Sub-space assisted multi-prior photon counting energy spectrum CT reconstruction method and device

Abstract

The invention provides a subspace-assisted multi-prior photon counting energy spectrum CT reconstruction method and device, and relates to the technical field of CT reconstruction. The method comprises the following steps: firstly, performing subspace decomposition on an energy spectrum CT image to obtain a noise characteristic graph tensor and an orthogonal basis; then selecting a reference image block on the noise feature map tensor, and extracting image blocks similar to the reference image block from all channels of the energy spectrum CT image to generate a non-local similar full-channel tensor group; constructing a global sparse regularization item about the non-local similar full-channel tensor group by using dictionary learning to obtain a de-noised non-local similar full-channel tensor group, and multiplying the de-noised non-local similar full-channel tensor group by the orthogonal basis to obtain a clean tensor image; then, according to TV regularization constraints of all channels of the clean tensor image and the energy spectrum CT image, a reconstruction model of double regularization constraints is constructed; and finally, updating the reconstruction model of the double regularization constraint by adopting an augmented Lagrangian multiplier method until a preset reconstruction quality requirement is met.

Description

Technical Field

[0001] The present invention relates to the technical field of CT reconstruction, and in particular to a subspace-assisted multi-priori photon counting energy spectrum CT reconstruction method and device. Background Art

[0002] Computed Tomography (CT) as a modern imaging technology has been widely used in medicine, industry and other fields. However, traditional CT technology uses X-rays at a single energy for imaging, and its ability to distinguish material components is limited. In order to break through the imaging limitations of traditional CT technology, spectral CT uses the difference in the attenuation coefficient of a material at different X-ray energies to obtain richer material information. It has the ability to identify and analyze materials that traditional CT technology does not have. As a new imaging technology, it is being gradually studied and used.

[0003] The key step in spectral CT reconstruction is to use the X-ray projection information at multiple energies collected by the detector to restore the distribution of different material information inside the scanned object. It directly determines the imaging quality of spectral CT and affects the analysis, processing and application of subsequent spectral CT reconstruction images. In order to obtain high-quality spectral CT imaging results, how to develop and design efficient spectral CT reconstruction algorithms has become a hot and difficult issue in the current research field of X-ray imaging, and has important research and application value.

[0004] In the field of spectral CT reconstruction, the traditional CT sparse regularization reconstruction method based on the single energy assumption can be directly applied to spectral CT reconstruction. For example, in 2012, Xu et al. used the total variation (TV) technique to reconstruct the region of interest (ROI) of spectral CT. In 2013, Zhao et al. proposed an iterative method based on a tight frame to improve the image quality of spectral CT reconstruction; in 2016, Zeng et al. introduced the concept of structure tensor and transformed TV regularization into a penalized weighted least squares method, achieving results that surpassed traditional methods. These techniques achieve powerful denoising effects by processing each channel image separately. However, they fail to fully utilize the high correlation between images of different channels in spectral CT, resulting in the loss of image texture and edge information.

[0005] In order to more deeply explore the prior information in the image, tensor-based methods have begun to play an increasingly important role in spectral CT reconstruction. In 2019, Xia et al. regarded similar image blocks in spectral CT images as third-order tensors and decomposed them into low-rank and sparse components through principal component analysis (PCA), providing an intuitive method to describe the non-local similarity features in spectral CT images. In 2021, Chen et al. further proposed a fourth-order non-local tensor decomposition model for spectral CT reconstruction. In 2022, Yu et al. proposed a frame set tensor nuclear norm to constrain the non-local similarity of spectral CT images, thereby significantly improving the image quality. Nevertheless, these non-local-based methods rely heavily on block matching operations to identify similar image blocks, which not only makes them sensitive to noise but also increases a significant computational burden. In the implementation of these algorithms, it is crucial to maintain a balance between reconstruction quality and computational efficiency.

