Computer-implemented method for evaluating set of projected images, evaluation device, computer program product and data carrier

Abstract

The invention relates to a computer-implemented method for evaluating a set of projection images when an X-ray device (1) performs maskless angiography in an examination region having a vascular structure, comprising the following steps: providing the recorded set of projection images, the set of projection images shows, in a number of projection images of the plurality of projection images, a flow of the contrast agent through the vascular structure over an elapsed flow time period; selecting a reference image from the plurality of projection images; the weight of each projection image is determined, the weight of a projection image different from the reference image is determined according to the similarity between the projection image and the reference image, and the weight of the reference image is not lower than the maximum weight of other projection images; at least one evaluation image is ascertained from the plurality of projection images, the projection images being used in accordance with the ascertained weights during the ascertainment.

Description

Technical Field

[0001] This invention relates to a computer-executed method and evaluation apparatus for evaluating a set of projected images from an X-ray device used in maskless angiography within an examination area containing vascular structures. Furthermore, this invention also relates to a computer program product and an electronically readable data carrier. Background Technology

[0002] Precise visualization of vascular structures, and in some cases, quantification of their static and flow parameters, is crucial for various medical applications. Particularly in the case of X-ray angiography, angiography typically uses a contrast agent introduced into the patient's vascular system and then delivered through the vascular structures. To enable the acquisition of evaluative images of vascular structures solely from a projection image set, i.e., two-dimensional X-ray images, a digital subtraction angiography (DSA) technique is proposed. In this method, at least one mask image without contrast agent is first acquired, showing the anatomical background, and then subtracted from a filled image containing contrast agent. 3D-DSA is also suggested. However, acquiring mask images also has some drawbacks, such as potentially amplifying noise, increasing dosage, and the possibility of the patient moving between images with and without contrast agent.

[0003] So-called maskless angiography, also known as kinetic or temporal dynamic imaging, provides temporal dynamics by inputting a contrast agent, obtaining an evaluative image of vascular structures without anatomical background. Because this method does not require subtracting a mask image, it can achieve a higher signal-to-noise ratio and reduce radiation dose and contrast agent usage. Examples of maskless angiography include digital variance angiography (DVA) and virtual mask angiography (VMA).

[0004] For example, Viktor Imre The article "Digital variance angiography asaparadigm shift in carbon dioxide angiography" by [Authors' Name] et al., *Radiological Research*, 54.7 (2019), pp. 428-436, describes DVA and provides further references. It suggests acquiring several typically underexposed projection images and deriving an evaluation image through statistical analysis. This involves creating a statistical image whose values ​​include the standard deviation, variance, and other time-derived parameters of X-ray attenuation for each pixel. Using standard deviation and variance as statistical values, the image values ​​contain motion-related functional information, improving the visualization of vascular structures. A DVA method is also described in document EP2628146A2.

[0005] Reference DE 10 2022 209 890 A1 describes a VMA method, specifically a method for generating virtual mask images. Multiple projected images of an object are detected using an angiography device. For at least a portion of all pixel locations, extreme pixel values ​​at each pixel location in the multiple projected images are determined, particularly by the maximum intensity projection (MIP) of iodine as a contrast agent over time. Finally, a virtual mask image is created based on the extreme pixel values ​​at each pixel location. Therefore, the virtual mask image has, at each pixel location, the absolute value or statistically formed minimum or maximum value of some or all pixel values ​​from all projected images at that pixel location. It can be specified that the denoised virtual mask image is obtained by adding pixel-by-pixel of at least two (optionally all) multiple images, including the virtual mask image, plus a basic weighting to more strongly emphasize the extreme pixel values.

[0006] However, the problem with this method is that patient movement also introduces dynamic changes over time and can cause artifacts in the generated evaluation images. Currently, there are no specific methods to reduce patient movement for maskless angiography, especially DVA and VMA. Known motion correction methods are based on 2D-2D registration between projected images. However, this two-dimensional method has limitations in terms of multi-layer motion and non-rigid three-dimensional motion. In other words, motion patterns inconsistent with single-layer translational motion within the image plane can lead to residual motion artifacts. Summary of the Invention

[0007] Therefore, the technical problem to be solved by the present invention is to improve the image quality of evaluation images for maskless angiography, especially in terms of patient movement.

