Method and device for image processing for an analysis device and analysis device, in particular for the detection of pathogens via nucleic acid amplification

Abstract

The invention relates to a method for image evaluation for an analysis device (100), in particular for an analysis device (100) for detecting pathogens via nucleic acid amplification. Image data representing image points of a detected object and predetermined reference data representing reference points of a reference object are read in. The reference points are transformed using a transformation parameter to obtain transformed reference points. The image points are assigned to the transformed reference points, and deviations between positions of the image points and positions of the transformed reference points are determined to obtain a position error representing the deviations. The transformation parameter is adjusted, and the steps of transforming, assigning, and determining are repeated if the position error is greater than a predetermined convergence threshold.Otherwise, a position signal indicating a position of the detected object is provided using the transformation parameter.

Description

Prior ArtThe invention is based on a method and apparatus for image processing for an analysis device and on an analysis device according to the preamble of the independent claims. The present invention also relates to a computer program.Modeling recognition algorithms is an important aspect in image processing. Particularly with the success of convolutional neural networks (CNNs) and recently also of transformer-based networks, the performance of object detectors has drastically increased. However, these networks are susceptible to undesirable and non-domain examples.Disclosure of the InventionAgainst this background, the approach presented here presents an improved method for image processing for an analysis device, and also an improved device which uses this method, an improved analysis device and finally a corresponding computer program according to the main claims. The measures listed in the dependent claims allow advantageous refinements and improvements of the device specified in the independent claim.The approach presented here advantageously makes it possible to improve a detection accuracy in connection with image processing and thus to reduce a number of erroneous detections. The approach can be used, for example, for analysis instruments, but also in other fields, such as, for example, in the automobile industry.A method for image evaluation for an analysis device is presented, wherein the method comprises a step of reading in image data representing image points of a detected object and predetermined reference data representing reference points of a reference object, wherein the reference points represent a derived geometry of the reference object. The method further comprises a step of transforming the reference points using at least one transformation parameter to obtain transformed reference points, and a step of associating the pixels with the transformed reference points. Furthermore, the method comprises a step of determining deviations between positions of the pixels and positions of the transformed reference points in order to obtain a position error representing the deviations, a step of adjusting the transformation parameter and repeating the steps of transforming, assigning and determining if the position error is greater than a predefined convergence threshold value, and a step of providing a position signal indicating a position of the detected object using the transformation parameter if the position error is less than the predefined convergence threshold value.The analytical device can be used in the field of microfluidics, for example, and can be designed to analyze samples contained in a cartridge, in particular for detecting pathogens via nucleic acid amplification. Advantageously, image data can be evaluated in the method, which can relate to the cartridge, for example. More specifically, the image data may represent a projection of the object from a three-dimensional space onto a two-dimensional plane. The detected object can have characteristic features which can be recognized as image points in the image data. The reference data can also have characteristic features as the reference points. The transformation parameter can advantageously be used to change the reference data, for example by translation, scaling and additionally or alternatively by rotation, and thus to approximate the image data. If the position error indicates, for example, an excessively large deviation in terms of amount between the image data and the reference data, the reference data are still approximated to the image data using the original or an adapted transformation parameter until the detected object can advantageously be unambiguously identified.The proposed method improves the recognition performance for objects with known geometric structures. In contrast to approaches that usually detect the object itself, a second post-processing step is used to determine the position of an object on the basis of the position of its components or the surrounding objects. This offers a high detection accuracy. Namely, since the position of the object can be determined by a plurality of instances or surrounding instances, there is less dependence on the recognition accuracy of a single instance. In addition, the approach is very robust. Even if the object itself cannot be recognized, e.g., by noise or occlusion, the position may be determined based on its environment. This reduces the number of false negative detections. This also means a high detection accuracy.For example, the method can comprise a step of deriving the reference points from a geometry of the reference object. A derivation of the reference points from the exact geometry of the reference object can significantly contribute to a high accuracy. Advantageously, the known geometry of the reference object can draw conclusions exactly about a geometry of the detected object.According to one embodiment, the steps of adjusting