Method and device for reconstruction of synchrotron radiation micro-ct based on decomposition and diffusion sampling

Abstract

The application discloses a method and device for reconstructing a synchrotron radiation micro-CT based on decomposition diffusion sampling, which collects a synchrotron radiation micro-CT image reconstructed at full projection angle and free from annular artifacts as a training data set, applies random differential equation disturbance at a set time to the training image, inputs the training image and the disturbed training image into a fractional matching network model to train the fractional matching network model, obtains the trained fractional matching network model, samples the image through the trained fractional matching network model, solves the sampled image through an augmented Lagrange multiplier algorithm, obtains a synchrotron radiation micro-CT image, carries out denoising diffusion processing on the synchrotron radiation micro-CT image through a denoising diffusion implicit model, obtains a starting value for next round sampling, iteratively samples and solves until a set iteration round is completed, and outputs a final round synchrotron radiation micro-CT image, and carries out denoising processing on the final round synchrotron radiation micro-CT image through a Tweedie equation to obtain a final synchrotron radiation micro-CT image. The application guarantees the accuracy of the reconstructed image.

Description

Technical Field

[0001] This invention relates to the field of radiomedical image processing technology, specifically to a synchrotron radiation micro-CT reconstruction method and apparatus based on decomposition diffusion sampling. Background Technology

[0002] Synchrotron radiation micro-CT, due to its high detector resolution and sensitivity, may introduce ring artifacts into the reconstructed images. Furthermore, its long scan time and high radiation dose often necessitate sparse angle sampling, which may introduce sparse angle artifacts and noise into the reconstructed images.

[0003] Currently, traditional methods address the annular artifact problem and the sparse angle reconstruction problem separately. For the annular artifact problem, filtering methods such as wavelet Fourier filtering or model-based iterative methods such as low-rank tensor decomposition are used in the projection domain to remove the annular artifacts, which appear as vertical stripes. However, these methods have limited stripe removal capabilities, and residual stripes can significantly interfere with the reconstructed image. For the sparse angle reconstruction problem, traditional methods utilize model-based iterative methods such as SART-TV and end-to-end deep learning methods. However, model-based iterative methods rely on pre-defined model priors, resulting in limited artifact removal capabilities. Deep learning algorithms depend on large amounts of paired labeled data, which requires significant time and effort to obtain in practice. While generative diffusion model-based methods, such as decomposition-diffusion sampling, can achieve high-quality sparse angle CT reconstruction without paired data, they do not consider the impact of annular artifacts. Annular artifacts that are not completely removed by traditional methods can cause pseudo-structures in the reconstructed image. Currently, algorithms that simultaneously remove sparse corner artifacts and ring artifacts are deep learning algorithms used in image post-processing, which require paired data, and such data is difficult to obtain in practice.

[0004] Therefore, it is essential to provide a sparse angle synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling to address the shortcomings of existing technologies in X-ray synchrotron radiation micro-CT image reconstruction. Summary of the Invention

[0005] Therefore, this invention provides a synchrotron radiation micro-CT reconstruction method and apparatus based on decomposition diffusion sampling. Through data priors and model priors, the CT image is updated on a clean CT image distribution manifold during the decomposition diffusion sampling constraint reconstruction process. The reconstructed image can effectively remove ring artifacts and sparse angular artifacts. The tensor low-rank model prior decomposes vertical stripes and projected images, ensuring the accuracy of the reconstructed image.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling, comprising:

[0007] Collect synchrotron radiation micro-CT images reconstructed under full projection angles and free of annular artifacts, and use the collected synchrotron radiation micro-CT images as a training dataset;

[0008] By applying a random differential equation perturbation at a set time to the training images in the training dataset, the perturbed training images are obtained.

[0009] The training images in the training dataset and the perturbed training images are input into the score matching network model; the score matching network model is trained based on the training images in the training dataset and the perturbed training images to learn the corresponding scores under the perturbation of the random differential equation at a set time, and the trained score matching network model is obtained.

[0010] Image sampling is performed using the trained score matching network model; the sampled images are then calculated using the Tweedie equation to obtain the training image manifold distribution image that satisfies the perturbation.

[0011] The sampled images are solved by the augmented Lagrange multiplier algorithm to obtain synchrotron radiation micro-CT images constrained by the model and detector measurement projection information;

[0012] The obtained synchrotron radiation micro-CT image is denoised and diffused using a denoising diffusion implicit model to obtain the starting value for the next round of sampling; based on the starting value for the next round of sampling, iterative sampling and solving are performed until the set number of iterations is completed, and the final round of synchrotron radiation micro-CT image is output.