[0008] To address the aforementioned technical problems, the present invention provides a computer-executed method, an evaluation apparatus, a computer program product, and an electronically readable data carrier according to the parallel claims. Advantageous improvements are provided by the dependent claims.

[0009] In a computer-executed method according to the present invention for evaluating a set of projected images of an X-ray device performing maskless angiography in an examination area having vascular structures, the method according to the present invention includes the following steps:

[0010] - Provides a set of captured projection images, which, as a subset of multiple projection images, show the flow of contrast agent through vascular structures over an elapsed flow period.

[0011] - Select a reference image from multiple projected images.

[0012] - Determine the weight for each projected image, wherein the weight of a projected image different from the reference image is determined based on the similarity between the projected image and the reference image, and the weight of the reference image is not lower than the maximum weight of other projected images.

[0013] - Determine at least one evaluation image from multiple projected images, wherein the projected images are used according to determined weights during the determination process.

[0014] At the start of this method, a set of projection images, such as multiple projection images from 20 to 40 projection images, particularly X-ray images, are provided, which have previously been recorded of the examination area. The patient is then injected with a contrast agent, so the projection images taken during the flow of the contrast agent can show the temporal dynamics of the contrast agent's inundation, filling, and outflow during this flow. However, the projection images can also show the patient's movement within the examination area. The projection images are two-dimensional X-ray images taken by an X-ray device, particularly an angiography device.

[0015] From this set of projection images, a reference image representing the baseline motion state is selected. As will be explained in detail, the reference image is especially a projection image where the vascular structure is clearly visible, as its purpose is to extract the vascular structure as well as possible, thus providing a good reference. Next, weights are assigned to the remaining projection images, representing the degree of matching between each other projection image and the reference image, such as the similarity of motion states. If these weights (also called motion weights) are taken into account when determining the evaluation images, particularly similar motion states will be emphasized, and artifacts arising from different motion states will be suppressed.

[0016] The underlying concept is that while both contrast agent flow and patient movement are temporal dynamic effects reflected in the projected image, the change in contrast agent primarily concerns intensity—an effect that appears in the background but is weaker than the actual patient movement, which causes hard edge shifts. In other words, image similarity, especially the similarity of motion states, clearly dominates the actual patient movement, and therefore the weights effectively indicate which motion state deviations exist. Nevertheless, it should be noted that to reinforce the dominance of actual movement (in contrast to changes in contrast agent concentration), in embodiments, projected image pixels located outside the vascular mask (representing the location of the vascular structure depicted in the projected image) can also be used to determine the weights in the embodiments. For example, the vascular mask can be determined using so-called vascular feature extraction, which may also include a tolerance range around the actually determined vascular orientation to better buffer movement, as will be described in detail below.

[0017] In summary, this invention introduces a weighting scheme to reduce motion artifacts in maskless angiography. This method significantly reduces the overall impact of temporal dynamics caused by patient movement. In particular, it reduces motion artifacts, thereby significantly improving image quality.

[0018] It should be noted that traditional digital subtraction angiography (DSA) employs weighted schemes to reduce motion artifacts in the face of numerous masked images. This invention combines the advantages of reduced motion artifacts with those of maskless angiography, particularly improving the signal-to-noise ratio and reducing dose. The recognition of the dominance of the patient's actual motion relative to temporal dynamics when considering the similarity of projected images allows this technique to shift from contrast-free techniques to those that determine the essential elements of the masked image.

[0019] Depending on the specific method used, such as VMA or DVA, different types of assessment images can be determined. Particularly suitable are assessment images, or at least one of them, that are images of vascular structures and have minimal background features. In principle, both VMA and DVA can yield assessment images comparable to digital subtraction angiography, where at least the anatomical background (and possibly other background elements) is substantially subtracted, i.e., only the vascular structures or the contrast agent within them and / or their dynamics are essentially displayed. According to the method of the invention, the quality of other types of assessment images, such as images reproducing dynamics through color encoding, can also be significantly improved.

