Scatter radiation correction methods

Abstract

The invention relates to a computer-implemented method for radiation scatter correction. The method comprises receiving (REC-1) a plurality of projection images (PI) depicting an object and receiving (REC-3) or determining (DET-2) a virtual three-dimensional reference image (VRI) depicting the object. The method includes determining (DET-3) at least one scattering model (SM) and subtracting (SUB) the at least one scattering model (SM) from the projection images (PI). This results in corrected projection images. The method comprises determining (DET-4) a corrected medical image (MI) based on the corrected projection images and comparing (COMP) the corrected image (MI) with the virtual reference image (VRI). The method includes checking (CHECK) whether a termination criterion has been met.If the termination criterion is met, the procedure includes providing (PROV) the corrected medical image (MI). If the termination criterion is not met, the procedure includes adjusting (ADA) the at least one scattering model (SM) and repeating the procedure steps based on the adjusted scattering model.

Description

[0001] The present invention relates to a method for correcting scattered radiation in a three-dimensional medical image, a system designed to carry out this method, a computer program product and a computer-readable storage medium.

[0002] It is known to determine or reconstruct a three-dimensional medical image based on a plurality of projection images. The three-dimensional medical image can depict an object. For this purpose, the object is positioned between an X-ray source and a detector. The projection images depict the object from different directions. Such a three-dimensional medical image is therefore based on an X-ray image of the object.

[0003] Furthermore, it is known that scattered radiation has a significant impact on the image quality of such three-dimensional medical images. Scattered radiation has a particularly significant impact on the image quality of three-dimensional medical images, which are determined based on projection images acquired in a cone beam geometry. In a cone beam geometry, the object is magnified onto the detector depending on its distance from the X-ray tube and its distance from the detector. Such projection images in X-ray imaging are typically acquired using a C-arm. Without correction, the scattered radiation can lead to streak artifacts, cupping artifacts, and / or smearing artifacts in the three-dimensional medical image. These artifacts can make a medical diagnosis based on the three-dimensional medical image difficult or impossible.For this reason, it is necessary to reduce and / or correct the scattered radiation.

[0004] It is well known to position an anti-scatter grid in front of a detector to filter out scattered radiation. However, the anti-scatter grid also filters out direct radiation that is not scattered radiation. This can result in a higher dose being required to achieve the desired image quality.

[0005] Alternatively, it is known to provide the greatest possible distance between the object and the detector to reduce scattered radiation in the projection images. However, due to the magnification in a cone beam geometry, this results in the area covered by the detector being comparatively small, requiring large and therefore expensive detectors.

[0006] Alternatively, it is known to remove or reduce scattered radiation in the projection images by applying a trained function to the projection images. However, such trained functions are prone to errors. This can lead to medically relevant structures being removed from the projection images, and therefore important medical facts are no longer recognizable or displayed in the three-dimensional medical image.

[0007] Alternatively, it is known that scattered radiation can be modeled by solving the Boltzmann transport equation and thus corrected in the projection images. However, this method is also prone to errors.

[0008] Alternatively, it is known to describe or model the scattered radiation using a low-frequency function. Depending on the materials depicted in the three-dimensional medical image and the actually measured or determined Hounsfield (acronym: HU) values, the function modeling the scattered radiation can be iteratively adjusted. However, this requires knowledge of the object's materials. Furthermore, this method is also error-prone and computationally intensive.

[0009] It is therefore the object of the present invention to provide a method by means of which scattered radiation in a three-dimensional medical image can be corrected reliably and without mechanical interventions in the measurement setup.

[0010] The object is achieved by a method for correcting scattered radiation in a three-dimensional medical image, by a device for correcting scattered radiation in a three-dimensional medical image, by a computer program product, and by a computer-readable storage medium according to the independent claims. Advantageous further developments are set out in the dependent claims and in the following description.

[0011] The inventive solution to the problem is described below with reference to both the claimed devices and the claimed method. Features, advantages, or alternative embodiments mentioned here are also applicable to the other claimed subject matter, and vice versa. In other words, the subject claims (which, for example, are directed to a device) can also be developed with the features described or claimed in connection with a method. The corresponding functional features of the method are implemented by corresponding subject modules.

[0012] Regardless of the grammatical gender of a particular term, persons with male, female or other gender identities are included.

[0013] The invention relates to a method for correcting scattered radiation in a three-dimensional medical image. The method comprises a method step of receiving a plurality of projection images. The projection images depict an object. The method comprises a method step of receiving or determining a virtual three-dimensional reference image. The virtual three-dimensional reference image depicts the object. The method comprises a method step of determining at least one scattering model. The method comprises a method step of subtracting the at least one scattering model from the plurality of projection images. In this case, a plurality of corrected projection images are determined. The method comprises a method step of determining a corrected three-dimensional medical image as a function of the plurality of corrected projection images.The method comprises a method step of comparing the corrected three-dimensional medical image with the virtual three-dimensional reference image, wherein a quality parameter is determined. The method comprises a method step of checking, based on the quality parameter, whether a termination criterion is met. If the termination criterion is met, the method comprises a method step of providing the corrected three-dimensional medical image. If the termination criterion is not met, the method comprises a method step of adapting the scattering model depending on the quality parameter and repeating the method steps based on the adapted scattering model.

