Computer-implemented method for identifying an object detected in multiple X-ray projection images

Abstract

The invention relates to a computer-implemented method for providing detection information (33) that at least describes whether at least one object (22) is identified in several X-ray projection images (20, 23), comprising: applying a function (41) to a received first X-ray projection image (20), at least one received second X-ray projection image (23), and received and / or determined relative arrangement information (28), wherein the function (41) is configured to identify a projection of at least one object (22) in a predetermined epipolar region (24) in the first X-ray projection image (20) and in the at least one second X-ray projection image (23), wherein the predetermined epipolar region (24) is arranged around an epipolar line (25) on which the projection of the at least one object (22) lies in the first X-ray projection image (20) and / or in the at least one second X-ray projection image (23);and, if the projection of the at least one object (22) is identified in the specified epipolar area (24) in the first X-ray projection image (20) and in at least one second X-ray projection image (23), providing corresponding detection information (33).

Description

[0001] The invention relates to a computer-implemented method for providing detection information. The detection information describes at least one object detected in several X-ray projection images. The invention further relates to a computer-implemented method for providing a learned function, a method for providing detection information, a data processing system for carrying out the computer-implemented method, and a computer program product for carrying out the computer-implemented method.

[0002] An X-ray imaging system can, for example, acquire individual two-dimensional X-ray projection images that depict at least part of a patient from different perspectives. For instance, two or more individual X-ray projection images can be acquired, with the relative position between the X-ray imaging system and the patient changing between acquisitions. Such individual X-ray projection images of the patient are often acquired using two-dimensional fluoroscopy. However, this method of acquiring X-ray projection images is less suitable for detecting small structures, such as a bone fragment chipped off after a fracture or a hemorrhage, compared to cross-sectional images acquired using computed tomography (CT).The reason for this is that such small structures are often difficult to detect in two-dimensional X-ray projection images. This can lead to incorrect or inappropriate treatment decisions, which may subsequently necessitate a costly follow-up CT scan.

[0003] However, a CT scan is more time-consuming than acquiring individual X-ray projection images and involves an additional radiation dose for the patient. It is therefore advisable to automatically determine whether several individual X-ray projection images provide indications that a three-dimensional follow-up examination might be beneficial, for example, using cone-beam computed tomography (CBCT), a C-arm X-ray imaging system for acquiring three-dimensional X-ray projection images, or CT. The initial search should therefore be for the small structure, i.e., the object, in several X-ray projection images.

[0004] The object of the invention is to provide a solution by means of which several X-ray projection images can be examined with regard to the presence of an object.

[0005] The problem is solved by the subject matter of the independent patent claims.

[0006] A first aspect of the invention relates to a computer-implemented method for providing detection information. The detection information describes, at least, whether at least one object is identified in multiple X-ray projection images. For example, it can describe whether the object is identified in the multiple X-ray projection images or not. Furthermore, the detection information can describe the at least one object identified and thus detected in multiple X-ray projection images. The detection information can, for example, describe that a specific object detected in a first X-ray projection image is also detected in at least a second X-ray projection image. In addition to the presence of the detected object, the detection information can also describe its position and / or orientation.The detection information might, for example, describe the fact that no object is identified in the multiple X-ray projection images. In this example, at least one object was not identified in the first X-ray projection image and / or in at least one second X-ray projection image. Providing the detection information can include signaling the presence or absence of the object. Furthermore, based on the provided detection information, the detected object can be displayed, for example, by marking, annotating, and / or highlighting it in the X-ray projection images.In a preferred example, the object is a small structure in a patient's body, such as a bone fragment, a hemorrhage, a foreign body, in particular an implant, a contrast agent, for example in a vessel, and / or some other type of object in which a person, for example a doctor, is interested and / or which is specified as an object of interest.

[0007] The computer-implemented procedure involves receiving a first X-ray projection image (one of several) and at least one second X-ray projection image (one of several). Numerous second X-ray projection images can be received. The received X-ray projection images depict at least a partially shared area of ​​the object being examined from different projection directions. Thus, the received X-ray projection images were acquired from different perspectives. Specifically, the X-ray projection images are two-dimensional projection images. The first X-ray projection image and the at least one second X-ray projection image describe or show, for example, a part of a patient's body from different perspectives, that is, from different projection directions.In this example, the body part is the area of ​​examination, and the patient is the object of examination. For this purpose, for instance, an X-ray imaging system capturing projection images may have been moved relative to the patient, and / or the patient may have been moved relative to a stationary X-ray imaging system, while the multiple projection images were being acquired as part of the computer-implemented procedure. The projection images can alternatively be referred to as X-ray images. Each received projection image may be a pre-processed projection image.

[0008] The computer-implemented procedure involves receiving and / or determining relative arrangement information. This information describes the relative arrangement of the various projection directions. It thus describes a relationship between the projection direction from which the first X-ray projection image was acquired and the projection direction from which at least one subsequent X-ray projection image was acquired. The relative arrangement information can be understood as information or data that describes the relative projection geometry of the different projection directions. For example, the relative arrangement information can be received when the X-ray imaging system detects movement of an X-ray projection image acquisition device relative to the patient.For example, a C-arm X-ray imaging system can, in addition to the individual X-ray projection images, specify the C-arm's position at which each image was acquired. Such C-arm positioning information, which can alternatively be referred to as angulation data, is received as relative arrangement information in one example within the computer-implemented procedure.

[0009] If the relative arrangement information is determined, a three-dimensional body part model, particularly a three-dimensional bone model, can be generated using machine learning methods, i.e., a trained machine learning model, based on the multiple received X-ray projection images. The relative projection geometry can then be derived from this three-dimensional body part model, thus determining the relative arrangement information. Alternatively or additionally, a three-dimensional model of the patient can be generated during the acquisition of each X-ray projection image using at least one camera that captures at least part of the patient. The relative arrangement information can then be estimated, and in particular determined, based on this three-dimensional model.Known methods for determining relative arrangement information can be used here.

[0010] It can therefore be assumed that when carrying out the computer-implemented procedure, at least two different X-ray projection images and information about the perspectives from which these X-ray projection images were captured relative to each other are available, so that, for example, calculations are possible to at least estimate, and in particular to find exactly, a given location in the first X-ray projection image, and especially in at least one second X-ray projection image.

[0011] The computer-implemented method involves applying a function to the received first X-ray projection image, at least one received second X-ray projection image, and the received and / or determined relative arrangement information as input data. The function can be a learned or trained function. Alternatively, the function can be an unlearned or untrained function, which can be understood, for example, as a rule-based algorithm. The learned function can be understood as a learned or trained machine learning model. The learned function can, for example, be a trained neural network module comprising at least one artificial neural network. In a preferred example, the trained neural network module comprises at least one convolutional neural network (CNN).

[0012] The function is designed to identify a projection of at least one object in a predefined epipolar region in the first X-ray projection image and in at least one second X-ray projection image. For example, when applying the function, a projection of the at least one object in a predefined epipolar region can be searched for, i.e., identified, in the first X-ray projection image and in at least one second X-ray projection image. The predefined epipolar region is arranged around an epipolar line on which the projection of the at least one object lies in the first X-ray projection image and / or in at least one second X-ray projection image.For example, in all X-ray projection images to which the function is applied, the search for the at least one object is performed in parallel, and in particular simultaneously, in order to identify the at least one object in these X-ray projection images or in at least individual X-ray projection images of these X-ray projection images. If, for example, the projection of the at least one object is identified or detected in the first X-ray projection image, that is, found, the epipolar line on which the detected projection lies is determined, and the corresponding epipolar region around the epipolar line is determined in at least one second X-ray projection image, and it is checked, for example, whether the projection of the at least one object, which was identified or detected in at least one second X-ray projection image, lies in this epipolar region.If, in another example or additionally in at least one second X-ray projection image, the projection of the at least one object is detected, that is, identified or found, the epipolar line on which the detected projection lies is determined, and the corresponding epipolar region around the epipolar line in the first X-ray projection image is determined. It is then checked, for example, whether the projection of the at least one object detected in the first X-ray projection image lies within this epipolar region. The projection of the at least one object in the respective X-ray projection image can be understood as a shadow of the object in the respective X-ray projection image.

[0013] The computer-implemented method includes providing the detection information as output data for the function. The provided detection information describes whether at least one object is identified in multiple X-ray projection images if the projection of that object is found in the specified epipolar region in the first X-ray projection image and in at least one second X-ray projection image. For example, if the projection of that object is found in the specified epipolar image region in both the first and at least one second X-ray projection image, the detection information is provided, describing at least the detection of that object and, if applicable, the detected object itself. The output might then state, for example, that the object was detected in multiple X-ray projection images.The detection information is suitable, for example, for verifying whether a chipped bone fragment is found in multiple X-ray images, thus confirming that this bone fragment is indeed present and not, for instance, a misinterpretation of the first X-ray image. Ultimately, an object is automatically detected in multiple X-ray images, allowing even smaller structures, such as chipped bone fragments, to be identified. However, a single detection does not automatically confirm the presence of the object; instead, the system first automatically determines, and specifically verifies, whether this finding—that is, the object—is present in multiple X-ray images, thereby confirming the detection.Detection information is only provided if an automatically detected object appears in multiple X-ray projection images. This allows individual X-ray projection images to be examined for the presence of an object.

[0014] One embodiment includes a function that comprises a first object detection algorithm and a second object detection algorithm, or that has been trained using the first and second object detection algorithms in a training process. Applying the function involves retrieving initial object information. This initial object information is obtained by applying the first object detection algorithm to the first X-ray projection image. The initial object information describes at least one object detected in the first X-ray projection image. The first object detection algorithm may, for example, include specifications regarding the type of object to search for in the first X-ray projection image. Alternatively or additionally, any object in the first X-ray projection image may be searched for.In a preferred example, a chipped bone fragment is detected as the at least one object in the first X-ray projection image. Detecting the at least one object means that a projection, and thus a shadow, of the object is detected in the X-ray projection image. The first object detection algorithm comprises at least one rule and / or procedure, the execution of which, based on at least the first X-ray projection image, allows the at least one object to be detected and thus identified. When applying the first object detection algorithm, the object can be detected and thus identified, for example, based on pattern recognition methods and / or segmentation, particularly threshold-based segmentation.

