A method, system, device and medium for reconstructing tomographic images
By preprocessing and constraint extraction of CT image reconstruction method, combined with the dual-domain image constraint reconstruction technology, the existing CT image reconstruction method has solved the problems of large computing resource consumption and difficulty in assumption of noise artifacts, and efficient CT image reconstruction is achieved.
Patent Information
- Application Number
- CN202411227515.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-03
- Publication Date
- 2025-05-27
- Estimated Expiration
- 2044-09-03
AI Technical Summary
The existing CT image reconstruction methods consume a lot of computing resources and take a long time, and deep learning-based methods require fine assumptions of noise artifacts during the imaging process, which is difficult and inefficient.
By obtaining the target projection data, input it into the trained image reconstruction model, and image reconstruction is performed. The trained image reconstruction model uses preprocessing full-angle projection data, extracts constrained images, and performs two-domain image constraint reconstruction to update the image reconstruction model parameters.
Effectively reduce the difficulty and computing resource consumption of CT image reconstruction, improve the efficiency of CT image reconstruction, and reduce the hypothetical demand for noise artifacts.
Smart Images
Figure CN119338932B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of CT image reconstruction, and in particular to a method, system, device and medium for reconstructing tomographic scan images. Background Art
[0002] With the development of science and technology, the imaging technology of low-dose CT (Low-dose CT, LDCT) images has become one of the popular research contents for reducing the X-ray radiation loss of patients.
[0003] Currently, traditional CT image reconstruction methods mainly obtain CT reconstructed images by performing multiple forward or back-projection iterative reconstructions and updates on scan data, which requires a large amount of computing resources and takes a long time. In addition, there are some CT image reconstruction methods based on deep learning. Such methods need to make fine assumptions about noise artifacts during the imaging process, which makes the CT image reconstruction difficult, requires a large amount of computing resources, and has low reconstruction efficiency.
[0004] Therefore, the problems existing in the prior art still need to be solved and optimized urgently. Summary of the Invention
[0005] An object of the present invention is to solve at least to some extent one of the technical problems existing in the related art.
[0006] To this end, an object of an embodiment of the present invention is to provide a method, system, device and medium for reconstructing tomographic scan images, which can effectively reduce the difficulty and required computing resources of CT image reconstruction and improve the efficiency of CT image reconstruction.
[0007] Another object of an embodiment of the present application is to provide a system for reconstructing tomographic scan images.
[0008] To achieve the above technical objectives, the technical solutions adopted in the embodiments of the present application include:
[0009] In a first aspect, an embodiment of the present application provides a method for reconstructing tomographic scan images, including:
[0010] Obtaining target projection data, where the target projection data is any one of full-angle chordogram data, sparse-angle chordogram data or limited-angle chordogram data;
[0011] Inputting the target projection data into a trained image reconstruction model to obtain an image reconstruction result;
[0012] Wherein, the trained image reconstruction model is obtained through the following steps:
[0013] Obtain full-angle projection data, and preprocess the full-angle projection data to obtain a back-projection image, where the back-projection image includes a full-angle back-projection image, a sparse-angle back-projection image, and a limited-angle back-projection image;
[0014] Perform constraint extraction on the back-projection image to obtain a constraint image, where the constraint image includes a full-angle constraint image, a sparse-angle constraint image, and a limited-angle constraint image with mutual constraint relationships;
[0015] Perform dual-domain image constraint reconstruction on the constraint image to obtain a dual-domain constraint reconstruction result;
[0016] According to the constraint image and the dual-domain constraint reconstruction result, perform parameter constraint update on the initialized image reconstruction model to obtain the trained image reconstruction model.
[0017] In addition, according to the tomographic image reconstruction method of the above embodiments of the present application, the following additional technical features may also be included:
[0018] Further, in an embodiment of the present application, the preprocessing of the full-angle projection data to obtain a back-projection image includes:
[0019] Perform first filtered back-projection on the full-angle projection data to obtain the full-angle back-projection image;
[0020] According to a preset sparse scanning multiple, perform sparse feature extraction on the full-angle projection data to obtain sparse-angle projection data, and according to a preset limited-angle scanning range, perform limited-angle feature extraction on the full-angle projection data to obtain limited-angle projection data;
[0021] Perform second filtered back-projection on the sparse-angle projection data to obtain the sparse-angle back-projection image;
[0022] Perform third filtered back-projection on the limited-angle projection data to obtain the limited-angle back-projection image.