and repeating may be performed when the position error is greater than a predetermined convergence threshold and further an iteration of the steps of adjusting and repeating is less than a predetermined minimum number. By adapting and repeating, an analysis accuracy can be advantageously improved.In the step of providing, the position signal can be provided if the position error is less than the predefined convergence threshold value and, furthermore, an iteration of the steps of adapting and repeating falls below a predefined minimum number. The position error can be smaller in magnitude than the convergence threshold value. Since all operations at this time are differentiable after t, a gradient descent can advantageously be carried out. This advantageously minimizes an expected result, so that an approximation can take place.The method may include a step of storing the position error. The storing step may be repeated so that a position change may be understood. For example, all results can be stored temporarily.The transformation parameter can be configured to effect a rotation, translation and additionally or alternatively a scaling of the reference points. Advantageously, the transformation parameter can represent one or more rotation angles, translation parameters and additionally or alternatively a scaling parameter for the reference data.Furthermore, the method may comprise a step of detecting the pixels in the image data as key points in response to the reading-in. This advantageously allows the object to be detected.According to one embodiment, in the step of assigning, the pixels can be assigned to the transformed reference points using the Hungarian algorithm or other optimization algorithms. In this way, known algorithms can be used.In a step of capturing, the image data can be captured as digital images, video images, radar images, lidar images, ultrasonic images, motion images and additionally or alternatively as thermal images using a sensor device. Advantageously, the image data can thus be classified, a semantic segmentation of the image data carried out and additionally or alternatively identify the detected object.This method can be implemented, for example, in software or hardware or in a mixed form of software and hardware, for example in a control device.The approach presented here furthermore creates a device which is designed to carry out, actuate or implement the steps of a variant of a method presented here in corresponding devices. This embodiment variant of the attachment in the form of a device also enables the object on which the attachment is based to be achieved quickly and efficiently.For this purpose, the device can have at least one arithmetic unit for processing signals or data, at least one memory unit for storing signals or data, at least one interface to a sensor or an actuator for reading in sensor signals from the sensor or for outputting data or control signals to the actuator, and / or at least one communication interface for reading in or outputting data which are embedded in a communication protocol. The computing unit can be, for example, a signal processor, a microcontroller or the like, wherein the memory unit can be a flash memory or a magnetic memory unit. The communication interface can be designed to read in or output data wirelessly and / or in a wired manner, wherein a communication interface that can read in or output wired data can read in this data, for example electrically or optically, from a corresponding data transmission line or output it into a corresponding data transmission line.In the present case, a device can be understood to mean an electrical device which processes sensor signals and outputs control and / or data signals as a function thereof. The device can have an interface which can be designed as hardware and / or software. In a hardware configuration, the interfaces can be part of a so-called system ASIC, for example, which contains various functions of the device. However, it is also possible for the interfaces to be dedicated, integrated circuits or to consist at least partially of discrete components. In the case of a software configuration, the interfaces can be software modules which are present, for example, on a microcontroller in addition to other software modules.Furthermore, an analysis device for analyzing a sample contained in a cartridge is presented, wherein the analysis device has a sensor device for capturing image data, wherein the image data depict a section of the cartridge, as well as a device in a previously mentioned variant for evaluating the image data.The analytical instrument can be used, for example, in the medical field and have, for example, at least one optical sensor as the sensor device. The cartridge can be designed, for example, as a microfluidic cartridge.A computer program product or computer program with program code which can be stored on a machine-readable carrier or storage medium such as a semiconductor memory, a hard disk memory or an optical memory and is used for carrying out, implementing and / or controlling the steps of the method according to one of the embodiments described above is also advantageous, in particular if the program product or program is executed on a computer or a device.Exemplary embodiments of the approach presented here are illustrated in the drawings and explained in more detail in the following description. It shows: FIG. 1 shows a schematic illustration of an exemplary embodiment of an analytical device; FIG. 2 shows a flow diagram of an exemplary embodiment of a method for image evaluation for an analysis device; FIG. 3 shows a flow diagram of an exemplary embodiment of a method according to an exemplary embodiment; FIG. 4 shows a block diagram of an apparatus according to an exemplary embodiment; FIG. 5 shows a schematic example illustration of captured image data according to an exemplary embodiment; FIG. 6 shows a schematic example illustration of captured image data according to an exemplary embodiment; FIG. 7 shows a schematic example representation of reference data according to an