Deep learning based limited angle projection ptychographic ct image reconstruction method and device

Abstract

The application discloses a deep learning-based limited-angle projection phase contrast CT image reconstruction method and device, which comprises the following steps: acquiring limited-angle projection data of X-ray phase contrast CT scanning through an equal-angle increment strategy; performing flat-field correction and dark-field correction preprocessing on the limited-angle projection data to obtain limited-angle projection data before phase recovery; performing preprocessing on the limited-angle projection data through a phase recovery strategy to obtain limited-angle projection data after phase recovery; performing CT reconstruction on the limited-angle projection data before and after phase recovery through a filtered back projection strategy to obtain CT images of the limited-angle projection before and after phase recovery; constructing a phase contrast CT intelligent image reconstruction network model and training; and performing reconstruction processing on the CT images of the limited-angle projection before and after phase recovery through the trained phase contrast CT intelligent image reconstruction network model to obtain reconstructed phase contrast CT images. The application can quickly and efficiently perform phase contrast CT image reconstruction, effectively suppress limited-angle reconstruction artifacts and improve image quality.

Description

Technical Field

[0001] This invention relates to the field of radiomedical image processing technology, specifically to a method and apparatus for reconstructing finite-angle projection phase-contrast CT images based on deep learning. Background Technology

[0002] Compared to traditional X-ray absorption CT imaging, X-ray phase-contrast CT imaging can produce higher contrast and spatial resolution for weakly absorbing samples, and has great application potential in biomedical and clinical imaging. Currently, long CT scan times and high radiation doses are significant obstacles to the application of X-ray phase-contrast CT imaging, while phase-contrast CT reconstruction based on finite angle projection is one of the feasible strategies to address these issues.

[0003] Traditional X-ray phase-contrast CT image reconstruction techniques can be mainly divided into three categories: (1) analytical methods based on the center slice theorem, such as the filtered back projection algorithm (FBP); (2) iterative reconstruction techniques based on precise mathematical modeling, including algebraic iterative reconstruction techniques (such as ART) and compressed sensing algorithms (such as the total variation minimization algorithm TV); and (3) reconstruction methods based on deep learning (such as FBPCovnet). Analytical methods rely on complete projection data to reconstruct high-quality phase-contrast CT images, which faces the problem of high radiation dose. Iterative reconstruction techniques have high computational costs, resulting in long reconstruction times, and usually rely on available prior knowledge (such as compressed sensing algorithms) to optimize the reconstruction of CT images, thus easily affecting their generalization ability. In addition, iterative reconstruction techniques have limited control over the quality of reconstructed images. With the development of deep learning technology, using deep neural networks to model and optimize the X-ray phase-contrast CT reconstruction problem from big data can more efficiently complete the phase-contrast CT reconstruction task, especially for its finite-angle CT reconstruction requirements. However, purely data-driven deep learning methods, such as DDNet and FBPCovnet, do not effectively incorporate the physical imaging characteristics of phase-contrast CT imaging, resulting in limited interpretability and reconstruction performance in finite-angle phase-contrast CT reconstruction.

[0004] Therefore, in view of the shortcomings of existing technologies, it is necessary to provide a deep neural network model that combines the physical imaging characteristics of X-ray phase-contrast CT to solve the shortcomings of existing technologies in the finite angle reconstruction of X-ray phase-contrast CT images. Summary of the Invention

[0005] To address this, the present invention provides a method and apparatus for finite-angle projection phase-contrast CT image reconstruction based on deep learning. This method utilizes the physical imaging characteristics of X-ray phase-contrast CT imaging to design the structure of a deep neural network, enabling not only rapid and efficient reconstruction of phase-contrast CT images but also effective suppression of finite-angle reconstruction artifacts and improvement of image quality.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a deep learning-based finite-angle projection phase-contrast CT image reconstruction method, comprising:

[0007] Finite-angle projection data of X-ray phase-contrast CT scans were obtained using an equal-angle incremental strategy.

[0008] The obtained finite angle projection data is preprocessed with flat field correction and dark field correction to obtain the finite angle projection data before phase recovery; the finite angle projection data before phase recovery is preprocessed with a phase recovery strategy to obtain the finite angle projection data after phase recovery.

[0009] The finite angle projection data before phase recovery and the finite angle projection data after phase recovery are respectively subjected to CT reconstruction through a filtering back projection strategy to obtain CT images before and after phase recovery of the finite angle projection.