[0013] The final synchrotron radiation micro-CT image was denoised using the Tweedie equation to obtain the final synchrotron radiation micro-CT image.

[0014] As a preferred embodiment of a synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling, in the process of obtaining the perturbed training images by applying a stochastic differential equation perturbation at a predetermined time to the training images in the training dataset, the expression of the stochastic differential equation at the predetermined time is:

[0015]

[0016] In the formula, x0 is the training image; s t σ is the drift function; t is the diffusion function; z is a Gaussian distribution.

[0017] As a preferred embodiment of a synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling, the loss function expression of the fractional matching network model is as follows:

[0018]

[0019] In the formula, For all possible samples, x t Minimize the expectation of x0. For score matching network models; This means that, given x0, with respect to x t The gradient of the log-conditional probability density.

[0020] As a preferred embodiment of a synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling, in the process of the fractional matching network model learning the corresponding fraction under the perturbation of the stochastic differential equation at a given time, the calculation formula for the corresponding fraction under the perturbation of the stochastic differential equation at the given time is as follows:

[0021]

[0022] In the formula, s θ For score matching network models; For x t The corresponding score.

[0023] As a preferred embodiment of a synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling, in the process of calculating the manifold distribution image of the sampled image using the Tweedie equation to obtain the manifold distribution image of the training image after perturbation, the expression of the manifold distribution image of the training image after perturbation is as follows:

[0024]

[0025] In the formula, To satisfy the perturbation of the training image manifold distribution image.

[0026] As a preferred scheme for synchrotron radiation micro-CT reconstruction based on decomposition diffusion sampling, in the process of solving the sampled image using the augmented Lagrange multiplier algorithm to obtain the synchrotron radiation micro-CT image constrained by the model and detector measurement projection information, the expression obtained by the conjugate gradient method according to the augmented Lagrange multiplier is as follows:

[0027]

[0028] In the formula, A is the CT projection matrix; A T The transpose projection operator is used; ρ is the regularization parameter; D z For the total variational operator in the Z direction; D z The transpose operator; y represents the projection information; S t+1 R1 represents the stripe term from the previous iteration.t+1 w1 is an auxiliary variable. t+1 It is a Lagrange multiplier.

[0029] As a preferred embodiment of a synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling, in the process of denoising the last round of synchrotron radiation micro-CT images using the Tweedie equation to obtain the final synchrotron radiation micro-CT image, the expression for the final synchrotron radiation micro-CT image is:

[0030]

[0031] In the formula, x0 is the final synchrotron radiation micro-CT image.

[0032] The present invention also provides a synchrotron radiation micro-CT reconstruction device based on decomposition-diffusion sampling, employing the above-described synchrotron radiation micro-CT reconstruction method based on decomposition-diffusion sampling, including:

[0033] The training dataset acquisition module is used to collect synchrotron radiation micro-CT images reconstructed under full projection angles and free of ring artifacts, and to use the collected synchrotron radiation micro-CT images as the training dataset.

[0034] The training image perturbation processing module is used to obtain perturbed training images by applying a random differential equation perturbation at a set time to the training images in the training dataset;

[0035] The score matching network model training module is used to input the training images in the training dataset and the perturbated training images into the score matching network model; the score matching network model is trained based on the training images in the training dataset and the perturbated training images to learn the corresponding scores under the perturbation of the random differential equation at a set time, and obtain the trained score matching network model.

[0036] The image sampling module is used to sample images using the trained score matching network model; the sampled images are then calculated using the Tweedie equation to obtain the training image manifold distribution image that satisfies the perturbation.

[0037] The sampled image solving module is used to solve the sampled image using the augmented Lagrange multiplier algorithm to obtain a synchrotron radiation micro-CT image constrained by the model and the detector measurement projection information;

[0038] The iterative sampling and solving module is used to perform denoising and diffusion processing on the obtained synchrotron radiation micro-CT image through a denoising and diffusion implicit model to obtain the starting value for the next round of sampling; based on the starting value for the next round of sampling, iterative sampling and solving are performed until the set number of iterations is completed, and the final round of synchrotron radiation micro-CT image is output.

[0039] The final synchrotron radiation micro-CT image acquisition module is used to denoise the last round of synchrotron radiation micro-CT images using the Tweedie equation to obtain the final synchrotron radiation micro-CT image.

