Image reconstruction method and device based on orthogonal basis material decomposition

Abstract

The application provides an image reconstruction method and device based on orthogonal basis material decomposition, obtains scanning parameters of a CT device; obtains initial projection data and energy spectrum data of a single group of energy spectrum of a to-be-measured object; obtains a basis function representing attenuation when a multi-color X-ray scans the to-be-measured object; based on the scanning parameters, the energy spectrum data, the initial projection data and each basis function, iteratively reconstructs a coefficient image of each basis function with the aim of minimizing the sum of inner products between different coefficient images of the basis functions; and generates a pseudo single-energy image of the to-be-measured object based on the optimized coefficient images of each basis function obtained after the iteration. The application can eliminate hardening artifacts and / or metal artifacts to improve image quality and reduce hardware cost required for image reconstruction.

Description

Technical Field

[0001] This invention relates to the field of CT imaging technology, and in particular to an image reconstruction method and apparatus based on orthogonal basis material decomposition. Background Technology

[0002] X-ray computed tomography (CT) technology can non-destructively reproduce the internal structural information of an object under inspection and has been widely used in medical diagnosis, industrial non-destructive testing, reverse engineering, security inspection, and other fields. Image reconstruction algorithms are a key technology affecting the quality of CT imaging. In the field of image reconstruction, there are many fast and effective algorithms, such as filtered back-projection (FBP) and algebraic reconstruction technique (ART). However, these traditional image reconstruction algorithms usually assume that the X-rays passing through the object are of a single energy, which does not conform to the actual situation of CT systems. This is because X-ray photons in actual CT systems do not have only one energy but rather an energy spectrum with a certain width. The energy spectrum of X-rays changes before and after passing through the object under inspection, resulting in beam hardening. Especially when the object under inspection contains high-density metal, images reconstructed using traditional image reconstruction algorithms will exhibit banded hardening artifacts and / or striped bright and dark metallic artifacts, reducing image quality and thus affecting the accuracy of medical diagnosis and / or industrial inspection.

[0003] To address the inconsistency between multi-spectral and single-energy image reconstruction, one approach in existing image reconstruction techniques involves introducing spectral information and reconstructing a distribution image using initial projection data from dual (multi)-spectral sources. This distribution image is then fused into a pseudo-single-energy image, thus achieving artifact correction. However, this approach often requires initial projection data from dual (multi)-spectral sources. The hardware cost of dual (multi)-spectral CT equipment used to acquire this initial projection data is typically high, limiting the development of CT imaging technology. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide an image reconstruction method and apparatus based on orthogonal basis material decomposition, so as to alleviate or partially alleviate the above-mentioned problems existing in related technologies.

[0005] In a first aspect, embodiments of the present invention provide an image reconstruction method based on orthogonal basis material decomposition. The method includes: acquiring scanning parameters of a CT device; wherein the CT device includes a radiation source, a turntable, and a detector composed of multiple detector units, and the scanning parameters include the distance from the radiation source to the center of the turntable, the distance from the radiation source to the detector, the number of detector units contained in the detector, and the size of each detector unit; acquiring initial projection data and energy spectrum data of a single energy spectrum of the object under test; wherein the object under test is composed of at least two basis materials, and the initial projection data is obtained by the CT device using multicolor X-rays. The image is obtained by projecting the object under test using a line scan; basis functions representing the attenuation during multicolor X-ray scanning of the object under test are acquired; wherein, each basis function corresponds one-to-one with the type of base material of the object under test, and each basis function reflects the change of the attenuation coefficient of the corresponding type of base material with X-ray energy; based on the scanning parameters, the energy spectrum data, the initial projection data, and each basis function, the coefficient images of each basis function are iteratively reconstructed with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions; based on the optimized coefficient images of each basis function obtained after iterative reconstruction, a pseudo-monoenergy image of the object under test is generated.

[0006] Secondly, embodiments of the present invention also provide an image reconstruction apparatus based on orthogonal basis material decomposition. The apparatus includes: a first acquisition module for acquiring scanning parameters of a CT device; wherein the CT device includes a radiation source, a turntable, and a detector composed of multiple detector units, and the scanning parameters include the distance from the radiation source to the center of the turntable, the distance from the radiation source to the detector, the number of detector units contained in the detector, and the size of each detector unit; and a second acquisition module for acquiring initial projection data and energy spectrum data of a single energy spectrum of an object under test; wherein the object under test is composed of at least two basis materials, and the initial projection data is obtained by scanning with multicolor X-rays by the CT device. The image is obtained after projection imaging of the object under test; the third acquisition module is used to acquire the basis functions representing the attenuation during multicolor X-ray scanning of the object under test; wherein, the basis functions correspond one-to-one with the base material types of the object under test, and each basis function reflects the change of the attenuation coefficient of the corresponding base material with X-ray energy; the reconstruction module is used to iteratively reconstruct the coefficient images of each basis function based on the scanning parameters, the energy spectrum data, the initial projection data, and each basis function, with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions; the generation module is used to generate a pseudo-monoenergy image of the object under test based on the optimized coefficient images of each basis function obtained after iterative reconstruction.

[0007] This invention provides an image reconstruction method and apparatus based on orthogonal basis material decomposition. First, the scanning parameters of a CT scanner are acquired. Then, the initial projection data and energy spectrum data of a single energy spectrum of the object under test, as well as the basis functions corresponding to various basis materials of the object, are acquired. Next, based on the scanning parameters, energy spectrum data, initial projection data, and each basis function, the coefficient images of each basis function are iteratively reconstructed with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions. Finally, based on the optimized coefficient images of each basis function obtained after iterative reconstruction, a pseudo-monoenergy image of the object under test is generated. Using this technique, given the scanning parameters of the CT scanner, only the initial projection data and energy spectrum data of a single energy spectrum of the object under test, as well as the attenuation during multicolor X-ray scanning of the object, are needed to achieve image reconstruction of the initial projection data. This eliminates hardening artifacts and / or metal artifacts, thereby improving image quality and enhancing the accuracy of medical diagnosis and / or industrial testing. Furthermore, the relevant data for a single energy spectrum can be obtained using a single-energy CT scanner, resulting in lower hardware costs. Therefore, the application of this technique can reduce the hardware costs required for image reconstruction, thus promoting the development of CT imaging technology to a certain extent.

[0008] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0009] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0010] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0011] Figure 1 This is a schematic flowchart of an image reconstruction method based on orthogonal basis material decomposition in an embodiment of the present invention;

[0012] Figure 2 This is a flowchart illustrating the process of reconstructing coefficient images by iterating three times using three base materials as examples in an embodiment of the present invention.

[0013] Figure 3The images shown in this invention are: experimental multicolor energy spectrum, mass attenuation coefficient of material as a function of X-ray energy, pseudo-monoenergy image of experimental phantom, and cross-sectional view of pseudo-monoenergy image at the central gray vertical line.

[0014] Figure 4 This is a noise-free multicolor projection data reconstructed image and its cross-section diagram in an embodiment of the present invention;

[0015] Figure 5 The above are the coefficient image, marker image, and pseudo-monoenergy image obtained by iterating noise-free multicolor projection data three times in this embodiment of the invention.

[0016] Figure 6 This is a region map of the labeled image of the noise-free multicolor projection data in the second and third iterations in an embodiment of the present invention.