[0015] Applying the function involves determining a second object information by applying a second object detection algorithm to at least one second X-ray projection image. In a preferred example, the second object detection algorithm differs from the first object detection algorithm; that is, there are two different object detection algorithms. It can be provided that the second object information is only determined if the first object information has been determined beforehand, that is, if at least one object could be detected in the first X-ray projection image.When applying the second object detection algorithm, for example, based on pattern recognition methods and / or segmentation, in particular threshold-based segmentation, the object already identified in the first X-ray projection image can be detected and thus identified in at least one second X-ray projection image.

[0016] The second object detection algorithm is applied not only to at least one second X-ray projection image, but also to the determined first object information and the received and / or determined relative arrangement information. Therefore, the second object detection algorithm not only has the second X-ray projection image in which the object is to be searched for, but also information about the already detected object and the relative arrangement between the received X-ray projection images.

[0017] The second object information describes that at least one object detected in the first X-ray projection image is also detected in at least one second X-ray projection image. The second object detection algorithm thus specifically searches for the object detected in the first X-ray projection image in at least one second X-ray projection image. This second object detection algorithm is designed to determine at least one three-dimensional model of the at least one object based on a projection of the at least one object in the first X-ray projection image. This three-dimensional model can alternatively be referred to as a three-dimensional object model.In a preferred example, several three-dimensional models are generated that take into account, for instance, different three-dimensional configurations of the object, which can be determined based on the object's projection in the first X-ray projection image. In other words, it is estimated or predicted what the object detected in the first X-ray projection image might or could look like in three dimensions.

[0018] The second object detection algorithm is designed to identify a projection of the at least one identified three-dimensional model in at least one other X-ray projection image within a predefined epipolar region. For this purpose, the projection of the at least one identified three-dimensional model can be searched for, for example, in at least one other X-ray projection image within a predefined epipolar region. In a preferred example, projections of several, in particular all, identified three-dimensional models are searched for within the predefined epipolar region. The predefined epipolar region is a subregion of the at least one other X-ray projection image in which the projection of the at least one identified three-dimensional model is expected.It is assumed that the three-dimensional shape of the object, i.e., the three-dimensional model, is at least estimated, in particular determined, based on the projection of the object in the first X-ray projection image, so that it can at least be estimated, in particular predicted, where and with what shape the object is expected in the second X-ray projection image, i.e., should be arranged, if the object in the second X-ray projection image is the object from the first X-ray projection image.

[0019] The second object detection algorithm comprises, for example, at least one rule and / or procedure which, when executed based on at least one second X-ray projection image, the first object information, and the relative arrangement information, determines whether a matching or correlated object to the object detected in the first X-ray projection image can also be detected in at least one second X-ray projection image, such that the object is then detected in both the first and at least one second X-ray projection image. Thus, the algorithm does not search for any arbitrary object in the second X-ray projection image, but rather for precisely the object detected in the first X-ray projection image. It therefore searches for corresponding or identical objects.

[0020] If the projection of the at least one identified three-dimensional model is found in at least one second X-ray projection image, then the at least one object is also detected in at least one second X-ray projection image. In this case, the second object information is determined, which describes that the at least one object detected in the first X-ray projection image is also detected in at least one second X-ray projection image. However, if, for example, it is determined that no projection of the at least one identified three-dimensional model is found in at least one second X-ray projection image within the specified epipolar region, then the at least one object is not detected in at least one second X-ray projection image.

[0021] If multiple three-dimensional models are identified, projections of these models are sought within the specified epipolar region. In one example, it is sufficient if a single projection of one of the identified three-dimensional models is found in at least one other X-ray projection image to assume that the object has also been detected.

[0022] The function's operation involves providing detection information that identifies at least one object in multiple X-ray projection images only if both the first and second object information are determined. By first searching for the projection of the at least one object in the first X-ray projection image and then verifying its presence by searching for it in the epipolar region of at least one second X-ray projection image, the object can be quickly and reliably searched for and detected in multiple X-ray projection images. This reliably prevents, for example, the erroneous interpretation of projections of different objects in the X-ray projection images as a single object.

[0023] Another embodiment involves determining an epipolar line on which the projection of at least one object in the first X-ray projection image lies. Thus, for an object detected in the first X-ray projection image, an epipolar line is determined for the at least one second X-ray projection image on which the projection of the at least one object lies. When applying the second object detection algorithm, the predefined epipolar region in the at least one second X-ray projection image is arranged around the determined epipolar line. The epipolar line indicates in which sub-region of the at least one second X-ray projection image the object detected in the first X-ray projection image should be searched.The epipolar line can be understood as a line along which a projection of an object must be arranged in multiple projection images if the projection images describe the object from different perspectives. This is achieved using epipolar geometry, a mathematical model that describes geometric relationships between different representations of the same object from different relative perspectives. Epipolar geometry allows us to establish a dependency between corresponding image points in different X-ray projection images, whereby the image points correspond if they describe a common object point of the image being represented in the different X-ray projection images.In other words, the system calculates where the object can be approximately located in the second X-ray projection image so that it is the same object detected in the first X-ray projection image. This calculation takes into account the epipolar line. By specifying the epipolar line, the search area in at least one of the second X-ray projection images is significantly restricted, but precisely to the area in which the projection of the object must lie. Objects detected outside the specified epipolar area around the epipolar line in the second X-ray projection image cannot be traced back to the object detected in the first X-ray projection image; that is, they do not correspond to it but are different from it.For example, the time required for the second object detection algorithm can be significantly reduced compared to a procedure where the entire second X-ray projection image is searched for the object. Furthermore, the number of false and / or false-positive detection results in the second X-ray image can be greatly reduced compared to a procedure where the entire second X-ray projection image is searched for the object.

[0024] Another embodiment includes the predefined epipolar region extending at least on one side of the epipolar line to a boundary line. The epipolar line and the boundary line run, for example, parallel to each other. In a preferred example, the predefined epipolar region extends on both sides around the epipolar line to a boundary line. Two boundary lines can then be provided, each running parallel to the epipolar line but spatially offset from it. A distance between the epipolar line and the boundary line is fixed. For example, it can be provided that a predefined number of pixels or image points lie between the epipolar line and the boundary line, thus defining a width of the predefined epipolar region at least on one side of the epipolar line.The fixed distance may, for example, depend on the accuracy with which the relative position of the patient to the X-ray imaging system can be determined during the acquisition of the X-ray projection images. Alternatively or additionally, inaccuracies in determining the epipolar line and / or locating the object in the first X-ray projection image can be taken into account in order to select a suitable distance between the epipolar line and the boundary line. The idea here is that this distance is sufficiently large to encompass the entire epipolar region in the X-ray projection image in which the object could be located if it is detectable in at least one subsequent X-ray projection image.

[0025] As an alternative to a fixed distance, the distance between the epipolar line and the boundary line can depend on the size of the projection of at least one identified three-dimensional model. The size of the projection could be, for example, a diameter, height, width, and / or depth of the three-dimensional model. For instance, the one-sided distance between the epipolar line and the boundary line could be at least half the size of the projection, for example, in the case of boundary lines positioned on both sides of the epipolar line, or the full size of the projection, for example, in the case of a boundary line positioned on only one side of the epipolar line. Furthermore, a tolerance range can be considered in addition to the size of the projection to specify the distance. If multiple three-dimensional models are identified, the distance can be specified taking into account the maximum size of the projection of the different three-dimensional models.The predefined epipolar region around the epipolar line can therefore be dynamically adapted to the object. For example, the dimensions of the predefined epipolar region are dimensioned based on the object. This enables reliable detection of the object in at least one second X-ray projection image, as a suitable sub-area of ​​the second X-ray projection image is searched.

[0026] According to an additional embodiment, after receiving the first X-ray projection image, at least one detection image area is determined by applying a detection image area determination algorithm to the first X-ray projection image. Thus, the detection image area can be determined even before applying the first object detection algorithm. The determined detection image area specifies at least one image area of ​​the first X-ray projection image for applying the first object detection algorithm. For example, the determined detection image area describes at least one image area of ​​the first X-ray projection image in which the at least one object is to be searched for when applying the first object detection algorithm. The first object detection algorithm is then applied to the determined detection image area.In a preferred example, the first object detection algorithm is applied only to the identified detection image area and, for example, not to other image areas in the first X-ray projection image. An image area can be understood as an area as described above in connection with the second X-ray projection image, that is, as a sub-area or sub-region of the first X-ray projection image.

[0027] In other words, the detection image area is used to pre-select at least one area of ​​interest in the first X-ray projection image, within which the object is then specifically searched. Alternatively, the detection image area can be understood as the anatomical search area for a detection task, that is, for applying the first object identification algorithm. For example, the detection image area is the area surrounding a bone where the bone borders soft tissue, the area of ​​a joint, and / or the area surrounding a fracture that can be automatically detected in the first X-ray projection image.

[0028] The detection area can be determined automatically, for example, by predefining at least one area of ​​interest, which is then identified in the first X-ray projection image. This means that the image areas in the first X-ray projection image that can be assigned to or correspond to the at least one area of ​​interest are identified. Alternatively or additionally, the detection area can be manually defined by a person, such as a physician or other medical personnel. This can be done, for example, by recording manual input via an input device that may be connected to the X-ray imaging system and / or a data processing system that executes the computer-implemented procedure. In a preferred example, the detection area is determined automatically.

