WO2008104922A1

WO2008104922A1 - Grid pattern determination system

Info

Publication number: WO2008104922A1
Application number: PCT/IB2008/050669
Authority: WO
Inventors: Willem M. J. M. Coene
Original assignee: Koninklijke Philips Electronics N.V.
Priority date: 2007-03-01
Filing date: 2008-02-25
Publication date: 2008-09-04

Abstract

An apparatus (195) for retrieving a grid pattern of a plurality of spots (113, 114) on a substrate (111) of a sensor carrier (110), wherein the plurality of spots (113, 114) comprise a plurality of reference spots (113) arranged for identifying the grid pattern based on an image of at least the plurality of reference spots (113, 114), the apparatus (195) comprising a first determination unit (163) adapted for performing a coarse gridding procedure for determining coarse information regarding the grid pattern based on an analysis of the image of the plurality of reference spots (113), and a second determination unit (165) adapted for performing a fine gridding procedure for determining the grid pattern based on the determined coarse information regarding the grid pattern.

Description

GRID PATTERN DETERMINATION SYSTEM

FIELD OF THE INVENTION

The invention relates to an apparatus for retrieving a grid pattern.

Further, the invention relates to a method of retrieving a grid pattern.

Moreover, the invention relates to a sensor carrier. Beyond this, the invention relates to a method of manufacturing a sensor carrier. BACKGROUND OF THE INVENTION

A biosensor assembly may comprise a membrane having a plurality of detection spots deposited on the substrate or membrane in accordance with a specific two-dimensional lattice of spots. The biosensor assembly may further comprise a light source for illuminating the membrane with light, wherein a fluorescence pattern originating from the spots may then be detected by a multi-pixel detector such as a CCD. On the CCD image, positions corresponding to all the possibly detectable spots of the membrane have to be reconstructed during image processing. EP 1,227,311 A2 discloses a method for imaging molecules contained in an array of discrete reaction sites on the surface of a solid support, comprising imaging the array and detecting a first molecule located on the solid support at a known position with respect to the array, by reference to the first molecule, aligning inspection windows in registration with the discrete reaction sites, and determining the amount of detectable signal in each window. The method is used to locate the reaction sites accurately on the array, and to correct for any misalignments.

However, the localization of the reaction sites on the array according to EP 1 ,227,311 A2 may lack accuracy and reliability. OBJECT AND SUMMARY OF THE INVENTION

It is an object of the invention to provide a sufficiently accurate and reliable grid pattern determination system.

In order to achieve the object defined above, an apparatus for retrieving a grid pattern, a method of retrieving a grid pattern, a sensor carrier, and a method of manufacturing a sensor carrier according to the independent claims are provided.

According to an exemplary embodiment of the invention, an apparatus for retrieving a grid pattern of a plurality of spots on a substrate of a sensor carrier is provided, wherein the plurality of spots comprise a plurality of reference spots arranged for identifying the grid pattern based on an image of the plurality of reference spots, the apparatus comprising a first determination unit adapted for performing a coarse gridding procedure for determining coarse information regarding the grid pattern based on an analysis of the image of at least the plurality of reference spots, and a second determination unit adapted for performing a fine gridding procedure for determining the grid pattern based on the determined coarse information regarding the grid pattern.

According to another exemplary embodiment of the invention, a method of retrieving a grid pattern of a plurality of spots on a substrate of a sensor carrier is provided, wherein the plurality of spots comprise a plurality of reference spots arranged for identifying the grid pattern based on an image of the plurality of reference spots, the method comprising performing a coarse gridding procedure for determining coarse information regarding the grid pattern based on an analysis of the image of at least the plurality of reference spots, and performing a fine gridding procedure for determining the grid pattern based on the determined coarse information regarding the grid pattern.

According to still another exemplary embodiment of the invention, a sensor carrier is provided comprising a substrate and a plurality of spots formed on the substrate in accordance with a grid pattern, wherein the plurality of spots comprise a plurality of reference spots arranged for enabling an identification of the grid pattern in accordance with a method having the above mentioned features.

According to yet another exemplary embodiment of the invention, a method of manufacturing a sensor carrier is provided, the method comprising forming a plurality of spots on a substrate in accordance with a grid pattern, and arranging a plurality of reference spots of the plurality of spots on the substrate for enabling an identification of the grid pattern in accordance with a method having the above mentioned features.

According to still another exemplary embodiment of the invention, a program element (e.g. a software routine, in source code or in executable code) is provided, which, when being executed by a processor, is adapted to control or carry out a method of retrieving a grid pattern of a plurality of spots on a substrate of a sensor carrier having the above mentioned features.

According to yet another exemplary embodiment of the invention, a computer-readable medium (e.g. a CD, a DVD, a USB stick, a floppy disk or a harddisk) is provided, in which a computer program is stored which, when being executed by a processor, is adapted to control or carry out a method of retrieving a grid pattern of a plurality of spots on a substrate of a sensor carrier having the above mentioned features. The grid estimation scheme according to embodiments of the invention can be realized by a computer program, that is by software, or by using one or more special electronic optimization circuits, that is in hardware, or in hybrid form, that is by means of software components and hardware components.

The term "plurality of spots" may particularly denote an array of several, for instance some hundred or some thousand, spots. Each of these spots may be a sensor portion of a sensor array. The spots may be arranged, for instance, in a matrix-like manner. Each or a part of the plurality of spots may comprise capture molecules or other probes which may be used during the actual sensor performance for identifying molecules of a sample, for instance by hybridization events. The plurality of spots may comprise a plurality of sensor spots which are (only) intended to use as probes during the sensor procedure. Additionally, the plurality of spots may comprise a plurality of reference spots.

The term "reference spot(s)" may particularly denote a particular subset of spots comprising a spot/spots formed at specifically selected positions on the array of spots and which serve as markers which allow to estimate a grid pattern according to which the plurality of spots are arranged. When the plurality of reference spots are arranged at an edge, more particularly at corners of the array of spots, their distance or pairwise distance may indicate or may allow to calculate back one or more parameters being characteristic for the pattern according to which the spots are arranged in a spot- depositing step or spot-printing step.

The term "corner marker" may particularly denote a reference spot arranged on an edge or, more precisely, on a corner of the arrangement of spots. Thus, corner-marker spots are an example for reference spots.

The term "grid pattern" may particularly denote a lattice structure according to which the spots are arranged on a substrate.

The term "coarse gridding" may particularly denote the fact that, according to exemplary embodiments of the invention, a rough approximation of gridding information may be determined based on a numerically simple and therefore fast procedure which can be performed with reasonable computational burden.

The term "fine gridding" may particularly denote the fact that, according to exemplary embodiments of the invention, refined or better resolved gridding information may be determined based on a fine tuning of the (previously determined rough) gridding parameters.

