Image correction procedures and medical X-ray equipment

Abstract

For particularly good image quality, a method for correcting artifacts caused by extrafocal radiation (=OFR) in an X-ray image of an object is provided, comprising the following steps: X-ray image of the object is acquired by an X-ray detector with a large number of detector pixels, the object being irradiated by X-ray radiation emitted from an X-ray tube; provision of a primary X-ray spectrum of the X-ray radiation and provision of an estimated or determined extrafocal radiation spectrum for the primary X-ray spectrum; determination of an image X-ray spectrum from the X-ray image for each image pixel; division of the image X-ray spectrum into a high-energy component and a low-energy component for each image pixel.Determining a weighting value for each image pixel by calculating the quotient between high-energy irradiation and total energy irradiation, and determining a weighted image from the weighting values ​​of the image pixels; providing a high-energy image from the high-energy components; processing the low-energy components for each image pixel with respect to the extrafocal radiation component using at least one image processing algorithm, taking into account image features and / or structures of the high-energy image; and recombination of the high-energy component and the processed low-energy component for each image pixel and creating a corrected X-ray image from this.

Description

[0001] The invention relates to a method for correcting artifacts produced by extrafocal radiation (OFR) in an X-ray image of an object according to claim 1 and a device for carrying out such a method according to claim 11.

[0002] Off-focal radiation (OFR), or X-rays emitted outside the central beam, is a well-known problem with X-ray tubes, which use electromagnetic fields to accelerate electrons onto an anode made of a material with a high atomic number (e.g., tungsten). A large proportion of the electrons are backscattered and accelerated again, but this time they strike the anode outside the focal spot / focal path. Some of these electrons interact with the anode, producing unfocused X-rays (e.g., about 10% of the emitted X-rays). In imaging, these cause slight blurring, which reduces the spatial and intensity resolution. This affects, for example, tomographic reconstructions and impairs the accuracy and sharpness of Hounsfield units. For tungsten, which is used in angiography and computed tomography, up to 50% of the electrons are backscattered.For materials with lower atomic numbers, such as those used in mammography, less than 40% are backscattered. In the . Fig. Figure 3 shows the scattering coefficient η in percent versus the electron energy E0 for several materials, e.g. tungsten (W) and copper (Cu).

[0003] To reduce extrafocal radiation, current techniques include, for example, improvements to the electromagnetic fields to increase the efficiency of X-ray tubes. Alternatively, an approach is known that attempts to remove extrafocal radiation from the acquired images using simulations.

[0004] German patent DE 10 2009 015032 A1 discloses a method for iterative extrafocal radiation correction in the reconstruction of CT images. German patent DE 102 22 702 A1 discloses a method for correcting the extrafocal radiation of an X-ray tube in computed tomography.

[0005] It is an object of the present invention to provide a method which enables improved imaging with regard to extrafocal radiation; furthermore, it is an object of the invention to provide an X-ray device suitable for carrying out the method.

[0006] The problem is solved according to the invention by a method for correcting artifacts produced by extrafocal radiation (OFR) in an X-ray image of an object according to claim 1 and of a medical X-ray device according to claim 11. Advantageous embodiments of the invention are the subject of the respective dependent claims.

[0007] The inventive method for correcting artifacts produced by extrafocal radiation (=OFR) in an X-ray image of an object comprises the following steps: recording the X-ray image of the object using an X-ray detector with a plurality of detector pixels, the object being irradiated by X-ray radiation emitted from an X-ray tube, providing a primary X-ray spectrum of the X-ray radiation and providing an estimated or determined extrafocal radiation spectrum for the primary X-ray spectrum, determining an image X-ray spectrum from the X-ray image for each image pixel, splitting the image X-ray spectrum into a high-energy component and a low-energy component for each image pixel, determining a weighting value for each by calculating the quotient between high-energy irradiation and total energy irradiation for each image pixel, and determining a weighted image from the weighting values ​​of the image pixels.Providing a high-energy image from the high-energy components, processing the low-energy components for each image pixel with respect to the extrafocal radiation component using at least one image processing algorithm, taking into account image features and / or structures of the high-energy image, and recombination of the high-energy component and the processed low-energy component for each image pixel and creating a corrected X-ray image from this.