[0029] The detection image area determination algorithm includes, for example, at least one rule or regulation that, when executed, determines the location of the detection image area within the first X-ray projection image. The detection image area determination algorithm can utilize information on typical detection tasks and thus a list of typical detection image areas, such as regions of interest. Ultimately, it can automatically determine where in the first X-ray projection image the object should be searched for. This reduces the computational effort and time required to execute the computer-implemented procedure and also leads to a pre-selection that can prevent irrelevant objects, such as intact bone segments, from being detected.

[0030] Another embodiment involves generating a three-dimensional body part model by applying a body part model determination algorithm to the first X-ray projection image after it has been received. The generated three-dimensional body part model describes a part of the body of the subject as depicted in the first X-ray projection image. For example, if the first X-ray projection image shows a patient's leg, the generated three-dimensional body part model describes this leg of the patient in three dimensions.

[0031] The body part model determination algorithm includes, for example, at least one rule and / or procedure, the execution of which generates and thus determines the three-dimensional body part model. In other words, a three-dimensional model of a patient's bone anatomy can be generated based on the first X-ray projection image. In one example, the body part model determination algorithm may be applied not only to the first X-ray projection image, but also to at least one second X-ray projection image in order to determine the three-dimensional body part model.

[0032] When applying the detection image area determination algorithm, the determined three-dimensional body part model is taken into account. It can therefore be assumed that when determining the detection image area, it is known which body part is described by the first X-ray projection image and what it is likely to look like in three dimensions. This allows for particularly quick and easy automatic identification of, for example, soft tissue around bones, areas of a joint, or areas around visible fractures that may be depicted in the three-dimensional body part model. Ultimately, this allows the detection image area to be determined quickly, as details of the body part described in the first X-ray projection image are considered.

[0033] Another embodiment involves determining a probability value after the second object information has been obtained. This probability value describes the likelihood that the respective projection of the object in the first X-ray projection image and in at least one subsequent X-ray projection image represents an object from a predefined group of objects of interest. In other words, it is checked whether the object detected in the first X-ray projection image is an object of interest, that is, a candidate for an object that is relevant, for example, when providing the detection information.The probability information can be determined, for example, by applying an algorithm for determining probability information. This algorithm may include at least one rule and / or procedure for which the probability is determined. The probability information can be understood as a kind of plausibility check, indicating whether an object is a plausible candidate for providing the detection information. For example, the group of objects of interest may include bone fractures, bone fragments, implants, medical instruments, catheters, catheter tips, other foreign bodies, contrast agents, and / or hemorrhages. Undamaged bones, for example, are not included in the group of objects of interest.In this example, the probability information can therefore describe, for example, how likely it is that the detected object in the first X-ray projection image shows a chipped bone fragment and not an intact bone.

[0034] Only if the determined probability information lies within a predefined probability range does the function provide the detection information that describes whether at least one object is identified in multiple X-ray projection images. Alternatively, the function provides the detection information depending on the determined probability information. Alternatively or additionally, the provided detection information includes the determined probability information. For example, a probability value of zero can be specified for a minimum probability that an object belongs to the group of objects of interest. A probability value of 1 or 100 can then describe a maximum probability that the object belongs to the group of objects of interest.In this example, the probability range can define at least a lower threshold and, in particular, an upper threshold, above and up to which the probability information is large enough to provide the detection information. Other definitions of the probability information and the probability range are possible.

[0035] By determining and considering probability information, it is possible to prevent the search for objects in the X-ray projection images that are not of interest at all, such as intact bone segments. This allows focusing on fixed or dynamically defined types of objects that belong to the group of objects of interest. This illustrates how, for example, a targeted search for chipped bone fragments is possible without detecting objects that, based on their position, are highly unlikely to be chipped bone fragments.

[0036] In another embodiment, a three-dimensional body part model is determined by applying a body part model determination algorithm to at least the first X-ray projection image. The determination of the three-dimensional body part model can be carried out as described above. The determined three-dimensional body part model thus describes a part of the body of the object under investigation, as depicted by the first X-ray projection image, in three dimensions. The at least one object is registered with the determined three-dimensional body part model. Therefore, for example, the at least one object described by the first object information is drawn into the determined three-dimensional body part model. The three-dimensional body part model determined here thus includes information about the detected object, for example, its position and / or orientation relative to other parts of the body part model.

[0037] Furthermore, for each object in the given group of objects of interest, at least one location within the body is specified. For example, this location is listed for each individual object of interest within the group. This might include, for instance, specifying areas around bones, but not within bones, for chipped bone fragments, whereas for contrast agents, it might be specified that they are expected to be found in blood vessels. Other and / or additional locations beyond these examples are possible.

[0038] When determining the probability information, consideration is given to whether an arrangement of the object registered in the determined three-dimensional body part model lies at least partially within at least one of the arrangement areas specified for at least one of the objects in the given group of objects of interest. Thus, based on the position and / or orientation of the detected object in the three-dimensional body part model, and therefore in the patient's body, an estimate is made of how likely an object of interest lies in this arrangement relative to the body part model, using the specified arrangement areas. This allows for a reliable assessment of whether a detected object should be further investigated, for example, by searching for it in the second X-ray projection image, or whether the object cannot or should not be considered further.This leads to the rejection of objects that are not sufficiently relevant, without requiring the entire process, in particular the provision of detection information, to be carried out for each of these objects. This results in an appropriate selection of potential objects.

[0039] The described approach is based, for example, on the understanding that a chipped bone fragment cannot be located within a bone, but only outside of it, yet always in close proximity. This is because soft tissue surrounds the bone, limiting its movement. Thus, locations within the patient's body are known where a given object of interest, such as the chipped bone fragment, can be found, as well as locations where it cannot. This information on the arrangement of specific types of objects within the body is then used to minimize the computational and time-related effort of the computer-implemented procedure, as objects can be identified and eliminated early on as irrelevant or of no interest.

[0040] Furthermore, one embodiment provides that, after determining the first object information, a third object detection algorithm is applied to the first X-ray projection image. This algorithm determines an epipolar line on which the projection of at least one object in the first X-ray projection image lies. It then determines whether at least one projection of another object corresponding to the first object is located within the predefined epipolar region around this epipolar line. The epipolar line described above can be determined when applying the third object detection algorithm to identify the epipolar line on which the detected object lies in the first X-ray projection image. Therefore, the epipolar line described here can correspond to the epipolar line described above.If at least one projection of the additional object is detected that lies along this epipolar line, no detection information is provided. Alternatively, detection information can be provided indicating that the at least one object is not identified in multiple X-ray projection images.

[0041] Alternatively, if at least one projection of the additional object has been determined, it can be checked whether the projection of the at least one object or the projection of the additional object can be assigned to the projection of the at least one identified three-dimensional model in at least one second X-ray projection image. If the projection of the at least one object or the projection of the additional object can be assigned to the projection of the at least one identified three-dimensional model in at least one second X-ray projection image, the detection information is provided.To assign the projection of at least one object, or the projection of the other object, to the projection of the at least one determined three-dimensional model in at least one second X-ray projection image, probability information is determined. This information describes the probability that the projection of the at least one object, or the projection of the other object, on the one hand, and the projection of the at least one determined three-dimensional model in at least one second X-ray projection image, on the other hand, describe the same object. If the probability information determined in this way lies within a predefined probability information range, the projection of the at least one object, or the projection of the other object, is considered to be assigned to the projection of the at least one determined three-dimensional model in at least one second X-ray projection image.However, if the determined probability information is not within the specified probability information range, the projection of the at least one object or the projection of the further object is considered not to be assigned to the found projection of the at least one determined three-dimensional model in the at least one second X-ray projection image.

[0042] In other words, a search is performed along the epipolar line in the first X-ray projection image to determine if there are any additional objects lying along the epipolar line besides the detected object. If this is the case, meaning that further objects are detected along the epipolar line, it cannot be ruled out, for example, that an object detected in the specified epipolar region in at least one second X-ray projection image can actually be assigned to the originally detected object and not to one of the other corresponding objects along the epipolar line. Therefore, in the case of multiple objects along the epipolar line, the object in the second X-ray projection image can no longer be uniquely assigned to an object in the first X-ray projection image.However, this is only the case if the object and the other object correspond to each other, meaning, for example, that they meet a predefined minimum similarity. This includes, for instance, the requirement that both objects can be assigned to a common type of object. This is the case, for example, if the object and the other object are both bone fragments. Therefore, if several objects of the same type are detected along the epipolar line in the first X-ray projection image, the object along the epipolar line in the second X-ray projection image can no longer be uniquely assigned to the object. The second object information would then be meaningless. It is therefore unnecessary to determine the detection information. The computer-implemented procedure is then either terminated directly or performed for a different object in the first X-ray projection image that is not located on the epipolar line.Alternatively, detection information is provided indicating that at least one object is not identified in multiple X-ray projection images. This is an additional safeguard to prevent the provision of meaningless, and potentially erroneous, detection information.

[0043] Another embodiment provides that if the projection of the at least one determined three-dimensional model is not found in at least one second X-ray projection image, no second object information is determined and thus no detection information is provided. Therefore, no detection information can be provided based solely on an object detected in the first X-ray projection image, unless the object detected in the first X-ray projection image is confirmed by a corresponding detection in at least one second X-ray projection image. The detected object is thus discarded. Alternatively, detection information is provided indicating that the at least one object is not identified in multiple X-ray projection images. This is an advantageous approach to prevent erroneous detection information.

[0044] One embodiment involves the function being implemented as either a learned or an unlearned function. If it is implemented as an unlearned function, it can, for example, be implemented as a rule-based algorithm. The learned function can, for example, be trained using the first object detection algorithm and the second object detection algorithm.