[0017] Figure 7 The figures shown are coefficient images, marker images, pseudo-monochromatic images, and region maps of the marker images obtained by adding morphological processing when iterating on noise-free multicolor projection data in this embodiment of the invention.

[0018] Figure 8 The graphs show the PSNR and NMAD of the pseudo-monochromatic image obtained by iterating noise-free multicolor projection data in an embodiment of the present invention as a function of the number of iterations.

[0019] Figure 9 The coefficient image, marker image, and pseudo-monochromatic image obtained by threshold segmentation and morphological processing respectively during the 20th iteration of noise-free multicolor projection data in this embodiment of the invention;

[0020] Figure 10 These are the cross-sections of the standard pseudo-monoenergy image, the pseudo-monoenergy image obtained by iterating noise-free multicolor projection data 20 times using only threshold segmentation, and the pseudo-monoenergy image obtained by adding morphological processing in the embodiments of the present invention.

[0021] Figure 11 This is a reconstructed image and its cross-section diagram of noisy multicolor projection data in an embodiment of the present invention;

[0022] Figure 12 The above are the coefficient image, marker image, and pseudo-monoenergy image obtained by iterating noisy multicolor projection data three times in this embodiment of the invention.

[0023] Figure 13 The graphs show the PSNR and NMAD of the pseudo-monochromatic image obtained by iterating on noisy multicolor projection data in an embodiment of the present invention as a function of the number of iterations.

[0024] Figure 14These are the coefficient image, marker image, and pseudo-monoenergy image obtained during the 10th iteration of noisy multicolor projection data in this embodiment of the invention.

[0025] Figure 15 These are cross-sectional views of the standard pseudo-monoenergy image and the pseudo-monoenergy image iterated 10 times over noisy multicolor projection data in the embodiments of the present invention.

[0026] Figure 16 This is a schematic diagram of an image reconstruction device based on orthogonal basis material decomposition in an embodiment of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Currently, to address the inconsistency between multi-spectral and single-energy image reconstruction, one approach in existing image reconstruction techniques involves introducing spectral information and reconstructing a distribution image using initial projection data from dual (multi)-spectral sources. This distribution image is then fused into a pseudo-single-energy image under a single energy level, thus achieving artifact correction. However, this approach often requires initial projection data from dual (multi)-spectral sources. The hardware cost of dual (multi)-spectral CT equipment used to acquire this initial projection data is typically high, limiting the development of CT imaging technology.

[0029] Based on this, the present invention provides an image reconstruction method and apparatus based on orthogonal basis material decomposition, which can alleviate or partially alleviate the above-mentioned problems existing in related technologies.

[0030] To facilitate understanding of this embodiment, a detailed description of an image reconstruction method based on orthogonal basis material decomposition disclosed in this invention will be provided first. This method is applicable to scenarios such as medical diagnosis and industrial inspection, and no specific scenario is limited here. See also Figure 1 As shown, the method may include the following steps:

[0031] Step S102: Obtain the scanning parameters of the CT equipment.

[0032] The CT equipment may include a radiation source and a turntable, as well as a detector composed of multiple detector units. The scanning parameters may include the distance from the radiation source to the center of the turntable, the distance from the radiation source to the detector, the number of detector units contained in the detector, and the size of each detector unit.

[0033] For CT equipment, since each X-ray forms a line segment (i.e., a path) between the emission source and the detector unit, when multicolor X-rays are projected onto the object under test, they form a set of paths. Each path in this set contributes to the projection imaging of the CT equipment. The initial projection data, as the image obtained after projection imaging, has the same number of pixels as the number of paths. Correspondingly, the scanning parameters of the CT equipment are actually the parameters of the CT equipment itself that can be directly obtained. What it affects is the contribution of each pixel in the initial projection data to the projection of the object under test along each path.

[0034] Step S104: Obtain the initial projection data and energy spectrum data of a single set of energy spectra of the object under test.

[0035] The object to be tested may be composed of at least two base materials, and the initial projection data is obtained by using a CT device to scan the object to be tested with multicolor X-rays to perform projection imaging.

[0036] In practical applications, single-energy spectral CT equipment can use multicolor X-rays to scan the object under test and perform projection imaging to obtain the initial projection data of the object under a single energy spectrum. The energy spectrum data is the known data of the single-energy spectral CT equipment.

[0037] Step S106: Obtain the basis function representing the attenuation during multicolor X-ray scanning of the object under test.

[0038] The aforementioned basis functions correspond one-to-one with the types of base materials of the object under test, and each basis function reflects the change of the attenuation coefficient of the corresponding type of base material with X-ray energy.

[0039] The aforementioned attenuation coefficient can be a linear attenuation coefficient, a mass attenuation coefficient, etc., and there is no limitation on this. For example, the aforementioned basis function can be selected as a function with X-ray energy as the variable and the attenuation coefficient as the function value; the selection method of the aforementioned basis function is not limited here.

[0040] Step S108: Based on the scanning parameters, energy spectrum data, initial projection data, and each basis function, the coefficient images of each basis function are iteratively reconstructed with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions.

[0041] Step S110: Based on the optimized coefficient images of each basis function obtained after iterative reconstruction, a pseudo-monoenergy image of the object under test is generated.

[0042] This invention provides an image reconstruction method based on orthogonal basis material decomposition. First, the scanning parameters of a CT scanner are acquired. Then, the initial projection data and energy spectrum data of a single energy spectrum of the object under test, as well as the basis functions corresponding to various basis materials of the object, are acquired. Next, based on the scanning parameters, energy spectrum data, initial projection data, and each basis function, the coefficient images of each basis function are iteratively reconstructed with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions. Finally, based on the optimized coefficient images of each basis function obtained after iterative reconstruction, a pseudo-monoenergy image of the object under test is generated. Using this technique, given the scanning parameters of the CT scanner, only the initial projection data and energy spectrum data of a single energy spectrum of the object under test, as well as the attenuation during multicolor X-ray scanning of the object, are needed to achieve image reconstruction of the initial projection data. This eliminates hardening artifacts and / or metal artifacts, thereby improving image quality and enhancing the accuracy of medical diagnosis and / or industrial testing. Furthermore, the relevant data for a single energy spectrum can be obtained using a single-energy CT scanner, resulting in lower hardware costs. Therefore, the application of this technique can reduce the hardware costs required for image reconstruction, thus promoting the development of CT imaging technology to a certain extent.

[0043] As one possible implementation, step S108 (i.e., iteratively reconstructing the coefficient images of each basis function based on scanning parameters, energy spectrum data, initial projection data, and each basis function, with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions) may include:

[0044] Step 1: Based on the energy spectrum data and each basis function, generate the energy spectrum sequence of the energy spectrum data and the decay sequence of each basis function; wherein, both the energy spectrum sequence and the decay sequence are composed of multiple discrete elements, and the element positions of the energy spectrum sequence correspond one-to-one with the element positions of the decay sequence.

[0045] Step 2: Based on the scanning parameters, generate the projection vector for the CT equipment to perform projection imaging of the object under test by scanning with multicolor X-rays; wherein, the projection vector represents the contribution of each pixel position of the coefficient image to the initial projection data.

[0046] For example, each element in a projection vector can be used to characterize the contribution of the corresponding pixel position of the coefficient image to the projection along the corresponding X-ray, thereby characterizing the contribution of each pixel position of the coefficient image to the projection along different X-rays through different projection vectors.