[0020] Several conceivable possibilities exist for determining a reference image. A suitable improvement of the present invention specifies that the projection image with the maximum contrast agent filling volume is selected as the reference image, particularly the projection image determined by the average image values ​​and / or chromatogram and / or according to user input, and / or the reference image is determined by using a trained selection function. The maximum contrast agent filling volume means that as many vascular structures as possible are clearly visible in the reference image. Since the purpose of angiography is to observe vascular structures, it is particularly suitable to use a reference image capable of particularly complete identification of vascular structures, as is the case with the maximum contrast agent filling volume. While the reference image can, in principle, be selected based on user input, it can also be automatically selected by a selection function that can be trained, i.e., based on machine learning. In the case of automatic selection by a selection function, the selection function may, for example, evaluate the average image values ​​and / or chromatogram, and / or include so-called vascular feature extraction, ultimately providing information on which features displayed in the image are vascular (because these features have particularly high "vascular feature values"). For example, the projection image showing the most vessels can be found, which essentially corresponds to the maximum contrast agent filling volume. If artificial intelligence is used, i.e., a trained selection function, vascular feature extraction can be achieved, but it can also be trained in other ways, such as based on the user's selection from different sets of projected images.

[0021] For possible vascular feature extraction, please refer to the article "Multiscale Vessel Enhancement Filtering" by Alejandro F. Frangi et al., Medical Image Computation and Computer-Aided Intervention—MICCAI'98, MICCAI 1998, Computer Science Lecture Notes, Vol. 1496, pp. 130-137, which introduces multiscale methods and mentions other conceivable methods.

[0022] In a particularly advantageous embodiment of the invention, the weights are determined during optimization to minimize the deviation function for the deviation image. The deviation image is calculated by subtracting the sum of other projected images different from the reference image, as determined by the test set of the weights, from the reference image. This ultimately attempts to reproduce the reference image from the other projected images, where the projected image with the highest similarity to the reference image contributes the most. This provides a natural method for describing similarity related to motion states and is also easily implemented algorithmically due to the abundance of known techniques for solving optimization problems. High-quality weights can be derived.

[0023] Particularly preferred here is that the deviation function is minimized in the form of a one-dimensional convex quadratic programming problem, the dimension of which is equal to the number of projected images minus one. It is suggested that the optimization problem be transformed into a quadratic programming problem; by reducing the dimension to the number of projected images minus one, the computational cost can be reduced by 100 to 1000 times. Simultaneously, the optimal weights can be robustly determined. This provides an efficient method for solving weighted problems. The optimization problem is transformed into a low-dimensional quadratic programming problem, which can be solved with less computation. Free quadratic programming solutions are already available, thus significantly reducing the workload required for implementation. In particular, it avoids the need to create a computationally efficient general algorithm for solving nonlinear constraint problems. Tests show that, according to the method of the present invention, the weights of other projected images can be robustly determined in a very low-cost, real-time compatible computational mode, for example, on a laptop computer used as a computing device.

[0024] In a particularly preferred embodiment of the invention, a weighted L2 norm is used as the bias function. During optimization, the L2 norm of each bias image used for the test set is minimized. Preferably, the quadratic programming can be expressed as the product of subtraction terms used for the bias images and quadratic and linear terms separated in the weight vector.

[0025] In the following text, let P be the number of pixels in the projected image D, the reference image R, and the deviation image, and let N be the number of projected images (the total number of projected images), such that besides the reference image R, there exist N-1 other projected images D. i Furthermore, let d, r, and a be the vectorized projected image, reference image, and deviation image, respectively (each image has P real-valued terms, i.e., image values). Let the i-th other projected image D... i The weights of the projections are α, where each weight is in the interval between 0 and 1, and the sum of the weights of the other projection images is 1. The optimization problem can be written as:

[0026] α = argmin(r - ∑ i α i d i ) T (r - ∑ i α i d i ) st ∑ i α i = 1 0 ≤ αi ≤ 1 (1)

[0027] Here, M is a P x(N-1) matrix that contains the vectorized projection images d as columns. i M = (d1...d N-1 ), and α is the mask weight α with (N-1) terms. ivector. As follows:

[0028] α = argmin (Mα - r) T (Mα - r) s.t. ∑ i α i = 1 0 ≤ α i ≤ 1 (2)