[0014] In the method step of receiving the plurality of projection images, the plurality of projection images are received via an interface. The plurality of projection images depict an object from different directions. In other words, two projection images of the plurality of projection images depict the object from two different directions. The object can be a human, an animal, or an object. The object can, in particular, be part of the human, part of the animal, or part of the object.

[0015] The projection images are captured primarily using X-ray imaging. The object is positioned between an X-ray tube or X-ray source and an X-ray detector. This arrangement is referred to below as the measurement arrangement. X-rays emitted by the X-ray tube penetrate the object, their intensity being attenuated depending on the object's material and the radiation's energy, and are detected by the detector.

[0016] The X-ray tube and / or detector are designed to rotate around the object within a specific angular range. This angular range can be, for example, 90°, 100°, 120°, 150°, 180°, 210°, 240°, 270°, 300°, 330°, or 360°.

[0017] The measurement setup, in particular, forms a cone beam geometry. In other words, the object is magnified onto the detector depending on its distance from the X-ray tube and its distance from the detector.

[0018] In the method step of receiving or determining the virtual three-dimensional reference image, the virtual three-dimensional reference image is received via the interface or determined via the computing unit. The virtual three-dimensional reference image depicts the object three-dimensionally. The virtual three-dimensional reference image is advantageously free of artifacts caused by scattered radiation, or the artifacts caused by scattered radiation are reduced in the virtual three-dimensional reference image.

[0019] Based on the plurality of projection images, a three-dimensional medical image can be determined that depicts the object in three dimensions. The virtual three-dimensional reference image depicts the object analogously to the three-dimensional medical image. In other words, the three-dimensional medical image and the virtual three-dimensional reference image are registered.

[0020] In the method step of determining the at least one scattering model, the at least one scattering model is determined using the computing unit. Alternatively, an initial scattering model can be retrieved from a database.

[0021] Each projection image of the plurality of projection images comprises a plurality of pixels arranged in a matrix. Each pixel comprises a pixel value that depends on an X-ray intensity measured at the location of the pixel. The at least one scattering model comprises an analog number of pixels arranged in an analog-sized matrix. Each pixel of the scattering model comprises a pixel value.

[0022] In the method step of subtracting the at least one scatter model from the plurality of projection images, the scatter model is subtracted from each projection image by means of the computing unit. In this process, the pixel value of the pixel of the scatter model is subtracted from each pixel value of a pixel of a projection image, which, according to its position in the matrix, corresponds to the position of the pixel in the matrix of the projection image.

[0023] The majority of corrected projection images are determined.

[0024] In the method step of determining the corrected three-dimensional medical image, the corrected three-dimensional medical image is determined based on the plurality of corrected projection images. The corrected three-dimensional medical image is determined analogously to the three-dimensional medical image. In particular, the corrected three-dimensional medical image can be determined using a filtered backprojection based on the plurality of corrected projection images.

[0025] In the process step of comparing the corrected three-dimensional medical image with the virtual three-dimensional reference image, the quality parameter is determined using the computing unit. The quality parameter indicates a similarity between the corrected three-dimensional medical image and the virtual three-dimensional reference image.

[0026] In the process step of checking whether the termination criterion has been met, the arithmetic unit determines whether the termination criterion has been met based on the quality parameter. The termination criterion can, for example, specify a target value for the quality parameter. The termination criterion is met when the quality parameter falls below the target value. Alternatively, the termination criterion can be met when the quality parameter exceeds the target value. Alternatively or additionally, in an iterative process as described below, the termination criterion can specify a maximum number of iterations. Alternatively or additionally, the termination criterion can specify a minimum difference between two quality parameters from consecutive iterations. If the difference between two quality parameters from two consecutive iterations is smaller than the minimum difference, the termination criterion is met.

[0027] The termination criterion can include one of the described termination criteria or an “or” combination of these termination criteria.

[0028] If the termination criterion is reached, the method comprises the step of providing the corrected three-dimensional medical image by means of the interface.

[0029] If the termination criterion is not met, in the method step of adapting the at least one scattering model by means of the computing unit, the scattering model is adapted as a function of the quality parameter. In particular, the at least one scattering model is advantageously adapted such that when the quality parameter is determined again, it comes closer to the termination criterion. The method steps are then repeated based on the adapted scattering model. In particular, the method steps of subtracting the at least one scattering model from the plurality of projection images, determining the corrected three-dimensional medical image, comparing the corrected three-dimensional medical image with the virtual three-dimensional medical image, and checking are carried out again based on the adapted scattering model.

[0030] In particular, these steps are executed iteratively until the termination criterion is reached.

[0031] The inventor has recognized that, based on a virtual three-dimensional reference image, the scattered radiation in the majority of projection images can be corrected, and in this way, the scattered radiation and the resulting artifacts in the three-dimensional medical image can also be corrected. The inventor has recognized that, in this way, an adaptation of the measurement setup is not necessary, and in this way, the measurement setup can be optimized with regard to dose and magnification. The inventor has also recognized that, with the described method, no knowledge of the object is necessary. The inventor has recognized that, with the described method, it can be ensured that medically relevant structures are not erroneously removed from the three-dimensional medical image during the correction process.