[0010] A phase-contrast CT intelligent image reconstruction network model is constructed; the phase-contrast CT intelligent image reconstruction network model includes an edge enhancement network model and a multi-scale reconstruction network model; the edge enhancement network model and the multi-scale reconstruction network model are trained by setting data respectively to obtain the trained phase-contrast CT intelligent image reconstruction network model;

[0011] The CT image before phase restoration and the CT image after phase restoration, projected with a finite angle, are input into the trained phase-contrast CT intelligent image reconstruction network model; the reconstructed phase-contrast CT image is obtained through reconstruction processing by the trained phase-contrast CT intelligent image reconstruction network model.

[0012] As a preferred embodiment of the deep learning-based finite-angle projection phase-contrast CT image reconstruction method, the calculation formula for the equal-angle increment strategy in the process of acquiring finite-angle projection data of X-ray phase-contrast CT scans is as follows:

[0013]

[0014] In the formula, Δd is the angle increment; N is the number of detector units corresponding to the maximum thickness of the sample under a single projection angle.

[0015] As a preferred embodiment of the deep learning-based finite-angle projection phase-contrast CT image reconstruction method, in the process of preprocessing the obtained finite-angle projection data with flat-field correction and dark-field correction to obtain the finite-angle projection data before phase restoration, the correction expressions for the flat-field correction and the dark-field correction are as follows:

[0016]

[0017] In the formula, I tomo To correct the anterior phase-contrast CT projection image; I flat I is the average image of the flat field image. dark This is the average image of the dark field image;

[0018] In the process of preprocessing the finite angle projection data before phase restoration using a phase restoration strategy to obtain the phase-restored finite angle projection data, the expression of the phase restoration strategy is:

[0019]

[0020] In the formula, (x,y) are the detector plane coordinates; (ξ,v) are the coordinates of (x,y) in the frequency domain. It is the light intensity recorded by the detector when the distance from the sample to the detector is D and the projection angle is θ; for The image after phase recovery; γ is a constant coefficient; λ is the X-ray wavelength; F and F -1 This refers to the Fourier transform and its inverse transform.

[0021] As a preferred embodiment of the deep learning-based finite-angle projection phase-contrast CT image reconstruction method, during the training of the edge enhancement network model and the multi-scale reconstruction network model using the set data, the edge enhancement network model is trained using the CT image before phase recovery and the real edge image data; the multi-scale reconstruction network model is trained using the CT image after phase recovery and the CT image after phase recovery of the full projection.

[0022] As a preferred embodiment of the deep learning-based finite-angle projection phase-contrast CT image reconstruction method, the loss function expression of the phase-contrast CT intelligent image reconstruction network model is as follows:

[0023]

[0024] In the formula, These represent pixel consistency loss, perceptual loss, and edge loss, respectively. α1 and α2 are used to control the weights of the perceptual loss and edge loss, respectively.

[0025] This invention also provides a deep learning-based finite-angle projection phase-contrast CT image reconstruction device, employing the deep learning-based finite-angle projection phase-contrast CT image reconstruction method described above, including:

[0026] The finite angle projection data acquisition module is used to acquire finite angle projection data of X-ray phase-contrast CT scans through an equal angle incremental strategy;

[0027] The projection data preprocessing module is used to perform flat field correction and dark field correction preprocessing on the obtained finite angle projection data to obtain the finite angle projection data before phase restoration; and to preprocess the finite angle projection data before phase restoration through a phase restoration strategy to obtain the finite angle projection data after phase restoration.

[0028] The CT image reconstruction module is used to perform CT reconstruction on the finite angle projection data before phase recovery and the finite angle projection data after phase recovery respectively through a filtering back projection strategy, so as to obtain the CT image before phase recovery and the CT image after phase recovery of the finite angle projection.

[0029] A phase-contrast CT intelligent image reconstruction network model construction and training module is used to construct a phase-contrast CT intelligent image reconstruction network model; the phase-contrast CT intelligent image reconstruction network model includes an edge enhancement network model and a multi-scale reconstruction network model; by setting data, the edge enhancement network model and the multi-scale reconstruction network model are trained respectively to obtain the trained phase-contrast CT intelligent image reconstruction network model.

[0030] The reconstructed phase-contrast CT image acquisition module is used to input the CT image before phase recovery and the CT image after phase recovery, both with finite angle projection, into the trained phase-contrast CT intelligent image reconstruction network model; and to obtain the reconstructed phase-contrast CT image through the reconstruction processing of the trained phase-contrast CT intelligent image reconstruction network model.