embodiment; and FIG. 8 shows a schematic example illustration for captured image data and reference data associated therewith according to an exemplary embodiment.In the following description of advantageous exemplary embodiments of the present invention, the same or similar reference numerals are used for the elements shown in the different figures and acting in a similar manner, wherein a repeated description of these elements is omitted.FIG. 1 shows a schematic illustration of an exemplary embodiment of an analysis device 100. The analysis device 100 in this exemplary embodiment is designed to analyze input samples, whereby PCR tests can be carried out, for example. For this purpose, a microfluidic cartridge 105 with, for example, a plastic housing and a microfluidic network for processing the sample can be inserted into a receiving region 110. In this exemplary embodiment, the analysis device 100 furthermore comprises an optional display 115 with a touch function. This allows settings to be manually input to the desired analysis process. In addition, the display 115 is configured merely by way of example in order to display analysis results.The analysis device 100 has a sensor device 120 for capturing image data, wherein the image data represents a section of the cartridge 105, and a device 125 for forming the image data. The device 125 is embodied, for example, as a control unit which is designed to actuate and / or carry out a method for image evaluation, as is described or at least mentioned, for example, in at least one of the figures described below.In other words, the analytical device 100 according to this exemplary embodiment is realized, which contains the components necessary for molecular biological analysis of a sample in a microfluidic cartridge 105, for example a DxC3 cartridge.There are several approaches that formulate the object of object recognition as keypoint recognition. Instead of directly generating a so-called bounding box for an object, key point methods output the semantically important key points of the components of the object. This includes, for example, face recognition by key points such as eyes, ears, nose, etc., or estimation of posture based on joint positions. Other works simply formulate object recognition as predicting a single keypoint of the object.Against this background, a method for recognizing objects by their instances (sub-objects) in images is proposed. In essence, this makes use of prior knowledge about the object, e.g. the geometric relationship of sub-objects, in order to increase the recognition accuracy. Such an object is, for example, a series of screws on a part or reagent chambers on medical devices. The object here is, for example, to identify reagent chambers on cartridges on the basis of the radiation intensity of the reagent located in at least some reagent chambers, as is illustrated below in FIGS. 5 to 8.Alternatively, it is used for detecting any desired objects whose arrangement among one another is known. These are, for example, screws on a mechanical part or electrical components such as resistors on a printed circuit board or any other task in which the geometry is known and a high detection accuracy is required.The approach described herein is used for analysis of data obtained from a sensor. The sensor acquires measurement values of the environment in the form of sensor signals, which can be provided, for example, by digital images, for example video, radar, LiDAR, ultrasonic, motion and thermal images.In other words, the sensor data is classified, the presence of objects in the sensor data is recognized, or semantic segmentation of the sensor data is performed, for example with respect to traffic signs, road surfaces, pedestrians and / or vehicles.The described approach is used here for determining one or more continuous values, that is to say for carrying out a regression analysis in the data, such as for example with regard to tracking an object.FIG. 2 shows a flow diagram of an exemplary embodiment of a method 200 for image evaluation for an analytical device, as was described, for example, in FIG. 1. The method 200 can therefore be carried out supportingly for a sample analysis of a sample contained in a cartridge and can be controlled, for example, by a device as likewise described in FIG. 1.The method 200 for image evaluation comprises a step 202 of reading in image data representing image points of a detected object, that is to say characteristic features in an image. More specifically, for this purpose, in an optional step 203 of capturing, the image data were captured as digital images, video images, radar images, lidar images, ultrasonic images, motion images and / or thermal images using a sensor device as described, for example, in FIG. 1, such that the method 200 classifies the image data, a semantic segmentation of the image data can be carried out and / or in order to be able to identify the captured object. For this purpose, for example, at least one piece of environmental information is captured and the object captured in a three-dimensional space is projected onto a two-dimensional image plane. In step 202 of reading in, predetermined reference data are also read in, the reference points of a reference object. The reference points are likewise to be understood as characteristic features of the reference object and are transformed in a step 204 of transforming using at least one transformation parameter in order to obtain transformed reference points. The transformation parameter is designed, for example, to effect a rotation, translation and / or scaling of the reference points. This means that the transformation parameter represents one or more rotation angles θx, θy, θz, the translation parameters x and y and / or a scaling parameter s for the reference data or for the at least one reference object. The aim of this is to adapt the reference data as much as possible to the image data in order to be able to identify the detected object. Furthermore, the method 200 comprises a step 206 of assigning the pixels to the transformed reference points, and a step 