[0006] At the same time, the rapid development of deep learning technology has brought new vitality to data-driven spectral CT reconstruction methods. In 2021, Wu et al. first introduced a U-NET-based network for sparse spectral CT reconstruction. In 2023, Chen et al. developed SOUL-NET, a network that combines sparse and low-rank properties specifically for spectral CT reconstruction. In 2024, Wang et al. proposed the HITI-NET model, an interpretable hybrid domain integrated transform iterative network that can simultaneously optimize spectral CT reconstruction and material decomposition processes. Although these methods have achieved remarkable results in image reconstruction performance, they usually need to process a large number of parameters and have a high dependence on the characteristics of the training data set. In addition, deep learning technology also faces challenges such as data overfitting, computing resource limitations, insufficient model interpretability, and data privacy and security. Summary of the invention

[0007] In view of the problems that photon counting spectral CT image reconstruction takes a long time, has poor reconstruction quality, and causes a surge in decomposition noise, which seriously affects the quality of material decomposition, the present invention provides a subspace-assisted multi-priori photon counting spectral CT reconstruction method and device.

[0008] The present invention provides a subspace-assisted multi-priori photon counting energy spectrum CT reconstruction method, comprising:

[0009] Step 1: Generate a spectral CT image according to projection data of spectral CT imaging, wherein the spectral CT image includes images of S channels;

[0010] Step 2: performing subspace decomposition on the energy spectrum CT image to obtain a noise feature map tensor and an orthogonal basis;

[0011] Step 3: Select a reference image block on the noise feature map tensor, and extract image blocks similar to the reference image block from all channels of the spectral CT image to generate a non-local similarity full-channel tensor group;

[0012] Step 4: construct a global sparse regularization term for the non-local similar full-channel tensor group using dictionary learning to obtain a denoised non-local similar full-channel tensor group, and multiply the group with the orthogonal basis to obtain a clean tensor image;

[0013] Step 5: constructing a reconstruction model with dual regularization constraints according to the TV regularization constraints of each channel of the clean tensor image and the spectral CT image;

[0014] Step 6: Use the augmented Lagrange multiplier method to update the dual regularization constrained reconstruction model until the preset reconstruction quality requirement is met.

[0015] Furthermore, in step 1, a simultaneous algebraic reconstruction algorithm SART is used to generate a spectral CT image.

[0016] Furthermore, step 3 specifically includes:

[0017] Step 3.1: Select a reference image block on the noise feature map tensor, extract an image block similar to the reference image block from one channel of the spectral CT image using a block matching algorithm based on image domain distance, and form a similar image block set with all the extracted image blocks;

[0018] Step 3.2: Change the position of the reference image block and repeat step 3.1 to obtain multiple different sets of similar image blocks;

[0019] Step 3.3: For all channels of the spectral CT image, perform steps 3.1 and 3.2;

[0020] Step 3.4: vectorize all sets of similar image blocks, and then stack the vectorization results in the order of channels to obtain the non-local similarity full-channel tensor group.

[0021] Furthermore, the global sparse regularization term is:

[0022]

[0023] in, Represents the i-th non-local similarity full channel tensor group The expanded matrix form is represents the Frobenius norm, represents an adaptive dictionary, represents the group sparsity coefficient, ||·|| p Indicates Lp Norm, 0 <p≤1, represents the weight matrix, represents the Hadamard product of two matrices, and τ represents the regularization parameter.

[0024] Furthermore, the reconstruction model of the dual regularization constraint is:

[0025]

[0026] Where A represents the system matrix, x s Represents spectral CT image The sth channel, p s represents the projection data corresponding to the sth channel, ||x s || TV Represents the TV norm of the sth channel, the Frobenius norm term Represents a clean tensor image Spectral CT images The difference between them, ρ, λ are regularization parameters, Represents a group of non-locally similar full-channel tensors The global sparse regularization term of .

[0027] Furthermore, step 6 specifically includes:

[0028] Introducing auxiliary variables The Lagrange multiplier Λ and the non-negative penalty parameter β rewrite the reconstruction model of the dual regularization constraint as:

[0029]

[0030] The alternating direction multiplier method is used to solve the above rewritten model; the iterative formula is:

[0031]

[0032] in, Represents the image reconstructed in this round of spectral CT.