[0023] Further, in an embodiment of the present application, the performing constraint extraction on the back-projection image to obtain a constraint image includes:
[0024] Perform prior information extraction on the back-projection image to obtain a prior image;
[0025] Perform chord diagram compensation on the prior image to obtain the constraint image.
[0026] Further, in an embodiment of the present application, the prior image includes a full-angle prior image, a sparse-angle prior image, and a limited-angle prior image, and the performing prior information extraction on the back-projection image to obtain a prior image includes:
[0027] Perform first information extraction on the full-angle back-projected image to obtain the full-angle prior image;
[0028] Perform second information extraction on the sparse-angle back-projected image to obtain the sparse-angle prior image;
[0029] Perform third information extraction on the limited-angle back-projected image to obtain the limited-angle prior image.
[0030] Further, in an embodiment of the present application, the performing chordogram compensation on the prior image to obtain the constrained image includes:
[0031] Obtain a first information mask and a second information mask. The first information mask is used to represent the sparse-angle information mask missing in the sparse-angle projection data relative to the full-angle projection data, and the second information mask is used to represent the limited-angle information mask missing in the limited-angle projection data relative to the full-angle projection data;
[0032] Perform mask reconstruction on the prior image according to the first information mask and the second information mask to obtain the constrained image.
[0033] Further, in an embodiment of the present application, the performing mask reconstruction on the prior image according to the first information mask and the second information mask to obtain the constrained image includes:
[0034] Perform first forward-projection compensation on the full-angle prior image according to the full-angle projection data to obtain the full-angle constrained image;
[0035] Perform second forward-projection compensation on the sparse-angle prior image according to the sparse-angle projection data and the first information mask to obtain the sparse-angle constrained image;
[0036] Perform third forward-projection compensation on the limited-angle prior image according to the limited-angle projection data and the second information mask to obtain the limited-angle prior image.
[0037] Further, in an embodiment of the present application, the performing dual-domain image constrained reconstruction on the constrained image to obtain the dual-domain constrained reconstruction result includes:
[0038] Perform chordogram domain restoration on the constrained image to obtain a chordogram restored image;
[0039] Perform image domain restoration on the chordogram restored image to obtain the dual-domain constrained reconstruction result.
[0040] Second, an embodiment of the present application provides a reconstruction system for tomographic images, including:
[0041] An acquisition module, configured to acquire target projection data, where the target projection data is any one of full-angle chordogram data, sparse-angle chordogram data, or limited-angle chordogram data;
[0042] A processing module, configured to input the target projection data into a trained image reconstruction model to obtain an image reconstruction result;
[0043] Wherein, the trained image reconstruction model is obtained through the following steps:
[0044] Acquire full-angle projection data, and perform preprocessing on the full-angle projection data to obtain a back-projection image, where the back-projection image includes a full-angle back-projection image, a sparse-angle back-projection image, and a limited-angle back-projection image;
[0045] Perform constraint extraction on the back-projection image to obtain a constraint image, where the constraint image includes a full-angle constraint image, a sparse-angle constraint image, and a limited-angle constraint image with mutual constraint relationships;
[0046] Perform dual-domain image constraint reconstruction on the constraint image to obtain a dual-domain constraint reconstruction result;
[0047] According to the constraint image and the dual-domain constraint reconstruction result, perform parameter constraint update on an initialized image reconstruction model to obtain the trained image reconstruction model.
[0048] In a third aspect, an embodiment of the present application further provides an electronic device, including:
[0049] At least one processor;
[0050] At least one memory, configured to store at least one program;
[0051] When the at least one program is executed by the at least one processor, the at least one processor implements the method in the first aspect above.
[0052] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, in which a program executable by a processor is stored, and the program executable by the processor is used to implement the method in the first aspect above when executed by the processor.