Thus, the coarse gridding may determine gridding information with a first accuracy level being smaller than a second accuracy level with which the gridding information is determined by the fine gridding procedure. The term "difference vector" may particularly denote a vector formed to connect two of the reference spots. Such a vector may be defined by a direction and an absolute value. It may in particular denote the two (horizontal and vertical) components of the difference vector that are of use for the fine gridding procedure, and not the sign of the difference vector. The term "unique" may particularly denote the fact that, according to exemplary embodiments of the invention, the reference spots are arranged according to a rule which prevents that different difference vectors connect two of the plurality of reference spots in an identical manner. Therefore, apart from a sign or a vector direction, each vector connecting a pair of reference spots may exist only one single time at the array. This may allow to unambiguously derive the grid pattern from the set of difference vectors, since a plurality of identical difference vectors yielding an identical positions on an autocorrelation image may securely prevented.

The term "autocorrelation" may particularly denote a mathematical tool which may be used in signal processing for analyzing the function or series of values. It may denote the cross-correlation of a signal with itself. In other words, it may denote the degree to which the return of a given series is related from period to period. Instead of a one-dimensional series of values, the auto-correlation can also be applied on a two- dimensional set of values, as applies for an image.

In the context of this application, the "substrate" may be made of any suitable material such as glass, plastics, or a semiconductor. The term "substrate" may be thus used to define generally the elements for layers that underlie a spot comprising layer or portions of interest. Also, the "substrate" may be any other base on which a structure is formed, for example a glass or metal layer.

The term "electromagnetic radiation" may particularly denote a beam of photons of any appropriate wavelength or any appropriate range of wavelengths. This may include the optical spectrum (for instance the range between 400 nm and 800 nm), but may also include electromagnetic radiation of other wavelengths such as UV, infrared, or even X-rays.

The term "dispenser device" may particularly denote any device for emitting or applying any substance to a specific region in space, particularly onto a defined surface portion of a substrate.

According to an exemplary embodiment of the invention, the gridding of an array of detection spots of a sensor array may be determined based on a two-step procedure. In a first step, the gridding parameters are estimated only very roughly, therefore in a very quick manner. These estimated gridding parameters may then be analyzed and refined in more detail by performing a fine gridding procedure which starts with the results of the rough approximation of the first step. This two-step procedure has turned out to be numerically very efficient to derive gridding parameters with high accuracy and with reasonable computational burden. According to an exemplary embodiment of the invention, it is possible to perform gridding for a sensor carrier (for instance a 2D-substrate). A possible application scenario is a sensor device comprising a liquid processing unit (producing the analyte molecules in one way or the other), a spotted array with capture probes (such as a sensor carrier), a reaction unit (for the reaction between spotted array and analyte molecules), and a detection unit for detection of the binding via binding specific signals. The spotted array with capture probes may be a disposable. Thus, a sensor carrier may denote a substrate used for sensing purposes or may denote a (spotted) array or an assay.

In the following, particularly two embodiments will be explained which may be used for realizing the above described two-step gridding procedure. According to a first exemplary embodiment of the invention, a sensor array and a method of reading out such a sensor array may be provided at which multiple reference spots such as corner markers are arranged according to a specific rule. According to this rule, no two difference vectors between any pair (2 -tup Ie) of reference spots may be identical to any other reference vector. By preventing double use of the same difference vector, it may be securely prevented that more than a single pair of corner markers contributes to the same peak in an autocorrelation function (ACF) of an image. Therefore, any undesired interference between two peaks in the ACF may be prevented allowing to accurately derive the grid pattern from the set of reference spots. The reference spots may be formed by probes carrying a fluorescence material. Therefore, in the absence of a sensor event, an image only of the reference spots by illumination with light and by detecting the fluorescence image may be obtained. On such an image, the positions of the reference spots may be easily identified. Using an autocorrelation algorithm or the like, it is possible to calculate back from the reference spots to the lattice parameters according to which the sensor spots and reference spots are arranged.

Therefore, according to an exemplary embodiment of the invention, a corner marker pattern may be provided for molecular diagnostic sensor arrays, allowing to improve gridding. Particularly, an efficient positioning of a limited number of corner markers on an assay may be provided in order to be able to reconstruct (CCD image) the 2D lattice that best fits.

Thus, according to an exemplary embodiment of the invention, a corner marker pattern for a molecular diagnostics sensor array optimized for gridding may be provided. In the lay-out of an assay (for instance with hexagonal lay-out), a number of so-called corner marker spots (also called reference spots) are included. These are positioned at the boundaries of the array, at well defined locations on the 2D grid. An embodiment of the invention relates to the issue how to optimize the pattern of corner markers in order to obtain the maximum gridding performance for the array. Such a gridding procedure may be based on the autocorrelation of the CCD image (preferably recorded before hybridization has taken place). Such optimized gridding may be advantageous because a limited number (typically less than 10) of corner markers are to be used (unlike in large-scale micro-arrays), and because the position inaccuracy for spot-deposition on the membrane (by ink-jet printing) is quite high, ranging from about 10% to 15% (a position error of 25 μm to 40 μm for an inter-spot distance of 254 μm). Thus, a method for retrieving the grid information comprising the lattice parameters and the angle made by two base vectors of a 2D grid may be provided, from a 2D gridding pattern of spots on the 2D lattice of spots, wherein

(i) any difference vector for any pair of spots of the gridding pattern pointing from a first spot of the spot pair to the second spot of the spot pair is a unique vector, apart from sign inversion, that is characteristic for that pair of spots, such that, for a number of N spots in the gridding pattern, there are a total of N(N-I )/2 unique difference vectors;

(ii) a quality- factor equal to Q = \ {^■= — _j + _, } is used to determine

1 1 the optimal gridding pattern of N corner markers. In such a method for retrieving the grid information, the 2D lattice may be a hexagonal lattice. Further, in such a method, the autocorrelation of the image containing only the intensity spots from the N corner marker spots may be used to generate the set of N(N-I )/2 unique difference vectors from which the grid information can be retrieved by means of a least-squares fitting procedure.

Thus, an improved or optimized method of positioning a limited number of corner markers at an assay may be provided in order to be able to reconstruct (based on a CCD image) the 2D lattice with the best fit-quality. According to a second exemplary embodiment of the invention, the coarse gridding procedure may be performed without any prior knowledge with the exception of the arrangement of the corner markers on the substrate. In other words, the rough gridding parameters are derived from the mere knowledge how the spots are arranged on the substrate, and therefore this method is referred to as brute-force method

Usually, the theoretical positions of all corner-markers are obviously known. The brute-force method is a kind of joint-spot-finding algorithm: it tries to find the most probable location of the pattern of the N (=8) corner-marker spots by performing a brute-force search over a well-defined range for each of the following 4 parameters (which are the 4 degrees of freedom in registration of the fluorescence image onto the CCD):

* a zoom- factor reflecting slight changes in magnification upon recording of the CCD-image(s); the zoom-factor can be reflected by the actual value of the hexagonal lattice parameter an, * a 2D translation- vector (or shift vector), which connects the origin of the CCD-image to the origin of the layout of the assay; it can be represented by the two components (T_x, T_y); the x-axis may be taken along the horizontal direction, and the y- axis along the vertical direction;

* a rotation of the layout with respect to the CCD-axes; it can be represented by an angle φ.