[0031] As a preferred embodiment of a deep learning-based finite-angle projection phase-contrast CT image reconstruction device, in the finite-angle projection data acquisition module, during the acquisition of finite-angle projection data from X-ray phase-contrast CT scans using the equal-angle increment strategy, the calculation formula for the equal-angle increment strategy is as follows:

[0032]

[0033] In the formula, Δd is the angle increment; N is the number of detector units corresponding to the maximum thickness of the sample under a single projection angle.

[0034] As a preferred embodiment of a deep learning-based finite-angle projection phase-contrast CT image reconstruction device, in the projection data preprocessing module, during the process of performing plan field correction and dark field correction preprocessing on the obtained finite-angle projection data to obtain the finite-angle projection data before phase restoration, the correction expressions for the plan field correction and the dark field correction are as follows:

[0035]

[0036] In the formula, I tomo To correct the anterior phase-contrast CT projection image; I flat I is the average image of the flat field image. dark This is the average image of the dark field image;

[0037] In the process of preprocessing the finite angle projection data before phase restoration using a phase restoration strategy to obtain the phase-restored finite angle projection data, the expression of the phase restoration strategy is:

[0038]

[0039] In the formula, (x,y) are the detector plane coordinates; (ξ,v) are the coordinates of (x,y) in the frequency domain. It is the light intensity recorded by the detector when the distance from the sample to the detector is D and the projection angle is θ; for The image after phase recovery; γ is a constant coefficient; λ is the X-ray wavelength; F and F -1 This refers to the Fourier transform and its inverse transform.

[0040] As a preferred embodiment of a deep learning-based finite-angle projection phase-contrast CT image reconstruction device, in the phase-contrast CT intelligent image reconstruction network model construction and training module, during the training of the edge enhancement network model and the multi-scale reconstruction network model using the set data, the edge enhancement network model is trained using the CT image before phase recovery and the real edge image data; the multi-scale reconstruction network model is trained using the CT image after phase recovery and the CT image after phase recovery of the full projection.

[0041] As a preferred embodiment of a deep learning-based finite-angle projection phase-contrast CT image reconstruction device, the loss function expression of the phase-contrast CT intelligent image reconstruction network model in the phase-contrast CT intelligent image reconstruction network model construction and training module is as follows:

[0042]

[0043] In the formula, These represent pixel consistency loss, perceptual loss, and edge loss, respectively. α1 and α2 are used to control the weights of the perceptual loss and edge loss, respectively.

[0044] This invention has the following advantages: It acquires finite-angle projection data from X-ray phase-contrast CT scans using an equal-angle incremental strategy; preprocesses the acquired finite-angle projection data with flat-field correction and dark-field correction to obtain the finite-angle projection data before phase recovery; preprocesses the finite-angle projection data before phase recovery using a phase recovery strategy to obtain the finite-angle projection data after phase recovery; performs CT reconstruction on the finite-angle projection data before and after phase recovery using a filtering back-projection strategy to obtain CT images before and after phase recovery of the finite-angle projection; constructs a phase-contrast CT intelligent image reconstruction network model; the phase-contrast CT intelligent image reconstruction network model includes an edge enhancement network model and a multi-scale reconstruction network model; trains the edge enhancement network model and the multi-scale reconstruction network model using set data to obtain a trained phase-contrast CT intelligent image reconstruction network model; inputs the CT images before and after phase recovery of the finite-angle projection into the trained phase-contrast CT intelligent image reconstruction network model; and reconstructs the phase-contrast CT image through the trained phase-contrast CT intelligent image reconstruction network model. Compared to traditional methods, this invention achieves higher reconstruction quality. It significantly improves image clarity and contrast, and demonstrates higher peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) in reconstructed image quality evaluation, as well as lower reconstruction error (MAE) evaluation. Compared to traditional methods, this invention offers superior reconstruction performance. It achieves good CT reconstruction results within the limited-angle scanning range of phase-contrast CT ([0, 150°], [0, 120°], and [0, 90°], effectively suppressing artifacts in limited-angle CT reconstruction and better preserving edge features and details. Furthermore, it possesses high practical application value. The X-ray phase-contrast CT intelligent reconstruction model can rapidly obtain high-quality CT images, exhibiting high reconstruction efficiency and strong generalization ability. This invention designs a deep neural network model MSFR-EEG (with a certain degree of interpretability) based on the physical imaging characteristics of X-ray phase-contrast CT, and provides a finite-angle reconstruction system for X-ray phase-contrast CT images. This system can quickly and efficiently reconstruct phase-contrast CT images, effectively suppress finite-angle reconstruction artifacts and improve image quality, and has high practical application value in the field of phase-contrast CT imaging. Attached Figure Description