[0032] According to one aspect of the invention, the quality parameter is based on a structural similarity value.

[0033] In particular, the corrected three-dimensional medical image and the virtual three-dimensional reference image comprise a plurality of voxels arranged in a three-dimensional matrix, each comprising a voxel value. The three-dimensional matrices of the voxels of the corrected three-dimensional medical image and the virtual three-dimensional reference image are identical in their dimensions.

[0034] The structural similarity value is known as the structural similarity index measure (SSIM). The structural similarity value is determined based on the mean voxel values ​​and the standard deviation of the voxel values ​​of the corrected three-dimensional medical image and the virtual three-dimensional reference image.

[0035] The inventor has recognized that the structural similarity value is a good and standardized measure to evaluate the quality of the corrected three-dimensional medical image.

[0036] According to an alternative, optional aspect of the invention, the quality parameter is based on an alternative distance measure.

[0037] For example, the quality parameter can be based on the mean square deviation between the voxel values ​​of two corresponding voxels of the corrected three-dimensional medical image and the virtual three-dimensional reference image. Corresponding voxels are arranged at the same position in the three-dimensional matrix. In particular, the quality parameter can be the square root of the sum of the square deviations of all voxels.

[0038] The inventor has recognized that various metrics are suitable for evaluating the quality of the corrected three-dimensional medical image. The inventor has recognized that, depending on the computing power and the type of quality assessment, the desired metric can be used to determine the quality parameter.

[0039] According to a further aspect of the invention, a scattering model is determined for each projection image of the plurality of projection images.

[0040] In particular, an individual scattering model is determined for each of the projection images. The individual scattering models may be at least partially identical. In particular, at least two scattering models from two different projection images may be different.

[0041] The inventor has recognized that this method allows for the flexible determination of different scattering models for different projection images. The inventor has recognized that this allows for the consideration of various conditions that may arise when the X-ray tube and detector rotate around the object. In particular, different conditions when capturing the different projection images from different directions can be taken into account by the different scattering models.

[0042] According to a further aspect of the invention, the at least one scattering model is based on a low-frequency signal or on basis functions.

[0043] In particular, the low-frequency signal or function can be designed as described in Trapp, P, Maier, J, Susenburger, M, Sawall, S, Kachelrieß, M. Empirical scatter correction: CBCT scatter artifact reduction without prior information. Med Phys. 2022; 1- 19. https: / / doi.org / 10.1002 / mp.15656.

[0044] The basis functions can in particular be so-called basis splines or B-splines.

[0045] The inventor recognized that the scattering model can model scattered radiation using simple functions. The inventor recognized that standard functions can be initially assumed for this purpose. This allows the same starting point for the scattering model to be selected for each measurement setup.

[0046] According to a further aspect of the invention, the plurality of projection images are projection images acquired during an intervention.

[0047] In other words, the projection images were captured during an intervention. In other words, the projection images were captured intra-interventionally.

[0048] The intervention was performed or carried out on the object. The intervention can, in particular, be a medical intervention. The intervention can, for example, be a catheter examination or a stent procedure, etc.

[0049] The inventor recognized that projection images acquired during an intervention are particularly affected by scattered radiation, since the dose during the intervention should be as low as possible, the projection images should be acquired as quickly as possible, and the geometry of the measurement setup is limited. The inventor recognized that these circumstances, in particular, necessitate subsequent scattered radiation correction.

[0050] According to a further aspect of the invention, the plurality of projection images depict the object from different directions in a cone beam geometry.

[0051] In other words, the object is projected onto the detector from a different direction for each projection image. In other words, the object is oriented differently relative to the detector and / or X-ray tube for each projection image.

[0052] Due to the cone beam geometry, the object is magnified onto the detector depending on its distance from the X-ray tube and its distance from the detector.

[0053] In particular, the majority of projection images may have been acquired using a C-arm.

[0054] The inventor recognized that projection images in a cone-beam geometry are particularly affected by scattered radiation. The inventor also recognized that to determine a three-dimensional medical image, the object must be projected from different directions.

[0055] According to a further aspect of the invention, the virtual three-dimensional reference image is a computed tomography image.

[0056] The computed tomography image was acquired using a computed tomography system. The computed tomography image depicts the object.

[0057] In particular, a three-dimensional medical image of the object was determined based on the plurality of projection images. The computed tomography image was specifically referenced with the three-dimensional medical image. In other words, the computed tomography image was aligned with the three-dimensional medical image in such a way that both depicted the object in the same way and with the same orientation.

[0058] In particular, the method for this purpose can comprise a method step of registering the computed tomography image with the three-dimensional medical image, wherein the virtual three-dimensional reference image is determined.

[0059] The computed tomography image may have been taken or recorded pre-interventionally or before the intervention.