Variation of each of these 4 parameters (an, T_x, T_y, and φ) over their ranges yields a large number of parameter-settings. For each of these settings, a simple quality-criterion is evaluated, which is the sum of the intensity values of the N (=8) pixels that are closest to the expected corner-marker positions for the current parameter- setting (the expected positions can be any real number, and the closest integer number is used for the allocation of the pixels of interest for a given setting). Mathematically, there is the criterion C given by (for N corner-markers) k=N

C[a_H , T_x, T_y ,q>] = ∑I(x_{k[aH ,}τ_{x ,}τ_{y ,Ψ] >}y_{k[aH ,}τ_{x ,}τ_{y ,Ψ]}) ■ ^τhe values x_k &nά y_k are the closest k=ϊ integer numbers, closest to the coordinates of the k-th corner-marker for the given parameter-setting (an, T_x, T_y, and φ). The parameter-setting with the largest value for C is chosen as the best one, thus defining the coarse grid, that is, the coarse positions of the N (=8) corner-marker spots. One or more most likely configurations of corner markers may be determined using a trial and error approach, thereby determining rough approximations of degrees of freedoms of the gridding. The only information needed for this is the printing position of the corner markers on the substrate relative to one another. Then, all combinations of zooming, rotating and 2D-shifting may be calculated and the most likely configuration may be determined as the one that maximizes the above criterion denoted C.

In a subsequent fine gridding step, the 2D lattice may be fitted, for instance using information regarding difference vectors of reference spots (corner markers) for determination of rotation and lattice parameters, and using the information regarding the absolute positions of the reference spots on the captured image for determination of absolute shift vector of the arrangement of spots relative to the CCD image. According to such an embodiment, a procedure for fine-gridding on a 2D hexagonal array is provided using corner marker spots or reference spots. Such a fine- gridding procedure may be realized as well as a two-stage procedure. In a first stage, the lattice parameters of the basic rectangular grid (for the case of the hexagonal lattice) are determined via a least-squares approach, together with a possible small mis-orientation (or misalignment) of the grid relative to the horizontal axis of the CCD-image. In a second stage, the absolute position of the grid is determined with respect to the origin of the CCD-image. These two steps together may yield the 2D lattice or grid at which the spots on the array are expected to be deposited (or printed).

Next, further exemplary embodiments of the apparatus will be explained. However, these embodiments also apply to the retrieving method, to the sensor carrier and to the manufacturing method.

The first determination unit may be adapted for determining a plurality of difference vectors between each pair of the plurality of reference spots based on an image captured of at least the plurality of reference spots. These difference vectors are a proper basis for obtaining rough gridding information.

The first determination unit may further be adapted for determining the plurality of difference vectors based on an autocorrelation analysis. Particularly, in a sensor array which does not have too many spots, autocorrelation may be a powerful tool for deriving difference vectors with high accuracy and with small numerical effort. The first determination unit may further be adapted for determining the plurality of difference vectors exclusively based on an a priori knowledge of position information of the plurality of reference spots on the substrate of the sensor carrier. This has the advantages that no further information apart from the relative orientation of the spots on the sensor carrier is needed to determine a rough approximation of the gridding parameters.The second determination unit may be adapted for determining the grid pattern of the arrangement of the plurality of spots on the sensor carrier based on the plurality of difference vectors as generated by the first determination unit, using a least squares fit of parameters being a rotation of the arrangement of spots relative to the captured image as well as lattice parameters of said grid pattern. The least squares fit may use also absolute locations of the plurality of reference spots in the captured image to be used in a further least squares fit of parameters being a translational shift of the arrangement of spots relative to the origin of the captured image. The second determination unit may be further adapted for determining the grid pattern by performing at least one of the group consisting of a procedure of eliminating outliers, and a procedure of refining a previously determined grid pattern by their center of masses. Therefore, outliers, that is to say features or corner marker spots which do not fit in the general picture or deviate more than a predetermined threshold value from their expected location relative to the other corner-marker spots, may be eliminated from the further analysis. This may suppress artifacts and may increase the accuracy, although less than the maximum number of corner-marker spots are being used. Additionally or alternatively, the center of masses may be taken into account for getting a better resolved gridding result.

Advantageously, the apparatus may comprise a quality estimation unit adapted for estimating a quality of the determined grid pattern. By calculating a quality factor, the reliability of the determining gridding pattern may be controlled. A quality factor Q may be used to determine the optimal gridding pattern of the corner markers.

In the following, further exemplary embodiments of the sensor carrier will be explained. However, these embodiments also apply to the apparatus, to the retrieving method and to the manufacturing method.

In this sensor carrier, the reference spots (some of which may be corner marker spots) may be positioned to enable the retrieving method to be performable with such a sensor carrier. For instance, this can be achieved by a distribution of the reference spots which rules out a double occurrence of the same difference vector. This may also be achieved by arranging the reference spots along a perimeter of the spot array.

The plurality of reference spots may be arranged on the substrate in such a manner that each difference vector between each pair of the plurality of reference spots is unique. This may securely prevent that more than a single pair of corner markers contributes to the same peak in the autocorrelation function of the image. By selecting a corner marker pattern with the additional property that no two pairs of corner markers are allowed to generate the same difference vectors, uncertainties or artefacts in the evaluation may therefore prevented. For instance, when more than one single pair of corner-marker spots or reference spots have the same difference vector, then there will be more than a single pair of corner-marker spots contributing to the same peak in the auto-correlation of the image; actually, as a result there could be a much less well- defined peak, from which it is more difficult to determine the center of the peak in the auto-correlation image; also, a lower number of difference vectors enters the fitting procedure (of the fine-gridding step), which might also lead to a loss of accuracy in the overall gridding procedure.

Additionally or alternatively, each of the plurality of reference spots may be arranged on an outer circumference of an array of the plurality of spots. With such a configuration, it may be securely ensured that the two-step procedure may be carried out and that unambiguous information is derivable from such a sensor carrier. This can be seen as follows. The brute-force variant of the coarse gridding procedure tries to recognize the pattern of the corner-marker spots from the original acquired image. No shift of this pattern of corner-markers should also be a valid pattern in the lay-out. This property is realized by the measure just described. The apparatus may comprise an imaging device adapted for capturing the image of the plurality of reference spots. Such an imaging device may be a multiple pixel detector, such as a CCD (charge coupled device) or a CMOS detector. In a calibration phase, the reference spots may be imaged on the surface of the detector. Based on this image, it is possible to correlate coordinates on the image to coordinates on the sensor carrier, i.e. to perform gridding. For this purpose, the positions of spots originating from the reference spots on the detector may be evaluated to calculate back the lattice parameters defining the arrangement of the spots.

The apparatus may further comprise an electromagnetic radiation source adapted for illuminating the plurality of (reference) spots. Such an electromagnetic radiation source may be an LED, a laser, or any other light emitting lamp. However, it is also possible to use electromagnetic radiation sources of other wavelengths, like infrared radiation, UV radiation or even X-rays.