[0040] As a preferred embodiment of a synchrotron radiation micro-CT reconstruction device based on decomposition diffusion sampling, in the training image perturbation processing module, during the process of obtaining the perturbed training image by applying a stochastic differential equation perturbation at a set time to the training images in the training dataset, the expression of the stochastic differential equation at the set time is:

[0041]

[0042] In the formula, x0 is the training image; s t σ is the drift function; t is the diffusion function; z is a Gaussian distribution.

[0043] As a preferred embodiment of the synchrotron radiation micro-CT reconstruction device based on decomposition diffusion sampling, the loss function expression of the fractional matching network model in the fractional matching network model training module is as follows:

[0044]

[0045] In the formula, For all possible samples, x t Minimize the expectation of x0. For score matching network models; This means that, given x0, with respect to x t The gradient of the log-conditional probability density.

[0046] As a preferred embodiment of a synchrotron radiation micro-CT reconstruction device based on decomposition diffusion sampling, in the fractional matching network model training module, during the process of the fractional matching network model learning the fractions corresponding to the perturbations of the stochastic differential equations at a set time, the calculation formula for the fractions corresponding to the perturbations of the stochastic differential equations at the set time is as follows:

[0047]

[0048] In the formula, s θ For score matching network models; For x t The corresponding score.

[0049] As a preferred embodiment of a synchrotron radiation micro-CT reconstruction device based on decomposition diffusion sampling, in the image sampling module, during the process of calculating the sampled image using the Tweedie equation to obtain the perturbed training image manifold distribution image, the expression for the perturbed training image manifold distribution image is as follows:

[0050]

[0051] In the formula, To satisfy the perturbation of the training image manifold distribution image.

[0052] As a preferred embodiment of a synchrotron radiation micro-CT reconstruction device based on decomposition diffusion sampling, in the sampled image solving module, during the process of solving the sampled image using the augmented Lagrange multiplier algorithm to obtain a synchrotron radiation micro-CT image constrained by the model and detector measurement projection information, the expression solved by the conjugate gradient method according to the augmented Lagrange multipliers is as follows:

[0053]

[0054]

[0055] In the formula, A is the CT projection matrix; A T The transpose projection operator is used; ρ is the regularization parameter; D z For the total variational operator in the Z direction; D z The transpose operator; y represents the projection information; S t+1 R1 represents the stripe term from the previous iteration. t+1 w1 is an auxiliary variable. t+1 It is a Lagrange multiplier.

[0056] As a preferred embodiment of the synchrotron radiation micro-CT reconstruction device based on decomposition diffusion sampling, in the final synchrotron radiation micro-CT image acquisition module, during the process of denoising the last round of synchrotron radiation micro-CT images using the Tweedie equation to obtain the final synchrotron radiation micro-CT image, the expression for the final synchrotron radiation micro-CT image is:

[0057]

[0058] In the formula, x0 is the final synchrotron radiation micro-CT image.

[0059] This invention has the following advantages: It collects synchrotron radiation micro-CT images reconstructed from the full projection angle and free of annular artifacts, and uses these collected synchrotron radiation micro-CT images as a training dataset; it obtains perturbed training images by applying a stochastic differential equation perturbation at a set time to the training images in the training dataset; it inputs the training images in the training dataset and the perturbed training images into a score matching network model; the score matching network model is trained based on the training images in the training dataset and the perturbed training images, learning the corresponding scores under the stochastic differential equation perturbation at the set time, thus obtaining a trained score matching network model; and it performs image processing using the trained score matching network model. The process involves sampling the image; calculating the manifold distribution of the training image after perturbation using the Tweedie equation; solving the sampled image using the augmented Lagrange multiplier algorithm to obtain a synchrotron radiation micro-CT image constrained by the model and detector measurement projection information; applying a denoising diffusion implicit model to the obtained synchrotron radiation micro-CT image to obtain the starting value for the next round of sampling; iterative sampling and solving based on the starting value for the next round of sampling until the set number of iterations is completed, outputting the final round of synchrotron radiation micro-CT image; and denoising the final round of synchrotron radiation micro-CT image using the Tweedie equation to obtain the final synchrotron radiation micro-CT image. Compared to current algorithms that simultaneously remove sparse angle artifacts and ring artifacts (which are image post-processing deep learning algorithms), the system proposed in this invention trains by learning the probability distribution of high-quality images, eliminating the need for paired low-quality data during training. Furthermore, the reconstruction results of this method are constrained by model priors and the original projection information, resulting in more accurate reconstruction results compared to post-processing methods and reducing the likelihood of pseudo-structures. Compared to existing ring artifact removal algorithms, such as model-based iterative algorithms, this invention proposes a system that simultaneously utilizes both data priors and model priors. Based on model prior constraints, the CT image update during reconstruction is performed on a clean CT image distribution manifold through decomposition and diffusion sampling constraints, resulting in superior ring artifact removal capabilities compared to traditional algorithms. Compared to current sparse angle CT reconstruction algorithms, for model-based iterative algorithms, this invention utilizes both data priors and model priors, leading to higher reconstruction quality. For end-to-end deep learning methods, training in this invention is achieved by learning the probability distribution of high-quality images, eliminating the need for paired data, and the reconstruction quality is superior due to the consideration of model priors and constraints from the original projection information. For diffusion-generated model methods, the proposed algorithm considers a ring artifact term in the objective function, merging ring artifact removal and sparse angle reconstruction into a single problem. The network simultaneously constrains both the projection and reconstructed images, preventing residual ring artifacts from traditional ring artifact removal algorithms from further affecting the reconstruction process and avoiding image pseudo-structure problems. Attached Figure Description