[0053] The advantages and beneficial effects of the present application will be partially given in the following description, partially will become obvious from the following description, or be understood through the practice of the present application:
[0054] A method, system, device, and medium for reconstructing tomographic scan images disclosed in an embodiment of the present application. In this method, target projection data is obtained, and the target projection data is any one of full-angle chordogram data, sparse-angle chordogram data, or limited-angle chordogram data. The target projection data is input into a trained image reconstruction model to obtain an image reconstruction result. The trained image reconstruction model is obtained through the following steps: obtaining full-angle projection data, and preprocessing the full-angle projection data to obtain a back-projection image, where the back-projection image includes a full-angle back-projection image, a sparse-angle back-projection image, and a limited-angle back-projection image; performing constraint extraction on the back-projection image to obtain a constraint image, where the constraint image includes a full-angle constraint image, a sparse-angle constraint image, and a limited-angle constraint image with mutually constrained relationships; performing dual-domain image constraint reconstruction on the constraint image to obtain a dual-domain constraint reconstruction result; and updating parameter constraints of an initialized image reconstruction model according to the constraint image and the dual-domain constraint reconstruction result to obtain the trained image reconstruction model. This method is based on constraint extraction of the back-projection image to obtain a full-angle constraint image, a sparse-angle constraint image, and a limited-angle constraint image with mutually constrained relationships, which does not require assumptions about noise artifacts, effectively reduces the difficulty and required computing resources for CT image reconstruction, and improves the efficiency of CT image reconstruction. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following introduces the accompanying drawings related to the technical solutions in the embodiments of the present application or the prior art. It should be understood that the accompanying drawings below are only for conveniently and clearly expressing some embodiments of the technical solutions in the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative efforts.
[0056] Figure 1 It is a schematic flowchart of a method for reconstructing tomographic scan images provided by an embodiment of the present application;
[0057] Figure 2 It is a schematic flowchart of the training process of an image reconstruction model provided by an embodiment of the present application;
[0058] Figure 3 It is a schematic structural diagram of a system for reconstructing tomographic scan images provided by an embodiment of the present application;
[0059] Figure 4 It is a schematic structural diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0060] Embodiments of the present application will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are only used to explain the present application, and should not be construed as a limitation of the present application. For the step numbers in the following embodiments, they are only set for the convenience of explanation, and no limitation is imposed on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.
[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.
[0062] Currently, traditional CT image reconstruction methods mainly obtain CT reconstructed images by performing multiple forward or backprojection iterative reconstructions on the scanned data. Although it can improve the quality of the reconstructed image to a certain extent, it has the disadvantages of requiring a large amount of computing resources and long operation time.
[0063] In addition, there are some CT image reconstruction methods based on deep learning. Specifically, supervised CT reconstruction methods usually need to generate a pseudo-paired data on a low-dose chordogram or image and then perform paired data training; self-supervised CT reconstruction methods usually need to make fine assumptions about noise artifacts during the imaging process. These methods are difficult to perform CT image reconstruction, require a large amount of computing resources, and have low reconstruction efficiency.
[0064] In view of this, the embodiments of the present invention provide a method for reconstructing tomographic scan images. This method obtains a full-angle constrained image, a sparse-angle constrained image, and a limited-angle constrained image with mutual constraint relationships based on the constraint extraction of the backprojection image. It does not need to make assumptions about noise artifacts, nor does it need to pair chordogram data or image data, which can effectively reduce the difficulty of CT image reconstruction and the required computing resources, and improve the efficiency of CT image reconstruction.
[0065] Referring to Figure 1 and Figure 2 , in the embodiments of the present application, a method for reconstructing tomographic scan images includes:
[0066] Step 110, obtaining target projection data, where the target projection data is any one of full-angle chordogram data, sparse-angle chordogram data, or limited-angle chordogram data;
[0067] Step 120, inputting the target projection data into a trained image reconstruction model to obtain an image reconstruction result;
[0068] In an embodiment of the present application, any one of the full-angle chordogram data, sparse-angle chordogram data, or limited-angle chordogram data to be recognized can be input into a trained image reconstruction model for image reconstruction. The trained image reconstruction model is used to complete a CT reconstruction task corresponding to the input chordogram data, so as to obtain an image reconstruction result corresponding to the input chordogram data.