[0047] Step 3: Based on the projection vector, initial projection data, energy spectrum sequence, and various decay sequences, the coefficient images of each basis function are iteratively reconstructed with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions.

[0048] As one possible implementation, step 1 above (i.e., generating the energy spectrum sequence of the energy spectrum data and the decay sequence of each basis function based on the energy spectrum data and each basis function) may include:

[0049] Step a1: Sample the energy spectrum data and each basis function using the same sampling parameters to obtain the energy spectrum sampling values ​​of the energy spectrum data and the basis function sampling values ​​of each basis function.

[0050] For example, the effective energy range of the energy spectrum data can be divided into multiple energy intervals of equal length; wherein the length of each energy interval is a preset length value; then, the energy spectrum data and each basis function are sampled in each energy interval to obtain the energy spectrum sampled value of the energy spectrum data corresponding to each energy interval and the basis function sampled value of each basis function corresponding to each energy interval.

[0051] Step a2: Combine all the energy spectrum sample values ​​to form the energy spectrum sequence, and combine the basis function sample values ​​of each basis function to form a decay sequence, thus obtaining the energy spectrum sequence and the decay sequence of each basis function.

[0052] As one possible implementation, step 3 above (i.e., based on the projection vector, initial projection data, energy spectrum sequence, and various decay sequences, iteratively reconstructing the coefficient images of each basis function with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions) may include:

[0053] (1) Initialize the coefficient images, first label image and second label image of each basis function.

[0054] In this context, the pixel position distribution of the coefficient image, the first marker image, and the second marker image of each basis function is the same. The first marker image represents the location region of the basis material on the corresponding coefficient image, and the second marker image represents the location region of the non-basis material on the corresponding coefficient image.

[0055] (2) For each iteration, perform the following steps:

[0056] A. Based on the above projection vector, the above preset length value, the energy spectrum sequence and each decay sequence, and the coefficient image of each basis function used in this iteration, determine the estimated data of the object under test corresponding to this iteration.

[0057] The estimated data mentioned above may include estimated projection data and weighting coefficients.

[0058] For example, for a certain iteration, after obtaining the coefficient images of each basis function used in that iteration, the estimated projection data and weighting coefficients of the object under test corresponding to that iteration can be calculated based on the above projection vector, the above preset length value, the energy spectrum sequence and the corresponding decay sequence, and the coefficient images of each basis function used in that iteration.

[0059] B. Based on the above scanning parameters, the above initial projection data, the estimated data of the object under test corresponding to this iteration, and the first and second labeled images of the corresponding basis functions used in this iteration, the coefficient images of the corresponding basis functions used in this iteration are reconstructed to obtain the reconstructed coefficient images of each basis function corresponding to this iteration.

[0060] C. Perform image segmentation on the reconstructed coefficient images of each basis function corresponding to this iteration to obtain the first and second labeled images of each basis function used in the next iteration.

[0061] D. Based on the reconstructed coefficient images of each basis function corresponding to this iteration and the first labeled images of each basis function used in the next iteration, determine the coefficient images of each basis function used in the next iteration.

[0062] (3) The iteration ends when the preset iteration end condition is met, and the reconstructed coefficient image of each basis function at the end of the iteration is determined as the optimized coefficient image of the corresponding basis function.

[0063] Among them, satisfying the preset iteration termination condition may include: the number of iterations reaches the preset number, and / or, the sum of the inner products between the coefficient images of different basis functions is less than a preset threshold.

[0064] As one possible implementation, step 3 above may further include: determining the sum of the inner products between the coefficient images of different basis functions corresponding to each iteration based on the coefficient images of each basis function used in each iteration.

[0065] For example, for each iteration, the sum of the inner products between the coefficient images of different basis functions corresponding to that iteration can be determined based on the coefficient images of each basis function used in that iteration; and so on, the sum of the inner products between the coefficient images of different basis functions corresponding to each iteration can be obtained in the same way, so that the reconstructed coefficient image with the smallest sum of the inner products between the coefficient images of different basis functions is selected as the optimal reconstructed coefficient image (i.e., the optimized coefficient image), so that the optimized coefficient image can be used for medical diagnosis, industrial inspection, etc. in the future.

[0066] For ease of understanding, the operation of steps 1 to 3 above will be described as an example using a specific application as an example below.

[0067] In practical applications, neglecting scattering, the projection data obtained after multicolor X-rays pass through an object (composed of at least two base materials) can be represented as an image composed of multiple pixels. Each pixel in this image can be represented in the following form:

[0068]

[0069] in, It is the set of paths formed by multicolor X-rays as they pass through an object, where L is the path of the i-th X-ray, and p i is a pixel generated by the projection of the i-th X-ray through the object (i.e., a pixel contained in the projection data), where E is the energy of the photon in the X-ray, μ n (E) is a basis function that depends only on the X-ray energy and is a known quantity; f n (x) represents the distribution coefficient of the corresponding basis function in the space of the detected object, which is an unknown quantity; N is the number of basis functions, the value of which is equal to the number of base materials of the object through which the multicolor X-rays pass; n is the index of the base material of the object through which the multicolor X-rays pass; S(E) is the normalized equivalent X-ray energy spectrum (i.e., energy spectrum data). When N≥2, the basis function μ n (E) can be selected as the mass attenuation coefficient of the nth base material, so that the change of the mass attenuation coefficient of the base material with X-ray energy can be reflected by the basis function.

[0070] According to equation (1) above, the problem of single-energy spectral CT multi-base material decomposition image reconstruction is transformed into the following problem: given the projection data and energy spectrum data, find the coefficient image f of each basis function. n (x), n=1, 2,..., N, N≥2.

[0071] Consider the discrete form of equation (1) above. In numerical implementation, the coefficient graph f of each basis function can be used. n (x), n = 1, 2, ..., N, are expressed in the form of a vector (i.e., a discrete image):

[0072] f n =(f n，1 f n，2 , ...f n，J ) T (2)

[0073] Among them, f n It is achieved by plotting the coefficients of the nth basis function f. n (x) is a discrete image obtained after sampling, where each element is f. n (x) is the sampled value on a single pixel; f n，j It is f n (x) is the sampled value at the j-th pixel, where J is the discrete image f.n Let R be the number of pixels, and T represent the transpose of the vector or matrix; let R be the number of pixels. i =(r i，1 r i，2 ,...,r i，J ) represents the projection vector corresponding to the i-th ray, where r i，j Represents a discrete image f n The contribution of the j-th pixel to the projection of the object along the i-th ray. The effective energy range of the X-ray spectrum is divided into M equal parts (i.e., M energy intervals of equal length), with the length of each part denoted as δ. Within each energy interval, the contribution of S(E) and μ... n S(E), n = 1, 2, ..., N are sampled, and S(E) and μ are compared. n (E) The sampled values ​​in each energy range are S(E) and μ. n (E) are their respective approximate values ​​within each energy range. Therefore, S(E) and μ n (E) is discretized into the following two vectors:

[0074] (S1, S2, ..., S m S M ) T , (μ n，1 μ n2 μ n2 , ..., μ n，m , ..., μ n，M ) T (3)

[0075] Among them, S m and μ n，m S(E) and μ respectively n (E), n = 1, 2, ..., N, the sampled values ​​in the m-th energy interval.