[0060] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0061] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0062] Figure 1 This is a schematic diagram of the synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling provided in Embodiment 1 of the present invention;

[0063] Figure 2 This is a schematic diagram of the specific implementation process of the synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling provided in Embodiment 1 of the present invention;

[0064] Figure 3 This is a schematic diagram of the results of the processing part of the present invention and the prior art in the synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling provided in Embodiment 1 of the present invention;

[0065] Figure 4 This is a schematic diagram of the architecture of the synchrotron radiation micro-CT reconstruction device based on decomposition diffusion sampling provided in Embodiment 2 of the present invention. Detailed Implementation

[0066] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0067] Example 1

[0068] See Figure 1 and Figure 2 Embodiment 1 of the present invention provides a synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling, comprising:

[0069] S1. Collect synchrotron radiation micro-CT images reconstructed under full projection angles and free of annular artifacts, and use the collected synchrotron radiation micro-CT images as a training dataset.

[0070] S2. By applying a random differential equation perturbation at a set time to the training images in the training dataset, the perturbed training images are obtained.

[0071] S3. Input the training images in the training dataset and the perturbated training images into the score matching network model; the score matching network model is trained based on the training images in the training dataset and the perturbated training images to learn the corresponding scores under the perturbation of the random differential equation at a set time, and obtain the trained score matching network model.

[0072] S4. Image sampling is performed using the trained score matching network model; the sampled images are then calculated using the Tweedie equation to obtain the training image manifold distribution image that satisfies the perturbation.

[0073] S5. Solve the sampled image using the augmented Lagrange multiplier algorithm to obtain a synchrotron radiation micro-CT image constrained by the model and the detector measurement projection information;

[0074] S6. The obtained synchrotron radiation micro-CT image is subjected to denoising and diffusion processing through a denoising diffusion implicit model to obtain the starting value for the next round of sampling; based on the starting value for the next round of sampling, iterative sampling and solving are performed until the set number of iterations is completed, and the final round of synchrotron radiation micro-CT image is output.

[0075] S7. The final synchrotron radiation micro-CT image is denoised using the Tweedie equation to obtain the final synchrotron radiation micro-CT image.

[0076] In this embodiment, in step S1, synchrotron radiation micro-CT images reconstructed under the full projection angle and free of annular artifacts are collected, and the collected synchrotron radiation micro-CT images are used as training datasets.

[0077] In this embodiment, in step S2, the perturbed training images are obtained by applying a random differential equation perturbation at a set time to the training images in the training dataset.

[0078] Specifically, the training images are perturbed by stochastic differential equations (SDEs) at different times, and then fed into the convolutional neural network along with the original training images.

[0079] The stochastic differential equation expression at the set time is:

[0080]

[0081] In the formula, x0 is the training image; s t σ is the drift function; t is the diffusion function; z is a Gaussian distribution.

[0082] In this embodiment, in step S3, the training images in the training dataset and the perturbated training images are input into the score matching network model; the score matching network model is trained based on the training images in the training dataset and the perturbated training images to learn the corresponding scores under the perturbation of the random differential equation at a set time, and obtain the trained score matching network model.

[0083] Specifically, the training images and the perturbed training images from the training dataset are input into the score matching network model. The score matching network model uses the input images and passes them through the score matching network s θ Learn the scores corresponding to SDE perturbations at different times.