According to an exemplary embodiment, the plurality of spots may comprise a plurality of detection spots adapted for detecting the presence of molecules to be detected. Such spots or sensor cells may include capture molecules, electric sensor cells, magnetic sensor cells, electrochemical sensor cells, etc. The plurality of spots are arranged in accordance with a specific grid pattern (for instance in a hexagonal manner with specific angular and distance parameters) which has to be recalculated on an image taken from the plurality of spots before a sensor event has taken place or also after a sensor event has taken place, in which case there are many more spots lighting up due to the binding of analyte molecules to capture probe molecules of the detection spots. The capture probe molecules do not carry a fluorescent label; the analyte molecules do so. The sensor element may take place at the detection spots.

For instance, the plurality of reference spots may comprise a fluorescence material. Then, in the absence of a sensor event, illumination of the sensor carrier with electromagnetic radiation may cause only the reference spots to emit light which can be detected by a detector such as a CCD array. Alternatively, the reference spots may comprise a highly reflective material so that electromagnetic radiation impinged on the reference spots is reflected and sent to the detector. The material of the reference spots should therefore be configured to cause an image only of the reference spots. The plurality of reference spots may comprise at least one reference spot located at an edge of the grid pattern, particularly at a corner of the grid pattern. In a one-dimensional analogon of such a two-dimensional scenario, the reference spots should be arranged at the beginning and the end of a line. Then, based on the distance of the reference spots and based on a knowledge of the number of spots of the sensor carrier, it may be possible to determine lattice parameters based on information regarding the reference spots.

The grid pattern may be a two-dimensional grid pattern, particularly a hexagonal grid pattern. On a hexagonal grid pattern, a plurality of edges are provided which are specifically appropriate candidates for positioning the reference spots.

At least a part of the plurality of spots may comprise capture molecules adapted for hybridizing with complementary molecules to be detected. For example, the capture molecules may be DNA molecules which are immobilized on the substrate thereby forming a sensor surface. When a sample to be analyzed is brought in contact with the surface of the sensor carrier, possibly present molecules in the sample which have a complementary sequence to the sequence of the capture molecules may selectively hybridize with the capture molecules, thereby forming double-stranded molecule complexes. The presence of such a detection event may be read out electrically, optically, etc.

The sensor carrier may be a biosensor or a molecular diagnostics sensor carrier. A biosensor may be a molecular probe, or particularly an array of a plurality of molecular probes, measuring the presence or concentration of biological molecules, biological structures, etc., by translating a biochemical interaction at the probe surface into a quantifiable physical signal such as light or an electric pulse. However, embodiments of the invention may be applied to any array- like sensor structure, for instance a gas sensor, a temperature sensor, or any molecular sensor. The aspects defined above and further aspects of the invention are apparent from the examples of embodiment to be described hereinafter and are explained with reference to these examples of embodiment. BRIEF DESCRIPTION OF THE DRAWINGS

The invention will be described in more detail hereinafter with reference to examples of embodiment but to which the invention is not limited. Fig. 1 illustrates a sensor assembly according to an exemplary embodiment of the invention.

Fig. 2 illustrates a hexagonal coordinate system. Fig. 3 illustrates an equivalent representation of the hexagonal coordinate system of Fig. 2 with rectangular coordinates.

Fig. 4 shows an example of a hexagonal lay-out with 392 spot positions.

Fig. 5 shows a candidate corner marker pattern according to an exemplary embodiment of the invention with N_CM ⁼ 8 corner markers for the hexagonal lay-out with 392 spot positions of Fig. 4.

Fig. 6 shows a diagram illustrating a quality-factor for a gridding pattern of corner markers as a function of the number of corner markers (for the lay-out of Fig. 4).

Fig. 7 shows a two-dimensional autocorrelation function map of peaks generated from a corner marker pattern of Fig.4.

Fig. 8 illustrates an apparatus for manufacturing a sensor carrier according to an exemplary embodiment of the invention.

Fig. 9 illustrates a sensor carrier according to an exemplary embodiment of the invention. Fig. 10 shows an example of an erroneous gridding procedure.

Fig. 11 shows an example of a correct gridding procedure.

Fig. 12 illustrates a first step in a gridding procedure, called coarse gridding procedure, according to an exemplary embodiment of the invention. Fig. 13 illustrates an evaluation criterion in the example of an 8 point set of corner-markers according to an exemplary embodiment of the invention.

DESCRIPTION OF EMBODIMENTS The illustration in the drawing is schematically. In different drawings, similar or identical elements are provided with the same reference signs.

In the following, referring to Fig. 1, a sensor arrangement 100 according to an exemplary embodiment of the invention will be explained. The sensor arrangement 100 comprises a sensor carrier 110 and a plurality of further components, including a readout apparatus 195.

The sensor carrier 110 comprises a substrate 111 which may be a membrane. On the substrate 111, a plurality of spots 113, 114 are formed in accordance with a specific grid pattern 112. At cross-sectional points of rows and columns of the grid pattern 112, the plurality of spots 113, 114 are formed. These spots 113, 114 comprise reference spots 113 and detection spots 114. The reference spots 113 a subset of which may also be denoted as corner markers are provided from a fluorescence material and are adapted and arranged for identifying the grid pattern 112 based on an image of the plurality of reference spots 113 which image may be captured by a CCD camera 150. In the middle area of the layout, there may be also reference spots 113 that are used for the purpose of identification and intensity calibration. These spots are also pre-labelled, thus containing a fluorophore already after deposition on the substrate.

Thus, reference spots 113 may include corner-marker spots and intensity- calibration-spots and identification spots (the last two together can be a joint class or category of spots).

According to an exemplary embodiment of the invention, the plurality of reference spots 113 are arranged at such positions of the matrix- like arrangement of the spots 113, 114 that each difference vector between each pair of the plurality of reference spots 113 is unique. In other words, the difference vectors may be formed by (virtually) connecting two reference spots 113, that is to say each pair of reference spots 113 forms a unique difference vector. When comparing the individual reference vectors, these should be different for each pair of reference spots 113.

As will be described in the following, the reference spots 113 are used to allow an image evaluation unit or central processing unit (CPU) 160 to calibrate or assign detection pixels on the CCD camera 150 to the spatial arrangement of the spots 113, 114 on the substrate 111. Such a correlation is useful for the evaluation of a sensor event. For this purpose, a light source 180 such as an LED emits a beam of light 181 which is impinged on the surface of the substrate 111. In the absence of a sensor event which will be described below in more detail, only the fluorescence reference spots 113 generate light spots on a surface of the CCD detector 150. With the knowledge that each pair of reference spots 113 provides only a unique reference vector, the CPU 160 is capable of calculating the lattice parameters, particularly to determine the lattice parameter "a" characterizing the grid pattern 112 in the case of the ideal undeformed hexagonal lattice, or two lattice parameters denoted a and b in case of a deformed lattice..

For this purpose, the apparatus 195 is adapted for retrieving the grid pattern 112 of the plurality of spots 113, 114 on the substrate 111 of the sensor carrier 110. The apparatus 195 comprises the determination unit 160 adapted for determining the plurality of unique difference vectors between each pair of the plurality of reference spots 113 on an image captured by the CCD camera 150 of the plurality of reference spots 113. The CPU 116 serves as a deriving unit for deriving the grid pattern 112 of the plurality of spots 113, 114 from the plurality of unique difference vectors, for instance by performing an autocorrelation algorithm. The ACF may benefit from using a set of unique difference vectors. However, the brute-force variant for the coarse gridding does not require that property.