[0076] Based on the above symbol definitions, the discrete form of a single-energy spectrum multicolor projection (i.e., the initial projection data of a single energy spectrum) can be represented as a discrete image composed of multiple discrete pixels. Each discrete pixel in this discrete image can be represented in the following form:

[0077]

[0078] After discretization, the problem of reconstructing multi-matrix material decomposition images in single-energy spectral CT is transformed into the following problem: Given p i , Find f n For n = 1, 2, ..., N, N ≥ 2, this is a problem of solving a system of nonlinear equations.

[0079] Solving the single-energy spectral CT multi-matrix material decomposition image reconstruction problem determined by the above equation (4) is actually solving for N≥2 coefficient images from a set of projection data. Since the solution to this problem is not unique, constraints need to be added to determine the true solution in a physical sense. Solving this problem in many practical detection situations requires satisfying the following two points: First, the material of the substance in the detected object is known, that is, the types of matrix materials contained in the detected object and their inherent parameters (such as density, attenuation coefficient, etc.) are known; Second, the substances in the detected object are not mixed, that is, when the substance is exactly the nth matrix material in the above equation (4), then the area occupied by this matrix material does not contain other types of matrix materials (that is, the coefficient image of this matrix material is orthogonal to the coefficient images of other types of matrix materials).

[0080] The orthogonality between the coefficient graphs of different matrix materials can be expressed by the following formula:

[0081] f n ·f k =0, n≠k (5)

[0082] Where · represents the inner product of two vectors.

[0083] Based on the above phenomena, the optimization problem for reconstructing orthogonal basis material decomposition images in single-energy spectral CT can be further proposed as follows: to make the coefficient images (i.e., discrete images) of different basis functions as close as possible to the orthogonality relationship. The objective of this optimization problem is expressed in the form of the following objective function:

[0084]

[0085] To facilitate the numerical realization of the objective function (6), an image O representing the location region of the nth base material on the corresponding coefficient image is introduced through image segmentation (indicative function). n The labeled image H of the position region on the corresponding coefficient image corresponding to the non-nth base material. n O n and H n with f n Homogeneous. Theoretically, H n =I f -O n , where I f Is with f n Vectors of the same type, all equal to 1; however, in practice, it is difficult to directly segment and obtain accurate O values ​​in CT images. n H n Therefore, O can be set through image segmentation. n H n The region is slightly larger than the actual location regions of the corresponding base material and the corresponding non-base material on the corresponding coefficient image. This is why, in the iterative solution of f... nWhen processing, the region can be divided into the following three regions: ① Region O that must belong to the nth base material. n ⊙(I f -H n ), ② Region O that may contain both the nth base material and other base materials n ⊙H n (Boundary portion), ③ The region that definitely does not belong to the nth base material (I) f -O n )⊙H n , where ⊙ represents the Hadamard product of two vectors.

[0086] In practical applications, the single-energy spectral CT multi-base material decomposition image reconstruction problem determined by the above formula (4) can be solved iteratively based on the objective function (6) to realize the iterative reconstruction of the coefficient images of each basis function, thereby generating a pseudo-single-energy image of the object under test based on the reconstructed images of each basis function obtained after iterative reconstruction.

[0087] In practical applications, the algorithm for iteratively reconstructing the coefficient images of each basis function (referred to as the iterative solution algorithm) can be pre-written into corresponding program code and saved so that the coefficient images of each basis function can be iteratively reconstructed by executing the program code through electronic devices.

[0088] In addition, for the convenience of data processing, the indices of the base materials in the above iterative solution algorithm can be sorted in descending order of their attenuation coefficients, that is, the attenuation coefficient of the nth base material is greater than that of the (n+1)th base material. Based on this, the pseudocode form corresponding to the program code of the iterative solution algorithm for the optimization problem proposed in equation (6) above is as follows:

[0089]

[0090]

[0091] For noisy images in practice, after performing the "update reconstructed coefficient image" operation in the pseudocode above, you can add denoising processing to the reconstructed coefficient image.

[0092] Figure 2 The process of iterating through three coefficient images using three base materials as examples is demonstrated. Figure 2First, initial projection data 211 and energy spectrum data 212 of a single set of energy spectra of the object under test (composed of three base materials) and basis functions of the three base materials of the object under test (at this time, the attenuation coefficients of the three base materials change with X-ray energy) 213 are acquired using a single-energy spectroscopy device. These data are used as fixed data 21 for reconstructing coefficient images in three iterations. In the first iteration, the fixed data 21 are used to perform existing algebraic reconstruction. The Technique (ART) algorithm directly reconstructs coefficient images 221, 222, and 223 of the three base materials of the test object as the reconstructed coefficient image data 22 for the first iteration. In the second iteration, the reconstructed coefficient image data 22 from the first iteration is segmented to obtain the corresponding labeled image data 23. Since only the image segmentation results of coefficient image 221 (i.e., labeled images 231 and 232) are available, two copies of labeled image 232 can be directly copied as the image segmentation results of coefficient image 222 (i.e., labeled images 233 and 234). Thus, labeled image data 23 is composed of labeled images 231, 232, 233, and 234. The reconstructed coefficient image for the second iteration is then calculated using the fixed data 21 and the labeled image data 23. Data 24 (i.e., coefficient images 241, 242, and 243 corresponding to coefficient images 221, 222, and 223, respectively); during the third iteration, the reconstructed coefficient image data 24 from the second iteration is first segmented to obtain the corresponding labeled image data 25. Since only the image segmentation results of coefficient image 241 (i.e., labeled images 251 and 252) and the image segmentation results of coefficient image 242 (i.e., labeled images 253 and 254) can be obtained, the labeled image data 25 can be composed of labeled images 251, 252, 253, and 254. Then, the reconstructed coefficient image data 26 from the third iteration (i.e., coefficient images 261, 262, and 263 corresponding to coefficient images 241, 242, and 243, respectively) is calculated using fixed data 21 and labeled image data 25.

[0093] To verify the effectiveness of the image reconstruction method based on orthogonal basis material decomposition, an oral cavity model was used in the experiment to exemplarily verify the effectiveness of the method in correcting metal artifacts. Both noise-free and Poisson noise data were processed in the experiment. To quantify the image quality, the peak signal-to-noise ratio (PSNR) and normalized mean absolute deviation (NMAD) between the resulting image (i.e., the pseudo-monotropic image, or reconstructed image) and the real image were calculated. The formulas for calculating PSNR and NMAD are as follows:

[0094]

[0095] Where Y j j = 1, ..., J is the sum of the estimated values. j = 1, ..., where J is the actual value.

[0096] For simplicity, the above experiments were limited to fan-beam CT systems so that the experimental process can be easily extended to cone-beam CT systems in the future.

[0097] The experimental procedure is described in the following parts:

[0098] (I) Experimental conditions.

[0099] The computer used in the experiment was equipped with an Intel Xeon Silver 4210R CPU, 64GB of system memory, and an NVIDIA RTX A6000 graphics card with 48GB of video memory. The experimental code was written in C++, CUDA, and MATLAB.