208 of determining deviations between positions of the pixels and positions of the transformed reference points in order to obtain a position error representing the deviations. For example, the pixels are assigned to the transformed reference points in the step 206 of assigning using the Hungarian algorithm or other optimization algorithms. If the position error is greater than a predefined convergence threshold value in terms of amount, the transformation parameters are adjusted in a step 210 and the steps 204, 206, 208 of transforming, assigning and determining are repeated. If, on the other hand, the position error is smaller in magnitude than the predefined convergence threshold value, a step 212 of providing a position signal indicating a position of the detected object is carried out using the original and / or the adjusted transformation parameter, which can also be referred to as a parametric description, for example. The position error is stored in a storage step 214 only optionally, for example after each repetition of the transformation, assignment and determination steps 204, 206, 208 in order to track a position change and thus the approach. For example, the position errors are temporarily stored on a memory unit, which is designed as a ring memory, for example.The method 200 optionally comprises a step 216 of detecting the pixels in the image data as key points in response to the reading-in 202 in order, for example, to carry out object recognition. Furthermore, the method 200 optionally comprises a step 218 of deriving the reference points from a geometry of the reference object before the step 202 of reading in.More specifically, according to one embodiment, the steps of adjusting and repeating are performed when the position error is greater than the predetermined convergence threshold and further an iteration of the steps 210 of adjusting and repeating is less than a predetermined minimum number. Furthermore, in step 212 of providing, the position signal is provided if the position error is less than the predefined convergence threshold value and, furthermore, the iteration of steps 210 of adapting and repeating falls below the predefined minimum number. Since all operations at this point in time are differentiable after t, for example, a gradient descent to t at the rate γ can be carried out. This minimizes the expected E and brings the transformed points closer to the detections on the image plane and thus causes an approximation.The method was developed within the scope of developing an algorithm for the detection of reagent chambers on a cartridge. This is necessary to evaluate the in particular microfluidic experiments within the chambers, since a known arrangement of the chambers is used to improve the detection accuracy.In other words, the approach presented here defines an object as a series of characteristic key points in the recognition phase of our method in the basis for the method, so that an exact geometry of the object is used as a priori for the post-processing.FIG. 3 shows a flow diagram of an exemplary embodiment of a method 300 according to an exemplary embodiment, which at least resembles the method described in FIG. 2. More specifically, a decision chain for the method described in FIG. 2 is described in FIG. 3.In a block 302, the decision chain is started, for example by acquiring the image data D. The image evaluation is then initialized by reading in the image data and the reference data in a block 304, and pixels of the image data and reference points of the reference data A are detected. Furthermore, in block 304, a reference data is determined. In a subsequent block 306, the reference data is changed using the transformation parameter t. This is done by the following function:In block 308, the image data D is assigned to the reference data A, so that the following assignment applies:At block 310, the position errors are then calculated using the following function: Thereafter, at block 312, a check is made to see if E has converged, so that (E<T) holds. If this is the case, the method ends with a block 314. If this is not the case, individual steps are repeated with a transformation parameter t updated in a block 316. In order to obtain the updated transformation parameter t, the following equation is applied to the transformation parameter:The heart of the described approach is the post-processing of the detections according to this embodiment. The computation model for the recognition itself may be any algorithm. The approach described herein is based on the assumption that the image of an object is / is a projection of the object from a 3D space onto the 2D image plane. In other words, if we know the geometry of an object G, the object can transform by a function with known transformations t to obtain the observed geometry of the object O as follows:The derivation of the geometry of O is not trivial, however, since every point from the image space of object I O would have to be assigned to the 2D projection of G.Therefore, the map g is estimated by restricting the range of g. In practice, for example, n repeating structures S O in the object itself would be selected if, for example, a single detector is used. As the number of n increases, the approximation g̅ g would approximate.If G is limited to a 2D space, the function must model the three degrees of freedom (rotation about the z-axis, translation in the x-y plane). For 3D spaces or higher spaces, this would scale accordingly. In addition to depth of field (DOF), a scaling parameter is proposed to model the depth of the image. The resulting transformation parameters t are the free parameters of the function g.Since we are interested in g and g̅, respectively, we use S O and the corresponding structures S G of G. The error E of g can be measured by the L1 norm between S O and S G :However, since this depends on the order of the elements in S O and S G the order of the elements in the two sets is first determined. Therefore, the L1 norm is used as a cost function to find the optimal matching between points from S O and S G.If G ∈ R is 2 the minima