More particularly, the CPU 160 comprises a first determination unit 163 adapted for performing a coarse gridding procedure for determining coarse information regarding the grid pattern based on an analysis of the image of the plurality of reference spots 113, and comprises a second determination unit 165 adapted for performing a fine gridding procedure for determining the grid pattern based on the determined coarse information regarding the grid pattern. After such a calibration, the CPU 160 has derived the 2D lattice parameters characterizing the 2D grid of spots. For detecting a sensor event, the detection spots 114 (which may comprise different capture molecules) may be brought in functional contact or fluidic communication with a fluidic sample so that hybridization events may occur between molecules to be detected and the capture molecules assigned to the spots 114. When, for instance, the detection molecules to be detected comprise fluorescence labels, this may result, after illumination of the sensor surface 111 by the light source 180, in a spot pattern on the CCD camera 150. To determine which spot on the image detector 150 is assigned to which detection spot 114 on the substrate 111, the previously derived gridding information is essential.

As will be described below in more detail, the CPU 160 may also calculate a quality factor Q indicative of the quality of the estimated gridding characteristic of the reference spots 113.

Furthermore, as can be taken from Fig. 1, the sensor arrangement 100 comprises a user interface 190 which allows a user to bidirectionally communicate with the apparatus 195. The user interface or input/output unit 190 comprises input elements like a keypad, a joystick, a trackball or even a microphone of a voice recognition system. Furthermore, the input/output device 190 may comprise a display device such as a cathode ray tube, an LCD device, a monitor, a TFT device, a plasma device, etc.

Next, corner markers and gridding will be explained in more detail.

The spots 113, 114 are printed on the membrane 111 according to a 2D grid 112 of lattice points. In the detection step, the array 110 is illuminated with a light source 180 such as an LED, and the induced fluorescence yields an image that is recorded on the CCD camera 150. Detection can be done prior to hybridization, or after hybridization. In the former case, only the spots 113 that are printed with fluorophore- labelled molecules (as are the corner marker spots) will light up in the image. The spots 113 detected on the CCD camera 150 are arranged according to the 2D lattice 112 with which the spots 113 of molecules have been printed on the membrane 111.

Gridding may be denoted as the processing step that is concerned with finding back from the spots 113 in the CCD image the equivalent 2D grid 112 on the CCD 150 that corresponds one-to-one with the 2D grid 112 with which the spots 113, 114 have been printed on the membrane 111 in the first place. In this sense, gridding may be considered to be similar or analogous to timing recovery in a receiver for (optical or magnetic) storage, with the difference there that many more bits along a track than spots 113, 114 on the membrane 111 are available for this purpose. Gridding may be the very first step in the overall signal processing chain, and can thus be considered as the mother of all detection.

In the following, a simplified view of a gridding procedure will be explained.

For reasons of simplicity, a 1 -dimensional case will be considered next, with two corner markers at ideal ID-lattice positions mi a and m₂ a, with "a" the lattice parameter, that is the closest distance between two consecutive lattice points. The measured positions as obtained from the CCD image are denoted Ri = mi a + Δi and R₂ = m₂ a + Δ₂. The positional inaccuracies Δi and Δ₂ consist of a common (and therefore irrelevant) shift Δ and a random shift denoted δi and δ₂ for the two corner markers. The difference in position is thus Δ R = R₂ - Ri =(m₂ - mi) a + (δ₂ - δi), from which the as- detected lattice parameter can be determined as:

AR a

ΪI Ϊ .I — πu - (1)

The error on the lattice parameter is then given by

Aa ^δ2 *' .

(2)

Next, a gridding procedure in two dimensions will be explained.

In the following, a scenario with a minimal set of three non-collinear corner markers will be discussed.

With three non-collinear corner markers, we can derive three different vectors Ri - R2, R2 - R3 and K₃ - Ri. Since they are not collinear, it is possible to derive two lattice parameters from them, in case the angle between the two base-vectors is known a priori; otherwise, it may be necessary to derive all three of these parameters. In the following, a general set of corner markers will be discussed.

With NcM corner markers, it is possible to derive NCM (NCM - 1) / 2 pairs of corner markers, and thus the same number of difference vectors, from which the lattice parameters can be fitted. This procedure will be explained in more detail below.

In the following, a use of autocorrelation (ACF) of the CCD image for gridding purposes will be discussed.

A convenient way to determine the 2D map of all difference vectors (corresponding with all pairs of corner markers) is by computation of the autocorrelation of the CCD-image (as obtained from the corner markers only). Denoting the CCD image intensity by I(R) with R the 2-dimensional position coordinate in the image, its autocorrelation is then given by Ai(R), defined as:

which can be efficiently computed via Fourier space (using F(ast)FTs), since:

where FT and FT^"1 represent forward and backward Fourier transforms, respectively.

Assuming now that the CCD image contains a set of peaks at the positions of the corner markers (only), then the autocorrelation of the CCD image also contains a set of peaks at positions that correspond to the difference vectors that connect one spot (corner marker) in the image to another spot (corner marker) in the image. This is similar to the so-called Patterson maps in X-ray crystallography.

Next, extra requirements on the pattern of corner markers will be discussed.

It is well possible that, for a given set of corner markers, at least two pairs of corner markers (say Si and S₂, and S3 and S₄) yield the same difference vector (ΔR), and will thus contribute to the ACF intensity at the very same position (apart from the position noise introduced by the printing step). Close to the nominal position of the ACF peak, for this case the contributions

/ SV R i AR i V tfoi R ! O₂ ) CiR \ f Sy- R I ΛR i 4 ⁱ ,S^v.j i R \ *j u/R

(5) are present where O₁ is the 2D position error vector of spot i (due to the printing process).

Conventionally, problems may occur during a gridding procedure. The fact that more than a single pair of corner markers contributes to the same peak in the ACF of the image leads to the following problems:

(i) instead of a single peak, there may be two peaks, so that, when searching for a single peak in the ACF, the other peak may be missed which may be as relevant as the other one; this could be solved by searching for two local peaks in the ACF instead of one, but this might become troublesome when occasionally a single peak results. (ii) the two peaks may interfere with each other, giving rise to less defined maxima, making it more cumbersome to identify the autocorrelation peaks.

In the following, a solution of such problems according to an exemplary embodiment of the invention will be discussed.

The above issues can be resolved by selection of a corner marker pattern with the additional property that no two pairs of corner markers are allowed to generate the same difference vector; in other words, every difference vector between two corner marker spots is unique in such case.

Next, a hexagonal lattice as described by a rectangular coordinate system will be discussed. In the case of an array with a hexagonal lay-out, a convenient way to index spots on the 2D grid is by using the hexagonal coordinate system as shown in an illustration 200 Fig. 2.