[0033] In a second aspect, the present invention provides a subspace-assisted multi-priori photon counting spectral CT reconstruction device, comprising:

[0034] An image generation module, used to generate a spectral CT image according to the projection data of spectral CT imaging, wherein the spectral CT image includes images of S channels;

[0035] A subspace decomposition module, used for performing subspace decomposition on the energy spectrum CT image to obtain a noise feature map tensor and an orthogonal basis;

[0036] A tensor group generation module, used for selecting a reference image block on the noise feature map tensor, extracting image blocks similar to the reference image block from all channels of the energy spectrum CT image to generate a non-local similarity full-channel tensor group;

[0037] A sparse module, used for constructing a global sparse regularization term about the non-local similar full-channel tensor group by using dictionary learning, obtaining a denoised non-local similar full-channel tensor group, and multiplying the denoised non-local similar full-channel tensor group with the orthogonal basis to obtain a clean tensor image;

[0038] A reconstruction module is used to construct a reconstruction model with dual regularization constraints according to the TV regularization constraints of each channel of the clean tensor image and the energy spectrum CT image, and to update the reconstruction model with dual regularization constraints by using an augmented Lagrange multiplier method until a preset reconstruction quality requirement is met.

[0039] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the first aspect when executing the program.

[0040] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in the first aspect.

[0041] Beneficial effects of the present invention:

[0042] The present invention aims at the problem that the reconstruction time of photon counting spectral CT image is long, the reconstruction quality is poor, the decomposition noise surges, and the quality of material decomposition is seriously affected. Based on the subspace decomposition technology, a multi-prior reconstruction method combining the non-local low-rank prior of the spectral feature map and the spatial TV regularization constraint of the spectral image is proposed to achieve high-precision image reconstruction of spectral CT; Different from the existing spectral CT reconstruction algorithm based on sparse optimization, the present invention combines the non-local structural sparse properties of the feature map corresponding to the spectral CT, selects a suitable reference image block in each round of iteration, extracts similar image blocks using the image domain distance, constructs a full-energy feature image tensor block group, integrates it into the reconstruction model to update the objective function, and performs an iterative strategy based on the alternating direction method on the updated objective function to obtain a higher quality reconstructed image. The present invention uses the non-local structural sparsity of the subspace feature image to explore the global-nonlocal correlation between the channels of the spectral image, combines the sparse prior of the gradient image in the channel, and designs an efficient optimization strategy to improve the computational efficiency and the quality of the reconstructed image. The experimental results verify the effectiveness of the proposed method in spectral CT reconstruction. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1A schematic flow chart of a subspace-assisted multi-priori photon counting spectral CT reconstruction method provided by an embodiment of the present invention;

[0044] Figure 2 An example of a digital simulation walnut body provided by an embodiment of the present invention;

[0045] Figure 3 Flow chart of a single iteration of spectral CT reconstruction based on subspace-assisted global-nonlocal sparsity constraints and spatial TV regularization constraints;

[0046] Figure 4 A comparison diagram of the local detail reconstruction results of the digital simulation walnut body using the method of the present invention and other existing methods provided in the embodiment of the present invention;

[0047] Figure 5 A comparison of material decomposition and reconstruction results on actual mouse data using the method of the present invention and other existing methods provided in an embodiment of the present invention;

[0048] Figure 6 A schematic structural diagram of a subspace-assisted multi-priori photon counting spectral CT reconstruction device provided by an embodiment of the present invention;

[0049] Figure 7 A structural block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] The present invention relates to a photon counting spectral CT reconstruction method based on subspace-assisted multiple priors (including global, non-local and local priors). First, a spectral CT image with global low-rank features is mapped to a low-dimensional feature image using subspace decomposition. Then, similar image blocks are extracted from the tensor feature image based on the image domain distance, and they are stacked to form a non-local full-channel tensor group. Next, the non-convex structure sparsity of adaptive dictionary learning is applied to characterize the non-local similarity of the tensor group. In addition, the spatial sparsity of the single-channel image is further characterized by utilizing total variation regularization. Finally, an efficient iterative solution algorithm is designed under the framework of the alternating direction method.