[0008] This method provides a way to correct image quality impairments in X-ray images caused by extrafocal radiation, resulting in exceptionally precise and high-quality images. High image quality leads to improved diagnostics and better patient care. The method exploits the fact that extrafocal radiation primarily contains low-energy photons, as the electrons that generate them have low energy. Therefore, every X-ray image contains a high-energy component that is essentially free of extrafocal radiation. By carefully dividing the energy components of the X-ray image, the severely impaired low-energy component can be processed to match the unaffected high-energy component in terms of image information.

[0009] I represents the respective radiation intensity. For the individual weighting values ​​w, 0 <w<1. Die physikalische Größe Bestrahlung H e is generally measured in units J·m -2 (radiant energy per surface; English: radiant fluence or radiant exposure).

[0010] According to one embodiment of the invention, the primary X-ray spectrum is calculated from the X-ray voltage, the X-ray current, the X-ray filtering, and the position of the image pixels, where u and v represent the positions of the image pixels in the pixel matrix of the X-ray detector. Such a calculation is known and can be performed easily. A particularly precise calculation ensures exceptionally good image quality.

[0011] In a further embodiment of the invention, the extrafocal radiation spectrum is determined by means of simulation, in particular Monte Carlo simulation, or by means of a machine learning algorithm. Monte Carlo simulations are capable of calculating even complex scenarios very precisely. Machine learning allows for fast and accurate calculations. Precise determination of the spectrum of the interfering extrafocal radiation is a prerequisite for optimal correction of the distortions in the X-ray image.

[0012] According to a further embodiment of the invention, the determination of an image X-ray spectrum for each image pixel is achieved by dividing the X-ray image into at least one bone density image D. b (u, v) and a waterproof image D w(u, v) and using the primary X-ray spectrum and the extrafocal radiation spectrum. This represents a particularly accurate and advantageous method of partitioning in medical imaging, as a human or animal body can be approximated by these two main components. Advantageous methods for partitioning the X-ray image into at least a bone density image and a water density image are provided, for example, by bone structure reconstructions based on detected image content or by a method using patient convexity conditions. In the latter, it is assumed that the patient shape is convex, smooth, and continuous (CSC), and this is used to estimate that the object (i.e., the patient or a part thereof) is composed of basic materials such as bone and water. Methods for such partitioning are generally state of the art.

[0013] The determination of the image X-ray spectrum is expediently carried out using the well-known polyenergetic Lambert-Beer law. Lambert-Beer's law describes the attenuation of radiation intensity relative to its initial intensity as it passes through an object, depending on the object's material properties. This is a well-known and proven method.

[0014] In a further embodiment of the invention, the first threshold value between high energy and low energy is selected such that the extrafocal radiation component of the high-energy component is negligible. This can be achieved, for example, by setting the extrafocal radiation component of the high-energy component below a limit value (e.g., a specific absolute value or a percentage, such as 2% or 5%). The limit value can be selected manually by an operator or set automatically. It is advantageous to select the lowest possible limit value while still ensuring that a significant high-energy component with sufficient relevant image information remains. These values ​​are based, for example, on calculated or simulated spectra.

[0015] Advantageously, the high-energy image, i.e., the image composed of the high-energy components for the individual image pixels according to the pixel matrix, can be represented using the weighted image W(u,v) and the X-ray image I(u,v) according to the following formula: I h (u,v) = I(u,v) · W(u, v). The weighted image is composed of a matrix, corresponding to the pixel matrix of the X-ray detector, of weight values ​​between 0 and 1. The high-energy image generally also shows an image of the object's structures without blurring caused by interfering extrafocal radiation.