The two base vectors a_H and bπ enclose an angle of 60°, and bπ is oriented along the vertical axis. A position vector R on the 2D-grid is given by R = i a_H + j bπ with i and j any integer numbers.

Equivalently, it is possible to describe the position vector R by means of a rectangular coordinate system as shown in an illustration 300 of Fig. 3, with respective base vectors denoted a_R and b_R. The position vector is then indexed by means of the integer numbers h and k as R = h a_R + k b_R with the restrictive condition that h + k must be even. The transformation from the hexagonal coordinates (i; j) towards the rectangular coordinates (h; k) is given by:

h VΛ, i k V.-.-.-. 2j \ i

(6) from which it is also clear that h + k is even, for any integers i and j. For the optimization procedure, the use of the rectangular lattice may yield a more simplified procedure, and its lattice parameters will be denoted simply by a ≡ a_R and b ≡ b_R.

The measured vectors will be denoted with respective coordinates along x (horizontal) and y (vertical) axes.

In the following, corner marker pattern and difference vectors will be explained.

A pattern with a number N_CM of corner markers will be considered. Each corner marker has position coordinates (on the rectangular lattice) denoted (h_l5 k^, with i = 0, 1 , ... , NCM - 1 • There are N_p = NCM (NCM - l)/2 essentially different pairs of corner markers, with their difference vector denoted Ri with coordinates (mi, ni), given by, for the 1-th pair with corner marker indices i and j :

"*t δ _ h_j i t$ ιγ_? A' .

(V)

Next, a least squares gridding procedure based on difference vectors measured as peaks from the ACF will be described.

It will be described how to derive the gridding information from a set of difference vectors as obtained from the corner marker spots (only). It will be considered only those corner marker patterns that give rise to the full number N_p = N_CM (N_CM - l)/2 of essentially different difference vectors, that is, no two difference vectors are allowed to be identical. This condition is required to have the maximum of N_p peaks in the autocorrelation of the as recorded CCD-image (with the corner marker pattern only).

Denoting the experimentally measured difference vector by R^ , and given the positioning error A₁ with components Ai _x and Ai _y due to the inaccuracy in the printing process, it is obtained:

ϊT. H, A₁

(8)

For each difference vector R^ , the corresponding coordinates on the grid are known, that is, (mi, ni), and it is possible to use this information to retrieve an

estimate for the grid parameters a and b (where the ^Λ-sign refers to the fact that these are estimates, that may differ from the real values a and b). In order to combine all the difference vectors, the least-squares sum of position information is considered which is given by:

,S'^~ > i [{/ ^■- t ItI : « . Ui 6 i ! ^£

y '. N i 2 ^■■■■ m i a ι ~ [ h i. tf ^■■■■ n i b i ~

(9) where the grid parameters a and b are to be measurements, given by the

set of difference vectors R^ (all the sums like V consider summation over all unique difference vectors, equal to N_p = N_CM (N_CM - l)/2). Least squares fitting yields

Y^{^} mi Jh.. a έ > — ^ t _f. tny i

(10)

Since:

(11) and

h h \ Λ_b

(12) it is possible to derive the fitting errors on the lattice parameters a and b as:

The overall quadratic gridding ; error is defined as:

Next, it is assumed that the positional errors on the difference vectors are uncorrelated such that (where < ... > denotes statistical averaging)

— p. x —^> q, .r '^■V. 1 J¹

(15) where δk,i represents the Kronecker delta (equal to 0, unless k = 1 and then equal to 1), and where Δ_x ² and Δ_y ² represent the average quadratic positional error (for a printed spot) along x and y axes respectively. Further, with Δ_av ² the overall average quadratic positional error (for a printed spot), it holds that

^■A ; A; -, ^

(16)

So finally, the relationship between the overall average quadratic positional error and the overall quadratic gridding error can be derived as:

Therefore, the quality factor Q of a given pattern of corner markers, the inverse of which reflects the amount of gridding information, can be defined as:

Given a certain amount of positional uncertainty due to the printing process, quantified by the parameter Δ_av ², the Q-factor determines how much of this uncertainty is coupled back into the overall gridding uncertainty, that is, uncertainty in the as-determined lattice parameters a and b.

Next, consecutive steps in an overall gridding procedure will be discussed. The following steps may be distinguished in the gridding procedure (with steps 4-5 referring to the fine gridding operation): Step-1 : compute autocorrelation of CCD-image (with corner markers only).

Step-2: locate the 2D-Map of nominal difference vectors, with proper zooming/scaling and rotation, in the ACF computed above. There is no origin problem, since the ACF applies to difference vectors, not to absolute position vectors. Step-3: determine the peak-positions in the ACF in the vicinity of the

(scaled and rotated) 2D-Map of nominal difference vectors.

Step-4: perform least-squares fitting using the above ACF-peak-positions as input for a least-squares fitting procedure. This step yields the 2D grid, but not yet its absolute position with respect to the CDD image. Step-5 : determine the required global shift of the grid by maximizing the averaged intensities at the grid-points for different positions of the grid.

In the following, a search for an optimal corner marker pattern will be explained, optimal in the sense of yielding the best gridding performance.

It is possible to allocate N_CM corner marker spots among the N_ed_ge spots located at the edges of the 2D array of spots. So, there are cNedge ₌ N edg^ ! _{( χ} ^

^J N^V CM U Λ^JN^V edge - ^JN^V CMV ) ' ^• possible allocation of corner markers along the edges of the array. An exemplary optimization scheme has been carried out for the hexagonal lay-out 400 as shown in Fig. 4. For the case that N_CM = 8, the optimal gridding pattern (without any extra constraints; later on, also practical constraints of the depositing device may be taken into account) 500 is shown in Fig. 5.

Also, when generating an optimized gridding pattern for a range of number of corner markers, it is possible to plot the quality factor as a function of N_CM, as is shown in Fig. 6 for a hexagonal layout. Fig. 6 illustrates a diagram 600 having an abscissa 601 along which a number of corner markers is plotted. Along an ordinate 602, the quality factor Q is plotted. A curved 603 shows the correlation between the number of corner markers and the quality factor. The plot 603 shows that going from 4 to 8 corner markers can improve the quality- factor by a factor of at least 2 (Q decreasing from 0.02 to 0.01). An improvement of a factor 2 in Q implies that it is possible to tolerate a factor 2 of more positional inaccuracy in the printing process. A straightforward line fitting procedure by just taking two corner markers along each edge would yield a result that is not better than the full-fledged ACF-procedure with N_CM = 4.

Finally, for the corner marker pattern of Fig. 5, the ACF with its 2x28 peaks is shown in an illustration 700 of Fig. 7 (only 28 peaks are really independent because of the point-inversion symmetry of the ACF).

In the following, referring to Fig. 8, a dispenser device 800 according to an exemplary embodiment of the invention will be explained.