[0060] The inventor recognized that a computed tomography image can be assumed to be virtually free of scattered radiation. The inventor recognized that the computed tomography image can be used as a virtual three-dimensional reference image. The inventor recognized that a computed tomography image of the object is often acquired pre-interventionally. The inventor recognized that this can be used as a virtual three-dimensional reference image, thus eliminating the need for additional recordings or the creation of a reference image. The inventor recognized that a potentially different orientation of the object in the computed tomography image and the three-dimensional medical image can be compensated for by registration.

[0061] According to a further aspect of the invention, the virtual three-dimensional reference image is determined based on a magnetic resonance imaging image.

[0062] The magnetic resonance imaging image is acquired using a magnetic resonance imaging system. The magnetic resonance imaging image depicts the object.

[0063] In particular, a three-dimensional medical image of the object was determined based on the plurality of projection images. In particular, the method for this can include a preliminary step of registering the magnetic resonance imaging image with the three-dimensional medical image, whereby the virtual three-dimensional reference image is determined.

[0064] The magnetic resonance imaging image comprises a plurality of voxels arranged in a matrix that corresponds to the matrix of the (corrected) three-dimensional medical image or has the same dimensions. Each voxel of the matrix comprises a voxel value. When determining the virtual three-dimensional reference image, a voxel value is derived for each voxel based on the voxel value of the magnetic resonance imaging image, which would be expected if an analog image of the object were acquired using X-rays.

[0065] In particular, the magnetic resonance imaging image may have been acquired pre-interventionally.

[0066] The inventor has recognized that the virtual three-dimensional reference image can be determined based on a magnetic resonance imaging image. The inventor has recognized that, based on the voxel values ​​of the magnetic resonance imaging image, voxel values ​​can be derived that would be expected in an image acquired using X-rays. The inventor has recognized that, in this way, a previously acquired magnetic resonance imaging image can serve as the basis for the virtual three-dimensional reference image, thus eliminating the need for an additional acquisition of the virtual three-dimensional reference image. The inventor has recognized that, by registering, a potentially deviating positioning of the object in the magnetic resonance imaging image from the three-dimensional medical image can be corrected.

[0067] According to a further aspect of the invention, the method comprises a method step of applying a first trained function to the magnetic resonance imaging image, whereby the virtual three-dimensional reference image is generated.

[0068] In general, a trained function mimics cognitive functions that humans associate with human reasoning. In particular, training based on training data allows the trained function to adapt to new circumstances and recognize and extrapolate patterns.

[0069] In general, parameters of a trained function can be adjusted through training. In particular, supervised training, semi-supervised training, unsupervised training, reinforcement learning, and / or active learning can be used. Furthermore, representation learning (an alternative term is "feature learning") can be used. In particular, the parameters of the trained functions can be adjusted iteratively through multiple training steps.

[0070] In particular, a trained function can comprise a neural network, a support vector machine, a random tree or a decision tree, and / or a Bayesian network, and / or the trained function can be based on k-means clustering, Q-learning, genetic algorithms, and / or association rules. In particular, a trained function can comprise a combination of several uncorrelated decision trees or an ensemble of decision trees (random forest). In particular, the trained function can be determined using XGBoosting (extreme gradient boosting). In particular, a neural network can be a deep neural network, a convolutional neural network, or a convolutional deep neural network.Furthermore, a neural network can be an adversarial network, a deep adversarial network, and / or a generative adversarial network. In particular, a neural network can be a recurrent neural network. In particular, a recurrent neural network can be a long-short-term memory (LSTM) network, in particular a gated recurrent unit (GRU). In particular, a trained function can comprise a combination of the described approaches. In particular, the approaches described here for a trained function are called the network architecture of the trained function.

[0071] In particular, the first trained function can be a transformer network or a vision transformer network or a convolutional network or a multilayer perceptron or a general adversarial network (acronym: GAN).

[0072] The first trained function is trained in particular by providing magnetic resonance imaging images of multiple objects as training input data, to which the first trained function is applied. Furthermore, corresponding X-ray-based three-dimensional images of the objects are provided as training output data. Each image of the training input data corresponds to an image of the training output data. Two corresponding images are registered to each other and depict the same object in the same position. When the first trained function is applied to an image of the training input data, an output image is determined. This output image is compared with the corresponding image of the training output data.The first trained function is then adjusted in such a way that when the first trained function is applied again, the image determined in this way better matches the corresponding image of the training output data.

[0073] The inventor has recognized that the virtual three-dimensional reference image can be determined by applying a first trained function to a corresponding magnetic resonance imaging image. The inventor has recognized that in this way, an image determined using a different imaging modality can be modified such that it resembles the three-dimensional medical image acquired using a different imaging modality.

[0074] According to a further aspect of the invention, the virtual three-dimensional reference image is determined by means of a heuristic algorithm based on the magnetic resonance imaging image.

[0075] In particular, the heuristic algorithm is applied to the magnetic resonance imaging image, thereby determining the virtual three-dimensional reference image. In particular, a heuristic algorithm known for this purpose can be applied to the magnetic resonance imaging image.

[0076] The inventor has recognized that, as an alternative to the first trained function, a heuristic algorithm can be used to determine the virtual three-dimensional reference image.