[0029] α = argmin α T M T Mα - 2r T Mα s.t. ∑ i α i = 1 0 ≤ α i ≤ 1 (3)

[0030] And

[0031] Q = M T M ∈ R N-1xN-1 and c = (-2r T M) T ∈ R N-1 (4)

[0032] Finally, the quadratic programming

[0033] α = argmin α T Qα + c T α s.t. ∑ i α i = 1 0 ≤ α i ≤ 1 (5)

[0034] This quadratic programming only involves matrices and vectors of dimension N - 1, i.e., the number of other projection images. The original projection images of dimension P do not need to be evaluated during the optimization process, where N - 1 is much smaller than P, i.e., N - 1 << P. Only when Q and c are pre-computed, the complete projection images need to be processed. By construction, Q is positive (semi)-definite, so the optimization problem is convex.

[0035] As shown in formula (4), both Q and c can be expressed as scalar products of the vectorized projection images d i of the projection image set and r:

[0036] For the quadratic term

[0037] where r,c = 1…N - 1(6)

[0038] And for the linear term

[0039] Where r = 1…N-1(7)

[0040] It can be said that, appropriately, in order to form the quadratic matrix of the quadratic term of the deviation function, the vector inner product of the vectorized other projected images is generated in the form of a quadratic programming problem, and in order to generate the vector of the linear term of the deviation function, the vector inner product of the vectorized reference image and the vectorized other projected images is generated in the form of a quadratic programming problem.

[0041] For this type of convex quadratic optimization problem, special algorithms already exist in the prior art, which are faster than general constrained nonlinear optimization algorithms. In the suitable design scheme of this invention, the optimization process is implemented using the Goldfarb-Idnani algorithm. In particular, the "Goldfarb-Idnani quadratic programming effective set method" can be used, which can be found in the toolbox called QuadProg. For a basic description of the Goldfarb-Idnani method, see the article "Anumerically stable dual method for solving strictly convex quadratic programs" by D. Goldfarb and A. Idnani, Mathematical Programming 27.1 (1983):1-33. In addition, for positive semi-finite problems, see the article "A dual-active-set algorithm for positive semi-definite quadratic programming" by NL Boland, Mathematical Programming 78 (1996):1-27.

[0042] This invention preferably uses optimization processes, particularly optimization processes for quadratic programming, but in principle, weights can also be determined in other ways within the framework of this invention. For example, it can be specified that the weights are determined at least in part as a function of the correlation and / or comparison between other projected images and the reference image. In principle, known correlations and / or comparisons based on the contrast of different images can be used. For example, it can be envisioned to determine the weights by normalizing the corresponding dimensions to 1.

[0043] Finally, the weight of the reference image itself needs to be determined. Reference images are naturally very similar, and according to this invention, an excessively high weight value is not desired, because although weights are assigned, the image content of many projected images should be largely considered when determining the evaluation image. Therefore, a particularly advantageous improvement of this invention stipulates that the weight of the reference image is determined as the maximum value of the weights of the other projected images. In the above form, the weight w of all projected images... i The following results can be obtained.

[0044]

[0045] Where, N R It is an indicator of the baseline image.

[0046] It should also be noted that the total weight (i.e., the weight of the reference image is also considered) does not necessarily need to be normalized. However, it is conceivable and appropriate to normalize the total weight after adding the weight of the reference image.

[0047] Depending on the method of determining at least one evaluation image, different weights can be used to ensure that the projected images in the projection image set participate to varying degrees. It is generally stipulated that weights can be used to form at least one weighted sum and / or to select projected images in the determination step of determining the evaluation image, wherein, in particular, only projected images with weights higher than the limit values ​​of each weight are selected. If a sum is formed, for example in the case of pixel-wise averaging or other combinations of projected images, weights can be introduced in principle, in a way known to be particularly easy in the sense of a weighted sum. However, these weights can also be used in other cases, such as when selecting specific projected images for the determination step, in which, for example, all projected images with weights higher than the limit values ​​are selected. Specifically, it can be stipulated that the limit values ​​can be determined by dividing 0.5 by the number of projected images, i.e., the formula is...