[0052] like Figure 1As shown, an embodiment of the present invention provides a subspace-assisted multi-priori photon counting spectral CT reconstruction method, comprising:

[0053] S101: generating a spectral CT image according to projection data of spectral CT imaging, wherein the spectral CT image includes images of S channels;

[0054] Specifically, after obtaining the imaging projection data, the parameter values ​​required by the reconstruction algorithm are selected, and the initial image can be a matrix of all 0s or all 1s; the projection data, algorithm parameters and initial image are processed through the simultaneous algebraic reconstruction algorithm (SART) to generate a spectral CT image. It consists of images of S channels, and the image of the sth channel is denoted by x s (1≤s≤S).

[0055] S102: Spectral CT image Perform subspace decomposition to obtain the noise feature map tensor and orthonormal basis E;

[0056] S103: In the noise feature map tensor Select the reference image block p i,j , from the spectral CT image Extracting image blocks similar to the reference image blocks in all channels to generate a non-local similarity full-channel tensor group;

[0057] S104: constructing a global sparse regularization term for the non-local similar full-channel tensor group using dictionary learning to obtain a denoised non-local similar full-channel tensor group, and multiplying the group with the orthogonal basis to obtain a clean tensor image;

[0058] Specifically, dictionary learning is used to characterize the structural sparsity of non-locally similar full-channel tensor groups, and sparse optimization theory is used to construct a global sparse regularization term.

[0059] S105: constructing a reconstruction model with dual regularization constraints according to the TV regularization constraints of each channel of the clean tensor image and the spectral CT image;

[0060] S106: Using the augmented Lagrange multiplier method to update the reconstruction model of the dual regularization constraint until the preset reconstruction quality requirement is met, such as the current round of reconstructed images have no significant changes compared with the previous round of reconstructed images. It should be noted that when the reconstruction is performed for the first time, the previous round of reconstructed images refers to the energy spectrum CT image generated in step S101.

[0061] The traditional spectral CT reconstruction algorithm is accompanied by strong noise artifacts, which seriously affect the identification of valuable information. The spectral CT reconstruction method provided by the embodiment of the present invention starts from the initial image, characterizes the non-local correlation of the feature image through the subspace decomposition technology during the reconstruction process, and improves the reconstruction model by using the constructed non-local similar full-channel tensor group, establishes a new reconstruction model, and improves the reconstruction accuracy and optimizes the computational efficiency by adopting the augmented Lagrange multiplier method, which is a continuous iteration and cycle mechanism.

[0062] Figure 2 This is an example of a digitally simulated walnut body. Figure 3 As shown, based on the above embodiment, the embodiment of the present invention adopts the following process to generate a non-local similarity full channel tensor group: first, in the noise feature map tensor Select a reference image block p i,j , using the graph-based blocking matching algorithm (GBM) on spectral CT images The first channel is extracted with p i,j There are N similar image blocks in total p and form a similar image block set; then, change the position of the reference image block to obtain a different similar image block set; then, refer to the above processing process for the first channel and repeat this process in different channels until the image block matching of all channels is completed; finally, vectorize the different similar image block sets and stack the vectorization results in the order of channels to form a non-local similarity full-channel tensor group.

[0063] On the basis of the above embodiments, the global sparse regularization term constructed in the embodiment of the present invention is:

[0064]

[0065] in, Represents the i-th non-local similarity full channel tensor group The expanded matrix form is represents the Frobenius norm, represents an adaptive dictionary, represents the group sparsity coefficient, ||·|| p Indicates L p Norm, 0 <p≤1, represents the weight matrix, represents the Hadamard product of two matrices, and τ represents the regularization parameter.

[0066] Based on the above embodiments, the reconstruction model of the dual regularization constraint is:

[0067]

[0068] Where A represents the system matrix, which is used to convert CT images into projection data; x s Represents spectral CT image The sth channel, p s Represents the projection data corresponding to the s-th channel, that is, the CT image at the s-th energy, with a total of S energies. ||x s || TV Represents the TV norm of the sth channel, the Frobenius norm term Represents a clean tensor image Spectral CT images The difference between them, ρ, λ are regularization parameters, Represents a group of non-locally similar full-channel tensors The global sparse regularization term is a non-convex low-rank regularization term.