[0069] Among them, the trained image reconstruction model is obtained through the following steps:
[0070] Step 130: Obtain full-angle projection data, and preprocess the full-angle projection data to obtain a back-projection image, where the back-projection image includes a full-angle back-projection image, a sparse-angle back-projection image, and a limited-angle back-projection image;
[0071] In some embodiments, the step 130 of preprocessing the full-angle projection data to obtain a back-projection image includes:
[0072] A1: Perform first filtered back-projection on the full-angle projection data to obtain the full-angle back-projection image;
[0073] A2: Extract sparse features from the full-angle projection data according to a preset sparse scanning multiple to obtain sparse-angle projection data, and extract limited-angle features from the full-angle projection data according to a preset limited-angle scanning range to obtain limited-angle projection data;
[0074] A3: Perform second filtered back-projection on the sparse-angle projection data to obtain the sparse-angle back-projection image;
[0075] A4: Perform third filtered back-projection on the limited-angle projection data to obtain the limited-angle back-projection image.
[0076] In an embodiment of the present application, a full-angle CT scan projection data set publicly available on the Internet can be used as the full-angle projection data in the embodiment of the present application, or, on the premise of obtaining the consent of the user, the computer tomography data of several users in the full angle can be used as the full-angle tomography in the embodiment of the present application.
[0077] It can be understood that the first filtered back-projection in step A1 can reconstruct the directly obtained full-angle projection data based on the Filter Back-Projection (FBP) algorithm. Specifically, step A1 can convert the full-angle projection data to the frequency domain through a one-dimensional Fourier transform, then enhance the high-frequency signals and suppress noise based on a filter, and convert the filtered data back to the spatial domain through an inverse Fourier transform to obtain the filtered full-angle projection data; then, perform a back-projection operation on the filtered full-angle projection data according to geometric information and normalize all the back-projection results to obtain a full-angle back-projection image.
[0078] It should be noted that the sparse feature extraction in step A2 can be based on the CT sparse scanning multiple to extract the full-angle projection data to obtain CT projection data of sparse angles (i.e., sparse-angle projection data); the limited-angle feature extraction in step A3 can be based on the CT limited-angle scanning range to extract the full-angle projection data to obtain CT projection data of limited angles. In addition, for steps A3 and A4, they are similar to the content of the aforementioned step A1 and can be simply deduced by analogy, so they will not be elaborated herein.
[0079] Step 140: Perform constrained extraction on the back-projection image to obtain a constrained image, where the constrained image includes a full-angle constrained image, a sparse-angle constrained image, and a limited-angle constrained image having mutual constraint relationships.
[0080] In some embodiments, step 140: Perform constrained extraction on the back-projection image to obtain a constrained image, including:
[0081] B1: Extract prior information from the back-projection image to obtain a prior image;
[0082] Furthermore, the prior image includes a full-angle prior image, a sparse-angle prior image, and a limited-angle prior image. Step B1: Extract prior information from the back-projection image to obtain a prior image, including:
[0083] B11: Extract first information from the full-angle back-projection image to obtain the full-angle prior image;
[0084] B12: Extract second information from the sparse-angle back-projection image to obtain the sparse-angle prior image;
[0085] B13: Extract third information from the limited-angle back-projection image to obtain the limited-angle prior image.
[0086] In the embodiments of the present application, the image reconstruction model includes a prior neural module, a chordogram compensation module, and a dual-domain neural module. Among them, the prior neural module includes three parallel prior neural network units, and each prior neural network unit corresponds to a back-projection image of a certain angle type. Specifically, the network structures of the three parallel prior neural network units in the embodiments of the present application may be the same. It may be a five-level Unet network with residual connections, or a Transformer network constructed based on an attention mechanism, etc. The present application does not limit the specific network structure of the prior neural network unit. While providing prior information for the subsequent chordogram compensation module and dual-domain neural module through mutual learning, the prior neural module mutually constrains, guides, and learns the three parallel prior neural network units based on a loss function, so that there is a mutual constraint relationship among the obtained full-angle constraint image, sparse-angle constraint image, and limited-angle constraint image, thereby achieving the effect of removing noise artifacts.