[0016] Processing or correcting the low-energy components, taking into account image features and / or structures detectable in the high-energy image(s), such as corners or edges, leads to a significant improvement in image quality. These image features and structures in the high-energy image can be identified, for example, through segmentation. Since the high-energy image contains few or no artifacts or blurring caused by extrafocal radiation, correcting and / or modeling the low-energy component based on the high-energy component is a particularly effective method for reducing or eliminating the influence of extrafocal radiation and thus optimizing image quality.

[0017] According to a further embodiment of the invention, at least one image processing algorithm for processing or correcting the low-energy components is an edge enhancement algorithm and / or a guided filter algorithm (smoothing filter that preserves edges). Such algorithms are known and proven in the processing of X-ray images and can be implemented quickly and easily. Any other algorithms that can improve the image can also be used.

[0018] According to a further embodiment of the invention, at least one partial step or step of the method is carried out using machine learning, for example the creation of the extrafocal beam spectrum, the image processing / correction of the low energy component or the division of the X-ray image.

[0019] The invention further comprises a medical X-ray device for carrying out a method described above, comprising an X-ray detector with a plurality of image pixels for recording an X-ray image of an object irradiated by X-rays, an X-ray tube with a cathode and an anode for generating X-rays which also generate extrafocal radiation, a control unit for controlling the steps of the method, a computing unit and an image processing unit with at least one algorithm for processing at least one X-ray image, in particular at least one algorithm for machine learning.

[0020] The invention and further advantageous embodiments according to the features of the dependent claims are explained in more detail below with reference to schematically illustrated exemplary embodiments in the drawing, without thereby limiting the invention to these exemplary embodiments. The drawing shows: Fig. 1 a sequence of steps of the procedure for correcting artifacts produced by extrafocal radiation; Fig. 2 a view of a typical energy spectrum for a tungsten anode; Fig. 3. A view of the scattering coefficient η as a function of the electron energy for different materials; and Fig. 4 A view of a device for carrying out the procedure.

[0021] In the Fig. Figure 1 shows the essential steps of the procedure for correcting artifacts caused by extrafocal radiation (OFR) in an X-ray image of an object. Since the electrons that generate the extrafocal radiation produce X-ray photons with a lower average energy, there is a high-energy component of the X-ray image that is essentially free of extrafocal radiation. The basic idea of ​​the procedure is to use the information from the (largely unaffected) high-energy component to correct the effects of the extrafocal radiation that negatively affect the low-energy component.

[0022] In the first step, an X-ray image I(u, v) is acquired. For this purpose, an object (e.g., a patient or part / organ of a patient) is illuminated by X-rays generated by an X-ray tube of an X-ray machine. This attenuation of the X-rays is characteristically detected by an X-ray detector and converted into an X-ray image. The X-ray detector consists of a matrix with rows and columns of detector pixels (u, v), where u and v represent the respective positions of the detector pixels. Each detector pixel generates one image pixel in the X-ray image.

[0023] In a second step, a primary X-ray spectrum I0(u, v, E) of the X-ray radiation from the X-ray tube and an extrafocal radiation spectrum corresponding to the primary X-ray spectrum are provided. This second step can also be performed before the first step, provided all X-ray parameters used to acquire the X-ray image are known. The primary X-ray spectrum I0(u, v, E) can be determined from the X-ray tube voltage, the tube current, the X-ray filtration, and the position of the image pixels, where u and v represent the position of the image pixels and E represents the energy. The extrafocal radiation spectrum can be estimated or calculated from the primary X-ray spectrum, for example, using simulation (e.g., Monte Carlo simulation), or determined using a machine learning algorithm. Both spectra can either be determined / calculated directly or previously determined, retrieved from memory, and provided.

[0024] In a third step 12, an image X-ray spectrum I is then generated for those image pixels of the X-ray detector which were used for imaging the X-ray image. S (u, v, E) is determined. This can be done in various ways. One particularly precise method is based on the use of Lambert-Beer's law, employing the primary X-ray spectrum, the extrafocal spectrum, and decomposing the X-ray image into basic materials, e.g., at least one bone densitometry image D. b (u, v) and a waterproof image D w (u, v). In particular, this is done for each pixel of the X-ray image.