[0084] The loss function expression for the score matching network model is as follows:

[0085]

[0086] In the formula, For all possible samples, x t Minimize the expectation of x0. For score matching network models; This means that, given x0, with respect to x t The gradient of the log-conditional probability density.

[0087] The formula for calculating the fraction corresponding to the perturbation of the stochastic differential equation at a given time is as follows:

[0088]

[0089] In the formula, s θ For score matching network models; For x t The corresponding score.

[0090] In this embodiment, in step S4, image sampling is performed using the trained score matching network model; the sampled image is then calculated using the Tweedie equation to obtain the training image manifold distribution image that satisfies the perturbation.

[0091] Specifically, the reconstruction process begins by initializing various variables, including the initial reconstruction image. Projection matrix A, iteration step size N, fringe S N Auxiliary variable R1 in the ALM solution processN Q N and Lagrange multipliers W1 N W2 N .

[0092] The sampling process of the diffusion model is carried out using a pre-trained score matching network model. In each sampling process, a clean image that satisfies the manifold distribution of the training image is first calculated using the Tweedie equation.

[0093] The expression for the manifold distribution image of the training image after the perturbation is:

[0094]

[0095] In the formula, To satisfy the perturbation of the training image manifold distribution image.

[0096] In this embodiment, in step S5, the sampled image is solved by the augmented Lagrange multiplier algorithm to obtain a synchrotron radiation micro-CT image constrained by the model and the detector measurement projection information;

[0097] Specifically, using the CT projection matrix A and the total variational (TV) operator D in the Z direction... z Calculate the projection matrix A used in the conjugate gradient method to solve the CT image subproblem. CG The calculation formula is as follows:

[0098]

[0099] In the formula, A T Here, ρ is the transpose projection operator; ρ is the regularization parameter. D z The transpose operator.

[0100] Using the projection information y and the fringe term S from the previous iteration t+1 and auxiliary variable R1 t+1 and Lagrange multipliers w1 t+1 Calculate the b required for solving the CT image subproblem using the conjugate gradient method. CG The calculation formula is as follows:

[0101]

[0102] According to A CG and b CG The expression for solving the CT image subproblem using the conjugate gradient method is as follows:

[0103]

[0104] In the formula, M represents the number of iterations of the conjugate gradient method.

[0105] Update auxiliary variable R1 using the soft thresholding method t+1 The subproblem has the following solution expression:

[0106]

[0107] In the formula, This is a soft threshold operator.

[0108] Using projection information y and auxiliary variable Q t+1 The fringe S in the projection of the low-rank tensor constraint is obtained through Tucker decomposition. t The solution process is as follows:

[0109]

[0110] In the formula, λ t This is the regularization parameter.

[0111] In this embodiment, in step S6, the obtained synchrotron radiation micro-CT image is subjected to denoising and diffusion processing through a denoising diffusion implicit model to obtain the starting value for the next round of sampling; based on the starting value for the next round of sampling, iterative sampling and solving are performed until the set number of iterations is completed, and the final round of synchrotron radiation micro-CT image is output.

[0112] Specifically, update the constraint stripe S t The auxiliary variable Q of size t The solution process is as follows:

[0113]

[0114] In the formula, Tensor Q t Expanded index values ​​in the second and third dimensions; This is the regularization parameter.

[0115] Update Lagrange multipliers w1 t w2 t Its expression is:

[0116]

[0117] w2 t ←w2 t+1 +β(Q t -S t )

[0118] In the formula, β is the regularization parameter.

[0119] The starting value of x for the next iteration is obtained through DDIM sampling. t-1 ,

[0120]

[0121] In the formula, η is the noise level controlling the sampling; ∈ is the noise coefficient.

[0122] In this embodiment, in step S7, the final synchrotron radiation micro-CT image is denoised using the Tweedie equation to obtain the final synchrotron radiation micro-CT image, the result of which is as follows. Figure 3 As shown.

[0123] Specifically, the expression for the final synchrotron radiation micro-CT image is:

[0124]

[0125] In the formula, x0 is the final synchrotron radiation micro-CT image.