[0077] According to a further aspect of the invention, the virtual three-dimensional reference image is determined based on a three-dimensional medical image. The three-dimensional medical image is based on the plurality of projection images.

[0078] In particular, the virtual three-dimensional reference image is derived from the three-dimensional medical image. The three-dimensional medical image is corrected by a first estimated scatter correction.

[0079] The inventor has recognized that the virtual three-dimensional reference image can be determined based on the three-dimensional medical image. Thus, it is not necessary to acquire another medical image. This saves time and money. Furthermore, it is not necessary to apply additional radiation to the subject, and thus a higher dose. The inventor has recognized that registration can be omitted if the virtual three-dimensional reference image is derived from the three-dimensional medical image. This also avoids potential errors caused by referencing.

[0080] According to a further aspect of the invention, the method comprises a method step of receiving or determining the three-dimensional image. The method further comprises a method step of applying a second trained function to the three-dimensional medical image, whereby the virtual three-dimensional reference image is determined.

[0081] In the method step of receiving the three-dimensional medical image, the three-dimensional medical image is received via the interface. The three-dimensional medical image is based on the plurality of projection images. In particular, the three-dimensional medical image was determined or reconstructed as a function of the plurality of projection images, as described above.

[0082] In the alternative method step of determining the three-dimensional medical image, the three-dimensional medical image is determined by a computing unit as a function of the plurality of projection images. In particular, the three-dimensional medical image can be determined by means of a filtered backprojection of the plurality of projection images.

[0083] The three-dimensional medical image depicts the object in three dimensions.

[0084] The second trained function can be designed, in particular, as described above with regard to the first trained function. The second trained function is, in particular, a Generative Adversarial Network (acronym: GAN).

[0085] The second trained function is designed to remove scattered radiation in the three-dimensional medical image.

[0086] The second trained function is trained by providing three-dimensional medical image data of various objects as training input data and three-dimensional medical image data corrected for scattered radiation as training output data. In this case, one image from the training input data and one image from the training output data always form a pair depicting the same object. The second trained function is applied to the training input data for training. The images thus determined are each compared with the training output data that corresponds pairwise to the training input data. Based on the comparison, the second trained function is adjusted such that when the second trained function is applied again, the images thus determined better match the training output data.

[0087] The inventor has recognized that known GANs can be used to correct scattered radiation to create or determine the virtual three-dimensional reference image. In particular, it is irrelevant whether smaller medically relevant structures of the object are not correctly represented in the virtual reference image determined in this way, since this does not correspond to the corrected three-dimensional medical image. The inventor has recognized that the described method can ensure that all structures are correctly represented in the corrected three-dimensional medical image.

[0088] According to a further aspect of the invention, the termination criterion is reached when the quality parameter falls below a target value.

[0089] In particular, falling below the target value indicates a particularly high similarity between the (corrected) three-dimensional medical image and the virtual reference image. The target value is predetermined. The quality parameter describes a difference between the virtual three-dimensional reference image and the corrected three-dimensional reference image.

[0090] Alternatively, the termination criterion may also be reached if the target value is exceeded if the quality parameter is defined otherwise.

[0091] Alternatively or additionally, the termination criterion may also be reached if, as described above, a maximum number of iterations or runs of the procedure have been carried out and / or if the quality parameter changes by less than a predetermined threshold between the iterations.

[0092] In particular, the different types of termination criteria may form a common termination criterion by terminating the procedure and providing the corrected three-dimensional medical image if at least one of the conditions is met.

[0093] The inventor recognized that the quality of the scattered radiation correction can be predetermined based on the termination criterion. The inventor recognized that it is possible to combine different termination criteria to avoid repeating the method infinitely if the quality parameter does not converge.

[0094] The invention also relates to a system for correcting scattered radiation in a three-dimensional medical image. The system comprises an interface and a computing unit. The interface and the computing unit are designed to perform the following method steps: - receiving a plurality of projection images, wherein the projection images depict an object, - Receiving or determining a virtual three-dimensional reference image, wherein the virtual three-dimensional reference image depicts the object, - Determine at least one scattering model, - subtracting the at least one scatter model from the plurality of projection images, whereby a plurality of corrected projection images is determined, - Determining a corrected three-dimensional medical image depending on the plurality of corrected projection images, - Comparing the corrected three-dimensional medical image with the virtual three-dimensional reference image, whereby a quality parameter is determined, - Check based on the quality parameter whether a termination criterion is met, if the termination criterion is met: - Providing the corrected three-dimensional medical image if the termination criterion is not met: - Adapting at least one scatter model depending on the quality parameter, - Repeat the procedure steps based on the adapted scattering model.

[0095] Such a system can, in particular, be configured to carry out the previously described method for scattered radiation correction in a three-dimensional medical image and its aspects. The system is configured to carry out this method and its aspects by arranging the interface and the computing unit to execute the corresponding method steps.

[0096] The invention also relates to a computer program product comprising a computer program and a computer-readable medium. A largely software-based implementation has the advantage that even previously used systems can be easily upgraded to operate in the described manner via a software update. Such a computer program product may, in addition to the computer program, optionally include additional components such as documentation and / or additional components, as well as hardware components such as hardware keys (dongles, etc.) for using the software.