[0100] The energy spectrum used in the experiment was obtained by simulation using the open-source software SpectrumGUI, such as... Figure 3 Figure (a) shows the simulated spectrum of a GE Maxiray 125 X-ray tube with a 0.1 mm copper filter added at a tube voltage of 140 kV. The 60-keV pseudo-monoenergy image of the phantom used in the experiment is shown below. Figure 3 As shown in (c) (the image display window is [0, 1] cm), -1 It contains three materials (i.e., base materials): water, bone (simulating teeth), and AgHg alloy (simulating tooth filling), each with a standard density ρ of 1 g / cm³. 3 1.92g / cm 3 and 12g / cm 3 The mass attenuation coefficients of the three materials as a function of X-ray energy can be obtained from the NIST website, such as... Figure 3 As shown in (b), AgHg is selected as the first basis material, bone as the second basis material, and water as the third basis material. When the mass attenuation coefficients of the three materials are used as basis functions, the densities of the three materials in the reconstructed image are the standard densities of the three materials. Figure 3 (d) shows a cross-sectional view of the 60-keV pseudo-monoenergy image of the phantom used, at the central gray vertical line (from top to bottom).

[0101] The scanning parameters of the CT system used in the experiment (i.e., single-energy spectral CT equipment) were set as follows: the distance from the X-ray source to the center of the turntable was 437 mm, and the distance from the X-ray source to the detector was 700 mm. The detector consisted of 1920 detector elements, each with a size of 0.127 mm. These parameters were designed to provide the single-energy spectral CT equipment with a suitable field of view (FOV). At this time, the FOV was 150 mm. In the experiment, the CT system rotated one revolution to perform multicolor X-ray scanning sampling at 720 angles on the phantom used in the experiment. The diameter of the water tray in the phantom used in the experiment was 84 mm.

[0102] (I) Noise-free data experiment

[0103] This section introduces experiments on processing noise-free multicolor projection data.

[0104] Figure 4 The image reconstruction results are shown by directly using the ART algorithm based on the multicolor projection data (i.e., the image reconstructed from the multicolor projection data) and the multicolor projection data reconstructed image in the above... Figure 3 The cross-sectional view at the central gray vertical line shown in (c) is as follows, Figure 4 Image (a) shows the reconstructed image from the multicolor projection data. Figure 4 (b) shows the cross-sectional view. Due to the polychromatic nature of the X-ray spectrum and the high attenuation of metals, there are obvious metal artifacts in the water and bone portions of the reconstructed image from the polychromatic projection data.

[0105] Figure 5 The image shows the coefficient image, marker image, and pseudo-monoenergy image obtained by iterating the noise-free multicolor projection data three times using the above iterative solution algorithm. Except for the image... The display window is [0, 0.05] cm. -1 Apart from that, the display window for all other images is [0, 1] cm. -1 In this experiment, a threshold was used. segmentation get Use threshold segmentation get Figure 5 The first row, from left to right, displays the coefficient images of AgHg, bone, and water obtained in the first iteration, along with a 60-keV pseudo-monoenergy image composed of the three coefficient images. Because the mass decay coefficients of AgHg differ significantly from those of bone and water, in... The grayscale values ​​of the AgHg region in the image differ significantly from those of the bone and water regions. Therefore, a relatively accurate labeled image can be obtained through simple global thresholding segmentation (e.g., Figure 5 In (As shown); because the bone coefficient image has large artifacts in this iteration, the labeled bone image has not yet been processed, so it is directly copied. Two copies serve as two marker images of the bone (e.g.) Figure 5 In (As shown).

[0106] Figure 5 The third row in the middle, from left to right, displays the coefficient images of AgHg, bone, and water obtained in the second iteration, as well as a 60-keV pseudo-monoenergy image composed of the three coefficient images. In this iteration, the coefficient image of bone (e.g.) Figure 5 In In the image shown, the bone region is already relatively clear, but the shadows within the bone region caused by metal artifacts cannot disappear, thus making the bone marker image (such as...) unclear. Figure 5 In The image shown is inaccurate. Furthermore, observation of the image reveals that adjusting the threshold cannot fundamentally improve the accuracy of the labeled image. In the pseudo-monoenergy image obtained in this iteration, metal artifacts in the water region remain severe.

[0107] Figure 5 The last line shows the coefficient images of AgHg, bone, and water obtained from the third iteration, as well as a 60-keV pseudo-monoenergy image composed of the three coefficient images. The labeled images of bone and water obtained in this iteration (e.g.) Figure 5 middle The image shown is inaccurate, but in the pseudo-monotropic image composed of labeled images, the artifacts in the water region are almost completely corrected, while some artifacts at the metal boundaries in the bone region remain.

[0108] Figure 6 The region maps of the labeled images in the 2nd and 3rd iterations of the noise-free multicolor projection data are given, where, Figure 6 Figure (a) shows the region breakdown of the labeled image in the second iteration. Figure 6 (b) in the figure gives the region map of the labeled image in the third iteration. The dark gray between the darkest gray and the lightest gray represents the region that belongs only to the base material 1 (AgHg), the darkest gray represents the region that belongs only to the base material 2 (bone), the lightest gray represents the region that may belong to either the base material 2 or the base material 3 (water), and black represents the region that belongs only to the base material 3. Figure 6 As can be seen from this, what affects the accuracy of the decomposition of the base material is that the portion outside the metal boundary that originally belonged to the bone region was allocated to the water region.

[0109] To address the aforementioned issues affecting the accuracy of base material decomposition, morphological dilation is applied when obtaining the labeled bone image from the segmented bone coefficient image. Specifically, after thresholding, a disk structure image with a radius of 2 is added to dilate the labeled image, resulting in the labeled image and coefficient image as shown below. Figure 7 As shown, the image display window is [0,1]cm. -1 Observe the labeled image (i.e. Figure 7 In ),as well as Figure 7 As shown in the segmented image (a), adding morphological processing can improve the segmentation of the portion outside the water boundary within the bone region in the bone-labeled image, while worsening the segmentation of the portion outside the bone boundary within the water region in the water-labeled image. These phenomena are observed in the coefficient images of bone and water (i.e., Figure 7 In This is reflected in ( ). and Figure 5 Compared to (c) in the middle, Figure 7 The metal artifacts in the bone region in the pseudo-monoenergy image shown in (b) have been further corrected.

[0110] To quantify the effectiveness of the iterative solution algorithm, the PSNR and NMAD of the pseudo-monoenergetic image obtained after each iteration were calculated. Figure 8 The graphs showing the PSNR and NMAD of a pseudo-monoenergy image obtained by iterating over noise-free multicolor projection data with the number of iterations are presented. Figure 8 Figure (a) shows the PSNR curve of a pseudo-monochromatic image obtained by iterating over noise-free multicolor projection data as a function of the number of iterations. Figure 8 Figure (b) shows the NMAD curves of pseudo-monoenergy images obtained by iterating over noise-free multicolor projection data as a function of iteration number. The dashed lines represent the PSNR and NMAD curves of pseudo-monoenergy images obtained by threshold segmentation alone as a function of iteration number, while the solid lines represent the PSNR and NMAD curves of pseudo-monoenergy images obtained by adding morphological processing as a function of iteration number. Observation Figure 8 It can be seen that the above iterative solution algorithm can converge within a finite number of iterations, and the addition of morphological processing does not improve the quality of the final pseudo-monoenergy image. This is because the morphological processing degrades the base material decomposition effect of the bone boundary.