[0045] One aspect of the invention relates to a computer-implemented method for providing a learned function. The learned function is suitable, for example, for providing detection information within the framework of the computer-implemented method described above, which at least describes whether at least one object is identified in several X-ray projection images. The computer-implemented method comprises receiving training input data that includes at least groups of received and / or artificially generated X-ray projection images, each describing a common examination area of ​​an object from different projection directions. The artificially generated X-ray projection images can be understood as synthetic or computer-generated X-ray projection images. A given group is, for example, a pair of two X-ray projection images or can comprise more than two X-ray projection images.Furthermore, in one example, the training input data for each group can include relative arrangement information that describes the relative arrangement of the different projection directions to each other. Alternatively or additionally, the training input data can include annotated images in which, for example, each image pixel is assigned a predefined category or class from a group of several categories or classes. The respective category or class from the group of several categories or classes can, for example, distinguish at least between object and non-object.Alternatively or additionally, the object can be identified by category or class as a bone fragment, hemorrhage, foreign body, in particular an implant, contrast medium, for example in a vessel, and / or as a different type of object in which a person, such as a doctor, is interested and / or which is specified as an object of interest.

[0046] The computer-implemented procedure involves receiving comparison output data, which can alternatively be referred to as training output data, where the received comparison output data depends on the training input data. The comparison output data includes, for each group, comparison detection information that describes at least whether at least one object is identified in multiple X-ray projection images of the group. The comparison detection information can be understood as detection information output during the training or learning of the function. Information described by the comparison detection information can correspond, in terms of the type or definition of the described information, to information described by the detection information.In one example, the comparison output data could include a generated three-dimensional image with or without at least one object for the respective group. The object could be, for example, the object described above, such as a bone fragment.

[0047] The computer-implemented method involves determining training image data by applying a previously untrained function to the training input data. Subsequently, the previously untrained function is trained based on a comparison of the training image data with the comparison output data, such that the learned function, generated through this training process, is capable of identifying at least one object actually present in at least one X-ray projection image. In this example, the computer-implemented method thus involves training a function based on at least the training input data and the comparison output data, such that the learned function is capable of identifying at least one object actually present in at least one X-ray projection image. The result of this process step is the learned function.The computer-implemented procedure also includes providing the learned function.

[0048] The training of the function can be based, in particular, on a comparison of the training mapping data and the training output data, which are provided here as comparison output data. The training mapping data has been provided by applying the function to the training input data. The comparison output data can be provided, for example, by applying an untrained function, different from the function that has not yet been trained, in particular by applying rule-based algorithms, to the training input data.At least the first and second object detection algorithms can be considered in an alternative example when training or learning the function, for example, by applying the first and / or the second object detection algorithm and comparing the resulting data, which can be understood as at least part of the comparison output data, with the training image data. The neural network module comprises at least one artificial neural network. The neural network module is a trained machine learning model. Furthermore, the other described algorithms, such as the third object detection algorithm, the detection image area determination algorithm, the algorithm for determining probability information, and / or the body part model determination algorithm, can be used to train or learn the function in a training process.The training process using at least one of the aforementioned algorithms can, for example, include a data comparison between training image data, provided by applying the function to training input data, and comparison output data, provided by applying at least one of the aforementioned algorithms to the training input data. In a preferred example, the received X-ray projection images and the relative arrangement information are fed into the neural network module. An output of the neural network module is, for example, the detection information and / or the information that no detection information can be determined. Furthermore, it can be provided that, in the case of determining the relative arrangement information, an algorithm for determining the relative arrangement information is considered when training the learned function, in particular the neural network module.In this example, only the received X-ray projection images are passed to the learned function, in particular fed into the neural network module.

[0049] In a preferred example, the neural network module used includes at least one generative neural network, which can be referred to as a Generative Adversarial Network (GAN).

[0050] The learned function, for example, the neural network module, was previously trained to perform the steps of the computer-implemented procedure that it could execute. The training of at least the part of the learned function that performs the task of the first object detection algorithm can be based on synthetic X-ray images virtually rendered from cross-sectional images (for example, computed tomography CT and / or cone-beam computed tomography CBCT). Then, for each synthetic image pair consisting of a cross-sectional image and a virtually rendered synthetic X-ray image, a ground truth segmentation exists. This describes, for example, whether and where bone fragments or other clinical findings such as hemorrhage or contrast agent accumulation might be present. Furthermore, real projection images can be annotated in conjunction with a corresponding volume image and then used in the training process.

[0051] Input data for training can include real and / or synthetically generated projection image pairs, optionally restricted (snipped out) to respective epipolar search areas.

[0052] The input data can also include an epipolar geometry between the X-ray projection images as an additional input parameter. Training input data or comparison output data, particularly data for verifying the learned function after or during training, can alternatively or additionally include an annotation of projection images, for example, X-ray projection images in which image areas are assigned to at least one class or category of object by means of the annotation.

[0053] In the architecture of the learned function, for example, in a network architecture of the neural network module, it can be taken into account that the generation function or convolution function accesses only data along and / or near the corresponding epipolar line in the other X-ray projection image of the multiple X-ray projection images. The input data can, for example, include associated three-dimensional X-ray projection image data or sections with and / or without bone fragments or other structures to be detected. If present, at least the bone fragments in the input data are at least partially annotated.

[0054] The training can, for example, include the following steps: A generating network, as a neural network module, is trained to generate a three-dimensional image with and without a bone fragment based on the X-ray projection image pairs (optionally, for example, spatially cropped along epipolar lines). A discriminator function within the neural network module's training is trained to distinguish between real and generated objects, especially bone fragments. In addition to (or as part of) the discriminator function, epipolar consistency with the X-ray projection image pairs is checked by the discriminator function via backprojection. A certain degree of inaccuracy should be allowed for consistency, particularly via a threshold value, since exact matches between the backprojected generated three-dimensional images and the projection images are unlikely.It could also simply be a matter of checking the consistency of certain image features in the rear projection.

[0055] Runtime input data includes: X-ray projection image pair, optionally limited to specific epipolar search areas (all relevant image areas are successively fed into the neural network module during different iterations of the generating neural network, or the architecture of the neural network includes the parameterizable consideration of the epipolar search areas), and / or the epipolar geometry between the two X-ray projection images.

[0056] The generating network can also be executed multiple times at runtime for the same epipolar search area, with varying noise being added to the input data. This allows different synthetically generated three-dimensional image sections to be produced in each iteration.

[0057] Runtime task data can include: Generated three-dimensional image section containing either no bone fragment or at least one generated bone fragment, and / or information on whether a bone fragment was generated or not (can alternatively or additionally be done via a separate detection function applied to the generated three-dimensional image section).

[0058] By backprojecting the generated three-dimensional image sections onto the projection image pairs, it is possible to check how well the generated three-dimensional objects, especially bone fragments, correspond to the X-ray projection image pairs. If the consistency exceeds a threshold, the generated object, especially the bone fragment, is included as a candidate in a results list of the neural network module.

[0059] In this example, candidate features cannot be identified in advance within a single two-dimensional X-ray projection image. Instead, three-dimensional objects, particularly bone fragments, can be generated by the generating neural network module. These objects are not individually detectable in the individual two-dimensional X-ray projection images. The filtering of candidate objects is performed analogously to the process described above. Since the method also functions without clearly visible and thus detectable objects in the individual views (i.e., in individual X-ray projection images), false positives are to be expected. However, these can be filtered out as described above, for example, by applying the third object detection algorithm.

[0060] Another aspect of the invention relates to a method for providing detection information, as described above. This method comprises acquiring the first of several X-ray projection images and acquiring at least one second of several X-ray projection images using an X-ray imaging system. The X-ray projection images depict at least a partially common examination area of ​​an object from different projection directions. The X-ray projection images are thus acquired from different perspectives. Subsequently, the computer-implemented method described above is carried out, and ultimately, for example, the detection information is provided if the condition for this is met. The computer-implemented method is carried out based on the previously acquired and subsequently received X-ray projection images.The process is therefore not limited to processing the received X-ray projection images, but can also include their acquisition. The process is carried out, for example, by the X-ray imaging system.

[0061] One embodiment of the method according to the invention provides that, after the detection information has been provided, a marker describing the at least one detected object is determined. The first X-ray projection image and / or the at least one second X-ray projection image are superimposed with the determined marker and displayed by means of a display device. The display device can be part of the X-ray imaging system or separate from it. For example, one of the X-ray projection images is displayed optically. However, it is displayed in such a way that the detected object is marked within it. The marker can, for example, be a boundary of the object and / or a box around the object, which is drawn into the respective X-ray projection image.The physician or medical staff can therefore view the potentially relevant object in one or more of the X-ray projection images and decide, for example, whether to acquire another X-ray projection image and / or whether other measures should be taken. The described display only occurs after the detection information has been provided, that is, when the detected object has been identified in multiple X-ray projection images. It may be sufficient, for example, if the object is detected in two of three X-ray projection images, for the detection information to be provided and the described display to appear. Ultimately, the object is displayed to allow the medical staff to make further manual decisions, which is a sensible step after the detection information has been provided.

[0062] Another embodiment of the method according to the invention comprises the following: after the detection information has been provided, a warning message is determined by evaluating the detection information, and the determined warning message is output by means of an output device, in particular the display device. In particular, the evaluation of the detection information can include identifying whether an object has been identified in the multiple X-ray projection images, whereby the warning message can be output depending on an evaluation result. The warning message includes the recommendation to acquire at least one further X-ray projection image from at least one further perspective.Alternatively or additionally, a three-dimensional X-ray projection image can be recommended, for example, a CT scan or an image acquired with a C-arm X-ray imaging system, where X-ray projection images are acquired at numerous positions along the C-arm. It can therefore be pointed out that more detailed X-ray projection images are useful before a diagnosis is made and / or a treatment plan is determined for the patient. The computer-implemented procedure can be repeated in an example based on the additional acquired X-ray projection images. Consequently, numerous reactions or further uses of the detection information are possible to particularly facilitate the evaluation of the X-ray projection images for the physician or medical staff.