[0097] In particular, the invention also relates to a computer program product comprising a computer program which is directly loadable into a memory of a system, with program sections for carrying out all steps of the above-described method for scattered radiation correction in a three-dimensional medical image and its aspects when the program sections are executed by the system.

[0098] In particular, the invention relates to a computer-readable storage medium on which program sections readable and executable by a system are stored in order to carry out all steps of the above-described method for correcting scattered radiation in a three-dimensional medical image and its aspects when the program sections are executed by the system.

[0099] The above-described properties, features, and advantages of this invention will become clearer and more understandable in conjunction with the following figures and their descriptions. The figures and descriptions are not intended to limit the invention and its embodiments in any way.

[0100] In different figures, identical components are provided with corresponding reference symbols. The figures are generally not to scale.

[0101] It shows Fig. 1 shows a first embodiment of a method for scattered radiation correction in a three-dimensional medical image, Fig. 2 shows a second embodiment of a method for scattered radiation correction in a three-dimensional medical image, Fig. 3 shows a first embodiment of a method step of determining a virtual three-dimensional reference image, Fig. 4 shows a second embodiment of a method step of determining a virtual three-dimensional reference image, Fig. 5 a system for scattered radiation correction in a three-dimensional medical image.

[0102] Fig. 1 shows a first embodiment of a method for scattered radiation correction in a three-dimensional medical image MI.

[0103] The method comprises a method step of receiving REC-1 a plurality of projection images PI. The plurality of projection images PI depict an object. Two projection images PI each depict the object from two different directions. In particular, the projection images PI are two-dimensional x-ray images of the object. The projection images PI can in particular have been acquired using a C-arm or an angled x-ray system or a computed tomography system. In particular, the majority of the projection images PI can have been acquired in a cone beam geometry. The object is arranged between a radiation source, in particular an x-ray source, and a detector, wherein the object is magnified on the detector depending on its distance from the detector and from the radiation source.

[0104] The object can be a human, an animal, or an object. In particular, the object can be a part of a human, an animal, or an object.

[0105] In an optional method step of receiving REC-2 or determining DET-1, a three-dimensional medical image MI can be received or determined. The three-dimensional medical image MI is based on the plurality of projection images PI. In particular, during determining DET-1, the three-dimensional medical image MI can be determined based on the plurality of projection images PI. For example, the three-dimensional medical image MI can be determined from the plurality of projection images PI by means of filtered backprojection. Alternatively, the three-dimensional medical image MI can be determined in advance as described based on the plurality of projection images PI. The three-dimensional medical image MI determined in this way can then be received, for example from a database. The three-dimensional medical image MI depicts the object three-dimensionally.

[0106] The method comprises a further method step of receiving REC-3 or determining DET-2 a virtual three-dimensional reference image VRI. The virtual three-dimensional reference image VRI depicts the object. In particular, the virtual three-dimensional reference image VRI and the three-dimensional medical image MI are registered to one another. In other words, the virtual three-dimensional reference image VRI and the three-dimensional medical image MI depict the object from the same perspective with the same orientation. The virtual three-dimensional reference image VRI comprises less scattered radiation than the three-dimensional medical image MI. The virtual three-dimensional reference image VRI can be determined or captured in advance and subsequently received for the method. Alternatively, the virtual three-dimensional reference image can be determined.For example, the virtual three-dimensional reference image VRI can be determined based on the three-dimensional medical image MI. Alternatively, the virtual three-dimensional reference image can be determined based on an alternative measurement of an image dataset of the object. In particular, the virtual three-dimensional reference image VRI can be a computed tomography image of the object. In particular, this computed tomography image can have been acquired using a computed tomography system with as little noise from scattered radiation as possible. Alternatively, the virtual three-dimensional reference image VRI can be based on a magnetic resonance imaging image or on the three-dimensional medical image MI.

[0107] The method comprises a further method step of determining DET-3 at least one scattering model SM. The scattering model SM describes an approximation of the scattered radiation. In particular, a separate scattering model SM can be determined for each projection image PI. The scattering model SM can be based, in particular, on a low-frequency polynomial or signal or a basis function.

[0108] The method comprises a further method step of subtracting SUB the at least one scattering model SM from the plurality of projection images PI. If a separate scattering model SM has been determined for each projection image PI, the corresponding scattering model SM is subtracted from each projection image PI. A plurality of corrected projection images is thereby determined.

[0109] Each of the projection images PI comprises a plurality of pixels arranged in a matrix or pixel matrix. Each pixel comprises a pixel value. The at least one scatter model SM also comprises a plurality of pixels arranged in a matrix whose dimensions correspond to the matrices of the projection images PI. Each of these pixels also comprises a pixel value. During subtraction SUB, the pixel values ​​corresponding to their position in the matrix are subtracted from one another.

[0110] The method further comprises a method step of determining DET-4 a corrected three-dimensional medical image MI as a function of the plurality of corrected projection images. In particular, the corrected three-dimensional medical image MI is determined, for example, by means of filtered backprojection based on the corrected projection images.