[0111] Figure 9 The image shows the coefficient image, labeled image, and pseudo-monoenergy image obtained after the 20th iteration of noise-free multicolor projection data, after thresholding only and after adding morphological processing. Figure 9The first row in the middle, from left to right, shows a marker image of each of the base material 1 and base material 2 obtained by threshold segmentation only when performing the 20th iteration on the noise-free multicolor projection data, as well as the coefficient image of base material 3 and the 60-keV pseudo monoenergetic image composed of the coefficient images; Figure 9 The second row, from left to right, shows a labeled image of each of the two base materials, 1 and 2, obtained after morphological processing was added to the noise-free multicolor projection data during the 20th iteration, as well as the coefficient image of base material 3 and a 60-keV pseudo-monoenergy image composed of the coefficient images. The image display window is [0, 1] cm. -1 .exist Figure 9 In the first row of images, the reconstruction results of the matrix material decomposition image at the water and bone boundaries are better; in the second row of results, the reconstruction results of the matrix material decomposition image at the AgHg and bone boundaries are better.

[0112] Figure 10 The image displays profiles and magnified views of the standard pseudo-monoenergetic image, the pseudo-monoenergetic image obtained after 20 iterations of threshold segmentation using noise-free multicolor projection data, and the pseudo-monoenergetic image obtained after morphological processing. Figure 10 (a) shows the profiles of the standard pseudo-monoenergy image, the pseudo-monoenergy image obtained by iterating through noise-free multicolor projection data 20 times using only threshold segmentation, and the pseudo-monoenergy image obtained after adding morphological processing. Figure 10 (b) shows Figure 10 (a) is a magnified view of the area at box 101. Observation shows that, compared with the pseudo-monoenergy image obtained by adding morphological processing, the pseudo-monoenergy image obtained by thresholding segmentation alone is more effective in artifact correction in the water region; while the pseudo-monoenergy image obtained by adding morphological processing is more effective in artifact correction in the bone region.

[0113] (II) Experiment with Noisy Data

[0114] This section describes an experiment on processing noisy multicolor projection data, where the noise level is equivalent to Poisson noise at 105 photons per ray. Figure 11 The image shown is a directly reconstructed image and its profile obtained by directly reconstructing noisy multicolor projection data using the ART algorithm. Figure 11 Image (a) shows the directly reconstructed image. Figure 11 (b) shows the directly reconstructed image in the above... Figure 3 As shown in the cross-section at the central gray vertical line in (c), it can be seen that there are severe metal artifacts and noise in the directly reconstructed image.

[0115] Figure 12The image shows the coefficient image, marker image, and pseudo-monoenergetic image after three iterations of the noisy multicolor projection data using the aforementioned iterative solution algorithm. The coefficient image of base material 1 is shown. The display window is [0, 0.8] cm. -1 Coefficient graphs of base material 2 and base material 3 The display window is [0, 0.05] cm. -1 The display window for the remaining images is [0, 1] cm. -1 In this experiment, after obtaining the coefficient image each time, the total variation (TV) image denoising algorithm was used to denoise the coefficient image. Image denoising processing is performed, and the parameters of the TV image denoising algorithm are selected (6, 3); a threshold is used. segmentation get Use threshold segmentation get Figure 12 The first row, from left to right, shows the coefficient images of base material 1 and base material 2 obtained after the first iteration, as well as the 60-keV pseudo-monoenergy image composed of the coefficient images. The coefficient image of base material 1... The grayscale values ​​of AgHg in the middle limb differ significantly from those of bone and water, thus allowing for relatively accurate labeling images to be obtained through simple global thresholding (e.g., Figure 12 In (As shown); because the bone coefficient image has large artifacts in this iteration, the labeled bone image has not yet been processed, so it is directly copied. Two copies serve as two marker images of the bone (e.g.) Figure 12 In (As shown).

[0116] Figure 12 The third row, from left to right, shows the coefficient images of matrix material 1 and matrix material 2 obtained after the second iteration, as well as the 60-keV pseudo-monoenergy image composed of the coefficient images. The coefficient image of bone obtained after this iteration (e.g.) Figure 12 In As shown in the image, the bone region is already relatively clear, and the noise reduction process has reduced the degree of metal artifacts to some extent, thus making the bone-labeled image (such as...) clearer. Figure 12 In The results shown are compared to those corresponding to noise-free multicolor projection data (as shown). Figure 5 In The image shown is closer to the true value, but the presence of metal artifacts still prevents the labeled image from being accurately obtained through simple thresholding.

[0117] Figure 12 The last row, from left to right, shows the coefficient images of matrix material 1 and matrix material 2 obtained after the third iteration, as well as the 60-keV pseudo-monoenergy image composed of the coefficient images. Although the coefficient images of bone and water obtained in this iteration (such as...) Figure 12 In (As shown) is inaccurate, but in cases such as Figure 12 In the pseudo-monochromatic image shown in (c), the artifacts in the water and bone regions are almost completely corrected. Because there is noise in the multicolor projection data, and the iteration process amplifies the noise, there is still obvious noise in the reconstruction result of the base material decomposition image of this iteration, despite the denoising process.

[0118] Figure 13 The graphs showing the PSNR and NMAD of the pseudo-monochromatic image obtained by iterating over the noisy multicolor projection data are presented as a function of the number of iterations. Figure 13 Figure (a) shows the PSNR of the pseudo-monochromatic image obtained by iterating over the noisy multicolor projection data as a function of the number of iterations. Figure 12 Figure (b) shows the NMAD curve of the pseudo-monochromatic image obtained by iterating over the noisy multicolor projection data as a function of the number of iterations. Observation Figure 12 It can be seen that the above iterative solution algorithm can converge within a finite number of iterations. However, due to the presence of noise and denoising processing, the quality of the final pseudo-monoenergy image is worse than that of the final pseudo-monoenergy obtained under the noise-free condition.

[0119] Figure 14 The image displays the coefficient image, marker image, and pseudo-monoenergy image obtained after the 10th iteration of the iterative solution algorithm for noisy multicolor projection data. The image display window is [0,1]cm. -1 , Figure 14 In the graph, (a), (b), and (c) represent the coefficient graphs of AgHg, bone, and water, respectively. Figure 14 In the figure, (d) represents the 60-keV pseudo-monoenergy image. The results of the base material decomposition image reconstruction demonstrate the effectiveness of the above iterative solution algorithm in correcting metal artifacts in noisy data. Figure 14 In the above, (d) indicates that the pseudo-monopotent image is in the above... Figure 3 The cross-sectional view at the central gray vertical line shown in (c) is compared to the one shown in the image. Figure 11 The image shown in (a) is a direct reconstruction image obtained by directly reconstructing the noisy multicolor projection data using the ART algorithm. Artifacts in the water area can be eliminated, but some noise remains.

[0120] Figure 15The diagrams show the cross-sections of the standard pseudo-monoenergy image and the pseudo-monoenergy image obtained after iterating through noisy multicolor projection data 10 times. Observation shows that the pseudo-monoenergy image obtained by iterating through noisy multicolor projection data using the above iterative solution algorithm has a better artifact correction effect in the bone and water regions.