[0063] One aspect of the invention relates to a data processing system. Unless otherwise specified, all steps of the computer-implemented method can be performed by the data processing system, which includes at least one data processing device. In particular, the at least one data processing device is configured or adapted to perform the steps of the computer-implemented method. For this purpose, the at least one data processing device can, for example, store a computer program containing instructions which, when executed by the at least one data processing device, cause the at least one data processing device to execute the computer-implemented method. The computer-implemented method can also be implemented wholly or partially in hardware. The terms "data processing system" and "at least one data processing device" may be used interchangeably here and in the following.This also applies to corresponding expressions derived from it.

[0064] In the event that the at least one data processing device comprises two or more data processing devices, certain steps performed by the at least one data processing device can also be understood as different data processing devices performing different steps or different parts of a step. In particular, it is not necessary for each data processing device to perform the steps. In other words, the execution of the steps can be distributed among the two or more data processing devices.

[0065] A data processing device can be understood, in particular, as a data processing device that contains a processing circuit. The data processing device can therefore, in particular, process data to perform arithmetic operations. This may also include operations to perform indexed accesses to a data structure, such as a lookup table (LUT), as well as a data processing process implemented in hardware.

[0066] The data processing device may, in particular, contain one or more computers, one or more microcontrollers, and / or one or more integrated circuits, for example, one or more application-specific integrated circuits (ASICs), one or more field-programmable gate arrays (FPGAs), and / or one or more systems on a chip (SoCs). The data processing device may also contain one or more processors, for example, one or more microprocessors, one or more central processing units (CPUs), one or more graphics processing units (GPUs), and / or one or more signal processors, in particular one or more digital signal processors (DSPs).The data processing device may also include a physical or virtual network of computers or other units of the aforementioned type.

[0067] In various embodiments, the data processing device includes one or more hardware and / or software interfaces and / or one or more storage units.

[0068] A storage unit can be volatile data storage, for example as dynamic random access memory (DRAM) or static random access memory (SRAM), or as non-volatile data storage, for example as read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory or flash EEPROM, ferroelectric random access memory (FRAM), or magnetoresistive random access memory.It can be designed as MRAM (magnetoresistive random access memory) or as phase-change random access memory, PCRAM (phase-change random access memory).

[0069] Receiving or receiving data or information within the meaning of the invention can, for example, include receiving or receiving the data, in particular by the data processing system from a sending entity, or reading the data from a data storage device, or receiving or receiving a data stream containing the data, or extracting the data from the data stream, and so on. Wired or wireless data transmission can be used for this purpose. In particular, the data transmission can take place between a hardware and / or software interface of the sending entity and a hardware and / or software interface of the data processing system.

[0070] Providing data or information within the meaning of the invention can, for example, include making the data or information available to the data processing system so that the system can access it. The provided data or information can be evaluated or analyzed, for example, by means of the data processing system and / or by applying an evaluation algorithm. Providing the data or information can enable it to be displayed as a graphical representation and / or at least contribute to the provision of a signal, for example, a control signal for another device, in particular an X-ray imaging system.

[0071] According to another aspect of the invention, a computer program with instructions is specified. When the instructions are executed by a data processing system, the instructions cause the data processing system to carry out a computer-implemented method according to the invention.

[0072] The instructions can be provided, for example, as program code. This program code can be provided, for example, as binary code or assembly language, and / or as source code in a programming language such as C, and / or as a program script, such as Python.

[0073] According to another aspect of the invention, a computer-readable storage medium is specified, in particular a physical and / or non-volatile computer-readable storage medium that stores a computer program according to the invention.

[0074] The computer program and the computer-readable storage medium are each computer program products containing the commands.

[0075] Each embodiment of the computer-implemented method gives rise to a corresponding embodiment of the method according to the invention that is not purely computer-implemented, for example, by including corresponding steps for generating the X-ray projection images or the steps following the provision of the detection information. The invention comprises combinations of the described embodiments. The embodiments of the computer-implemented method apply, where applicable, to the data processing system and the computer program product.

[0076] As described above, the learned function, specifically the neural network module, can be considered a trained machine learning model (MLM). More generally, a trained MLM can replicate cognitive functions that connect humans to other human minds. Specifically, by training on training data, the MLM can adapt to new circumstances and detect and extrapolate patterns. Another term for a trained MLM is "trained function."

[0077] In general, the parameters of a multi-level marketing (MLM) system can be adjusted or updated through training. This can involve supervised training, semi-supervised training, unsupervised training, reinforcement learning, and / or active learning. Furthermore, representational learning, also known as feature learning, can be employed. Specifically, the parameters of MLMs can be adjusted iteratively through multiple training steps. In particular, a specific loss function, also referred to as the cost function, can be minimized during training. When training an artificial neural network (ANN), the backpropagation algorithm can be used.

[0078] An MLM can, in particular, include an ANN, a support vector machine, a decision tree, and / or a Bayesian network, and / or the MLM can be based on k-means clustering, Q-learning, genetic algorithms, and / or association rules. Specifically, an ANN can be or include a deep neural network, a convolutional neural network (CNN), or a convolutional deep neural network. Furthermore, an ANN can be an adversarial network, a deep adverse network, and / or a generative adverse network (GAN).

[0079] Regardless of the grammatical gender of a particular term, persons with male, female or other gender identities are included.

[0080] Further features and combinations of features of the invention will become apparent from the figures and their description, as well as from the claims. In particular, further embodiments of the invention need not necessarily include all features of any one of the claims. Further embodiments of the invention may have features or combinations of features not mentioned in the claims.

[0081] The figures show: Fig. 1 a schematic representation of an X-ray imaging system; Fig. 2 a schematic representation of a first X-ray projection image; Fig. 3 a schematic representation of a second X-ray projection image; Fig. 4 a schematic representation of a given epipolar area in a first X-ray projection image; Fig. 5. A schematic representation of a signal flow graph of a method for providing detection information; Fig. 6 a schematic representation of an epipolar line in several X-ray projection images; and Fig. 7 a schematic representation of an artificial neural network.

[0082] The invention is explained in more detail below with reference to specific embodiments and associated schematic drawings. In the figures, identical or functionally equivalent elements may be designated with the same reference numerals. The description of identical or functionally equivalent elements is not necessarily repeated with respect to different figures.

[0083] Fig. Figure 1 schematically shows an exemplary embodiment of an X-ray imaging system 1 according to the invention. The X-ray imaging system 1 comprises a source unit 3 with an X-ray source, a detector unit 4 with an X-ray detector, and a control system 7, which is configured to control the X-ray source and the X-ray detector in order to generate X-ray images that depict an X-ray object 6, for example, a patient.

[0084] The X-ray imaging system 1 can, for example, comprise a patient table 5 on which the X-ray object 6 is arranged. The X-ray imaging system 1 comprises a data processing system 9 according to the invention, which is adapted to execute a computer-implemented method according to the invention. The following may describe some functions and process steps that are executed by the control system 7, while other functions and process steps that are executed by the data processing system 9 are described. It should be noted that the functions and process steps may also be distributed differently in alternative embodiments.

[0085] For example, control system 7 can adjust various imaging parameters of X-ray imaging system 1, including exposure parameters such as the peak voltage of the X-ray source, the tube current of the X-ray source, and / or the X-ray pulse duration. Control system 7 can also adjust other imaging parameters, such as the filter material and / or thickness of an X-ray filter (e.g., a copper filter), by inserting or removing the corresponding X-ray filter from the beam path. Furthermore, control system 7 can adjust other imaging parameters, such as the size of the collimator aperture of an X-ray collimator. For example, control system 7 can insert or remove the X-ray collimator from the beam path.Control system 7 can, for example, adjust further imaging parameters such as the gain factor of the X-ray detector. Control system 7 can also, for instance, introduce or remove an anti-scattering grid from the beam path.

[0086] In some implementations, the X-ray imaging system 1 includes a display device 8, wherein the control system 7 is configured to drive the display device 8 to display an X-ray image 12, a processed X-ray image 12, and / or an image dependent on the processed X-ray image 12. For example, if a medical procedure is performed using fluoroscopy, the X-ray images are generated as a sequence of successive images so that an operator can monitor or supervise the medical procedure.

[0087] In some implementations, the X-ray imaging system 1 is designed as a fluoroscopy system, specifically as a C-arm fluoroscopy system, in which the source unit 3 and the detector unit 4 are mounted opposite each other on a C-arm 2, which can be rotated about various axes. These movements are referred to as angular and orbital motion, respectively. In some embodiments, the patient table 5 and the C-arm 2 can be positioned relative to each other by corresponding translational movements of the C-arm 2 and / or the patient table 5, in addition to the rotational movement of the C-arm 2. Consequently, the position and / or orientation of the X-ray source with respect to the X-ray object 6 and the position and / or orientation of the X-ray detector with respect to the X-ray object 6 can be precisely adjusted to the desired image perspective.

[0088] In some embodiments, the X-ray imaging system 1 may, for example, include the data processing system 9, databases 10, 11 for storing and / or reading data during the course of the computer-implemented method according to the invention and / or for other purposes.

[0089] Fig. Figure 2 shows a first X-ray projection image 20. The first X-ray projection image 20 can be the X-ray image 12 described above. The first X-ray projection image 20 depicts a body part 21 of the patient, here, for example, a bone. The first X-ray projection image 20 is a single two-dimensional X-ray image 12. An object 22 was detected in the first X-ray projection image 20. To highlight this object 22, a marking 27 in the form of a box is sketched around the object 22.