[0069] Furthermore, the augmented Lagrange multiplier method is used to update the model shown in the above formula (2). The specific formula is as follows:

[0070]

[0071] Introducing auxiliary variables The Lagrange multiplier Λ and the non-negative penalty parameter β rewrite the above formula (3) into an unconstrained optimization model:

[0072]

[0073] The reconstruction model is decomposed into four sub-problems. The alternating direction multiplier method is used to minimize the above unconstrained optimization model. The iterative formula is as follows:

[0074]

[0075] in, Represents the image reconstructed in this round of spectral CT. Figure 3 A flowchart of a single iteration of spectral CT reconstruction based on subspace-assisted global-nonlocal sparse constraints and spatial TV regularization constraints is provided in an embodiment of the present invention.

[0076] In view of the problem that the reconstruction time of photon counting spectral CT images is long, the reconstruction quality is poor, the decomposition noise surges, and the quality of material decomposition is seriously affected, the method of the present invention is based on subspace decomposition technology, and proposes a multi-prior reconstruction method that combines the non-local low-rank prior of the spectral feature map and the spatial TV regularization constraint of the spectral image to achieve high-precision image reconstruction of spectral CT. Different from the existing spectral CT reconstruction algorithm based on sparse optimization, the present invention combines the non-local structural sparse properties of the feature map corresponding to the spectral CT, selects appropriate reference image blocks in each round of iteration, extracts similar image blocks using the image domain distance, constructs a full-energy feature image tensor block group, integrates it into the reconstruction model to update the objective function, and performs an iterative strategy based on the alternating direction method on the updated objective function to obtain a higher quality reconstructed image. The present invention uses the non-local structural sparsity of the subspace feature image to explore the global-nonlocal correlation between the channels of the spectral image, combines the sparse prior of the gradient image in the channel, and designs an efficient optimization strategy to improve the computational efficiency and the quality of the reconstructed image.

[0077] Figure 4 The results of local detail reconstruction on a digital simulation walnut body using the method of the present invention and other existing methods are shown. Figure 5 The following is a comparison diagram of the material decomposition and reconstruction results of the method of the present invention and other existing methods on actual mouse data. Other existing methods mainly include filtered back projection (FBP) technology, total variation (TV) method, SBM_L0, FTNN and ITS_TV. FBP is a traditional CT image analytical reconstruction algorithm; TV is an iterative reconstruction technology that establishes image priors by constraining the gradient of each channel image; SBM_L0 is a spectrum reconstruction method that combines subspace (BM3D) denoising technology and intra-channel gradient L0 norm constraints; FTNN is a spectrum CT reconstruction technology based on tensor frame set norm; ITS_TV is a multi-prior spectrum reconstruction method that combines tensor CANDECOMP / PARAFAC (CP) decomposition with tensor sparse regularization and intra-channel total variation (TV) regularization. The relevant parameters involved in each method are shown in Table 1. The experimental results verify the effectiveness of the proposed method in spectral CT reconstruction.

[0078] Table 1 Related parameters involved in each algorithm

[0079]

[0080] Corresponding to the above method, such as Figure 6 As shown, an embodiment of the present invention further provides a subspace-assisted multi-priori photon counting spectral CT reconstruction device, including: an image generation module, a subspace decomposition module, a tensor group generation module, a sparse module and a reconstruction module.

[0081] The image generation module is used to generate a spectral CT image according to the projection data of spectral CT imaging, and the spectral CT image includes images of S channels; the subspace decomposition module is used to perform subspace decomposition on the spectral CT image to obtain a noise feature map tensor and an orthogonal basis; the tensor group generation module is used to select a reference image block on the noise feature map tensor, and extract image blocks similar to the reference image block from all channels of the spectral CT image to generate a non-local similar full-channel tensor group; the sparse module is used to construct a global sparse regularization term about the non-local similar full-channel tensor group by using dictionary learning to obtain a denoised non-local similar full-channel tensor group, and multiply it with the orthogonal basis to obtain a clean tensor image; the reconstruction module is used to construct a reconstruction model with dual regularization constraints according to the TV regularization constraints of each channel of the clean tensor image and the spectral CT image, and update the reconstruction model with dual regularization constraints by using the augmented Lagrangian multiplier method until the preset reconstruction quality requirements are met.