[0025] Methods for separating bone density and water density images are generally known. For example, a method can be used that employs patient convexity conditions, assuming that the patient shape is convex, smooth, and continuous (CSC). This is then used to estimate that the object (i.e., the patient or a part thereof) is composed of basic materials such as bone and water. A more detailed description of this method can be found below under CSC Methods.

[0026] In a fourth step 13, the image X-ray spectrum for each image pixel is split into a high-energy component and a low-energy component. The first threshold, which separates the high energy from the low energy, is chosen such that the extrafocal radiation component of the high-energy component is negligible, for example, compared to the extrafocal radiation component in the low-energy component. The first threshold can be preset, selected by an operator, or set automatically. For example, the first threshold 18 can be set at the energy level (see, for example, the x-axis in Fig. 2) are drawn where the percentage of the extrafocal radiation spectrum 20 compared to the total spectrum 21 (primary energy spectrum 19 plus extrafocal radiation spectrum 20) falls below a threshold, e.g., 5%, 2%, or 1%, or when the absolute value of the extrafocal radiation or the normalized extrafocal radiation falls below a threshold. An example of the first energy threshold 18 is given in the Fig. 2 shown. It is advantageous to choose the lowest possible first threshold, while still maintaining a significant high-energy component with sufficient relevant image information. In the case of the Fig. In the example shown for a tungsten anode of an X-ray tube, the first threshold 18 is at approximately 65 keV: the low-energy component extends up to 64 keV, and the high-energy component begins at 65 keV. Here, Φ(E) is the radiant power, i.e., radiant energy per unit time. The bremsstrahlung peaks can be disregarded, for example, as absolute values. A typical total spectrum 21 is shown, the sum of a typical primary radiation spectrum 19 and a typical extrafocal radiation spectrum 20.

[0027] In a fifth step 14, a weighting value w is assigned to each of the corresponding image pixels. uv Determined by calculating the quotient between high-energy irradiation and total energy irradiation, e.g., using a calculation unit. The physical quantity irradiance H e is generally measured in units J·m -2(radiant energy per unit area; English: radiant fluence or radiant exposure). The weighting values ​​w uv lie between 0 and 1 (0< w uv <1) and indicate the proportion of radiant energy per surface of the quasi-extrafocal-radiation-free high-energy component to the total radiant energy per surface. From the weighting values ​​w uv A weighting image W (u, v) can then be determined for the individual image pixels.

[0028] In a sixth step (15), a high-energy image is then provided for each image pixel using the high-energy components. This can also be represented as follows: I h (u, v) = I(u, v) · W(u, v). The high-energy image shows a representation of the object's features and structures without blurring or artifacts caused by extrafocal radiation and is therefore of high image quality.

[0029] In a seventh step (16), the low-energy component of the image X-ray spectrum is corrected for the extrafocal radiation component, taking into account the high-energy image created in the sixth step (15). Image processing algorithms and / or correction algorithms are used for this purpose. Machine learning algorithms can also be employed. The image processing algorithms can be, for example, an edge enhancement algorithm and / or a guided filter algorithm (a smoothing filter that preserves edges) or another known optimization algorithm. The low-energy component is processed, for example, so that the resulting structures correspond to the structures in the high-energy image, e.g., with regard to edges, blurring, and other known negative effects of extrafocal radiation on image quality. The low-energy component can be processed with respect to the extrafocal radiation component, e.g.,The image is modeled according to the high-energy component. Information extracted from the high-energy image, such as identified structures or features, is used to correct the respective low-energy components. Using the existing high-energy image as a template, the negative effects and artifacts of extrafocal radiation can be reduced particularly effectively in the low-energy components as well.

[0030] In the eighth step, the processed, corrected low-energy component and the (unprocessed) high-energy component are then recombined, resulting in an improved, corrected X-ray image in which the negative effects of extrafocal radiation are significantly reduced. The image quality is thus considerably higher, and even with strong extrafocal radiation from an X-ray source, accurate diagnoses and / or treatments can be ensured based on the X-ray images. The effort required for this procedure is minimal, while the improvement in image quality is substantial.