of the error function with respect to t are numerically sought. Otherwise, a gradient descent is performed in m iterations to approximate an optimal solution with:Although this problem has been formulated as the detection of O in an image, the position of all sub-objects is determinable by t. For example, if we are interested in an object G i with a known relationship to its environment G S we can construct G from G i and G S. After solving for t in G, we can determine the position of O i by t.The processing steps can be divided in summary essentially into two sequences:First, a detector is used to locate characteristic structures as key points within the image. These can be either the objects of interest themselves, such as, for example, the reagent chambers on a cartridge, or partial regions of the objects which are to be detected, for example, resistances in the determination of the position of a printed circuit board. The detector can be any desired computer model, but also, for example, a neural network and / or non-neural methods, e.g., using hand-produced features or statistical classifiers.In the second stage, detected positions are used. In this case, each real object in an image becomes a transformation by rotation, translation and scaling of an ideal object. The real object is defined by its characteristic points D and the ideal object by its characteristic points A, which corresponds to the reference points. These points are the positions of part objects such as the screws on a part or the reagent chambers on a cartridge. It is possible, merely optionally, to formulate the recognition of sub-objects as a target.The post-processing begins by initializing D with the detected points and A with positions defined by a 2D projection of the ideal object. In addition, the translation parameters t are defined, which can comprise the rotation angles θ x, θ y, θ z, the translation parameters x and y, and a scaling parameter s. These parameters are used to transform A to A T.Next, the injective function is defined that associates each point in D with a point in A. For creating this assignment, the Hungarian algorithm is used, for example, but other combinatorial optimization algorithms can also be used here. The position error between the detections and the corresponding transformed positions is calculated below. The error function fE used depends on the respective application. For example, the L1 spacing is used.If the error is below a convergence threshold or a minimum number of iterations, we end the post-processing and use the convergence threshold as a parametric description of the position of the detected object, as we can calculate the position of the sub-objects or the associated object itself.If the error is above the threshold and below the minimum number of iterations, processing continues. Since all operations at this time are differentiable after t, we can perform a gradient descent to t at the rate y. This minimizes the expected E and brings the transformed points closer to the detections on the image plane.In summary, according to this exemplary embodiment, an image evaluation is shown. Here, the transformation parameters t are changed in such a way as to minimize a distance between the positions of the pixels and the positions of the reference points.FIG. 4 shows a block diagram of an apparatus 125 according to an exemplary embodiment as was described, for example, in FIG. 1. The device 125 is designed to control and / or carry out a method for image evaluation for an analysis device, as described, for example, in at least one of FIGS. 2 to 3.For this purpose, the device 125 has a reading-in unit 400 for reading in image data 401; D, which represent image points of a detected object, and for reading in predetermined reference data 402; A, which represent reference points of a reference object, wherein the reference points represent a derived geometry of the reference object. Furthermore, the apparatus 400 has a transformation unit 403, an assignment unit 404, a determination unit 406, an adaptation unit 408 and a provision unit 410. Optionally, the apparatus 125 has a derivation unit 411, which is designed to derive the reference points from the geometry of the reference object.The transformation unit 403 is configured to transform one of the reference points using at least one transformation parameter 412 in order to obtain transformed reference points. The assignment unit 404 is designed to assign the pixels to the transformed reference points. The determination unit 406 is configured to determine deviations between positions of the pixels and positions of the transformed reference points in order to obtain a position error representing the deviations. Furthermore, the adaptation unit 408 is designed to adapt the transformation parameter 412 and to cause a repetition of the steps of transforming, associating and determining if the position error is greater in absolute value than a predefined convergence threshold value. The provision unit 410 is configured to provide a position signal 414 indicating a position of the detected object using the original or the adjusted transformation parameter 412 if the position error is smaller in absolute value than the predefined convergence threshold value.The following FIGS. 5 to 8 each represent exemplary snapshot during a method for image evaluation, as described, for example, in at least one of FIGS. 2 to 3.FIG. 5 shows a schematic example illustration of captured image data 401; D according to an exemplary embodiment. The image data 401; D are acquired, for example, by a sensor device, as described, for example, in FIG. 1 as part of an analytical device. Alternatively, the image data can be captured by image sensors which are used, for example, in the automobile industry. The image data 401; D here represent an image or a projection of a three-dimensional object onto a two-dimensional image plane, which is also described as a captured object 502. The detected object 502 has a plurality of image points 504, which can be characteristic features of the object 502, for example. The