8 is a schematic view and a diagram of shadowing effects of objects in and outside the ear tissue.
DETAILED DESCRIPTION OF THE INVENTION
The invention relates to a biometric identification of individuals and, in particular, to an animal ear image capturing apparatus, an image processing device and a method for biometric identification of a test object such as an individual laboratory animal used in medical testing.
DEFINITIONS
In animal experiments, it is common to perform measurements, take samples, inject substances or otherwise treat animals at scheduled intervals. Multiple animals coexist unconstrained in the same cage to maintain their natural behaviour which requires social interaction. A study is a collection of animals divided in groups where animals in each group are treated similarly and effects of substances are compared by comparing average effects on animals between groups. In practice one sampling is performed as follows:
1. An animal is taken out of the cage. 2. The animal is identified with the ear image processing system described in this invention. The animal is held in one hand while the ear sensor is managed with the other hand. 3. An image of the blood vessels of the ear is captured by using the ear sensor and processed. Optionally a fixture holding the animal's ear is used. 4. The animal is weighted, treated or a sample is aspirated.
5. The animal is put back in either the same cage or a different cage.
6. Start over with step 1. "Next animal".
In this specification the term laboratory animal includes mouse, rat, and guinea pig.
In this specification, the expression "blood vessel pattern" includes the pattern of veins, arteries, capillaries and other blood vessels.
With algorithm is meant a mathematical expression with parameters obtained from the processed image. Algorithms for vein pattern recognition have been described. An example is given in Sang-Kyun Im et al, J. Korean Phys. Soc, 38 (3), 268 (2001), which describes a biometric identification for hand veins.
Any algorithm can be used as long as the algorithm can be used to identify the blood vessel pattern in the ear of a laboratory animal. It is specifically important to identify those structures in the blood vessel pattern which make a unique identification of each individual possible. According to embodiments of the invention, the position and orientation of the branching points of the blood vessel pattern, the total length of line segments at a certain orientation for line segments of a certain orientation and finally decompose the line segments in different length may be identified.
Another algorithm is used for extraction and quantification of the structures of the blood vessel pattern from the captured image of the individual's ear. Optionally a translation, a rotation and a scale invariant representation may be done out of the extracted pattern.
A schematic view of an apparatus 1 for biometric identification of a laboratory animal is shown in Figure 3. In one embodiment the apparatus 1 comprises a slot or a compartment 3 for insertion of an ear of an animal to be identified. The side walls of the compartment comprise plates 4a, 4b transparent to light. At one side of the compartment 3, a light source 5 provides light of a suitable nature as explained in further detail below. At the other side, an optical system 7 directs light transmitted through the plates 4a, 4b and an interposed ear to a digital image sensor 9, suitably of the CCD (charge coupled device) or CMOS (complementary metal oxide semiconductor) type. The image sensor 9 captures an image. A control unit 11 controls the exposure of the image sensor 9 and stores captured images in an internal memory 18. The apparatus 1 may communicate with a computer 15 through a communication interlace 17. The communication may be through a USB port using cable, IR light or short range radio. The computer 15 may in turn communicate with or incorporate a database 13.
In another embodiment, the capturing apparatus incorporates the necessary computer power and database.
Generally, the identification method is performed as follows. An animal's ear is positioned between the plates 4a, 4b. When positioned correctly, as sensed by the device or by the operator, an image is captured by the image sensor 9. The image is compressed by the control unit 11 or sent directly to the computer 15 for compression. Then the computer 15 compares the captured image with images in the database 13 in which the images are coupled to identities of individual animals. A match of the captured image with a stored image results in identification of the animal.
Figure 4 shows the plates 4a, 4b, the optical system 7, and image sensor 9 according to an embodiment of the invention. One plate 4b is movable to clamp an ear 6 in a suitable position. Figure 5 shows the plates 4a, 4b clamped together as a fixture with an interposed ear 6. Incident light 2a is transmitted through the plate 4a, the ear 6, the plate 4b, the optical system 7, and finally impinging on the image sensor 9.
The image in Figure 1 shows an ear which is fixed between two transparent plates. The image in Figure 2 shows an ear which is not fixed. Figures 1 and 2 are images depicting an ear from one and the same rat.
An ear may be considered a double curved surface. In order to improve the reliability, the ear 6 is flattened between the two plates 4a, 4b as is shown in figure 5. This avoids imaging errors otherwise resulting from curved surfaces which are not always inserted in the same way in the
slot and with the same curvature as most animals can move their ears. The clamping of the plates 4a, 4b is controlled by the operator and also triggers the exposure of the ear. When the distance between the plates 4a, 4b has reached a predetermined value, approximately equal to the ear thickness, and when this value is stable, at the same time as other sensors indicate that the operator wants an identification, the imaging process is started with the selected illuminations.
It may be desired to avoid clamping of the animal's ear. In another embodiment the geometry of the ear is measured and a flattened, normalised ear image is calculated on basis of the measurements. One alternative is to illuminate the ear with a known line or dot pattern impinging at an angle to the ear relative to the observation with the optical system. The distortion of the known line or dot pattern may be mapped unambiguously to the shape of the ear. If the known line or dot pattern has a diverging or converging shape as a function of the distance from the light source or lens, it is possible to calculate both the curvature of and the distance to the exposed ear surface. Hence, it is possible to determine both imaging errors and scale factors. Data from these measurements are also suitable as a trigger for the identification process.