[0031] Additionally, the process can be performed iteratively by repeating steps three through eight multiple times. This is particularly helpful because extrafocal radiation from neighboring detector pixels can affect image quality. Algorithms based on expectation maximization or other optimization algorithms can also be used for this purpose.

[0032] In the Fig.Figure 4 shows a medical X-ray device 29, which is designed to perform the procedure. The medical X-ray device 29 has an X-ray detector 30 with a multitude of image pixels for recording an X-ray image of an object 33 through which X-rays are irradiated, and an X-ray tube 31 with a cathode and an anode for generating X-rays, which also produce extrafocal radiation. The X-ray detector 30 and the X-ray tube 31 can, for example, be mounted by means of an adjustable C-arm 32. The medical X-ray device also has a control unit 34 for controlling the steps of the procedure, a processing unit 35, and an image processing unit 36 ​​with at least one algorithm for processing at least one X-ray image. A typical medical X-ray device 29 of this type can be, for example, a CBCT (cone beam CT), a fluoroscopy X-ray device, an angiography X-ray device, and / or a CT scanner.

[0033] Alternatively or in addition to pixel-by-pixel processing, neighboring image pixels can also be included. For example, when calculating the ratio between high-energy irradiance and total energy irradiance, neighboring image pixels (or the irradiance incident on them) can be included instead of a pixel-by-pixel calculation. Neighboring image pixels can also be taken into account when correcting the low-energy components.

[0034] The first threshold can be spatially variable to adapt to different areas where significantly different extrafocal radiation is expected. Accordingly, multiple thresholds can be used for different areas.

[0035] When using an X-ray detector with spectral resolution, instead of calculating the pixel-by-pixel X-ray spectra, these can simply be extracted directly from the X-ray detector, for example, in the case of a multi-layer X-ray detector or a photon-counting X-ray detector. Examples of such X-ray devices include computed tomography scanners with photon-counting or multi-layer fan-beam detectors.

[0036] Procedures using patient convexity conditions (CSC procedure): This approach is initially based on the following conditions: Since, in general, objects under investigation are positioned in the X-ray beam with the shortest possible path length (e.g., no other avoidable objects in the way), it can be assumed that the patient shape is convex, smooth, and continuous (CSC). It is also assumed that the imaging properties of the object under investigation can be adequately described by base materials (mat) such as bone (bone) and water (water). From the acquired X-ray images, the line integrals P of the linear attenuation can then be determined using intensity normalization and a log transformation. The line integrals per pixel can be calculated as the sum of the expected material attenuation coefficients µ̅. mat and the path length l matThe CSC conditions are used because mapping from the line integrals to the path lengths is not uniquely solvable. This approach can be implemented as follows: 1. Taking an X-ray projection image l raw ∈ ℝ 2 2. Performing intensity normalization: I=IrawI0 3. Performing a log transformation, e.g., P = -log(I), this yields the curve integrals of the total attenuation for all pixels. 4. Determination of the expected linear attenuation coefficients as a function of the X-ray tube parameters, e.g. μmat¯=E[μmat(E)], where E is the photon energy 5. Decomposition of the line integrals into known, expected linear attenuation coefficients and associated path lengths: P = ∑ mat µ̅ mat · l mat , where the path length l mat ∈ℝ 2; subject to the restriction that the total path length or object thickness l = Σ mat l mat If the CSC conditions are met, then P = µ bone · l bone + µ water · l water The decomposition can be performed using well-known optimizers, such as gradient descent. Initially, the path length can be set to zero for all materials except one.

[0037] Regardless of the grammatical gender of a particular term, persons with male, female or other gender identities are included.