[0126] In summary, this invention collects synchrotron radiation micro-CT images reconstructed from the full projection angle without annular artifacts, and uses these collected synchrotron radiation micro-CT images as a training dataset. By applying a stochastic differential equation perturbation at a set time to the training images in the training dataset, perturbed training images are obtained. The training images in the training dataset and the perturbed training images are input into a score matching network model. The score matching network model is trained based on the training images in the training dataset and the perturbed training images, learning the corresponding scores under the stochastic differential equation perturbation at the set time, thus obtaining a trained score matching network model. The trained score matching network model is then used to process images... The process involves sampling; calculating the sampled image using the Tweedie equation to obtain a training image manifold distribution image that satisfies the perturbation; solving the sampled image using the augmented Lagrange multiplier algorithm to obtain a synchrotron radiation micro-CT image constrained by the model and detector measurement projection information; performing denoising and diffusion processing on the obtained synchrotron radiation micro-CT image using a denoising diffusion implicit model to obtain the starting value for the next round of sampling; iteratively sampling and solving based on the starting value for the next round of sampling until the set number of iterations is completed, outputting the final round of synchrotron radiation micro-CT image; and denoising the final round of synchrotron radiation micro-CT image using the Tweedie equation to obtain the final synchrotron radiation micro-CT image. Compared to current algorithms that simultaneously remove sparse angle artifacts and ring artifacts (which are image post-processing deep learning algorithms), the system proposed in this invention trains by learning the probability distribution of high-quality images, eliminating the need for paired low-quality data during training. Furthermore, the reconstruction results of this method are constrained by model priors and original projection information, resulting in more accurate reconstruction results compared to post-processing methods and reducing the likelihood of pseudo-structures. Compared to existing ring artifact removal algorithms, such as model-based iterative algorithms, this invention proposes a system that simultaneously utilizes both data priors and model priors. Based on model prior constraints, the CT image update during reconstruction is performed on a clean CT image distribution manifold through decomposition and diffusion sampling constraints, resulting in superior ring artifact removal capabilities compared to traditional algorithms. Compared to current sparse angle CT reconstruction algorithms, for model-based iterative algorithms, this invention utilizes both data priors and model priors, leading to higher reconstruction quality. For end-to-end deep learning methods, training in this invention is achieved by learning the probability distribution of high-quality images, eliminating the need for paired data, and the reconstruction quality is superior due to the consideration of model priors and constraints from the original projection information. For diffusion-generated model methods, the proposed algorithm considers a ring artifact term in the objective function, merging ring artifact removal and sparse angle reconstruction into a single problem. The network simultaneously constrains both the projection and reconstructed images, preventing residual ring artifacts from traditional ring artifact removal algorithms from further affecting the reconstruction process and avoiding image pseudo-structure problems.

[0127] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0128] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0129] Example 2

[0130] See Figure 4 Embodiment 2 of the present invention provides a synchrotron radiation micro-CT reconstruction device based on decomposition diffusion sampling, comprising:

[0131] The training dataset acquisition module 001 is used to collect synchrotron radiation micro-CT images reconstructed under the full projection angle and free of annular artifacts, and to use the collected synchrotron radiation micro-CT images as the training dataset.

[0132] The training image perturbation processing module 002 is used to obtain the perturbed training image by applying a random differential equation perturbation at a set time to the training images in the training dataset;

[0133] The score matching network model training module 003 is used to input the training images in the training dataset and the perturbated training images into the score matching network model; the score matching network model is trained based on the training images in the training dataset and the perturbated training images to learn the corresponding scores under the perturbation of the random differential equation at a set time, and obtain the trained score matching network model.

[0134] Image sampling module 004 is used to sample images using the trained score matching network model; the sampled images are calculated using the Tweedie equation to obtain the training image manifold distribution image that satisfies the perturbation.

[0135] The sampled image solving module 005 is used to solve the sampled image using the augmented Lagrange multiplier algorithm to obtain a synchrotron radiation micro-CT image constrained by the model and the detector measurement projection information.

[0136] The iterative sampling and solving module 006 is used to perform denoising and diffusion processing on the obtained synchrotron radiation micro-CT image through a denoising and diffusion implicit model to obtain the starting value for the next round of sampling; based on the starting value for the next round of sampling, iterative sampling and solving are performed until the set number of iterations is completed, and the final round of synchrotron radiation micro-CT image is output.

[0137] The final synchrotron radiation micro-CT image acquisition module 007 is used to denoise the last round of synchrotron radiation micro-CT images using the Tweedie equation to obtain the final synchrotron radiation micro-CT image.

[0138] In this embodiment, in the training image perturbation processing module 002, during the process of obtaining the perturbed training image by applying a stochastic differential equation perturbation at a set time to the training images in the training dataset, the expression of the stochastic differential equation at the set time is:

[0139]

[0140] In the formula, x0 is the training image; s t σ is the drift function; t is the diffusion function; z is a Gaussian distribution.