[0087] It can be understood that step B11 may be to use the full-angle back-projection image as the input of the first prior neural network unit, and use the output image of the first prior neural network unit as the full-angle prior image; steps B12 and B13 are similar to the content of the foregoing step B11 and can be simply deduced by analogy.
[0088] Specifically, the loss function of the prior neural module in the embodiments of the present application can be expressed as:
[0089]
[0090] Among them, is the loss function of the prior neural module, is the full-angle prior image; is the sparse-angle prior image; is the limited-angle prior image.
[0091] B2. Perform chordogram compensation on the prior image to obtain the constraint image.
[0092] Further, the step B2 of performing chordogram compensation on the prior image to obtain the constraint image includes:
[0093] Step B21. Obtain a first information mask and a second information mask. The first information mask is used to represent the sparse-angle information mask missing in the sparse-angle projection data relative to the full-angle projection data, and the second information mask is used to represent the limited-angle information mask missing in the limited-angle projection data relative to the full-angle projection data;
[0094] Step B22. Perform mask reconstruction on the prior image according to the first information mask and the second information mask to obtain the constraint image.
[0095] Further, step B22, masking and reconstructing the prior image according to the first information mask and the second information mask to obtain the constrained image, includes:
[0096] Step B221, performing a first forward projection compensation on the full-angle prior image according to the full-angle projection data to obtain the full-angle constrained image;
[0097] Step B222, performing a second forward projection compensation on the sparse-angle prior image according to the sparse-angle projection data and the first information mask to obtain the sparse-angle constrained image;
[0098] Step B223, performing a third forward projection compensation on the limited-angle prior image according to the limited-angle projection data and the second information mask to obtain the limited-angle prior image.
[0099] In the embodiment of the present application, after the prior neural module outputs three parallel full-angle prior images, sparse-angle prior images, and limited-angle prior images, first, a first information mask corresponding to the sparse-angle prior image and a second information mask corresponding to the limited-angle prior image can be obtained; then, step B221 may be to perform a forward projection (Forward Projection, FP) on the full-angle prior image, which may be specifically implemented based on the first pre-customized FPLayer network layer in the chordogram compensation module to obtain the full-angle chordogram data corresponding to the full-angle prior image; then, based on the original full-angle projection data, compensation is performed on the obtained full-angle chordogram data, which may be specifically implemented by performing a concatenation (Concat) operation on the full-angle chordogram data and the full-angle projection data, so as to obtain the full-angle constrained image.
[0100] It can be understood that step B222 may be to perform a forward projection on the sparse-angle prior image based on the second pre-customized FPLayer network layer in the chordogram compensation module. After obtaining the sparse-angle chordogram data, the sparse-angle projection data and the first information mask are used to perform a substitution operation on the sparse-angle chordogram data to improve the effectiveness of the chordogram data and obtain the sparse-angle constrained image. Specifically, the equivalent expression form of the second forward projection compensation in step B222 may be:
[0101]
[0102] where p is the sparse-angle constrained image; A is a backpropagation-enabled forward projection operation; is the sparse-angle prior image; p sv is the sparse-angle projection data; M svIt is a first information mask, that is, a sparse angle information mask in which the sparse angle projection data is missing relative to the full angle projection data.
[0103] It should be noted that, for step B223, the contents of the other aforementioned steps B222 are similar and can be simply deduced by analogy, so this application will not go into details here.
[0104] Step 150, performing dual-domain image constrained reconstruction on the constrained image to obtain a dual-domain constrained reconstruction result;
[0105] In some embodiments, the step 150 of performing dual-domain image constrained reconstruction on the constrained image to obtain a dual-domain constrained reconstruction result includes:
[0106] C1. Perform chord diagram domain restoration on the constrained image to obtain a chord diagram restored image;
[0107] C2. Perform image domain restoration on the chord diagram restoration image to obtain the dual-domain constrained reconstruction result.
[0108] In an embodiment of the present application, the dual-domain constrained reconstruction result includes a full-angle reconstruction result, a sparse angle reconstruction result, and a limited-angle reconstruction result. The dual-domain neural module includes three parallel dual-domain network units, each of which includes a chord diagram domain network unit and an image domain network unit, and each dual-domain network unit corresponds to a constrained image of an angle type. Specifically, the chord diagram domain network unit and the image domain network unit can both be five-layer Unet networks with residual connections. The dual-domain neural module constrains, guides, and learns the three parallel dual-domain network units based on the loss function, so that the obtained full-angle reconstruction result, sparse angle reconstruction result, and limited-angle reconstruction result have a mutually constrained relationship.