[0045] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0046] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0047] Figure 1 This is a schematic diagram of the deep learning-based finite-angle projection phase-contrast CT image reconstruction method provided in Embodiment 1 of the present invention;

[0048] Figure 2 This is a schematic diagram illustrating the specific implementation process of the finite-angle projection phase-contrast CT image reconstruction method based on deep learning provided in Embodiment 1 of the present invention;

[0049] Figure 3 This is a schematic diagram of the multi-scale reconstruction network model module architecture in the deep learning-based finite-angle projection phase-contrast CT image reconstruction method provided in Embodiment 1 of the present invention.

[0050] Figure 4 This is a schematic diagram showing some results of the intelligent reconstruction method for phase-contrast CT images based on deep learning in Embodiment 1 of the present invention, as well as the existing technology.

[0051] Figure 5 This is a schematic diagram of the architecture of the finite-angle projection phase-contrast CT image reconstruction device based on deep learning provided in Embodiment 2 of the present invention. Detailed Implementation

[0052] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] Example 1

[0054] See Figure 1 and Figure 2 Embodiment 1 of the present invention provides a method for reconstructing finite-angle projection phase-contrast CT images based on deep learning, comprising:

[0055] S1. Obtain finite angle projection data of X-ray phase-contrast CT scans using an equal-angle incremental strategy;

[0056] S2. Perform flat field correction and dark field correction preprocessing on the obtained finite angle projection data to obtain the finite angle projection data before phase recovery; perform phase recovery strategy preprocessing on the finite angle projection data before phase recovery to obtain the finite angle projection data after phase recovery.

[0057] S3. The finite angle projection data before phase recovery and the finite angle projection data after phase recovery are respectively subjected to a filtering back projection strategy for CT reconstruction to obtain the CT image before phase recovery and the CT image after phase recovery of the finite angle projection.

[0058] S4. Construct a phase-contrast CT intelligent image reconstruction network model; the phase-contrast CT intelligent image reconstruction network model includes an edge enhancement network model and a multi-scale reconstruction network model; train the edge enhancement network model and the multi-scale reconstruction network model respectively by setting data to obtain the trained phase-contrast CT intelligent image reconstruction network model.

[0059] S5. Input the CT image before phase recovery and the CT image after phase recovery, projected with a finite angle, into the trained phase-contrast CT intelligent image reconstruction network model; reconstruct the phase-contrast CT image through the trained phase-contrast CT intelligent image reconstruction network model.

[0060] In this embodiment, in step S1, finite angle projection data I of X-ray phase-contrast CT scan is obtained through an equal-angle incremental strategy. tomo Where, the finite angle is the projection angle θ < 180°.

[0061] The calculation formula for the equal-angle incremental strategy is as follows:

[0062]

[0063] In the formula, Δd is the angle increment; N is the number of detector units corresponding to the maximum thickness of the sample under a single projection angle.

[0064] In this embodiment, in step S2, the obtained finite angle projection data is preprocessed with flat field correction and dark field correction to obtain the finite angle projection data before phase recovery; the finite angle projection data before phase recovery is preprocessed with a phase recovery strategy to obtain the finite angle projection data after phase recovery.

[0065] Specifically, based on the flat-field images I acquired during X-ray phase-contrast CT scanning flat and dark field image I dark For the obtained projection data I tomo I is obtained by performing flat field correction and dark field correction, and then φ is obtained by performing phase recovery on I using a phase recovery strategy;

[0066] The correction expressions for the flat field correction and the dark field correction are as follows:

[0067]

[0068] In the formula, I tomo To correct the anterior phase-contrast CT projection image; I flat I is the average image of the flat field image. dark This is the average image of the dark field image;

[0069] The expression for the phase recovery strategy is:

[0070]

[0071] In the formula, (x,y) are the detector plane coordinates; (ξ,v) are the coordinates of (x,y) in the frequency domain. It is the light intensity recorded by the detector when the distance from the sample to the detector is D and the projection angle is θ; for The image after phase recovery; γ is a constant coefficient; λ is the X-ray wavelength; F and F -1 This refers to the Fourier transform and its inverse transform.