The dispenser device 800 is capable of manufacturing a sensor carrier 110 according to an exemplary embodiment of the invention. The dispenser device 800 generates the spots 113, 114 to generate a pattern of detection spots 114 and reference spots 113 on the surface of the membrane 111. For this purpose, the dispenser device 800 comprises a tip 801 having an internal cavity through which spot material from containers 802 may be emitted on specific positions of the surface of the substrate 111. For instance, one of the containers 802 may comprise a fluorescence material which is used by a CPU or other control unit 810 to generate reference spots 113. A two-dimensional motion mechanism 815 allows to perform a relative motion in a two-dimensional manner between the substrate 110 and the tip 801. By taking this measure, the surface of the substrate 111 may be scanned by the tip 801 to deposit suitable material for forming the respective reference spots 113 or detection spots 114 on the surface of the substrate 111. In the CPU 810, a predetermined pattern is stored at which positions of the two-dimensional array reference spots 113 shall be spotted, and at which positions detection spots 114 shall be spotted. Fig. 9 illustrates the sensor carrier 110 in more detail.

Particularly, specific different types of spots are indicated in Fig. 9. PCR control spots are denoted with reference numeral 901 and are adapted for controlling a polymerase chain reaction which may be initiated in a separate reaction chamber prior to having the sample fluid streaming over the sensor surface of the sensor carrier 110. Corner marker spots 901 may be used for identifying a gridding scheme, i.e. a scheme according to which the spots of the sensor carrier 110 are arranged. Furthermore, an intensity calibration portion 904 located in a centre of the array of the spots has the purpose for controlling or monitoring the printing procedure by which the capture molecules from which the spots of Fig. 9 are formed are printed on the substrate 111.

Moreover, background spots 902 and detection spots 114 are shown as well.

In the following, some gridding procedures according to exemplary embodiments of the invention will be described in more detail. Fig. 10 shows an array 1000 corresponding to a conventional algorithm for gridding.

As can be taken from Fig. 10, the gridding result is erroneous so that no proper detection is possible with such an approach.

In contrast to this, Fig. 11 shows an example of a two-step gridding of a full array using an algorithm as explained above.

Therefore, as can be taken from Fig. 10 and Fig. 11, it can be observed that gridding with a hexagonal layout with 392 spots may be erroneous applying conventional approaches. A reason why is mostly because other spots than corner marker spots are wrongly identified as corner markers. This is typical for a single spot finding algorithm. In contrast to this, by a joint multiple spot finding algorithm (plus extra refinement), that is to say by a two-step gridding procedure, such problems can be overcome.

In the context of an autocorrelation function based algorithm, this may work particularly fine for a sparse array with relatively few spots (as is the case in the situation prior to the sensor event, like the hybridization reaction). However, problems may occur with full arrays, when too many spots are on (as may occur after the sensor event has taken place).

According to an exemplary embodiment of the invention, a two-step algorithm for gridding may be performed. In a first step, it may be possible to find a rough grid by a brute force looping over zoom, rotation and shifts (in x-direction and in y-direction). In a second step, such a grid may then be refined in a procedure much alike the fine-gridding procedure of the steps 4-5 discussed above.

In this context, it is possible to refine previously found positions by the centre of masses. Outliers may be identified. All difference vectors may be computed. A three-stage procedure may be applied including determining a rotation angle Θ, determining lattice parameters a and b, and determining shifts D_x and D_y.

Fig. 12 illustrates different images regarding the first step of the gridding procedure, also known as coarse gridding, using a brute-force searching solution. Coarse gridding is recognizing the pattern of the set of corner-marker spots from the image, from which the positions of the individual corner-marker spots can then subsequently be derived. Such a procedure using a joint-spot finding algorithm avoids that some bright neighboring spots are mistakenly considered as a corner-marker spot.

An image 1200 shows zoom looping only. An image 1210 shows rotation looping only. An image 1220 shows shift (x) looping only. An image 1230 shows shift (y) looping only. In all images 1200, 1210, 1220, 1230 extreme cases and a final case are shown.

In an illustration 1300 of Fig. 13, an example of an 8 point set of corner markers is given, illustrating an evaluation criterion. In accordance with the formula shown in Fig. 13, a search for a setting may be performed that maximizes C. This may involve for instance 9 million trial settings. In the following, a fine-gridding procedure according to an exemplary embodiment of the invention will be explained.

In a first stage of this gridding procedure, relative grid positions are determined as they result from the brute-force coarse gridding of above

First, it will be described how to derive the gridding information from a set of difference vectors as obtained from the corner marker spots (only). Using the above described nomenclature, in order to combine all the difference vectors, the least- squares sum of position information are considered to be given by:

S¹ ^ ¹ ! i?^ ^ ^■■■■ i rn i a i'i ^■>^■:^■■ θ - n/b r-mθ ι\²

I

\ V^ \ Rι ,_t ^■■■■ [ mta >mθ \ m^'b i-a>θ \}^d

¹ (20)

Least-squares fitting, first for Θ, starts from:

ά^(l " (21) which yields (in the approximation of small misplacement errors during printing):

,?. _:f

(22)

Using the (approximate) relation that a = V3 b , the above equation for Θ

becomes decoupled from the least-squares equations for a and b , that is:

from which Θ can be determined since it must be a small angle (resolving the uniqueness problem of the tan- function with respect to addition of π). Thus, with the above equation, also cos Θ and sin Θ are uniquely determined. A positive angle Θ implies a counter-clockwise mis-orientation of the (rectangular) grid with respect to a horizontal axis in the CCD image.

Next, the set of equations for a and b will be derived from: d^²

P

(24) and as- I i

^⁶ (25)

The solutions to the above equations (for this most general case of a grid with slight mis-orientation over an angle Θ) are given by:

Next, the second stage of the fme-gridding procedure will be explained, namely the determination of an absolute position of the grid.

Determination of the absolute position of the grid is realized via a second least-squares procedure based on the set of position vectors of the corner markers. Denoting the position vectors of the corner markers as measured from the CCD image (in units of CCD pixels) by rk with the two Cartesian components given by rk,_x and rk,_y, with k = 1, ..., N_CM, denoting the Cartesian coordinates of the (unknown) absolute position of the grid by (D_x, D_y), and further knowing that each k-th corner marker has integer coordinates on the grid given by (pk, qk), the absolute position can be determined by a solution of the least-squares problem with functional given by:

,S-> V, ?7_, r — { fq α i o.^ θ ~~ (p b >m β \ D_:r > h

Z_^*' ^rs,v ^■■■■ { pi (i ^U⁾ θ \ fβ b < o>-. θ \ D_υ i h (27)

The solutions of the equations obtained by setting the derivatives of S² with respect to D_x and D_y equal to zero, are given by (using also the previously obtained

lattice parameters α , b and the mis-orientation Θ):

(28)

Next, consecutive steps in a practical procedure for fme-gridding will be outlined.

In a first step of fϊne-gridding, position vectors of the corner marker spots Tk will are measured as they result from the coarse gridding operation.

For each of the corner marker spots, the Cartesian coordinates of the center of the spot on the 2D grid of the CCD-image needs to be determined, preferably with sub-pixel accuracy. This yields a set of 2D-coordinates (r_k,_x, r_k,_y) for k = 1, ..., N_CM with NcM the number of corner marker spots. For each spot, its ideal integer coordinates on the rectangular lattice (as derived from the hexagonal lattice) are known and are given by (pk, qk) (with p and q integer numbers).