[0121] The above experiments verified the effectiveness of the image reconstruction method based on orthogonal basis material decomposition in eliminating hardening artifacts and / or metal artifacts in images. This method can obtain the required projection and energy spectrum data using only a single-energy spectral CT device, thereby achieving basis material decomposition image reconstruction and eliminating hardening and / or metal artifacts. Compared to using dual-energy (multi-energy) spectral CT devices to acquire dual-energy (multi-energy) spectral projection data for image basis material decomposition reconstruction, this method requires lower hardware costs, thus reducing the overall hardware cost of basis material decomposition image reconstruction and potentially promoting the development of CT imaging technology. Furthermore, this image reconstruction method based on orthogonal basis material decomposition can be applied not only to single-energy spectral CT orthogonal basis material decomposition image reconstruction but also to dual-energy (multi-energy) spectral CT orthogonal basis material decomposition image reconstruction.

[0122] Based on the above-described image reconstruction method based on orthogonal basis material decomposition, this invention also provides an image reconstruction apparatus based on orthogonal basis material decomposition. (See attached image reconstruction apparatus.) Figure 16 As shown, the device may include the following modules:

[0123] The first acquisition module 1602 is used to acquire scanning parameters of a CT device; wherein, the CT device includes a radiation source and a turntable and a detector composed of multiple detector units, and the scanning parameters include the distance from the radiation source to the center of the turntable, the distance from the radiation source to the detector, the number of detector units contained in the detector, and the size of each detector unit;

[0124] The second acquisition module 1604 is used to acquire the initial projection data and energy spectrum data of a single energy spectrum of the object under test; wherein, the object under test is composed of at least two base materials, and the initial projection data is obtained by the CT device after scanning the object under test with multicolor X-rays to perform projection imaging;

[0125] The third acquisition module 1606 is used to acquire the basis functions representing the attenuation when scanning the object under test with multicolor X-rays; wherein, the basis functions correspond one-to-one with the base material types of the object under test, and each basis function reflects the change of the attenuation coefficient of the corresponding base material with X-ray energy.

[0126] The reconstruction module 1608 is used to iteratively reconstruct the coefficient images of each basis function based on the scanning parameters, the energy spectrum data, the initial projection data, and each basis function, with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions;

[0127] The generation module 1610 is used to generate a pseudo-monoenergy image of the object under test based on the optimized coefficient images of each basis function obtained after iterative reconstruction.

[0128] Using the aforementioned image reconstruction device based on orthogonal basis material decomposition, after knowing the scanning parameters of the CT equipment, it is only necessary to acquire the initial projection data and energy spectrum data of a single energy spectrum of the object under test, as well as the attenuation of the object under test during multicolor X-ray scanning. This allows for image reconstruction of the initial projection data, thereby eliminating hardening artifacts and / or metal artifacts, thus improving image quality and enhancing the accuracy of medical diagnosis and / or industrial testing. Furthermore, the relevant data of a single energy spectrum can be obtained using a single-energy spectral CT device, resulting in lower hardware costs. Therefore, the application of the above technology can reduce the hardware costs required for image reconstruction, thereby promoting the development of CT imaging technology to a certain extent.

[0129] The reconstruction module 1608 described above can also be used to: generate an energy spectrum sequence of the energy spectrum data and a decay sequence of each basis function based on the energy spectrum data and each basis function; wherein the energy spectrum sequence and the decay sequence are both composed of discrete multiple elements, and the element positions of the energy spectrum sequence correspond one-to-one with the element positions of the decay sequence; generate a projection vector for projection imaging of the object under test by a CT device using multicolor X-rays based on the scanning parameters; wherein the projection vector represents the contribution of each pixel position of the coefficient image to the initial projection data; and iteratively reconstruct the coefficient images of each basis function based on the projection vector, the initial projection data, the energy spectrum sequence, and each decay sequence, with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions.

[0130] The reconstruction module 1608 described above can also be used to: sample the energy spectrum data and each basis function according to the same sampling parameters to obtain the energy spectrum sampling value of the energy spectrum data and the basis function sampling value of each basis function; form the energy spectrum sequence by combining all the energy spectrum sampling values, and form an attenuation sequence by combining the basis function sampling values ​​of each basis function to obtain the energy spectrum sequence and the attenuation sequence of each basis function.

[0131] The reconstruction module 1608 described above can also be used to: divide the effective energy range of the energy spectrum data into multiple energy intervals of equal length; wherein the length of each energy interval is a preset length value; sample the energy spectrum data and each basis function in each energy interval to obtain the energy spectrum sampling value of the energy spectrum data corresponding to each energy interval and the basis function sampling value of each basis function corresponding to each energy interval.

[0132] The aforementioned reconstruction module 1608 can also be used for:

[0133] Initialize the coefficient image, first label image and second label image of each basis function; wherein, the pixel position distribution of the coefficient image, first label image and second label image of each basis function is the same, the first label image represents the location region of the base material on the corresponding coefficient image, and the second label image represents the location region of the non-base material on the corresponding coefficient image;

[0134] For each iteration, the following operations are performed: Based on the projection vector, the preset length value, the energy spectrum sequence, each decay sequence, and the coefficient images of each basis function used in this iteration, the estimated data of the object under test corresponding to this iteration is determined; wherein, the estimated data includes estimated projection data and weighting coefficients; Based on the scanning parameters, the initial projection data, and the estimated data of the object under test corresponding to this iteration, as well as the first and second labeled images of the corresponding basis functions used in this iteration, the coefficient images of the corresponding basis functions used in this iteration are reconstructed to obtain the reconstructed coefficient images of each basis function corresponding to this iteration; Image segmentation is performed on the reconstructed coefficient images of each basis function corresponding to this iteration to obtain the first and second labeled images of each basis function used in the next iteration; Based on the reconstructed coefficient images of each basis function corresponding to this iteration and the first labeled images of each basis function used in the next iteration, the coefficient images of each basis function used in the next iteration are determined;

[0135] The iteration ends when the preset iteration termination condition is met, and the reconstructed coefficient image of each basis function at the end of the iteration is determined as the optimized coefficient image of the corresponding basis function; wherein, the preset iteration termination condition is met includes: the number of iterations reaches a preset number, and / or, the sum of the inner products between the coefficient images of different basis functions is less than a preset threshold.

[0136] The reconstruction module 1608 described above can also be used to: determine the sum of the inner products between the coefficient images of different basis functions corresponding to each iteration based on the coefficient images of each basis function used in each iteration.

[0137] The image reconstruction device based on orthogonal basis material decomposition provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned image reconstruction method based on orthogonal basis material decomposition. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0138] Unless otherwise specifically stated, the relative steps, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention.

[0139] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0140] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0141] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An image reconstruction method based on orthogonal basis material decomposition, characterized in that, The method includes: The scanning parameters of a CT device are obtained; wherein the CT device includes a radiation source and a turntable, and a detector composed of multiple detector units, and the scanning parameters include the distance from the radiation source to the center of the turntable, the distance from the radiation source to the detector, the number of detector units contained in the detector, and the size of each detector unit; The initial projection data and energy spectrum data of a single energy spectrum of the object under test are acquired; wherein the object under test is composed of at least two base materials, and the initial projection data is obtained by the CT device after scanning the object under test with multicolor X-rays to perform projection imaging; Obtain basis functions representing the attenuation when scanning an object with multicolor X-rays; wherein, the basis functions correspond one-to-one with the base material types of the object under test, and each basis function reflects the change of the attenuation coefficient of the corresponding base material with X-ray energy; Based on the scanning parameters, the energy spectrum data, the initial projection data, and each basis function, the coefficient images of each basis function are iteratively reconstructed with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions. Based on the optimized coefficient images of each basis function obtained after iterative reconstruction, a pseudo-monoenergy image of the object under test is generated.