[0109] It can be understood that, for the full-angle constrained image in the constrained image, step 150 can be to input the full-angle constrained image into the corresponding dual-domain network unit. Specifically, the full-angle constrained image can be first input into the chord diagram domain network unit to obtain the chord diagram domain network unit output; then, the chord diagram domain network unit output is FBP reconstructed based on a pre-customized FBPLayer network layer to obtain the chord diagram restored image; then, the chord diagram restored image is input into the image domain network unit to obtain the dual-domain constrained reconstruction result corresponding to the full-angle projection data.
[0110] It should be noted that for sparse angle constraint images and limited angle constraint images, their content is similar to that of the aforementioned full angle constraint images, and can be simply inferred by analogy, so this application will not elaborate further. It is worth mentioning that for the pre-customized FPLayer network layer and FBPLayer network layer, they can be deep learning custom network layers. Specifically, the FPLayer network layer is used to implement the forward projection operation, and the FBPLayer network layer is used to implement the FBP reconstruction operation, and they can be specifically constructed based on the deep learning framework used, which will not be elaborated in this application.
[0111] Specifically, the loss function of the dual-domain neural module in the embodiments of this application can be expressed as:
[0112]
[0113] Among them, is the loss function of the dual-domain neural module, is the full angle reconstruction result; is the sparse angle reconstruction result; is the limited angle reconstruction result.
[0114] Step 160, according to the constraint image and the dual-domain constraint reconstruction result, perform parameter constraint update on the initialized image reconstruction model to obtain the trained image reconstruction model.
[0115] In the embodiments of this application, step 160 may be to calculate the loss function regarding image consistency based on the constraint image and the dual-domain constraint reconstruction result, and it can be specifically expressed as:
[0116]
[0117] Among them, L rc is the loss function regarding image consistency; μ ld is the full angle projection data.
[0118] It can be understood that the parameter constraint update may be to determine the loss value of model training based on the target loss function obtained by summing the loss function of the prior neural module, the loss function of the dual-domain neural module, and the loss function regarding image consistency, and use the backpropagation algorithm to update the initialized image reconstruction model. After iterating several rounds, the trained image reconstruction model can be obtained. The specific iteration discussion can be preset in advance, or it is considered that the training is completed when the accuracy of the model reaches the requirement.
[0119] Next, a reconstruction system for tomographic images proposed according to the embodiments of this application will be described in detail with reference to the accompanying drawings.
[0120] Refer to Figure 3, a reconstruction system for tomographic scan images proposed in an embodiment of the present application includes:
[0121] An acquisition module 101, configured to acquire target projection data, where the target projection data is any one of full-angle chordogram data, sparse-angle chordogram data, or limited-angle chordogram data;
[0122] A processing module 102, configured to input the target projection data into a trained image reconstruction model to obtain an image reconstruction result;
[0123] Among them, the trained image reconstruction model is obtained through the following steps:
[0124] Acquire full-angle projection data, and preprocess the full-angle projection data to obtain a back-projection image, where the back-projection image includes a full-angle back-projection image, a sparse-angle back-projection image, and a limited-angle back-projection image;
[0125] Perform constraint extraction on the back-projection image to obtain a constraint image, where the constraint image includes a full-angle constraint image, a sparse-angle constraint image, and a limited-angle constraint image with mutual constraint relationships;
[0126] Perform dual-domain image constraint reconstruction on the constraint image to obtain a dual-domain constraint reconstruction result;
[0127] According to the constraint image and the dual-domain constraint reconstruction result, perform parameter constraint update on the initialized image reconstruction model to obtain the trained image reconstruction model.
[0128] It can be understood that the content in the above method embodiments is applicable to the system embodiments of the present application. The functions specifically implemented by the system embodiments of the present application are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those in the above method embodiments.
[0129] Referring to Figure 4 , an embodiment of the present application further provides an electronic device, including:
[0130] At least one processor 201;
[0131] At least one memory 202, configured to store at least one program;
[0132] When the at least one program is executed by the at least one processor 201, the at least one processor 201 implements the above method embodiments.