[0082] It should be noted that the device provided in the embodiment of the present invention is for implementing the above method. The specific functions thereof can be referred to the above method embodiment and will not be described in detail here.

[0083] Figure 7 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 7 As shown, the electronic device may include: a processor (processor) 701, a communication interface (Communications Interface) 702, a memory (memory) 703 and a communication bus 704, wherein the processor 701, the communication interface 702, and the memory 703 communicate with each other through the communication bus 704. The processor 701 can call the logic instructions in the memory 703 to execute a subspace-assisted multi-prior photon counting spectral CT reconstruction method, which includes: generating a spectral CT image according to the projection data of spectral CT imaging, wherein the spectral CT image includes an image of S channels; performing subspace decomposition on the spectral CT image to obtain a noise feature map tensor and an orthogonal basis; selecting a reference image block on the noise feature map tensor, and extracting image blocks similar to the reference image block from all channels of the spectral CT image to generate a non-local similar full-channel tensor group; using dictionary learning to construct a global sparse regularization term on the non-local similar full-channel tensor group to obtain a denoised non-local similar full-channel tensor group, and multiplying the non-local similar full-channel tensor group with the orthogonal basis to obtain a clean tensor image; constructing a reconstruction model with dual regularization constraints according to the TV regularization constraints of the clean tensor image and each channel of the spectral CT image; and updating the reconstruction model with dual regularization constraints using an augmented Lagrangian multiplier method until a preset reconstruction quality requirement is met.

[0084] In addition, when the logic instructions in the above-mentioned memory 703 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0085] The embodiment of the present invention further provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the methods provided by the above-mentioned method embodiments, for example, including: generating a spectral CT image according to projection data of spectral CT imaging, wherein the spectral CT image includes images of S channels; performing subspace decomposition on the spectral CT image to obtain a noise feature map tensor and an orthogonal basis; selecting a reference image block on the noise feature map tensor, and obtaining a noise feature map tensor from the energy spectrum CT image. Image blocks similar to the reference image blocks are extracted from all channels of the spectral CT image to generate a non-local similar full-channel tensor group; a global sparse regularization term about the non-local similar full-channel tensor group is constructed by using dictionary learning to obtain a denoised non-local similar full-channel tensor group, which is multiplied with the orthogonal basis to obtain a clean tensor image; a reconstruction model with dual regularization constraints is constructed according to the TV regularization constraints of each channel of the clean tensor image and the spectral CT image; and the reconstruction model with dual regularization constraints is updated by using the augmented Lagrangian multiplier method until the preset reconstruction quality requirement is met.

[0086] The embodiment of the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method provided by the above-mentioned method embodiments is implemented, for example, including: generating a spectral CT image according to projection data of spectral CT imaging, wherein the spectral CT image includes an image of S channels; performing subspace decomposition on the spectral CT image to obtain a noise feature map tensor and an orthogonal basis; selecting a reference image block on the noise feature map tensor, and extracting image blocks similar to the reference image block from all channels of the spectral CT image to generate a non-local similar full-channel tensor group; constructing a global sparse regularization term on the non-local similar full-channel tensor group using dictionary learning to obtain a denoised non-local similar full-channel tensor group, and multiplying the denoised non-local similar full-channel tensor group with the orthogonal basis to obtain a clean tensor image; constructing a reconstruction model with dual regularization constraints according to the TV regularization constraints of each channel of the clean tensor image and the spectral CT image; and updating the reconstruction model with dual regularization constraints using the augmented Lagrangian multiplier method until the preset reconstruction quality requirement is met.