[0072] In this embodiment, in step S3, the finite angle projection data before phase recovery and the finite angle projection data after phase recovery are respectively subjected to CT reconstruction through a filtering back projection strategy to obtain the CT image before phase recovery and the CT image after phase recovery of the finite angle projection.

[0073] Specifically, the projection data (i.e., I and φ) before and after phase recovery are subjected to filtered back projection (FBP) for CT reconstruction to obtain CT images before and after phase recovery with finite angle projection.

[0074] In this embodiment, in step S4, a phase-contrast CT intelligent image reconstruction network model is constructed; the phase-contrast CT intelligent image reconstruction network model includes an edge enhancement network model and a multi-scale reconstruction network model; the edge enhancement network model and the multi-scale reconstruction network model are trained by setting data respectively to obtain the trained phase-contrast CT intelligent image reconstruction network model.

[0075] Specifically, the phase-contrast CT intelligent image reconstruction network model (multiscale reconstruction framework based on edge-enhanced guidance, MSFR-EEG) utilizes the two physical imaging characteristics of X-ray phase-contrast CT imaging, namely "multiscale" and "edge enhancement", to design the network model, namely the multiscale reconstruction subnetwork (MSRSN) and the edge-enhanced subnetwork (EESN).

[0076] Among them, such as Figure 3 As shown, the multi-scale reconstruction network model MSRSN includes an edge attention module (EAT) and an edge aggregation module (EAG). EAT establishes attention weights of different sizes for edge information at different scales on phase-contrast CT images, optimizing the edges of the reconstructed image and improving its quality and accuracy. EAG effectively utilizes structural information encoded at different scales on phase-contrast CT images, enhancing the model's ability to encode edge and structural features at different scales.

[0077] During the training of the Edge Enhancement Network Model EESN and the Multiscale Reconstruction Network Model MSRN, the Edge Enhancement Network Model EESN is trained using the CT images before phase retrieval and the real edge image data; the Multiscale Reconstruction Network Model MSRN is trained using the CT images after phase retrieval and the CT images after phase retrieval of the full projection.

[0078] In this embodiment, in step S5, the CT image before phase recovery and the CT image after phase recovery, projected with a finite angle, are input into the trained phase-contrast CT intelligent image reconstruction network model; the reconstructed phase-contrast CT image is obtained through reconstruction processing by the trained phase-contrast CT intelligent image reconstruction network model.

[0079] Specifically, firstly, EESN is used to extract the edge prior of the phase-contrast CT image. Then, the EAT and EAG modules guide MSRSN to perform optimized reconstruction, finally outputting a high-quality phase-contrast CT image. The results are as follows: Figure 4 As shown in Table 1, the quantitative evaluation scores are as follows;

[0080]

[0081] Table 1 Quantitative Evaluation Table

[0082] In summary, this invention acquires finite-angle projection data from X-ray phase-contrast CT scans using an equal-angle incremental strategy; preprocesses the acquired finite-angle projection data with flat-field and dark-field corrections to obtain the finite-angle projection data before phase restoration; preprocesses the finite-angle projection data before phase restoration using a phase restoration strategy to obtain the finite-angle projection data after phase restoration; performs CT reconstruction on the finite-angle projection data before and after phase restoration using a filtering back-projection strategy to obtain CT images before and after phase restoration of the finite-angle projection; constructs a phase-contrast CT intelligent image reconstruction network model, which includes an edge enhancement network model and a multi-scale reconstruction network model; trains the edge enhancement network model and the multi-scale reconstruction network model using set data to obtain a trained phase-contrast CT intelligent image reconstruction network model; inputs the CT images before and after phase restoration of the finite-angle projection into the trained phase-contrast CT intelligent image reconstruction network model; and reconstructs the phase-contrast CT image using the trained phase-contrast CT intelligent image reconstruction network model. Compared to traditional methods, this invention achieves higher reconstruction quality. It significantly improves image clarity and contrast, and demonstrates higher peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) in reconstructed image quality evaluation, as well as lower reconstruction error (MAE) evaluation. Compared to traditional methods, this invention offers superior reconstruction performance. It achieves good CT reconstruction results within the limited-angle scanning range of phase-contrast CT ([0, 150°], [0, 120°], and [0, 90°], effectively suppressing artifacts in limited-angle CT reconstruction and better preserving edge features and details. Furthermore, it possesses high practical application value. The X-ray phase-contrast CT intelligent reconstruction model can rapidly obtain high-quality CT images, exhibiting high reconstruction efficiency and strong generalization ability. This invention designs a deep neural network model MSFR-EEG (with a certain degree of interpretability) based on the physical imaging characteristics of X-ray phase-contrast CT, and provides a finite-angle reconstruction system for X-ray phase-contrast CT images. This system can quickly and efficiently reconstruct phase-contrast CT images, effectively suppress finite-angle reconstruction artifacts and improve image quality, and has high practical application value in the field of phase-contrast CT imaging.