[0133] Similarly, it can be understood that the content in the above method embodiments is applicable to the device embodiments of the present application. The functions specifically implemented by the device embodiments of the present application are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those in the above method embodiments.
[0134] The embodiments of the present application also provide a computer-readable storage medium, which stores a program executable by a processor 201. The program executable by the processor 201, when executed by the processor 201, is used to implement the above method embodiments.
[0135] Similarly, the content in the above method embodiments is applicable to this computer-readable storage medium embodiment. The functions specifically implemented by this computer-readable storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those in the above method embodiments.
[0136] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order mentioned in the operation diagrams. For example, depending on the functions / operations involved, two consecutive blocks shown may actually be executed substantially simultaneously, or the blocks can sometimes be executed in the reverse order. Additionally, the embodiments presented and described in the flowcharts of the present application are provided by way of example for the purpose of providing a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logical flows presented herein. Alternative embodiments are contemplated, where the order of various operations is changed and where sub-operations described as part of a larger operation are executed independently.
[0137] Furthermore, although the present application is described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the functions and / or features may be integrated in a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It can also be understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present application. Rather, given the attributes, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the modules will be understood within the ordinary skills of an engineer. Therefore, those skilled in the art can implement the present application as set forth in the claims without undue experimentation. It can also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present application, which is determined by the full scope of the appended claims and their equivalents.
[0138] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method of the embodiment of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.
[0139] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in combination with an instruction execution system, apparatus, or device.
[0140] More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection part with one or more wirings (electronic device), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber device, and portable compact disc read-only memory (CDROM). Additionally, a computer-readable medium can even be paper or other suitable media on which a program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or otherwise processing it as appropriate, and then storing it in a computer memory.
[0141] It should be understood that various parts of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), and the like.
[0142] In the foregoing description of this specification, the descriptions with reference to the terms "one embodiment / example", "another embodiment / example", or "certain embodiments / examples", etc., mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.
[0143] Although the embodiments of the present application have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present application. The scope of the present application is defined by the claims and their equivalents.
[0144] The above has specifically described the preferred embodiments of the present application, but the present application is not limited to the embodiments. Those skilled in the art can make various equivalent deformations or substitutions without departing from the spirit of the present application, and these equivalent deformations or substitutions are all included within the scope defined by the claims of the present application.
Claims
1. A method for reconstructing a tomographic image, characterized in that: include: Acquire target projection data, wherein the target projection data is any one of full-angle chord diagram data, sparse-angle chord diagram data, or limited-angle chord diagram data; Inputting the target projection data into a trained image reconstruction model to obtain an image reconstruction result; The trained image reconstruction model is obtained by the following steps: Acquire full-angle projection data, and pre-process the full-angle projection data to obtain a back-projection image, wherein the back-projection image includes a full-angle back-projection image, a sparse-angle back-projection image, and a limited-angle back-projection image; Performing constraint extraction on the back-projection image to obtain a constrained image, wherein the constrained image includes a full-angle constrained image, a sparse angle constrained image, and a limited-angle constrained image having a mutual constraint relationship; Performing dual-domain image constrained reconstruction on the constrained image to obtain a dual-domain constrained reconstruction result; According to the constrained image and the dual-domain constrained reconstruction result, updating the parameter constraints of the initialized image reconstruction model to obtain the trained image reconstruction model; The step of performing constraint extraction on the back-projection image to obtain a constrained image comprises: Extracting prior information from the back-projected image to obtain a prior image; Performing chord diagram compensation on the prior image to obtain the constrained image; The prior image includes a full-angle prior image, a sparse angle prior image and a limited-angle prior image, and the prior information extraction of the back-projection image to obtain the prior image includes: Performing first information extraction on the full-angle back-projection image to obtain the full-angle priori image; Performing second information extraction on the sparse angle back-projection image to obtain the sparse angle prior image; Extracting third information from the angle-limited back-projection image to obtain the angle-limited priori image; The performing chord diagram compensation on the prior image to obtain the constrained image includes: Acquire a first information mask and a second information mask, wherein the first information mask is used to characterize a sparse angle information mask where the sparse angle projection data is missing relative to the full angle projection data, and the second information mask is used to characterize a limited angle information mask where the limited angle projection data is missing relative to the full angle projection data; The priori image is mask reconstructed according to the first information mask and the second information mask to obtain the constrained image.