Figure 6A shows a straight light pattern 14a impinging on an animal's ear 6. The pattern 14a may be produced by a light pattern source 10 comprising a light source such as a laser and an optical system as indicated schematically in figure 3. Figure 6B shows the produced dot pattern 14b reflected from the ear 6 (or transmitted through the ear) as observed in the direction of the light pattern 14a. As the dots are equidistant this is not useful. Instead the ear should be observed at an angle to the direction of the light pattern 14a, as is shown in figure 6C. As may be seen, the dots 14 b resulting from the straight light pattern 14a are not equidistant, and from their positions the appearance of the ear in a flattened state may be calculated.
In one embodiment a light source 5 with white light is used together with a colour image sensor 9 having red, green and blue channels. The red and green channels are extracted. To improve the visibility of the blood vessels an image produced by subtracting the green channel from the red channel (or vice versa) is used. The red blood contained in the blood vessels absorbs green light to a much greater degree than red light.
Figure 7A is a red channel image, while figure 7B is a green channel image. Figure 7C is an image in which the green channel is subtracted from the red channel. It may be seen that the blood vessels are visible in white. (An image with the red channel subtracted from the green
channel would have shown the blood vessels as dark.) To improve the visibility even further, the gain of the red and green channels may be adjusted before subtraction.
In another embodiment a light source with a narrower colour spectrum may be used to improve the visibility even further. Furthermore, it is possible to use a light source with monochrome light at one or more specific wavelengths and a monochrome image sensor 9. For example, monochrome light in the green part of the visible spectrum would be suitable.
In a further embodiment, the ear 6 is also illuminated with light 2b (figures 4 and 5) impinging on the same side as the image sensor 9, such that reflected light is captured in an image. This image will predominantly contain information about an area on and close to the surface on which the light is reflected, for example artifacts produced by dark hairs and skin pigmentation. By subtracting this image from the image resulting from the light 2a transmitted through the ear, the capability of identifying blood vessels is improved. It also makes it possible to deselect areas which are not suitable for identification due to hair growth and skin pigmentation, or other information contained in the image resulting from the reflection.
In order to further facilitate the identification, the correct exposure values (time and/or aperture) should be set for each exposure. Since a typical ear is not rectangular like a conventional light sensor, an area of the image is typically not covered by the ear, but empty and letting through all the light or reflecting no light. In one embodiment a first exposure is made and the edge structure of the ear is detected. Then the overexposed area outside the ear is ignored and correct exposure values are set on the basis of the actual ear surface.
The edge structure is also useful in the identification process. The edge structure by itself is an identification parameter. Thus, a number of images may be excluded from the blood vessel image analysis on the basis of the edge structure. The edge structure may also be used to adjust the rotation of the image relative to the images stored in the database 13. The edge structure also contains information about suitable areas containing identifiable blood vessels, as the blood vessel structure typically is too fine with no discernible veins and arteries close to the edge of the ear. These sorting methods reduce the amount of data that must be processed.
In another embodiment, the contrast of the captured image is investigated as a further way to deselect areas. Figure 8 shows a cross section part of an ear 6 with a blood vessel 8, and an object such as a dark hair 16 or the like located in the air. A diagram of the light intensity is shown to the right. As may be seen the object located in the air produces a sharp shadow (high derivatives), whereas the blood vessel 8 located in the ear tissue produces a more
blurred or softer shadow. Dark objects enclosed in tissue cannot produce the same intensity variation as dark objects in non-spreading media (air). Thus, areas with sharp shadows may be deselected because the area lying behind is probably shadowed by an object not relevant to the identification.
The present invention also provides an ear sensor processing system designed to process an image of visible blood vessels of an ear of an individual. The ear sensor processing system comprises
a. said light source 5 and optionally a fixture 4a, 4b for holding the ear 6 of the animal and an image sensor 9 for capturing a contrasted image of the blood vessels 8 of the animal's ear 6;
b. electronic means 15 with software and a first algorithm which identifies critical blood vessel patterns in the captured contrasted image and convert it to a searchable parameterized image representation;
c. software with a second algorithm 15 which can compare and recognize a parameterized image representation from a collection of images in a database 13;
d. a central database 13 with identification data of each individual animal optionally with experimental data.
The present invention also provides a method for identification of an individual and particular an individual laboratory animal in medical testing, the method comprising:
a. capturing an image of the blood vessels of an ear 6 of the laboratory animal;
b. converting said image to a parameterized searchable image representation;
c. identifying the animal by comparing images from a collection in a database 13.
The present invention also provides software and an algorithm which can extract and identify unique structures of a blood vessel pattern of an animal's ear, the algorithm featuring the elements:
a. extraction of the structures of the blood vessel pattern from a captured image;
b. positioning and orientation of branching points of the blood vessel pattern in the image;
c. determination of the total length of line segments at a certain orientation for line segments of a certain orientation;
d. decomposition of the line segments in different lengths.
The structure may be a vein pattern, an arterial pattern or a combined vein/arterial blood vessel pattern of the animal's ear.
The disclosed method can be used in identifying changes in the blood vessel pattern in a laboratory animal in connection with medical treatment.
It is a principal object of the present invention to avoid the disadvantages of the related art. The apparatus, methods and animal ear image processing system thus have many advantages in relation to described prior art presented above. The method is non- invasive, fast and simple and gives an accurate identification of the laboratory animal in the group. In contrast to implants and ear clip, it does not hurt the laboratory animal.
The invention is not restricted to the described detailed embodiments. Modifications as well as other aspects thereof may occur to those skilled in the art. The scope of the protection is only limited by the appended claims.