[0048] Where μ = 0.5(9)

[0049] By selecting images in this way, a sufficient number of projected images can usually be chosen. However, if the limit ensures that the number of images selected is too small, various specific methods can be conceived within the framework of this invention.

[0050] For example, it can be stipulated that if less than a minimum number of projected images, especially 10% to 30%, are selected based on a preset extreme value, then either the extreme value is reduced until the minimum number of projected images is selected, or the weights are re-determined, using a renormalization technique to broaden the distribution of the set weights. Thus, the limit value can initially be reduced by decreasing μ in equation (9) to ensure that at least the minimum number of projected images are selected.

[0051] However, in this case, a particularly advantageous alternative specifies the use of renormalization techniques. In principle, known renormalization techniques can be used here, such as the Tikhonov renormalization or variations thereof known in linear regression. Specifically, when minimizing in quadratic programming form, the invention can specify that, when minimizing the deviation function in convex quadratic programming form, the identity matrix is ​​multiplied by the renormalization coefficients and added to the quadratic term matrix.

[0052] In this case, the optimization problem, as shown in Equation (5), can be written as:

[0053]

[0054] For example, parameters (renormalization coefficients) You can choose based on experience, or you can increase the selection until you reach the minimum value.

[0055] The following will explain some specific applications in more detail, especially those related to specific types of evaluation images.

[0056] Specifically, it can be stipulated that statistical images of digital variable angiography (DVA) are determined as at least one of at least one evaluation image, wherein weights are applied when forming the mean over time pixel by pixel and when determining the statistical values ​​(especially the variance and / or standard deviation during the run) of the image values ​​used as statistical images. This variant of DVA can also be called DWVA (“Digital Weighted Angiography”). Formulaically, DWVA images can be determined pixel by pixel as

[0057] in,

[0058] For use in VMA, it can be specified that at least one of at least one evaluation image is determined when using a virtual mask image, wherein weights are used for extreme value projection over time when selecting a projection image, and / or additional weights as relative to the basic weights are used to determine the denoised virtual mask image and / or to determine the evaluation image as a weighted sum of the projection image and the subtraction image of the virtual mask image.

[0059] Therefore, when using iodine as a contrast agent, such as in time-maximum projection (MIP), when determining the original mask through time-maximum projection, some projected images can be excluded by selecting only those with motion phases close to the reference image. For example, as described above, 0.5 divided by the number of projected images can be used as the limit value for selection. If the denoising mask is determined as described in document DE 10 2022 209 890 A1, motion weights can be added as additional weights to the weighted sum by multiplication of the basic weights. Regardless of the determination of the denoising mask image, if the VMA evaluation image is determined as a sum of subtracted images, motion weights can also be used to form a weighted sum. Adding weights when using virtual mask images can also be referred to as WVMA (“Weighted Virtual Mask Angiography”).

[0060] Motion weights can also be used to improve image quality in other situations, such as displaying evaluation images to visually reflect the facts. For example, it can be specified that at least one of the evaluation images is a blood flow information image, in which the region is colored pixel-by-pixel based on the time it takes for the contrast agent to reach the region, excluding projection images with weights below a certain threshold. This visualization technique is known as "iFlow." Better image quality can also be obtained by excluding projection images with significant deviations in motion. It should be noted that this blood flow information image can, of course, also be determined based on the aforementioned DWVA and / or WVMA images.

[0061] In a particularly preferred embodiment of the invention, a motion correction process is performed on the projected images before determining the weights. This process, in order to determine the modification parameters, includes, in particular, rigid or affine 2D-2D registration between the projected images. This approach further improves image quality, reducing motion artifacts in this step and increasing the number of projected images with comparable motion states. It is particularly advantageous that 2D-2D registration is performed at least between the reference image and other projected images, subsequently eliminating some motion differences between the reference image and other projected images and achieving greater similarity.

[0062] In a favorable improvement, multiple reference images are used, and at least one evaluation image is determined for each reference image. The final evaluation image is selected during the selection of at least one evaluation image based on a trained evaluation function and / or image quality values, particularly the maximum vascular feature value and / or user selection information. Instead of a single reference frame, weights can also be determined for multiple reference images. This method is particularly computationally efficient when solving optimization problems in the form of quadratic programming, subsequently requiring only the substitution of terms in Q and c and minimal computation to solve the quadratic programming problem. For example, various DWVA and / or WVMA images can be identified, and the optimal image can be selected. This selection can be performed manually by the user or automatically, for example, through a trained judgment function. This trained judgment function can, for example, operate based on image quality metrics such as the maximum vascular feature value and / or on annotations.