[0083] It should be noted that the method of this disclosure embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this disclosure embodiment, and the multiple devices will interact with each other to complete the method described.

[0084] It should be noted that the above description describes some embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0085] Example 2

[0086] See Figure 5 Embodiment 2 of the present invention provides a finite-angle projection phase-contrast CT image reconstruction device based on deep learning, comprising:

[0087] The finite angle projection data acquisition module 001 is used to acquire finite angle projection data of X-ray phase-contrast CT scans through an equal angle incremental strategy;

[0088] The projection data preprocessing module 002 is used to perform flat field correction and dark field correction preprocessing on the obtained finite angle projection data to obtain the finite angle projection data before phase restoration; and to preprocess the finite angle projection data before phase restoration through a phase restoration strategy to obtain the finite angle projection data after phase restoration.

[0089] The CT image reconstruction module 003 is used to perform CT reconstruction on the finite angle projection data before phase recovery and the finite angle projection data after phase recovery respectively through a filtering back projection strategy, so as to obtain the CT image before phase recovery and the CT image after phase recovery of the finite angle projection.

[0090] The phase-contrast CT intelligent image reconstruction network model construction and training module 004 is used to construct a phase-contrast CT intelligent image reconstruction network model; the phase-contrast CT intelligent image reconstruction network model includes an edge enhancement network model and a multi-scale reconstruction network model; by setting data, the edge enhancement network model and the multi-scale reconstruction network model are trained respectively to obtain the trained phase-contrast CT intelligent image reconstruction network model.

[0091] The reconstructed phase-contrast CT image acquisition module 005 is used to input the CT image before phase recovery and the CT image after phase recovery, both with finite angle projection, into the trained phase-contrast CT intelligent image reconstruction network model; and to obtain the reconstructed phase-contrast CT image through the reconstruction processing of the trained phase-contrast CT intelligent image reconstruction network model.

[0092] In this embodiment, in the finite angle projection data acquisition module 001, during the process of acquiring finite angle projection data of X-ray phase-contrast CT scans through the equal angle increment strategy, the calculation formula of the equal angle increment strategy is as follows:

[0093]

[0094] In the formula, Δd is the angle increment; N is the number of detector units corresponding to the maximum thickness of the sample under a single projection angle.

[0095] In this embodiment, in the projection data preprocessing module 002, during the process of performing flat field correction and dark field correction preprocessing on the obtained finite angle projection data to obtain the finite angle projection data before phase restoration, the correction expressions for the flat field correction and the dark field correction are as follows:

[0096]

[0097] In the formula, I tomo To correct the anterior phase-contrast CT projection image; I flat I is the average image of the flat field image. dark This is the average image of the dark field image;

[0098] In the process of preprocessing the finite angle projection data before phase restoration using a phase restoration strategy to obtain the phase-restored finite angle projection data, the expression of the phase restoration strategy is:

[0099]

[0100] In the formula, (x,y) are the detector plane coordinates; (ξ,v) are the coordinates of (x,y) in the frequency domain. It is the light intensity recorded by the detector when the distance from the sample to the detector is D and the projection angle is θ; for The image after phase recovery; γ is a constant coefficient; λ is the X-ray wavelength; F and F -1 This refers to the Fourier transform and its inverse transform.

[0101] In this embodiment, in the phase-contrast CT intelligent image reconstruction network model construction and training module 004, during the training of the edge enhancement network model and the multi-scale reconstruction network model using the set data, the edge enhancement network model is trained using the CT image before phase recovery and the real edge image data; the multi-scale reconstruction network model is trained using the CT image after phase recovery and the CT image after phase recovery of the full projection.