[0090] Fig. Figure 3 shows the same body part 21 from a different perspective, depicted by a second X-ray projection image 23. The second X-ray projection image 23 is also a single two-dimensional X-ray image 12. In the second X-ray projection image 23, the object 22 from the first X-ray projection image 20 is also detected in the sketched example, which is highlighted here by marker 27. It becomes clear that the object 22 is located here in a predefined epipolar region 24, which is defined by an epipolar line 25. The predefined epipolar region 24 is bounded, for example, by two boundary lines 26 parallel to the epipolar line 25.

[0091] Fig. Figure 4 shows the first X-ray projection image 20, however, the specified epipolar region 24 around the epipolar line 25 is drawn in this image. This corresponds to an epipolar region 24 in which the Fig. The object 22 shown in the diagram is arranged (not sketched here).

[0092] Fig. Figure 5 shows the procedural steps of a method for providing detection information 33. The detection information 33 describes, at least, whether at least one object 22 is identified and thus detected in several X-ray projection images 20, 23. The detection information 33 can also describe the object 22. The method can, for example, be based on the X-ray projection images 20, 23, as they appear in Fig. 2, Fig. 3 to Fig. The procedures shown in Figure 4 are carried out as follows. For example, in a process step S1, the first X-ray projection image 20 and the second X-ray projection image 23 can first be acquired using the X-ray imaging system 1. The X-ray projection images 20 and 23 depict at least a partially shared examination area of ​​an object from different projection directions; that is, they have been acquired from different perspectives. The examination area is, for example, a body part, and the object of examination is the patient. More than one second X-ray projection image 23 can be acquired and considered. Here, the procedure is described purely by way of example using exactly one first X-ray projection image 20 and exactly one second X-ray projection image 23.

[0093] In process step S2, the first X-ray projection image 20 and the second X-ray projection image 23 are received. Process steps S2, S3, S4, S5, S6, S7, S10 and S11 can be understood as part of a computer-implemented procedure, which can be included in an example of the procedure.

[0094] In process step S3, a relative arrangement information 28 is received and / or determined. This describes a relative arrangement of the different projection directions to each other.

[0095] The method can include applying a function 41 to the received first X-ray projection image 20, at least one received second X-ray projection image 23, and the received and / or determined relative arrangement information 28 as input data. The function 41 can be configured to identify a projection of at least one object 22 in a given epipolar region 24. Thus, when applying the function 41, a projection of at least one object 22 in the given epipolar region 24 can be searched for and identified in the first X-ray projection image 20 and in at least one second X-ray projection image 23. The given epipolar region 24 is, for example, arranged around an epipolar line 25 on which the projection of the at least one object 22 lies in the first X-ray projection image 20 and / or in the at least one second X-ray projection image 23.The procedure can include providing detection information 33 as output data for function 41. The provided detection information 33 describes that at least one object 22 is identified in several X-ray projection images 20, 23 if the projection of the at least one object 22 in the specified epipolar region 24 is identified in the first X-ray projection image 20 and in at least one second X-ray projection image 23. Therefore, if the projection of the at least one object 22 is found in the specified epipolar region 24 in the first X-ray projection image 20 and in at least one second X-ray projection image 23, corresponding detection information 33 can, for example, be provided.

[0096] An embodiment of the method is described in detail below. In this embodiment, it is provided that in a process step S4, a first object information 29 is determined by applying a first object detection algorithm 30 to the first X-ray projection image 20. The first object information 29 describes the at least one object 22 that is detected in the first X-ray projection image 20.

[0097] In a process step S5, a second object information 31 can be determined by applying a second object detection algorithm 32, different from the first object detection algorithm 30, to the second X-ray projection image 23, the first object information 29, and the relative arrangement information 28. The second object information 31 describes that the at least one object 22 detected in the first X-ray projection image 20 is also detected in the second X-ray projection image 23. Applying the second object detection algorithm 32, for example, involves determining at least one three-dimensional model of the at least one object 22 based on a projection of this object 22 in the first X-ray projection image 20. A projection of the at least one determined three-dimensional model is then sought in the at least one second X-ray projection image 23 within the specified epipolar region 24.If the projection of the at least one determined three-dimensional model is found in the second X-ray projection image 23, the object 22 is considered to be also detected in the at least one second X-ray projection image 23.

[0098] In process step S6, it can be checked whether both the first object information 29 and the second object information 31 are determined. If this is the case, for example, in process step S7, detection information 33 is provided, which describes that at least one object 22 is identified in several X-ray projection images 20, 23. However, if, for example, only the first object information 29 or not even the first object information 29 is determined, a rejection 34 can occur; that is, the process can be terminated, for example, without providing detection information 33, or detection information 33 is provided, which describes that at least one object 22 is not identified in several X-ray projection images 20, 23.

[0099] After the detection information 33 has been provided, for example in a process step S8 a marking 27 describing the detected object 22 can be determined and the first X-ray projection image 20 and / or the second X-ray projection image 23 can be superimposed with this marking 27 and displayed by means of a display device 8. Thus, for example, an image as in Fig. 2 or Fig. 3 will be displayed.

[0100] In a process step S9, after the detection information 33 has been provided, a warning message 35 can be determined, for example, by evaluating the detection information 33 and output by means of an output device 36, in particular by means of the display device 8. The warning message 35 includes, for example, the recommendation to acquire at least one further X-ray projection image 20, 23 from at least one further perspective and / or a three-dimensional X-ray projection image.

[0101] It may be provided that, after receiving the first X-ray projection image 20 in process step S2, at least one detection image area 37 is determined in process step S10 by applying a detection image area determination algorithm 38 to the first X-ray projection image 20. This algorithm determines at least one detection image area 37 and specifies at least one image area of ​​the first X-ray projection image 20 for applying the first object detection algorithm 30. Subsequently, for example, the first object detection algorithm 30 is applied to the determined detection image area 37 in process step S4; in particular, it is applied only to the determined detection image area 37.

[0102] For example, in process step S11, a three-dimensional body part model 39 can be determined by applying a body part model determination algorithm 40 to at least the first X-ray projection image 20, and in particular also to the second X-ray projection image 23. The determined three-dimensional body part model 39 describes a part of the body part 21 depicted in at least the first X-ray projection image 20 in three dimensions. When applying the detection image area determination algorithm 38 in process step S10, the determined three-dimensional body part model 39 can be taken into account.

[0103] Alternatively or additionally, the object 22 described by the first object information 29 can be registered in the three-dimensional body part model 39. Furthermore, a group of objects of interest may be known, for each of which a specific location within the patient's body is also defined. Probability information can then be determined, describing the probability that the respective projection of object 22 in the first X-ray projection image 20 and in at least one second X-ray projection image 23 represents an object 22 from the defined group of objects of interest. Only if this determined probability information lies within a defined probability range is the detection information 33 provided, for example, in process step S7. This information describes that at least one object 22 is identified in several X-ray projection images 20, 23.Otherwise, it may be rejected as insufficiently plausible. Alternatively, function 41 can specify the provided detection information 33 depending on the determined probability information. Alternatively or additionally, the provided detection information 33 can include the determined probability information.

[0104] When determining this probability information, it can also be taken into account whether an arrangement of the object 22 depicted in the determined three-dimensional body part model 39 lies at least partially within at least one of the arrangement areas that is specified for at least one of the objects 22 of the given group of objects of interest. The three-dimensional body part model 39 can therefore be considered.

[0105] It may be provided that, after the determination of the first object information 29, a third object detection algorithm (not shown here) is applied to the first X-ray projection image 20. The epipolar line 25 is then determined, on which the projection of the at least one object 22 in the first X-ray projection image 20 lies. It is determined whether, within the specified epipolar area 24 around the epipolar line 25, at least one projection of another object 22, corresponding to the at least one object 22, is located. If at least one projection of the other object 22 is determined, it may be provided that the detection information 33 is either not provided or that the detection information 33 is provided, which describes that the at least one object 22 is not identified in multiple X-ray projection images 20, 23. In the former case, for example, the rejection 34 then occurs.

[0106] Alternatively, if at least one projection of the further object 22 is determined, it can be checked whether the projection of the at least one object 22 or the projection of the further object 22 can be assigned to the found projection of the at least one determined three-dimensional model in the at least one second X-ray projection image 23.

[0107] If the projection of the at least one object 22 or the projection of the further object 22 can be assigned to the found projection of the at least one determined three-dimensional model in the at least one second X-ray projection image 23, the detection information 33 is provided in this example, which describes that the at least one object 22 is identified in several X-ray projection images 20, 23.

[0108] If the projection of the at least one determined three-dimensional model is not found in at least one second X-ray projection image 23, that is, if no second object information 31 can be determined in process step S5, then no detection information 33 is provided in process step S7, or detection information 33 is provided that describes that the at least one object 22 is identified in several X-ray projection images 20, 23. In the first case, for example, rejection 34 takes place.

[0109] It is provided that at least the first object detection algorithm 30 and the second object detection algorithm 32 are included by the function 41, in particular by a neural network module which includes at least one artificial neural network, or that the function 41 is taught or trained using at least the first object detection algorithm 30 and the second object detection algorithm 32.

[0110] Fig. Figure 6 illustrates the determination of the epipolar area 24, in which the object 22 is searched for in the second X-ray projection image 23 and / or further objects 22 are searched for in the first X-ray projection image 20. Fig. Figure 6 shows the X-ray imaging system 1 in two different relative positions with respect to the actual object 22, and thus the patient. It becomes clear that the object 22, determined from different perspectives, is positioned in the first X-ray projection image 20 as well as in the second X-ray projection image 23. Due to the geometry, the projections in the individual X-ray projection images 20 and 23 are arranged on a common line, which is the epipolar line 25. The epipolar line 25 is therefore a line on which the projection of the object 22 lies in the first X-ray projection image 20.