2. The method according to claim 1, characterized in that, Based on the scanning parameters, the energy spectrum data, the initial projection data, and each basis function, the coefficient images of each basis function are iteratively reconstructed with the objective of minimizing the sum of the inner products between the coefficient images of different basis functions, including: Based on the energy spectrum data and each basis function, an energy spectrum sequence of the energy spectrum data and a decay sequence of each basis function are generated; wherein, both the energy spectrum sequence and the decay sequence are composed of multiple discrete elements, and the element positions of the energy spectrum sequence correspond one-to-one with the element positions of the decay sequence. Based on the scanning parameters, a projection vector is generated for projecting the object under test using multicolor X-rays scanned by a CT scanner; wherein, the projection vector characterizes the contribution of each pixel position in the coefficient image to the initial projection data; Based on the projection vector, the initial projection data, the energy spectrum sequence, and each decay sequence, the coefficient images of each basis function are iteratively reconstructed with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions.

3. The method according to claim 2, characterized in that, Based on the energy spectrum data and each basis function, the energy spectrum sequence of the energy spectrum data and the decay sequence of each basis function are generated, including: The energy spectrum data and each basis function are sampled using the same sampling parameters to obtain the energy spectrum sampled values ​​of the energy spectrum data and the basis function sampled values ​​of each basis function. The energy spectrum sequence is formed by combining all the energy spectrum sample values, and the basis function sample values ​​of each basis function are combined into a decay sequence to obtain the energy spectrum sequence and the decay sequence of each basis function.

4. The method according to claim 3, characterized in that, The energy spectrum data and each basis function are sampled using the same sampling parameters, including: The effective energy range of the energy spectrum data is divided into multiple energy intervals of equal length; wherein the length of each energy interval is a preset length value; Within each energy range, the energy spectrum data and each basis function are sampled to obtain the energy spectrum sampled values ​​corresponding to each energy range and the basis function sampled values ​​corresponding to each energy range.

5. The method according to claim 4, characterized in that, Based on the projection vector, the initial projection data, the energy spectrum sequence, and each decay sequence, the coefficient images of each basis function are iteratively reconstructed with the objective of minimizing the sum of the inner products between the coefficient images of different basis functions, including: Initialize the coefficient image, first label image and second label image of each basis function; wherein, the pixel position distribution of the coefficient image, first label image and second label image of each basis function is the same, the first label image represents the location region of the base material on the corresponding coefficient image, and the second label image represents the location region of the non-base material on the corresponding coefficient image; For each iteration, perform the following steps in AD: A. Based on the projection vector, the preset length value, the energy spectrum sequence, each decay sequence, and the coefficient image of each basis function used in this iteration, determine the estimated data of the object under test corresponding to this iteration; wherein, the estimated data includes estimated projection data and weighting coefficients; B. Based on the scanning parameters, the initial projection data, the estimated data of the object under test corresponding to this iteration, and the first and second labeled images of the corresponding basis functions used in this iteration, the coefficient images of the corresponding basis functions used in this iteration are reconstructed to obtain the reconstructed coefficient images of each basis function corresponding to this iteration. C. Perform image segmentation on the reconstructed coefficient images of each basis function corresponding to this iteration to obtain the first and second labeled images of each basis function used in the next iteration; D. Based on the reconstructed coefficient images of each basis function corresponding to this iteration and the first labeled images of each basis function used in the next iteration, determine the coefficient images of each basis function used in the next iteration. The iteration ends when the preset iteration termination condition is met, and the reconstructed coefficient image of each basis function at the end of the iteration is determined as the optimized coefficient image of the corresponding basis function; wherein, the preset iteration termination condition is met includes: the number of iterations reaches a preset number, and / or, the sum of the inner products between the coefficient images of different basis functions is less than a preset threshold.

6. The method according to claim 5, characterized in that, Based on the projection vector, the initial projection data, the energy spectrum sequence, and each decay sequence, the coefficient images of each basis function are iteratively reconstructed with the objective of minimizing the sum of the inner products between the coefficient images of different basis functions. This also includes: Based on the coefficient graphs of each basis function used in each iteration, the sum of the inner products between the coefficient graphs of different basis functions corresponding to each iteration is determined.

7. An image reconstruction device based on orthogonal basis material decomposition, characterized in that, The device includes: The first acquisition module is used to acquire the scanning parameters of the CT equipment; wherein, the CT equipment includes a radiation source and a turntable and a detector composed of multiple detector units, and the scanning parameters include the distance from the radiation source to the center of the turntable, the distance from the radiation source to the detector, the number of detector units contained in the detector, and the size of each detector unit; The second acquisition module is used to acquire initial projection data and energy spectrum data of a single energy spectrum of the object under test; wherein, the object under test is composed of at least two base materials, and the initial projection data is obtained by the CT device after scanning the object under test with multicolor X-rays to perform projection imaging; The third acquisition module is used to acquire the basis functions that represent the attenuation when scanning the object under test with multicolor X-rays; wherein, the basis functions correspond one-to-one with the base material types of the object under test, and each basis function reflects the change of the attenuation coefficient of the corresponding base material with X-ray energy; The reconstruction module is used to iteratively reconstruct the coefficient images of each basis function based on the scanning parameters, the energy spectrum data, the initial projection data, and each basis function, with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions; The generation module is used to generate a pseudo-monoenergy image of the object under test based on the optimized coefficient images of each basis function obtained after iterative reconstruction.

8. The apparatus according to claim 7, characterized in that, The reconstruction module is also used for: Based on the energy spectrum data and each basis function, an energy spectrum sequence of the energy spectrum data and a decay sequence of each basis function are generated; wherein, both the energy spectrum sequence and the decay sequence are composed of multiple discrete elements, and the element positions of the energy spectrum sequence correspond one-to-one with the element positions of the decay sequence. Based on the scanning parameters, a projection vector is generated for projecting the object under test using multicolor X-rays scanned by a CT scanner; wherein, the projection vector characterizes the contribution of each pixel position in the coefficient image to the initial projection data; Based on the projection vector, the initial projection data, the energy spectrum sequence, and each decay sequence, the coefficient images of each basis function are iteratively reconstructed with the goal of minimizing the sum of the inner products between the coefficient images of different basis functions.

9. The apparatus according to claim 8, characterized in that, The reconstruction module is also used for: The energy spectrum data and each basis function are sampled using the same sampling parameters to obtain the energy spectrum sampled values ​​of the energy spectrum data and the basis function sampled values ​​of each basis function. The energy spectrum sequence is formed by combining all the energy spectrum sample values, and the basis function sample values ​​of each basis function are combined into a decay sequence to obtain the energy spectrum sequence and the decay sequence of each basis function.

10. The apparatus according to claim 9, characterized in that, The reconstruction module is also used for: The effective energy range of the energy spectrum data is divided into multiple energy intervals of equal length; wherein the length of each energy interval is a preset length value; Within each energy range, the energy spectrum data and each basis function are sampled to obtain the energy spectrum sampled values ​​corresponding to each energy range and the basis function sampled values ​​corresponding to each energy range.

Images