[0063] In addition to the method described above, the present invention also relates to an evaluation apparatus for evaluating a set of projected images of an X-ray device performing maskless angiography in an examination area with vascular structures, wherein the evaluation apparatus comprises:

[0064] - Input interface for receiving a set of captured projection images, wherein the set of projection images comprises a certain number of projection images from multiple projection images showing the flow of contrast agent through the vascular structure during the elapsed flow period.

[0065] - Selection unit, used to select a reference image from multiple projected images.

[0066] - A determining unit is used to determine the weight for each projected image, wherein the weight of a projected image different from the reference image is determined based on the similarity between the projected image and the reference image, and the weight of the reference image is not lower than the maximum weight of other projected images.

[0067] - A determination unit for determining at least one evaluation image from multiple projected images, wherein the projected images are used according to determined weights during the determination process.

[0068] All descriptions relating to the method of the present invention can be applied analogously to the evaluation apparatus of the present invention, and vice versa; therefore, the evaluation apparatus can also obtain the aforementioned advantages.

[0069] The evaluation apparatus may have at least one processor and at least one memory. Functional units may consist of hardware and / or software to perform the various steps of the method according to the invention. In addition to the aforementioned functional units, selection units, setting units, and determining units, other functional units for other steps may also be provided, such as a correction unit for motion correction via 2D-2D registration, and / or sub-units, such as an optimization sub-unit for minimizing the deviation function during optimization, as sub-units of the setting unit. The evaluation apparatus may also have an output interface through which at least one determined evaluation image can be output.

[0070] Of particular advantage is that the evaluation device can be part of the control unit of the X-ray equipment that takes a set of projected images. Therefore, especially as part of the control unit of the X-ray equipment, the X-ray equipment can have an evaluation device according to the invention. In this case, the interface can be an internal interface.

[0071] The preferred X-ray equipment is an angiography system and / or an X-ray system with a C-arm, on which the X-ray detector and X-ray source are arranged opposite each other. This type of C-arm X-ray equipment is commonly used in angiography laboratories and has the advantage of allowing for easy adjustment of different projection orientations of the projected image.

[0072] The computer program product according to the invention can be directly uploaded to the memory of the evaluation apparatus and has such a program device that, when the computer program product is executed, the evaluation apparatus performs the steps of the method according to the invention. The computer program can be stored on an electronically readable data carrier according to the invention, the data carrier including control information stored thereon, the control information including at least one computer program according to the invention and designed to configure the evaluation apparatus to perform the steps of the method according to the invention when the data carrier is used by the evaluation apparatus. The data carrier can be a non-transitory data carrier, such as an optical disc. Attached Figure Description

[0073] Further advantages and details of the invention will be obtained from the following embodiments, taken in conjunction with the accompanying drawings. In the drawings:

[0074] Figure 1 A flowchart illustrating an embodiment of the method according to the present invention is shown.

[0075] Figure 2 The schematic diagram of the X-ray equipment is shown, and

[0076] Figure 3 The functional structure of the control device of an X-ray equipment is shown. Detailed Implementation

[0077] Figure 1 A flowchart illustrating an embodiment of the method according to the invention is shown. In step S1, a set of projection images is provided, comprising multiple projection images of the patient's examination area, referred to as the number of projection images, through which contrast agent flows during the angiography imaging process. Here, no mask images are taken, meaning maskless angiography is performed. For example, there may be 20 to 40 projection images.

[0078] In step S2, a reference image is selected from the projected images in the projected image set. The projected image with the maximum contrast agent filling volume is selected as the reference image. This can be done manually by the user, but is preferably done through a specially trained selection function that can extract vascular features and / or evaluate the average intensity (image value) and / or the chromatograms therein, thereby selecting the projected image with the most image components identified as vessels.