object 502 is recognizable by the pixels 504. Within the scope of a method for image evaluation, as described, for example, in at least one of FIGS. 2 to 3, these image points 504 are matched to reference points of reference data and a detection accuracy for the detected object 502 is thereby improved. This means that, for example, data which are used are stored in a memory. In this case, the detected object 502 is already detected, for example, so that the correct reference data are selected for the comparison.FIG. 6 shows a schematic example illustration of captured image data 401; D according to an exemplary embodiment, which correspond to the image data described in FIG. 5, for example. In this case, the positions 600 of the pixels 504 of the detected object 502 are shown visualized, which are used for the further course of the method.FIG. 7 shows a schematic example illustration of reference data 402; A according to an embodiment, which comprise a plurality of reference points 702. The positions of the reference points 702 are also defined as key points. These reference points 702 are used in a method for image evaluation, as described, for example, in at least one of FIGS. 2 and / or 4, are matched to the image data and, if appropriate, adapted by rotation, translation and / or scaling. For this purpose, the reference data 402; A and the image data are superimposed, for example. More specifically, positions 704 of the reference points 702 are compared with the positions of the pixels.FIG. 8 shows a schematic example illustration for captured image data and reference data associated therewith according to an exemplary embodiment. More specifically, according to this exemplary embodiment, the reference data 700 is superimposed on the image data 401; D, so that position errors between the reference points 702 and the pixels 502 become visible. The reference data can be changed or adapted using a transformation parameter with respect to a rotation angle θ, a translation parameter x and y and / or a scaling parameter s.According to this embodiment, for the sake of simplicity, only a rotation about a z-axis and a translation in the x-y plane are visually displayed.

Claims

Method (200) for image evaluation for an analysis device (100), wherein the method (200) comprises the following steps: reading in (202) image data (401; D) representing image points (504) of a detected object (502) and reading in (202) predetermined reference data (402; A) representing reference points (702) of a reference object, wherein the reference points represent a derived geometry of the reference object; transforming (204) the reference points (702) using at least one transformation parameter (412) in order to obtain transformed reference points (702); assigning (206) the image points (504) to the transformed reference points (702); determining (208) deviations between positions (600) of the pixels (504) and positions (704) of the transformed reference points (702) in order to obtain a position error representing the deviations; adjusting (210) the transformation parameter (412) and repeating the steps (204, 206, 208) of transforming, assigning and determining if the position error is greater than a predefined convergence threshold value; and providing (212) a position signal (414) indicating a position of the detected object (502) using the transformation parameter (412) if the position error is less than the predefined convergence threshold value.Method (200) according to claim 1, comprising a step (218) of deriving the reference points from a geometry of the reference object.The method (200) of any preceding claim, wherein the steps (210) of adjusting and repeating are performed when the position error is greater than a predetermined convergence threshold and further an iteration of the steps (210) of adjusting and repeating is less than a predetermined minimum number.Method (200) according to one of the preceding claims, wherein in the step (212) of providing the position signal (414) is provided if the position error is less than the predetermined convergence threshold value and furthermore an iteration of the steps (210) of adapting and repeating falls below a predetermined minimum number.Method (200) according to one of the preceding claims, having a step (214) of storing the position error.Method (200) according to one of the preceding claims, wherein the transformation parameter (412) is configured to cause a rotation, translation and / or scaling of the reference points (702).Method (200) according to one of the preceding claims, having a step (216) of detecting the image points (504) in the image data (401; D) as key points in response to the reading-in (202).Method (200) according to one of the preceding claims, wherein in the step (206) of assigning the pixels (504) are assigned to the transformed reference points (702) using the Hungarian algorithm or other optimization algorithms.Method (200) according to one of the preceding claims, having a step (203) of capturing the image data (401; D) as digital images, video images, radar images, lidar images, ultrasonic images, motion images and / or thermal images using a sensor device (120).Device (125) which is configured to execute and / or actuate the steps (202, 203, 204, 206, 208, 210, 212, 214, 216, 218) of the method (200) according to one of the preceding claims in corresponding units (400, 403, 404, 406, 408, 410, 411).Analysis device (100) for analyzing a sample contained in a cartridge (105), in particular for detecting pathogens via nucleic acid amplification, wherein the analysis device (100) has the following features: a sensor device (120) for capturing image data (401; D), wherein the image data (401; D) map a section of the cartridge (105); and a device (125) according to claim 10 for evaluating the image data (401; D).Computer program which is configured to execute and / or control the steps (202, 203, 204, 206, 208, 210, 212, 214, 216, 218) of the method (200) according to one of Claims 1 to 9.A machine readable storage medium having stored thereon the computer program of claim 12.

Images