[0111] When applying, for example, the second object detection algorithm 32, the predefined epipolar region 24 can be chosen such that it is arranged around the determined epipolar line 25. In one example, the predefined epipolar region 24 extends at least on one side of the epipolar line 25 up to a boundary line 26. The distance between the epipolar line 25 and the boundary line 26 can be fixed or at least dependent on a size of the projection of the determined three-dimensional model. In a preferred example, boundary lines 26 are found on both sides of the epipolar line 25, as shown in Fig. 3 and Fig. 4 is outlined as an example.

[0112] A computer-implemented method for providing a learned function 41 may be provided. This method can be understood as a learning or training procedure for learning or training a function 41 as the learned function 41. The computer-implemented method includes receiving training input data, which comprises at least groups of received and / or artificially generated X-ray projection images 20, 23, each describing a common area of ​​investigation of an object from different projection directions. Furthermore, it may include receiving comparison output data that depends on the training input data, wherein the comparison output data for each group includes comparison detection information that describes at least whether at least one object 22 is identified in several X-ray projection images 20, 23 of the group.The computer-implemented method can include determining training image data by applying a previously unlearned function 41 to the training input data. It can also include learning the previously unlearned function 41 based on a comparison of the training image data with the comparison output data, such that the learned function 41 is trained to identify at least one object 22 actually present in at least one X-ray projection image 20, 23. The computer-implemented method also includes providing the learned function 41.

[0113] Fig. Figure 7 shows an embodiment of an artificial neural network, ANN, 800, which can be encompassed by the function 41 trained as a learned function 41. The ANN 800 comprises nodes 820, ..., 832 and edges 840, ..., 842, where each edge 840, ..., 842 is a directed connection from a first node 820, ..., 832 to a second node 820, ..., 832. In general, the first node 820, ..., 832 and the second node 820, ..., 832 are distinct nodes 820, ..., 832. However, it is also possible for the first node 820, ..., 832 and the second node 820, ..., 832 to be identical. In Fig.For example, edge 840 is a directed connection from node 820 to node 823, and edge 842 is a directed connection from node 830 to node 832. An edge 840, ..., 842 from a first node 820, ..., 832 to a second node 820, ..., 832 is also called an incoming edge for the second node 820, ..., 832 and an outgoing edge for the first node 820, ..., 832.

[0114] In this example, the nodes 820, ..., 832 of ANN 800 can be arranged in layers 810, ..., 813, where the layers can have an intrinsic order introduced by the edges 840, ..., 842 between the nodes 820, ..., 832. In particular, the edges 840, ..., 842 can only exist between adjacent layers of nodes. In the example shown, there is an input layer 810 consisting only of nodes 820, ..., 822 with no incoming edges, an output layer 813 consisting only of nodes 831, 832 with no outgoing edges, and hidden layers 811, 812 between the input layer 810 and the output layer 813. In general, the number of hidden layers 811, 812 can be chosen arbitrarily. For a multilayer perceptron (MLP), this number is at least one. The number of nodes is 820, ..., 822 within the input layer 810 usually refers to the number of input values ​​of the artificial neural network 800, and the number of nodes 831, 832 within the output layer 813 usually refers to the number of output values ​​of the artificial neural network 800.

[0115] In particular, each node 820, ..., 832 of the artificial neural network 800 can be assigned a real number as its value. Here, x denotes... (n) iThe value of the i-th node 820, ..., 832 of the n-th layer 810, ..., 813. The values ​​of nodes 820, ..., 822 of the input layer 810 correspond to the input values ​​of the artificial neural network 800. The values ​​of nodes 831, 832 of the output layer 813 correspond to the output value of the artificial neural network 800. Furthermore, each edge 840, ..., 842 can have a weight, which is a real number. In particular, the weight is a real number within the interval [-1, 1] or within the interval [0, 1]. Here, w denotes (m,n) i,j the weight of the edge between the i-th node 820, ..., 832 of the m-th layer 810, ..., 813 and the j-th node 820, ..., 832 of the n-th layer 810, ..., 813. Furthermore, the abbreviation w (n) i,j for the weight w (n,n+1) i,jdefined. To calculate the output values ​​of neural network 800, the input values ​​are propagated through neural network 800. Specifically, the values ​​of nodes 820, ..., 832 of the (n+1)th layer 810, ..., 813 can be calculated based on the values ​​of nodes 820, ..., 832 of the nth layer 810, ..., 813 as xj(n+1)=f(∑ixi(n)wi,j(n)).

[0116] In this context, the function f is referred to as the transfer function or activation function. Well-known transfer functions include step functions, sigmoid functions (for example, the logistic function, the generalized logistic function, the hyperbolic tangent, the arctangent function), the error function, the smoothstep function, and rectifier functions. The transfer function is used, for example, for normalization. Specifically, the values ​​are propagated layer by layer through neural network 800, with the values ​​of input layer 810 being given by the input of neural network 800. The values ​​of the first hidden layer 811 can be calculated based on the values ​​of input layer 810 of neural network 800, the values ​​of the second hidden layer 812 can be calculated based on the values ​​of the first hidden layer 811, and so on.

[0117] To determine the values ​​w (m,n)i,j To define the edges, the neural network 800 must be trained with training data. The training data includes, in particular, training input data and training output data (denoted as t). i In a training step, the neural network 800 is applied to the training input data to generate computed output data. Specifically, the training data and the computed output data comprise a number of values ​​equal to the number of nodes in the output layer. A comparison between the computed output data and the training data is used to recursively adjust the weights within the neural network 800 (backpropagation algorithm). Specifically, the weights are modified according to the following formula. wi,j'(n)=wi,j(n)−γδj(n)xi(n), where γ is a predefined learning rate, and the numbers δ (n) j can be calculated recursively as δj(n)=(∑kδk(n+1)wj,k(n+1))f'(xi(n)wi,j(n)) based on b (n+1) j , if the (n+1)th layer is not output layer 813, and δj(n)=(xj(n+1)−tj(n+1))f'(xi(n)wi,j(n)), if the (n+1)th layer is output layer 813, where f' is the first derivative of the activation function, and t (n+1) j The comparison training value for the j-th node of the output layer is 813.

[0118] Overall, the examples demonstrate the detection of bone fragments via epipolar constraints. Based on several two-dimensional X-ray images (X-ray projection images 20, 23) from different perspectives, the detection of faintly visible structures (object 22) is performed using epipolar geometry. In particular, anatomical spatial relationships are taken into account in order to filter out a large proportion of false positives.

[0119] The procedure requires a first and second X-ray image 12 (X-ray projection images 20, 23) depicting at least a partially shared part of an anatomy from different projection perspectives. The procedure may include the following steps: - Receipt of at least two X-ray images (X-ray projection images 20, 23). - Receipt or determination of the relative projection geometry (relative arrangement information 28) between the X-ray images (X-ray projection images 20, 23). Receipt: for example, from the angulation data of a C-arm; Determination: for example, by creating an artificial intelligence-based three-dimensional bone model (shape model) based on the X-ray images (X-ray projection images 20, 23) and deriving the relative projection geometry from it. Another possibility for determination: camera tracking of the patient and the resulting creation of a three-dimensional model of the patient at the acquisition times of the X-ray images (X-ray projection images 20, 23), as well as of the X-ray machine. - Optional creation of a three-dimensional model of the bone anatomy based on the X-ray images (X-ray projection images 20, 23), determination of anatomical search areas for the detection task (detection image area 37). The anatomical search areas can be, for example, a close radius in the soft tissue around bones, or the areas of a joint, or the radius of a fracture directly visible in the two-dimensional X-ray projection images 20, 23. - Image-based detection of possible fragments in the first X-ray projection image 20 (candidate, here object 22). - Search in the second X-ray projection image 23 in the epipolar area 24 around the epipolar line 25 of the candidate (object 22) for corresponding features to this candidate (object 22) that can be understood as correspondence candidates. - Optional search along the epipolar lines 25 in the first X-ray projection image 20 to see if there are any other possible corresponding features to the correspondence candidate, in addition to the original candidate (object 22). - Discard 34 candidates (objects 22) for which no corresponding candidates were found or for which other corresponding features existed. Optionally, filter out plausible candidates (objects 22) using the three-dimensional model. Bone fragments or other finds are only to be expected in certain areas of the anatomy, for example, in the area of ​​joints. In particular, they are not to be expected inside bones in three-dimensional space, but only outside, yet in close proximity, to the bones. - Display of the candidates (objects 22) as an overlay or marker 27 in the two-dimensional X-ray projection images 20, 23. The physician now has, for example, three options: discarding 34 the candidates (objects 22); taking at least one additional two-dimensional X-ray projection image 20, 23, which has a new perspective on the anatomy or an increased radiation dose to improve image quality, and repeating the procedure using the additional X-ray projection image 20, 23; three-dimensional X-ray imaging, for example computed tomography (CT) or cone beam computed tomography (CBCT).

[0120] The challenge compared to known methods that involve searching for corresponding features is that the shape of the unknown bone fragment, i.e., object 22, must first be estimated. This means that, based on object 22 found in the first X-ray projection image 20, a model is first created to determine the size and shape that a corresponding object, in particular a bone fragment, might have. This leads to the determination of the three-dimensional model.

[0121] For example, a diameter and / or a preferred direction of the projection in the first X-ray projection image 20 can be detected, and from this, the shape and / or size of the object 22, for example, the bone fragment, can be derived. Furthermore, depending on the image contrast of the detected projection, the thickness of the object 22 (bone fragment) along the projection direction can also be estimated. Backprojections of this estimated three-dimensional shape of the object 22 (bone fragment) with the projection geometry of the second X-ray projection image 23 are determined, taking into account the position along the epipolar line 25 in the second X-ray projection image 23 (this changes the positioning and size of the estimated three-dimensional object 22 in space, and slightly the direction of transmission).In the second X-ray projection image 23, a projection is sought along the epipolar line 25 which is compatible with the back projection (for example in shape / size / image contrast).