[0141] In this embodiment, the loss function expression of the score matching network model in the score matching network model training module 003 is as follows:

[0142]

[0143] In the formula, For all possible samples, x t Minimize the expectation of x0. For score matching network models; This means that, given x0, with respect to x t The gradient of the log-conditional probability density. In this embodiment, in the score matching network model training module 003, during the process of the score matching network model learning the score corresponding to the perturbation of the stochastic differential equation at a set time, the calculation formula for the score corresponding to the perturbation of the stochastic differential equation at the set time is:

[0144]

[0145] In the formula, s θ For score matching network models; For xt The corresponding score.

[0146] In this embodiment, in the image sampling module 004, during the process of calculating the sampled image using the Tweedie equation to obtain the perturbated training image manifold distribution image, the expression for the perturbated training image manifold distribution image is:

[0147]

[0148] In the formula, To satisfy the perturbation of the training image manifold distribution image.

[0149] In this embodiment, in the sampled image solving module 005, during the process of solving the sampled image using the augmented Lagrange multiplier algorithm to obtain a synchrotron radiation micro-CT image constrained by the model and detector measurement projection information, the expression solved by the conjugate gradient method according to the augmented Lagrange multiplier is as follows:

[0150]

[0151] In the formula, A is the CT projection matrix; A T The transpose projection operator is used; ρ is the regularization parameter; D z For the total variational operator in the Z direction; D z The transpose operator; y represents the projection information; S t+1 R1 represents the stripe term from the previous iteration. t+1 w1 is an auxiliary variable. t+1 It is a Lagrange multiplier.

[0152] In this embodiment, in the final synchrotron radiation micro-CT image acquisition module 007, during the process of denoising the last round of synchrotron radiation micro-CT images using the Tweedie equation to obtain the final synchrotron radiation micro-CT image, the expression for the final synchrotron radiation micro-CT image is:

[0153]

[0154] In the formula, x0 is the final synchrotron radiation micro-CT image.

[0155] It should be noted that the information interaction and execution process between the modules / units of the above system are based on the same concept as the method embodiment in Embodiment 1 of this application, and the resulting technical effects are the same as those in the method embodiment of this application. For details, please refer to the description in the method embodiment shown above in this application, and it will not be repeated here.

[0156] Example 3

[0157] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium storing program code of a synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling. The program code includes instructions for executing the synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling of Embodiment 1 or any possible implementation thereof.

[0158] Computer-readable storage media can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives, SSDs).

[0159] Example 4

[0160] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0161] The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor can execute the synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling according to Embodiment 1 or any possible implementation thereof.

[0162] Specifically, a processor can be implemented in hardware or software. When implemented in hardware, the processor can be a logic circuit, an integrated circuit, etc. When implemented in software, the processor can be a general-purpose processor that reads software code stored in memory. This memory can be integrated into the processor or located outside the processor and exist independently.

[0163] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0164] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0165] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A synchrotron radiation micro-CT reconstruction method based on decomposition-diffusion sampling, characterized in that, include: Collect synchrotron radiation micro-CT images reconstructed under full projection angles and free of annular artifacts, and use the collected synchrotron radiation micro-CT images as a training dataset; By applying a random differential equation perturbation at a set time to the training images in the training dataset, the perturbed training images are obtained. The training images in the training dataset and the perturbed training images are input into the score matching network model; the score matching network model is trained based on the training images in the training dataset and the perturbed training images to learn the corresponding scores under the perturbation of the random differential equation at a set time, and the trained score matching network model is obtained. Image sampling is performed using the trained score matching network model; the sampled images are then calculated using the Tweedie equation to obtain the training image manifold distribution image that satisfies the perturbation. The sampled images are solved by the augmented Lagrange multiplier algorithm to obtain synchrotron radiation micro-CT images constrained by the model and detector measurement projection information; The obtained synchrotron radiation micro-CT image is subjected to denoising and diffusion processing using a denoising diffusion implicit model to obtain the starting value for the next round of sampling. Based on the starting value of the next round of sampling, iterative sampling and solving are performed until the set number of iterations is completed, and the final round of synchrotron radiation micro-CT image is output. The final synchrotron radiation micro-CT image was denoised using the Tweedie equation to obtain the final synchrotron radiation micro-CT image.

2. The synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling according to claim 1, characterized in that, In the process of obtaining the perturbed training images by applying a stochastic differential equation perturbation at a predetermined time to the training images in the training dataset, the expression of the stochastic differential equation at the predetermined time is: In the formula, x0 is the training image; s t σ is the drift function; t is the diffusion function; z is a Gaussian distribution.