[0102] In this embodiment, in the phase-contrast CT intelligent image reconstruction network model construction and training module 004, the loss function expression of the phase-contrast CT intelligent image reconstruction network model is:

[0103]

[0104] In the formula, These represent pixel consistency loss, perceptual loss, and edge loss, respectively. α1 and α2 are used to control the weights of the perceptual loss and edge loss, respectively.

[0105] It should be noted that the information interaction and execution process between the modules / units of the above system are based on the same concept as the method embodiment in Embodiment 1 of this application, and the resulting technical effects are the same as those in the method embodiment of this application. For details, please refer to the description in the method embodiment shown above in this application, and it will not be repeated here.

[0106] Example 3

[0107] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium storing program code for a deep learning-based finite-angle projection phase-contrast CT image reconstruction method. The program code includes instructions for executing the deep learning-based finite-angle projection phase-contrast CT image reconstruction method of Embodiment 1 or any possible implementation thereof.

[0108] Computer-readable storage media can be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives, SSDs).

[0109] Example 4

[0110] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0111] The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor can call the program instructions to execute the deep learning-based finite-angle projection phase-contrast CT image reconstruction method of Embodiment 1 or any possible implementation thereof.

[0112] Specifically, a processor can be implemented in hardware or software. When implemented in hardware, the processor can be a logic circuit, an integrated circuit, etc. When implemented in software, the processor can be a general-purpose processor that reads software code stored in memory. This memory can be integrated into the processor or located outside the processor and exist independently.

[0113] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0114] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0115] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for reconstructing finite-angle projection phase-contrast CT images based on deep learning, characterized in that, include: Finite-angle projection data of X-ray phase-contrast CT scans were obtained using an equal-angle incremental strategy. The obtained finite angle projection data is preprocessed with flat field correction and dark field correction to obtain the finite angle projection data before phase recovery; the finite angle projection data before phase recovery is preprocessed with a phase recovery strategy to obtain the finite angle projection data after phase recovery. The finite angle projection data before phase recovery and the finite angle projection data after phase recovery are respectively subjected to CT reconstruction through a filtering back projection strategy to obtain CT images before and after phase recovery of the finite angle projection. A phase-contrast CT intelligent image reconstruction network model is constructed; the phase-contrast CT intelligent image reconstruction network model includes an edge enhancement network model and a multi-scale reconstruction network model; the edge enhancement network model and the multi-scale reconstruction network model are trained by setting data respectively to obtain the trained phase-contrast CT intelligent image reconstruction network model; The CT image before phase restoration and the CT image after phase restoration, projected with a finite angle, are input into the trained phase-contrast CT intelligent image reconstruction network model; the reconstructed phase-contrast CT image is obtained through reconstruction processing by the trained phase-contrast CT intelligent image reconstruction network model.

2. The method for reconstructing finite-angle projection phase-contrast CT images based on deep learning according to claim 1, characterized in that, In the process of acquiring finite-angle projection data of X-ray phase-contrast CT scans using the aforementioned equal-angle increment strategy, the calculation formula for the equal-angle increment strategy is as follows: In the formula, Δd is the angle increment; N is the number of detector units corresponding to the maximum thickness of the sample under a single projection angle.

3. The method for reconstructing finite-angle projection phase-contrast CT images based on deep learning according to claim 2, characterized in that, In the process of preprocessing the obtained finite angle projection data with flat field correction and dark field correction to obtain the finite angle projection data before phase recovery, the correction expressions for flat field correction and dark field correction are as follows: In the formula, I tomo To correct the anterior phase-contrast CT projection image; I flat I is the average image of the flat field image. dark This is the average image of the dark field image; In the process of preprocessing the finite angle projection data before phase restoration using a phase restoration strategy to obtain the phase-restored finite angle projection data, the expression of the phase restoration strategy is: In the formula, (x,y) are the detector plane coordinates; (ξ,v) are the coordinates of (x,y) in the frequency domain. It is the light intensity recorded by the detector when the distance from the sample to the detector is D and the projection angle is θ; for The image after phase recovery; γ is a constant coefficient; λ is the X-ray wavelength; F and F -1 This refers to the Fourier transform and its inverse transform.

4. The deep learning-based finite-angle projection phase-contrast CT image reconstruction method according to claim 3, characterized in that, During the training of the edge enhancement network model and the multi-scale reconstruction network model using the specified data, the edge enhancement network model is trained using the CT image before phase recovery and the real edge image data; the multi-scale reconstruction network model is trained using the CT image after phase recovery and the CT image after phase recovery of the full projection.