[0079] In optional step S3, a first motion correction is performed, thereby performing rigid or affine 2D-2D registration between the other projected images and the reference image to reduce differences in motion states between these images, and to perform the first step to avoid motion artifacts. It should be noted that, in principle, it is also conceivable that this correction step has been performed before the reference image is selected, wherein all projected images are also moved to the reference motion state through 2D-2D registration.

[0080] In step S4, the weights of each projected image in the projected image set are determined based on the similarity between other projected images and the reference image. First, the weights for all other projected images, excluding the reference image, are determined. Here, an optimization process is used to minimize the deviation function of the deviation image, which is calculated by subtracting the sum of other projected images, which differ from the reference image and are determined by the weighted test set, from the reference image. The test set of weights that minimize the value of the deviation function is determined as the motion weights.

[0081] Here, optimization is performed based on quadratic programming according to formula (5). Q and c (formula (4)) are pre-calculated by analyzing scalar products according to formulas (6) and (7). Then, actual optimization is performed according to the Goldfarb-Idnani algorithm. A bias image is determined for the test set of motion weights, and the L2 norm is used as a bias metric for analysis.

[0082] After determining the weights of the other projected images, the weights of the reference image are added according to formula (8), which is the maximum value of the weights of the other projected images.

[0083] In step S5, at least one evaluation image is determined using the weights set in step S4. Here, the weights can be used to determine the selection of the projected image in the step, i.e., only the projected image with a weight higher than the limit value is selected (see formula (9)), or they can be used for summation, i.e., to form a weighted sum.

[0084] If too few projected images are selected when using the limit value, especially if the number of projected images is less than the minimum (e.g., 5 or 6), then step S4 is repeated to broaden the weight distribution. For this purpose, a renormalization term is added to the quadratic share of the optimization problem according to formula (10). This can be easily implemented because the optimization process can be performed computationally efficiently. Alternatively, the limit value can be reduced, for example, by decreasing μ until the minimum value is reached.

[0085] In step S5, for example, a DWVA image (see Equation (11)) or a WVMA image can be determined as at least one of at least one evaluation image. In the latter case, an initial mask image is first determined by extreme value projection of the projected image selected according to the limit value, and then a denoising virtual mask image is determined by a weighted sum of the mask image and the projected image. The weights consist of basic weights and motion weights, with the basic weights significantly weighted on the mask image, and the motion weights defined by the motion weights determined as additional weights in step S4. Finally, the WVMA image is determined by a weighted sum of the subtraction images formed by the projected image and the denoising mask image, respectively, using the weights from step S4.

[0086] These two evaluation images can form the basis of the flow information image, in which the flow information is displayed in color-coded form, and in order to determine the flow information, the projection image to be used can be selected based on weights according to the limit value or another limit value.

[0087] Finally, at least one definite evaluation image can be output.

[0088] Even though this embodiment describes a single reference image, multiple reference images can be selected, and weights and evaluation images can be determined for each reference image. The best image can be selected from these evaluation images, for example, manually, by a decision function, and / or based on image quality metrics.

[0089] Figure 2 A schematic diagram of an X-ray device 1 is shown, which includes a C-arm 2 with opposing X-ray sources 3 and X-ray detectors 4. The C-arm 2, fixed to a support 5, is movable, allowing for different projection geometries to be set for a patient positioned on a patient table 6.

[0090] The operation of X-ray equipment 1 is controlled by control device 7, which includes an evaluation device according to the present invention, the functional structure of which will be referred to... Figure 3 A detailed description is provided. The control device 7 includes a memory 8, which is also used in the evaluation device 9 according to the present invention.

[0091] The recording unit 10 of the control device 7 controls the recording operation, particularly the projected images of the projected image set, and can provide them to the selection unit 12 via the input interface 11 inside the evaluation device 9 to select a reference image according to step S2. In the optional correction unit 13, motion correction can be performed via 2D-2D registration according to optional step S3. The setting unit 14 is designed to determine weights according to step S4 and also has an optimization subunit 15 accordingly. In the determination unit 16, at least one evaluation image can be determined according to step S5.

[0092] At least one definitive evaluation image can be provided through the output interface 17 of the evaluation device 9, such as the output interface of the display unit 18 of the control device 7, and output to the display device of the X-ray equipment 1. Of course, the definitive evaluation image can also be stored in the memory 8 or otherwise transferred / stored.