3. The synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling according to claim 2, characterized in that, The loss function expression for the score matching network model is: In the formula, For all possible samples, x t ..., minimize the expectation of x0; For score matching network models; This means that, given x0, with respect to x t The gradient of the log-conditional probability density.

4. The synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling according to claim 3, characterized in that, In the process of the score matching network model learning the scores corresponding to the perturbations of the stochastic differential equations at a given time, the formula for calculating the scores corresponding to the perturbations of the stochastic differential equations at the given time is as follows: In the formula, s θ For score matching network models; For x t The corresponding score.

5. The synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling according to claim 4, characterized in that, In the process of calculating the manifold distribution image of the sampled image using the Tweedie equation to obtain the perturbation-perfected training image, the expression for the perturbation-perfected training image manifold distribution image is: In the formula, To satisfy the perturbation of the training image manifold distribution image.

6. The synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling according to claim 5, characterized in that, In the process of solving the sampled image using the augmented Lagrange multiplier algorithm to obtain the synchrotron radiation micro-CT image constrained by the model and detector measurement projection information, the expression obtained by the conjugate gradient method based on the augmented Lagrange multipliers is as follows: In the formula, A is the CT projection matrix; A T For CT transpose projection operator; ρ represents the regularization parameter D. z For the total variational operator in the Z direction; D z The transpose operator; y represents the projection information; S t+1 R1 represents the stripe term from the previous iteration. t+1 w1 is an auxiliary variable. t+1 It is a Lagrange multiplier.

7. The synchrotron radiation micro-CT reconstruction method based on decomposition diffusion sampling according to claim 6, characterized in that, In the process of denoising the final synchrotron radiation micro-CT images using the Tweedie equation to obtain the final synchrotron radiation micro-CT image, the expression for the final synchrotron radiation micro-CT image is: In the formula, x0 is the final synchrotron radiation micro-CT image.

8. A synchrotron radiation micro-CT reconstruction device based on decomposition-diffusion sampling, employing the synchrotron radiation micro-CT reconstruction method based on decomposition-diffusion sampling as described in any one of claims 1-7, characterized in that, include: The training dataset acquisition module is used to collect synchrotron radiation micro-CT images reconstructed under full projection angles and free of ring artifacts, and to use the collected synchrotron radiation micro-CT images as the training dataset. The training image perturbation processing module is used to obtain perturbed training images by applying a random differential equation perturbation at a set time to the training images in the training dataset; The score matching network model training module is used to input the training images in the training dataset and the perturbated training images into the score matching network model; the score matching network model is trained based on the training images in the training dataset and the perturbated training images to learn the corresponding scores under the perturbation of the random differential equation at a set time, and obtain the trained score matching network model. The image sampling module is used to sample images using the trained score matching network model; the sampled images are then calculated using the Tweedie equation to obtain the training image manifold distribution image that satisfies the perturbation. The sampled image solving module is used to solve the sampled image using the augmented Lagrange multiplier algorithm to obtain a synchrotron radiation micro-CT image constrained by the model and the detector measurement projection information; The iterative sampling and solving module is used to perform denoising and diffusion processing on the obtained synchrotron radiation micro-CT image through a denoising and diffusion implicit model to obtain the starting value for the next round of sampling; based on the starting value for the next round of sampling, iterative sampling and solving are performed until the set number of iterations is completed, and the final round of synchrotron radiation micro-CT image is output. The final synchrotron radiation micro-CT image acquisition module is used to denoise the last round of synchrotron radiation micro-CT images using the Tweedie equation to obtain the final synchrotron radiation micro-CT image.

9. The synchrotron radiation micro-CT reconstruction device based on decomposition diffusion sampling according to claim 8, characterized in that, In the training image perturbation processing module, during the process of obtaining the perturbed training image by applying a stochastic differential equation perturbation at a set time to the training images in the training dataset, the expression of the stochastic differential equation at the set time is: In the formula, x0 is the training image; s t σ is the drift function; t is the diffusion function; z is a Gaussian distribution.

10. The synchrotron radiation micro-CT reconstruction device based on decomposition diffusion sampling according to claim 9, characterized in that, In the score matching network model training module, the loss function expression of the score matching network model is: In the formula, For all possible samples, x t Minimize the expectation of x0. For score matching network models; This means that, given x0, with respect to x t The gradient of the log-conditional probability density.

Images