[0087] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A subspace-assisted multi-prior photon counting spectral CT reconstruction method, characterized in that: include: Step 1: Generate a spectral CT image according to projection data of spectral CT imaging, wherein the spectral CT image includes images of S channels; Step 2: performing subspace decomposition on the energy spectrum CT image to obtain a noise feature map tensor and an orthogonal basis; Step 3: Select a reference image block on the noise feature map tensor, and extract image blocks similar to the reference image block from all channels of the spectral CT image to generate a non-local similarity full-channel tensor group; Step 4: construct a global sparse regularization term for the non-local similar full-channel tensor group using dictionary learning to obtain a denoised non-local similar full-channel tensor group, and multiply the group with the orthogonal basis to obtain a clean tensor image; Step 5: constructing a reconstruction model with dual regularization constraints according to the TV regularization constraints of each channel of the clean tensor image and the spectral CT image; Step 6: Use the augmented Lagrange multiplier method to update the dual regularization constrained reconstruction model until the preset reconstruction quality requirement is met.

2. The subspace-assisted multi-a priori photon counting spectral CT reconstruction method according to claim 1, characterized in that: In step 1, a simultaneous algebraic reconstruction algorithm SART is used to generate a spectral CT image.

3. The subspace-assisted multi-a priori photon counting spectral CT reconstruction method according to claim 1, characterized in that: Step 3 specifically includes: Step 3.1: Select a reference image block on the noise feature map tensor, extract an image block similar to the reference image block from one channel of the spectral CT image using a block matching algorithm based on image domain distance, and form a similar image block set with all the extracted image blocks; Step 3.2: Change the position of the reference image block and repeat step 3.1 to obtain multiple different sets of similar image blocks; Step 3.3: For all channels of the spectral CT image, perform steps 3.1 and 3.2; Step 3.4: vectorize all sets of similar image blocks, and then stack the vectorization results in the order of channels to obtain the non-local similarity full-channel tensor group.

4. The subspace-assisted multi-priori photon counting spectral CT reconstruction method according to claim 1, characterized in that: The global sparse regularization term is: in, Represents the i-th non-local similarity full channel tensor group The expanded matrix form is represents the Frobenius norm, represents an adaptive dictionary, represents the group sparsity coefficient, ||·|| p Indicates L p Norm, 0 <p≤1, represents the weight matrix, represents the Hadamard product of two matrices, and τ represents the regularization parameter.

5. The subspace-assisted multi-priori photon counting spectral CT reconstruction method according to claim 1, characterized in that: The reconstruction model of the dual regularization constraint is: Where A represents the system matrix, x s Represents spectral CT image The sth channel, p s represents the projection data corresponding to the sth channel, ||x s || TV Represents the TV norm of the sth channel, the Frobenius norm term Represents a clean tensor image Spectral CT images The difference between them, ρ, λ are regularization parameters, Represents a group of non-locally similar full-channel tensors The global sparse regularization term of .

6. The subspace-assisted multi-priori photon counting spectral CT reconstruction method according to claim 1, characterized in that: Step 6 specifically includes: Introducing auxiliary variables The Lagrange multiplier Λ and the non-negative penalty parameter β rewrite the reconstruction model of the dual regularization constraint as: The alternating direction multiplier method is used to solve the above rewritten model; the iterative formula is: in, Represents the reconstructed spectral CT image in this round.

7. A subspace-assisted multi-priori photon counting spectral CT reconstruction device, characterized in that: include: An image generation module, used to generate a spectral CT image according to the projection data of spectral CT imaging, wherein the spectral CT image includes images of S channels; A subspace decomposition module, used for performing subspace decomposition on the energy spectrum CT image to obtain a noise feature map tensor and an orthogonal basis; A tensor group generation module, used for selecting a reference image block on the noise feature map tensor, extracting image blocks similar to the reference image block from all channels of the energy spectrum CT image to generate a non-local similarity full-channel tensor group; A sparse module, used for constructing a global sparse regularization term about the non-local similar full-channel tensor group by using dictionary learning, obtaining a denoised non-local similar full-channel tensor group, and multiplying the denoised non-local similar full-channel tensor group with the orthogonal basis to obtain a clean tensor image; A reconstruction module is used to construct a reconstruction model with dual regularization constraints according to the TV regularization constraints of each channel of the clean tensor image and the energy spectrum CT image, and to update the reconstruction model with dual regularization constraints by using an augmented Lagrange multiplier method until a preset reconstruction quality requirement is met.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images