[0111] The method further comprises a method step of comparing COMP the corrected three-dimensional medical image MI with the virtual three-dimensional reference image VRI. In particular, both the corrected three-dimensional medical image MI and the virtual three-dimensional reference image VRI comprise a plurality of voxels arranged in a three-dimensional matrix or voxel matrix. The dimensions of these two voxel matrices are identical. Each of the voxels comprises a voxel value. During the comparison, the voxel values ​​of two voxels corresponding to each other according to their positions in the voxel matrices are compared. A quality parameter is determined in the process.

[0112] For example, the quality parameter can depend on a squared distance between the voxel values. Alternatively, the quality parameter can be based on a structural similarity value.

[0113] The method includes a further step of checking whether a termination criterion has been met based on the quality parameter. The termination criterion can be met, in particular, if the quality parameter exceeds or falls below a predefined value.

[0114] The method comprises a method step of providing PROV the corrected three-dimensional medical image MI when the termination criterion is met. In particular, the corrected three-dimensional medical image MI can be provided via an interface SYS.IF. In particular, the corrected three-dimensional medical image MI can be provided to a database and / or a display unit for displaying the three-dimensional medical image MI of a person.

[0115] If the termination criterion is not met, the method comprises the method steps of adapting ADA of at least one scattering model SM depending on the quality parameter and repeating the method steps based on the adapted scattering model SM.

[0116] When repeating the method steps, the method steps of subtracting (SUB), determining (DET-4) the corrected three-dimensional medical image (MI), comparing (COMP), and checking (CHECK) based on the adjusted scattering model (SM) are performed again. Based on the repeated checking, it is determined again whether the steps need to be repeated with a further adjusted scattering model (SM), or whether the termination criterion has been met and the thus determined corrected three-dimensional projection image is provided.

[0117] Fig. 2 shows a second embodiment of a method for scattered radiation correction in a three-dimensional medical image MI.

[0118] It is the schematic process of the Fig. 1. The process steps of comparing COMP, checking CHECK, subtracting SUB and determining DET-4 of the corrected three-dimensional medical image MI are to be carried out according to the description. Fig. 1 trained.

[0119] Fig. 3 shows a first embodiment of a method step of determining DET-2 a virtual three-dimensional reference image VRI.

[0120] The virtual three-dimensional reference image VRI is based on a magnetic resonance imaging image. When determining DET-2 the virtual three-dimensional reference image VRI, a first trained function is applied to the magnetic resonance imaging image in an application step APP-1. The virtual three-dimensional reference image VRI is generated. The first trained function is configured as described above. In particular, the first trained function can comprise a transformer network, a convolutional network, a generative adversarial neural network, or multilayer perceptrons. In particular, the first trained function can have been trained using reinforcement learning.

[0121] Alternatively, the virtual three-dimensional reference image VRI can be determined using a heuristic algorithm based on the magnetic resonance imaging image.

[0122] Fig. 4 shows a second embodiment of a method step of determining DET-2 a virtual three-dimensional reference image VRI.

[0123] In particular, the virtual three-dimensional reference image VRI is based on the three-dimensional medical image MI. In particular, the method step of receiving REC-2 or determining DET-1 of the three-dimensional medical image MI is not optional.

[0124] The method then comprises a method step of applying APP-2 a second trained function to the three-dimensional medical image MI. The virtual three-dimensional reference image VRI is determined. The second trained function is configured as described above. In particular, the second trained function comprises a generative adversarial neural network (GAN). In particular, the second trained function is trained to remove noise or scattered radiation from the three-dimensional medical image MI.

[0125] Fig. Figure 5 shows a system SYS for scattered radiation correction in a three-dimensional medical image MI.

[0126] The system SYS shown for scattered radiation correction in a three-dimensional medical image MI is designed to carry out a method according to the invention for scattered radiation correction in a three-dimensional medical image MI. The system SYS comprises an interface SYS.IF, a computing unit SYS.CU, and a storage unit SYS.MU.

[0127] The system SYS can, in particular, be a computer, a microcontroller, or an integrated circuit (IC). Alternatively, the system SYS can be a real or virtual computer network (a technical term for a real computer network is "cluster," a technical term for a virtual computer network is "cloud"). The system SYS can be implemented as a virtual system that runs on a computer or a real computer network or a virtual computer network (a technical term is "virtualization").

[0128] The SYS.IF interface can be a hardware or software interface (e.g., a PCI bus, USB, or Firewire). The SYS.CU processing unit can comprise hardware and / or software components, such as a microprocessor or a so-called FPGA (Field Programmable GateWave). The SYS.MU storage unit can be implemented as a temporary random access memory (RAM) or as a permanent mass storage device (hard drive, USB stick, SD card, solid state disk (SSD)).

[0129] The SYS.IF interface can, in particular, comprise a plurality of sub-interfaces that execute different method steps of the respective method according to the invention. In other words, the SYS.IF interface can be designed as a plurality of SYS.IF interfaces. The SYS.CU computing unit can, in particular, comprise a plurality of sub-computation units that execute different method steps of the respective method according to the invention. In other words, the SYS.CU computing unit can be designed as a plurality of SYS.CU computing units.