2. The method for reconstructing a tomographic image according to claim 1, characterized in that: The preprocessing of the full-angle projection data to obtain a back-projection image includes: Performing a first filtered back-projection on the full-angle projection data to obtain the full-angle back-projection image; According to a preset sparse scanning multiple, sparse feature extraction is performed on the full-angle projection data to obtain sparse angle projection data, and, according to a preset limited angle scanning range, limited angle feature extraction is performed on the full-angle projection data to obtain limited angle projection data; Performing a second filtered back-projection on the sparse angular projection data to obtain the sparse angular back-projection image; A third filtered back-projection is performed on the limited-angle projection data to obtain the limited-angle back-projection image.
3. The method for reconstructing a tomographic image according to claim 1, characterized in that: The step of performing mask reconstruction on the prior image according to the first information mask and the second information mask to obtain the constrained image includes: According to the full-angle projection data, performing a first front projection compensation on the full-angle priori image to obtain the full-angle constrained image; According to the sparse angle projection data and the first information mask, performing a second front projection compensation on the sparse angle prior image to obtain the sparse angle constrained image; According to the angle-limited projection data and the second information mask, a third front projection compensation is performed on the angle-limited a priori image to obtain the angle-limited a priori image.
4. The method for reconstructing a tomographic image according to claim 1, characterized in that: The performing dual-domain image constrained reconstruction on the constrained image to obtain a dual-domain constrained reconstruction result includes: Performing chord diagram domain restoration on the constrained image to obtain a chord diagram restored image; Perform image domain restoration on the chord diagram restoration image to obtain the dual-domain constrained reconstruction result.
5. A tomographic image reconstruction system, characterized in that: include: An acquisition module, used for acquiring target projection data, wherein the target projection data is any one of full-angle chord diagram data, sparse-angle chord diagram data or limited-angle chord diagram data; A processing module, used for inputting the target projection data into a trained image reconstruction model to obtain an image reconstruction result; The trained image reconstruction model is obtained by the following steps: Acquire full-angle projection data, and pre-process the full-angle projection data to obtain a back-projection image, wherein the back-projection image includes a full-angle back-projection image, a sparse-angle back-projection image, and a limited-angle back-projection image; Performing constraint extraction on the back-projection image to obtain a constrained image, wherein the constrained image includes a full-angle constrained image, a sparse angle constrained image, and a limited-angle constrained image having a mutual constraint relationship; Performing dual-domain image constrained reconstruction on the constrained image to obtain a dual-domain constrained reconstruction result; According to the constrained image and the dual-domain constrained reconstruction result, updating the parameter constraints of the initialized image reconstruction model to obtain the trained image reconstruction model; The step of performing constraint extraction on the back-projection image to obtain a constrained image comprises: Extracting prior information from the back-projected image to obtain a prior image; Performing chord diagram compensation on the prior image to obtain the constrained image; The prior image includes a full-angle prior image, a sparse angle prior image and a limited-angle prior image, and the prior information extraction of the back-projection image to obtain the prior image includes: Performing first information extraction on the full-angle back-projection image to obtain the full-angle priori image; Performing second information extraction on the sparse angle back-projection image to obtain the sparse angle prior image; Extracting third information from the angle-limited back-projection image to obtain the angle-limited priori image; The performing chord diagram compensation on the prior image to obtain the constrained image includes: Acquire a first information mask and a second information mask, wherein the first information mask is used to characterize a sparse angle information mask where the sparse angle projection data is missing relative to the full angle projection data, and the second information mask is used to characterize a limited angle information mask where the limited angle projection data is missing relative to the full angle projection data; The priori image is mask reconstructed according to the first information mask and the second information mask to obtain the constrained image.
6. An electronic device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 4.
7. A computer-readable storage medium storing a program executable by a processor, characterized in that: The program executable by the processor is used to implement the method according to any one of claims 1 to 4 when executed by the processor.