In a second step of fϊne-gridding, the difference vectors Rk are calculated. In this step, all the difference vectors are computed, given by:

'^'

(29)

At the same time, the ideal difference vectors on the rectangular lattice are also computed, yielding:

(30)

In a third step of fine-gridding, mis-orientation angle Θ is derived. With, as input, the set of 2D coordinates of the difference vectors (both the measured ones and the ideal ones on the rectangular lattice) as obtained in the second step, the angle Θ can be computed from equation (23). In a forth step of fine-gridding, lattice parameters a and b are determined.

Using the above determined value of Θ together with the set of 2D coordinates of the difference vectors (both the measured ones and the ideal ones on the

rectangular lattice), the lattice parameters a and b can be determined using equation (26).

In a fifth step of fine-gridding, the overall shift-vectors of the Grid (D_x, Dy) are determined.

Using all previously computed information, the overall shift vector can be derived from equation (29). This completely specifies the expected position of all spots on the array.

It should be noted that the term "comprising" does not exclude other elements or features and the "a" or "an" does not exclude a plurality. Also elements described in association with different embodiments may be combined. It should also be noted that reference signs in the claims shall not be construed as limiting the scope of the claims.

Claims

CLAIMS:

1. An apparatus (195) for retrieving a grid pattern of a plurality of spots (113, 114) on a substrate (111) of a sensor carrier (110), wherein the plurality of spots (113, 114) comprise a plurality of reference spots (113) arranged for identifying the grid pattern based on an image of at least the plurality of reference spots (113, 114), the app aratus (195) comprising a first determination unit (163) adapted for performing a coarse gridding procedure for determining coarse information regarding the grid pattern based on an analysis of the image of the plurality of reference spots (113); a second determination unit (165) adapted for performing a fine gridding procedure for determining the grid pattern based on the determined coarse information regarding the grid pattern.

2. The apparatus (195) of claim 1, wherein the first determination unit (163) is adapted for determining, and particularly for providing at its output, a plurality of difference vectors between each pair of reference spots (113) out of the plurality of reference spots (113) based on the image captured of at least the plurality of reference spots (113), said plurality of difference vectors being provided as an input for the second determination unit (165).

3. The apparatus (195) of claim 2, wherein the first determination unit (163) is adapted for determining, and particularly for providing at its output, the plurality of difference vectors based on an autocorrelation analysis of the image captured of the plurality of reference spots (113), said plurality of difference vectors being provided as an input for the second determination unit (165).

4. The apparatus (195) of claim 2, wherein the first determination unit (163) is adapted for determining, and particularly for providing at its output, the plurality of difference vectors exclusively based on position information of the plurality of reference spots (113) on the substrate (111) of the sensor carrier (110).

5. The apparatus ( 195) of claim 4, wherein the first determination unit (163) is further adapted to retrieve the position information of the plurality of reference spots (113) on the substrate (111) of the sensor carrier (110) by means of a joint-spot finding approach using a parameter- looping method, where four relevant parameters, being a zoom or magnification of the image, and a rotation of the image, and a two-dimensional translation of the image relative to a pattern of the reference spots (113), are extensively varied over their relevant parameter ranges, wherein a set of four parameters that matches with, particularly which optimizes, a selection criterion is defined to be representative as output of said first determination unit (163).

6. The apparatus (195) of claim 2, wherein the second determination unit (165) is adapted for determining the grid pattern of the arrangement of the plurality of spots (113, 114) on the sensor carrier (110) based on the plurality of difference vectors as generated by the first determination unit (163), using a least squares fit of parameters being a rotation of the arrangement of spots (113, 114) relative to the captured image as well as lattice parameters of said grid pattern.

7. The apparatus (195) of claim 6, wherein the least squares fit uses also absolute locations of the plurality of reference spots (113) in the captured image to be used in a further least squares fit of parameters being a two-dimensional translational shift of the arrangement of spots (113, 114) relative to the captured image.

8. The apparatus (195) of claim 1, wherein the second determination unit (165) is further adapted for determining the grid pattern by performing at least one of the group consisting of eliminating outliers among the reference spots (113), and refining a previous set of locations of a plurality of reference spots (113) by their centers-of-mass.

9. The apparatus (195) of claim 1, comprising a quality estimation unit adapted for estimating a quality of the determined grid pattern.

10. The apparatus ( 195) of claim 1 , comprising an imaging device (150) adapted for capturing the image of the plurality of reference spots (113).

11. The apparatus ( 195) of claim 1 , comprising an electromagnetic radiation source (180) adapted for illuminating the plurality of reference spots (113).

12. A method of retrieving a grid pattern of a plurality of spots (113, 114) on a substrate (111) of a sensor carrier (110), wherein the plurality of spots (113, 114) comprise a plurality of reference spots (113) arranged for identifying the grid pattern based on an image of at least the plurality of reference spots (113, 114), the method comprising performing a coarse gridding procedure for determining coarse information regarding the grid pattern based on an analysis of the image of the plurality of reference spots (113); performing a fine gridding procedure for determining the grid pattern based on the determined coarse information regarding the grid pattern.

13. A sensor carrier (110), comprising a substrate (111); a plurality of spots (113, 114) formed on the substrate (111) in accordance with a grid pattern; wherein the plurality of spots (113, 114) comprise a plurality of reference spots (113) arranged for enabling an identification of the grid pattern in accordance with a method of claim 11.

14. The sensor carrier ( 110) of claim 13 , wherein the plurality of reference spots (113, 114) are arranged on the substrate (111) in such a manner that each difference vector between each pair of the plurality of reference spots (113) is unique.

15. The sensor carrier ( 110) of claim 13 , wherein each of the plurality of reference spots (113) is arranged on an outer circumference of an array of the plurality of spots (113, 114).

16. The sensor carrier ( 110) of claim 13 , wherein the plurality of spots (113, 114) comprises a plurality of detection spots (114) adapted for detecting the presence of molecules to be detected.

17. The sensor carrier ( 110) of claim 13 , wherein the plurality of reference spots (113) comprise a fluorescence material.

18. The sensor carrier ( 110) of claim 13 , wherein the plurality of reference spots (113) comprise at least one reference spot (113) located at an edge of the grid pattern, particularly at a corner of the grid pattern.

19. The sensor carrier ( 110) of claim 13 , wherein the grid pattern is a two-dimensional grid pattern, particularly a hexagonal grid pattern with close packing.

20. The sensor carrier (110) of claim 13, wherein at least a part of the plurality of spots (113, 114) comprises capture molecules adapted for hybridizing with complementary molecules to be detected.

21. The sensor carrier ( 110) of claim 13 , adapted as at least one of the group consisting of a biosensor carrier and a molecular diagnostics sensor carrier.

22. A method of manufacturing a sensor carrier (110), the method comprising forming a plurality of spots (113, 114) on a substrate (111) in accordance with a grid pattern; arranging a plurality of reference spots (113) of the plurality of spots

(113, 114) on the substrate (111) for enabling an identification of the grid pattern in accordance with a method of claim 11.