[0130] Where not explicitly stated, but appropriate and in accordance with the spirit of the invention, individual embodiments, individual aspects, or features thereof may be combined or interchanged without departing from the scope of the present invention. Advantages of the invention described with reference to one embodiment also apply to other embodiments without explicit mention, where transferable. QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited non-patent literature

[0000] Trapp, P, Maier, J, Susenburger, M, Sawall, S, Kachelrieß, M. Empirical scatter correction: CBCT scatter artifact reduction without prior information. Med Phys. 2022; 1- 19. https: / / doi.org / 10.1002 / mp.15656

[0043]

Claims

[1] Computer-implemented method for scattered radiation correction in a three-dimensional medical image (MI) comprising the following method steps: - receiving (REC-1) a plurality of projection images (PI), wherein the projection images depict an object, - Receiving (REC-3) or determining (DET-2) a virtual three-dimensional reference image (VRI), wherein the virtual three-dimensional reference image (VRI) depicts the object, - Determining (DET-3) at least one scattering model (SM), - subtracting (SUB) the at least one scattering model (SM) from the plurality of projection images (PI), whereby a plurality of corrected projection images is determined, - Determining (DET-4) a corrected three-dimensional medical image (MI) depending on the plurality of corrected projection images, - comparing (COMP) the corrected three-dimensional medical image (MI) with the virtual three-dimensional reference image (VRI), whereby a quality parameter is determined, - Check (CHECK) based on the quality parameter whether a termination criterion is met, if the termination criterion is met: - Providing (PROV) the corrected three-dimensional medical image (MI) if the termination criterion is not met: - adapting (ADA) at least one scatter model (SM) depending on the quality parameter, - Repeat the procedure steps based on the adapted scattering model (SM). [2] The method of claim 1, wherein the quality parameter is based on a structural similarity value. [3] Method according to one of the preceding claims, wherein a scattering model (SM) is determined for each projection image (PI) of the plurality of projection images (PI). [4] Method according to one of the preceding claims, wherein the at least one scattering model (SM) is based on a low-frequency signal or on basis functions. [5] Method according to one of the preceding claims, wherein the plurality of projection images (PI) are projection images (PI) acquired during an intervention. [6] Method according to one of the preceding claims, wherein the plurality of projection images (PI) image the object from different directions in a cone beam geometry. [7] Method according to one of the preceding claims, wherein the virtual three-dimensional reference image (VRI) is a computed tomography image. [8] Method according to one of claims 1 to 6, wherein the virtual three-dimensional reference image (VRI) is determined based on a magnetic resonance imaging image. [9] A method according to claim 8, further comprising the following method step: - Applying (APP-1) a first trained function to the magnetic resonance imaging image, thereby generating the virtual three-dimensional reference image (VRI). [10] The method according to claim 8, wherein the virtual three-dimensional reference image (VRI) is determined by means of a heuristic algorithm based on the magnetic resonance imaging image. [11] The method of any one of claims 1 to 6, wherein the virtual three-dimensional reference image (VRI) is determined based on a three-dimensional medical image, the three-dimensional medical image being based on the plurality of projection images. [12] A method according to claim 11, further comprising the following method steps: - Receiving (REC-2) or determining (DET-1) the three-dimensional medical image, - Applying (APP-2) a second trained function to the three-dimensional medical image, thereby determining the virtual three-dimensional reference image (VRI). [13] Method according to one of the preceding claims, wherein the termination criterion is reached when the quality parameter falls below a target value. [14] System (SYS) for scattered radiation correction in a three-dimensional medical image, comprising an interface (SYS.IF) and a computing unit (SYS.CU), wherein the interface (SYS.IF) and the computing unit (SYS.CU) are designed to carry out the following method steps: - receiving (REC-1) a plurality of projection images (PI), wherein the projection images depict an object, - Receiving (REC-3) or determining (DET-2) a virtual three-dimensional reference image (VRI), wherein the virtual three-dimensional reference image (VRI) depicts the object, - Determining (DET-3) at least one scattering model (SM), - subtracting (SUB) the at least one scattering model (SM) from the plurality of projection images (PI), whereby a plurality of corrected projection images is determined, - Determining (DET-4) a corrected three-dimensional medical image (MI) depending on the plurality of corrected projection images, - comparing (COMP) the corrected three-dimensional medical image (MI) with the virtual three-dimensional reference image (VRI), whereby a quality parameter is determined, - Check (CHECK) based on the quality parameter whether a termination criterion is met, if the termination criterion is met: - Providing (PROV) the corrected three-dimensional medical image (MI) if the termination criterion is not met: - adapting (ADA) at least one scatter model (SM) depending on the quality parameter, - Repeat the procedure steps based on the adapted scattering model (SM). [15] Computer program product comprising a computer program which can be loaded directly into a memory (SYS.MU) of a system (SYS), with program sections to carry out all steps of the method according to one of claims 1 to 13 when the program sections are executed by the system (SYS). [16] Computer-readable storage medium on which program sections readable and executable by a system (SYS) are stored in order to carry out all the steps of the method according to one of claims 1 to 13 when the program sections are executed by the system (SYS).

Images