5. The deep learning-based finite-angle projection phase-contrast CT image reconstruction method according to claim 4, characterized in that, The loss function expression of the phase-contrast CT intelligent image reconstruction network model is as follows: In the formula, These represent pixel consistency loss, perceptual loss, and edge loss, respectively; α1 and α2 are used to control the weights of perceptual loss and edge loss, respectively.

6. A deep learning-based finite-angle projection phase-contrast CT image reconstruction device, employing the deep learning-based finite-angle projection phase-contrast CT image reconstruction method according to any one of claims 1-5, characterized in that, include: The finite angle projection data acquisition module is used to acquire finite angle projection data of X-ray phase-contrast CT scans through an equal angle incremental strategy; The projection data preprocessing module is used to perform flat field correction and dark field correction preprocessing on the obtained finite angle projection data to obtain the finite angle projection data before phase restoration; and to preprocess the finite angle projection data before phase restoration through a phase restoration strategy to obtain the finite angle projection data after phase restoration. The CT image reconstruction module is used to perform CT reconstruction on the finite angle projection data before phase recovery and the finite angle projection data after phase recovery respectively through a filtering back projection strategy, so as to obtain the CT image before phase recovery and the CT image after phase recovery of the finite angle projection. A phase-contrast CT intelligent image reconstruction network model construction and training module is used to construct a phase-contrast CT intelligent image reconstruction network model; the phase-contrast CT intelligent image reconstruction network model includes an edge enhancement network model and a multi-scale reconstruction network model; by setting data, the edge enhancement network model and the multi-scale reconstruction network model are trained respectively to obtain the trained phase-contrast CT intelligent image reconstruction network model. The reconstructed phase-contrast CT image acquisition module is used to input the CT image before phase recovery and the CT image after phase recovery, both with finite angle projection, into the trained phase-contrast CT intelligent image reconstruction network model; and to obtain the reconstructed phase-contrast CT image through the reconstruction processing of the trained phase-contrast CT intelligent image reconstruction network model.

7. The deep learning-based finite-angle projection phase-contrast CT image reconstruction device according to claim 6, characterized in that, In the finite angle projection data acquisition module, during the process of acquiring finite angle projection data of X-ray phase-contrast CT scans using the equal-angle increment strategy, the calculation formula for the equal-angle increment strategy is as follows: In the formula, Δd is the angle increment; N is the number of detector units corresponding to the maximum thickness of the sample under a single projection angle.

8. The deep learning-based finite-angle projection phase-contrast CT image reconstruction device according to claim 7, characterized in that, In the projection data preprocessing module, during the process of performing flat field correction and dark field correction preprocessing on the obtained finite angle projection data to obtain the finite angle projection data before phase restoration, the correction expressions for the flat field correction and the dark field correction are as follows: In the formula, I tomo To correct the anterior phase-contrast CT projection image; I flat I is the average image of the flat field image. dark This is the average image of the dark field image; In the process of preprocessing the finite angle projection data before phase restoration using a phase restoration strategy to obtain the phase-restored finite angle projection data, the expression of the phase restoration strategy is: In the formula, (x,y) are the detector plane coordinates; (ξ,v) are the coordinates of (x,y) in the frequency domain. It is the light intensity recorded by the detector when the distance from the sample to the detector is D and the projection angle is θ; for The image after phase recovery; γ is a constant coefficient; λ is the X-ray wavelength; F and F -1 This refers to the Fourier transform and its inverse transform.

9. The deep learning-based finite-angle projection phase-contrast CT image reconstruction device according to claim 8, characterized in that, In the phase-contrast CT intelligent image reconstruction network model construction and training module, during the training of the edge enhancement network model and the multi-scale reconstruction network model using the set data, the edge enhancement network model is trained using the CT image before phase recovery and the real edge image data; the multi-scale reconstruction network model is trained using the CT image after phase recovery and the CT image after phase recovery of the full projection.

10. The deep learning-based finite-angle projection phase-contrast CT image reconstruction device according to claim 9, characterized in that, In the phase-contrast CT intelligent image reconstruction network model construction and training module, the loss function expression of the phase-contrast CT intelligent image reconstruction network model is as follows: In the formula, These represent pixel consistency loss, perceptual loss, and edge loss, respectively; α1 and α2 are used to control the weights of perceptual loss and edge loss, respectively.

Images