[0038] The invention can be summarized as follows: For particularly good image quality, a method for correcting artifacts produced by extrafocal radiation (=OFR) in an X-ray image of an object is provided, comprising the following steps: X-ray image of the object is acquired by an X-ray detector with a plurality of detector pixels, the object being irradiated by X-ray radiation emitted from an X-ray tube; provision of a primary X-ray spectrum of the X-ray radiation and provision of an estimated or determined extrafocal radiation spectrum for the primary X-ray spectrum; determination of an image X-ray spectrum from the X-ray image for each image pixel; division of the image X-ray spectrum into a high-energy component and a low-energy component for each image pixel.Determining a weighting value for each image pixel by calculating the quotient between high-energy irradiation and total energy irradiation, and determining a weighted image from the weighting values ​​of the image pixels; providing a high-energy image from the high-energy components; processing the low-energy components for each image pixel with respect to the extrafocal radiation component using at least one image processing algorithm, taking into account image features and / or structures of the high-energy image; and recombination of the high-energy component and the processed low-energy component for each image pixel and creating a corrected X-ray image from this.

Claims

[1] Method for correcting artifacts produced by extrafocal radiation (=OFR) in an X-ray image of an object by the following steps: • Recording of the X-ray image I(u, v) of the object by an X-ray detector with a large number of detector pixels, which object is irradiated by X-ray radiation emitted from an X-ray tube, • Provision of a primary X-ray spectrum I0(u, v,E) of the X-ray radiation and provision of an estimated or determined extrafocal radiation spectrum for the primary X-ray spectrum, • Determination of an image X-ray spectrum I S (u, v, E) from the X-ray image for each image pixel, • Division of the image X-ray spectrum I S (u, v,E) into a high-energy component and a low-energy component for each image pixel, • Determination of a weighting value w uvby calculating the quotient between high-energy irradiance and total energy irradiance for each image pixel and determining a weighted image W(u, v) from the weighted values ​​w uv the image pixel • Providing a high-energy image I h (u, v) from the high-energy components of the individual image pixels • Processing of the low-energy components for each image pixel with respect to the extrafocal radiation component using at least one image processing algorithm, taking into account image features and / or structures of the high-energy image, and • Recombination of the high-energy component and the processed low-energy component for each image pixel and creation of a corrected X-ray image. [2] Method according to claim 1, wherein for the division of the image X-ray spectrum for each image pixel into a high-energy component and a low-energy component, a first threshold value between high energy and low energy is selected such that the extrafocal radiation component of the high-energy component is negligible. [3] Method according to claim 1 or 2, wherein the primary X-ray spectrum is calculated from the X-ray voltage, the X-ray current, the filtering of the X-ray radiation and the position of the image pixels, where u, v represent the positions of the image pixels. [4] Method according to one of the preceding claims, wherein the extrafocal radiation spectrum is determined by simulation, in particular Monte Carlo simulation, or by a machine learning algorithm. [5] Method according to one of the preceding claims, wherein the determination of an image X-ray spectrum for each image pixel is carried out using a division of the X-ray image into at least one bone density image and one water density image and using the primary X-ray spectrum and the extrafocal radiation spectrum. [6] Method according to claim 5, wherein the division of the X-ray image into at least one bone density image and one water density image is carried out using patient convexity conditions or by means of bone structure reconstructions based on detected image content. [7] Method according to claim 5, wherein the determination of the image X-ray spectrum is carried out using the Lambert-Beer law. [8] Method according to any one of claims 2 to 7, wherein the first threshold between high energy and low energy is selected such that the extrafocal radiation of the high energy component falls below a limit value. [9] Method according to any of the preceding claims, wherein the at least one image processing algorithm is an edge enhancement algorithm and / or a guided filter algorithm. [10] Method according to any of the preceding claims, wherein at least one step is performed using machine learning. [11] Medical X-ray apparatus for carrying out a method according to any one of claims 1 to 10, comprising • An X-ray detector with a large number of image pixels for capturing an X-ray image of an object irradiated by X-rays. • An X-ray tube with a cathode and an anode for generating X-ray radiation, which also produces extrafocal radiation, • A control unit for controlling the steps of the process, • A unit of calculation, and • An image processing unit with at least one algorithm for processing at least one X-ray image. [12] Medical X-ray device according to claim 11, comprising at least one machine learning algorithm.

Images