Claims

1. A computer-executed method for evaluating a set of projected images of an X-ray apparatus (1) performing maskless angiography in an examination area having vascular structures, the method comprising the steps of: - Provides a set of captured projection images, which, as a subset of multiple projection images, show the flow of contrast agent through vascular structures over an elapsed flow period. - Select a reference image from multiple projected images. - Determine the weight for each projected image, wherein the weight of a projected image different from the reference image is determined based on the similarity between the projected image and the reference image, and the weight of the reference image is not lower than the maximum weight of other projected images. - Determine at least one evaluation image from multiple projected images, wherein the projected images are used according to determined weights during the determination process.

2. The method according to claim 1, characterized in that, The projection image with the maximum contrast agent filling volume, in particular the average value of the projection image and / or the chromatogram and / or the projection image determined based on user input, is selected as the reference image, and / or the reference image is determined by using a trained selection function.

3. The method according to claim 1 or 2, characterized in that, In the optimization process, the weights are determined to minimize the deviation function for the deviation image. The deviation image is calculated by subtracting the sum of other projected images different from the reference image from the reference image, as determined by the test set of the weights.

4. The method according to claim 3, characterized in that, The deviation function is minimized in the form of a one-dimensional convex quadratic programming problem, where the dimension is equal to the number of projected images minus one.

5. The method according to claim 4, characterized in that, The L2 norm is used as the bias function.

6. The method according to claim 5, characterized in that, Quadratic programming is expressed by the product of subtraction terms and the separation of quadratic and linear terms in the weight vector.

7. The method according to claim 5 or 6, characterized in that, To form a quadratic matrix of the quadratic term of the deviation function, the vector inner product of the vectorized other projected images is generated in the form of a quadratic programming problem. And to generate a vector of the linear term of the deviation function, the vector inner product of the vectorized reference image and the vectorized other projected images is generated in the form of a quadratic programming problem.

8. The method according to any one of claims 4 to 7, characterized in that, The optimization process is implemented using the Goldfarb-Idnani algorithm.

9. The method according to any one of the preceding claims, characterized in that, The weights are used when generating at least one weighted sum and / or when selecting a projection image for the determination step in the process of determining the evaluation image, wherein, in particular, only projection images that are above the extreme values ​​of the corresponding weights are selected.

10. The method according to claim 9, characterized in that, If fewer than the minimum number of projected images, especially 10% to 30%, are selected based on a preset extreme value, then either the extreme value is reduced until the minimum number of projected images is selected, or the weights are redefined, using a renormalization technique to broaden the distribution of the determined weights.

11. The method according to any one of the preceding claims, characterized in that, Before determining the weights, a motion correction process is performed on the projected images to determine the modification parameters, including rigid or affine 2D-2D registration between the projected images.

12. The method according to any one of the preceding claims, characterized in that, Multiple benchmark images are used, and at least one evaluation image is determined for each benchmark image, wherein the final evaluation image is selected in the process of selecting at least one evaluation image based on a trained evaluation function and / or image quality values, especially the largest vascular feature value and / or user selection information.

13. An evaluation device (9) for evaluating a set of projected images of an X-ray device (1) performing maskless angiography in an examination area with vascular structures, wherein, The evaluation device (9) includes: - Input interface (11) for receiving a set of projected images, the set of projected images showing the flow of contrast agent through the vascular structure during an elapsed flow period, using a certain number of projected images from a set of multiple projected images. - Selection unit (12) is used to select a reference image from multiple projected images. - Determining unit (14), used to determine the weight for each projected image, wherein the weight of a projected image different from the reference image is determined based on the similarity between the projected image and the reference image, and the weight of the reference image is not lower than the maximum weight of other projected images, and - Measurement unit (16) for measuring at least one evaluation image from multiple projected images, wherein the projected images are used according to determined weights during the measurement process.

14. A computer program product, when run in an evaluation apparatus (9), causes the evaluation apparatus (9) to perform the steps of the method according to any one of claims 1 to 12.

15. An electronically readable data carrier on which a computer program according to claim 14 is stored.

Images