[0122] As a further possibility, function 41 for detecting candidates for bone fragments (objects 22) can also be trained to specify, in addition to the location of the detected candidate (object 22), a size (and possibly further shape parameters) of the bone fragment (object 22). These parameters can then be scaled, deformed and / or adjusted accordingly, depending on the position along the epipolar line 25 in the second X-ray projection image 23.

[0123] In summary, the invention relates, in one example, to a computer-implemented method and a method for providing detection information 33 that describes an object 22 detected in several X-ray projection images 20, 23, comprising: Receiving a first X-ray projection image 20 and at least a second X-ray projection image 23, acquired from different perspectives; determining a first object information 29 describing an object 22 detected in the first X-ray projection image 20; determining a second object information 31 describing that the object 22 is also detected in the second X-ray projection image 23, wherein at least one three-dimensional model of the object 22 is determined based on a projection of the object 22 in the first X-ray projection image 20, and a projection of the at least one determined three-dimensional model in the second X-ray projection image 23 is sought in a specified epipolar area 24, wherein at least if the projection is found, the object 22 is also detected in the second X-ray projection image 23;and if the first object information 29 and the second object information 31 are determined, provide the detection information 33.;

Claims

[1] Computer-implemented method for providing detection information (33) that describes at least whether at least one object (22) is identified in several X-ray projection images (20, 23), comprising: - Receiving a first X-ray projection image (20) of the several X-ray projection images (20, 23) and at least a second X-ray projection image (23) of the several X-ray projection images (20, 23), wherein the received X-ray projection images (20, 23) depict at least a partially common examination area of ​​an object under investigation from different projection directions; - Receiving and / or determining relative arrangement information (28) that describes a relative arrangement of the different projection directions to each other; - Applying a function (41) to the received first X-ray projection image (20), the at least one received second X-ray projection image (23) and the received and / or determined relative arrangement information (28) as input data, wherein the function (41) is configured to identify a projection of at least one object (22) in a given epipolar area (24) in the first X-ray projection image (20) and in at least one second X-ray projection image (23), wherein the specified epipolar area (24) is arranged around an epipolar line (25) on which the projection of the at least one object (22) in the first X-ray projection image (20) and / or in the at least one second X-ray projection image (23) lies; and providing the detection information (33) as output data of the function (41), wherein the provided detection information describes that the at least one object (22) is identified in several X-ray projection images (20, 23) if the projection of the at least one object (22) in the specified epipolar area (24) was identified in the first X-ray projection image (20) and in the at least one second X-ray projection image (23). [2] Computer-implemented method according to claim 1, wherein the function (41) comprises a first object detection algorithm (30) and a second object detection algorithm (32) and includes the application of the function (41): - Determining initial object information (29) by applying the initial object detection algorithm (30) to the initial X-ray projection image (20), wherein the first object information (29) describes at least one object (22) that is detected in the first X-ray projection image (20); - Determining a second object information (31) by applying the second object detection algorithm (32) to the at least one second X-ray projection image (23), the first object information (29) and the relative arrangement information (28), wherein the second object information (31) describes that the at least one object (22) that was detected in the first X-ray projection image (20) is also detected in at least one second X-ray projection image (23), wherein the second object detection algorithm (32) is configured to determine at least one three-dimensional model of the at least one object (22) based on a projection of the at least one object (22) in the first X-ray projection image (20) and to identify a projection of the at least one determined three-dimensional model in at least one second X-ray projection image (23) in a specified epipolar area (24), wherein at least if the projection of the at least one determined three-dimensional model is identified in the at least one second X-ray projection image (23), then at least one object (22) is also detected in the at least one second X-ray projection image (23); and - only if both the first object information (29) and the second object information (31) are determined, to provide the detection information (33) which describes that at least one object (22) is identified in several X-ray projection images (20, 23). [3] Computer-implemented method according to claim 2, wherein an epipolar line (25) is determined on which the projection of the at least one object (22) in the first X-ray projection image (20) lies and when applying the second object detection algorithm (32) the predetermined epipolar area (24) in the at least one second X-ray projection image (23) is arranged around the determined epipolar line (25). [4] Computer-implemented method according to claim 3, wherein the predetermined epipolar region (24) extends at least on one side of the epipolar line (25) to a boundary line (26), wherein a distance between the epipolar line (25) and the boundary line (26) is fixed or at least dependent on a size of the projection of the determined three-dimensional model. [5] Computer-implemented method according to any one of claims 2 to 4, wherein after receiving the first X-ray projection image (20) at least one detection image area (37) is determined by applying a detection image area determination algorithm (38) to the first X-ray projection image (20), wherein the determined at least one detection image area (37) specifies at least one image area of ​​the first X-ray projection image (20) for the application of the first object detection algorithm (30). [6] Computer-implemented method according to claim 5, wherein after receiving the first X-ray projection image (20) a three-dimensional body part model (39) is determined by applying a body part model determination algorithm (40) to at least the first X-ray projection image (20), wherein the determined three-dimensional body part model (39) describes a body part (21) described by the first X-ray projection image (20) in three dimensions, where the determined three-dimensional body part model (39) is taken into account when applying the detection image area determination algorithm (38). [7] Computer-implemented method according to any one of claims 2 to 6, wherein after determining the second object information (31) a probability information is determined which describes a probability with which the respective projection of the object (22) in the first X-ray projection image (20) and in at least one second X-ray projection image (23) is an object (22) from a given group of objects of interest, where the function (41) provides the detection information (33), which describes that at least one object (22) is identified in several X-ray projection images (20, 23), only if the determined probability information lies within a specified probability range, or the function (41) specifies the provided detection information (33) depending on the determined probability information, and / or the provided detection information (33) includes the determined probability information. [8] Computer-implemented method according to claim 7, wherein - a three-dimensional body part model (39) is determined by applying a body part model determination algorithm (40) to at least the first X-ray projection image (20), wherein the determined three-dimensional body part model (39) describes in three dimensions a part of a body of the object of investigation depicted by the first X-ray projection image (20), and the object (22) is registered with the determined three-dimensional body part model (39); and - for the objects (22) in the given group of objects of interest, at least one arrangement area in the body is specified, where, when determining the probability information, it is taken into account whether an arrangement of the object (22) registered in the determined three-dimensional body part model (39) lies at least partially in at least one of the arrangement areas that is specified for at least one of the objects (22) in the given group of objects of interest, or not. [9] Computer-implemented method according to any one of claims 2 to 8, wherein after determining the first object information (29) a third object detection algorithm is applied to the first X-ray projection image (20), wherein an epipolar line (25) is determined on which the projection of the at least one object (22) in the first X-ray projection image (20) lies and it is determined whether in a specified epipolar area (24) around the epipolar line (25) at least one projection of a further object (22) corresponding to the at least one object (22) is arranged, wherein if the at least one projection of the further object (22) is determined, - no detection information (33) is provided or the detection information (33) provided describes that at least one object (22) is not identified in multiple X-ray projection images (20, 23); or - it is checked whether the projection of the at least one object (22) or the projection of the further object (22) can be assigned to the found projection of the at least one determined three-dimensional model in the at least one second X-ray projection image (23), wherein if the projection of the at least one object (22) or the projection of the further object (22) can be assigned to the found projection of the at least one determined three-dimensional model in the at least one second X-ray projection image (23), the detection information (33) is provided which describes that the at least one object (22) is identified in several X-ray projection images (20, 23). [10] Computer-implemented method according to any one of claims 2 to 9, wherein if the projection of the at least one identified three-dimensional model is not found in at least one second X-ray projection image (23), no second object information (31) is determined and the detection information (33) is provided which describes that the at least one object (22) is not identified in several X-ray projection images (20, 23). [11] Computer-implemented method according to one of the preceding claims, wherein the function (41) is configured as a learned function (41) or as an unlearned function (41). [12] Computer-implemented method for providing a learned function (41), comprising: - Receiving training input data comprising at least groups of received and / or artificially generated X-ray projection images (20, 23) that each describe a common area of ​​investigation of an object from different projection directions; - Receiving comparison output data that depend on the training input data, wherein the comparison output data for each group includes comparison detection information that describes at least whether at least one object (22) is identified in several X-ray projection images (20, 23) of the group; - Determining training image data by applying an unlearned function (41) to the training input data; - Learning the not yet learned function (41) based on a comparison of the training image data with the comparison output data, such that the learned function (41) is trained to identify at least one object (22) actually present in at least one X-ray projection image (20, 23); - Provide the learned function (41). [13] Method for providing detection information (33) that describes at least one object (22) detected in several X-ray projection images (20, 23), comprising: - Acquisition of a first X-ray projection image (20) of the several X-ray projection images (20, 23) and at least a second X-ray projection image (23) of the several X-ray projection images (20, 23) using an X-ray imaging system (1), - wherein the X-ray projection images (20, 23) depict at least a partially common examination area of ​​an object under investigation from different projection directions; - Performing a computer-implemented method according to one of claims 1 to 11 on the captured X-ray projection images (20, 23). [14] Method according to claim 13, wherein after the detection information (33) is provided, a marking (27) describing the at least one detected object (22) is determined and the first X-ray projection image (20) and / or the at least one second X-ray projection image (23) is superimposed with the determined marking (27) by means of a display device (8). [15] Method according to claim 13 or 14, wherein, after the detection information (33) has been provided, a warning message (35) is determined and output by means of an output device (36), in particular a display device (8), by evaluating the detection information (33), wherein the warning message (35) includes the recommendation to acquire at least one further X-ray projection image (20, 23) from at least one further perspective and / or a three-dimensional X-ray projection image. [16] Data processing system (9) configured to execute a computer-implemented method according to any one of claims 1 to 12. [17] Computer program product comprising instructions which, when executed by a data processing system (9), cause the data processing system (9) to execute a computer-implemented method according to any one of claims 1 to 12.

Images