Image reconstruction method, device and computer equipment

Abstract

The application relates to an image reconstruction method, device and computer equipment. The method comprises the following steps: acquiring projection data obtained through medical imaging scanning; preprocessing the projection data to obtain preprocessed projection data; inputting the preprocessed projection data into a trained data optimization model to obtain optimized projection data; and obtaining a reconstructed image according to the optimized projection data. The method can reduce the complexity of image reconstruction and improve the speed of image reconstruction.

Description

Technical Field

[0001] This application relates to the field of medical image processing technology, and in particular to an image reconstruction method, apparatus and computer equipment. Background Technology

[0002] Clinical reconstruction methods for positron emission tomography (PET) include filtered back projection (FBP) and ordered subsets expectation maximum (OSEM).

[0003] In FBP, the projection data (profiles) at different angles obtained from scanning are first transformed to the frequency domain, filtered by a pre-designed filter, and then transformed back to the spatial domain for backprojection to obtain the reconstructed image. However, because the filter is sensitive to noise, the reconstructed image obtained by the FBP method is prone to artifacts. OSEM has a higher noise suppression level than FBP, but due to its iterative reconstruction, the reconstruction complexity is higher and the iteration process is more time-consuming, resulting in a slower image reconstruction speed.

[0004] Therefore, current image reconstruction techniques suffer from slow reconstruction speed. Summary of the Invention

[0005] Therefore, it is necessary to provide a faster image reconstruction method, apparatus, computer device, computer-readable storage medium, and computer program product to address the aforementioned technical problems.

[0006] Firstly, this application provides an image reconstruction method, including:

[0007] Acquire projection data obtained from medical imaging scans;

[0008] The projection data is preprocessed to obtain preprocessed projection data;

[0009] The preprocessed projection data is input into the trained data optimization model to obtain optimized projection data;

[0010] The reconstructed image is obtained based on the optimized projection data.

[0011] In one embodiment, the method further includes:

[0012] Obtain a first sample dataset; the first sample dataset includes sample projection data and filtered projection data corresponding to the sample projection data, wherein the filtered projection data is obtained based on different filtering parameters;

[0013] The data optimization model to be trained is iteratively trained based on the first sample dataset to obtain the trained data optimization model.

[0014] In one embodiment, obtaining the reconstructed image based on the optimized projection data includes:

[0015] The optimized projection data is back-projected to obtain the reconstructed image.

[0016] In one embodiment, the method further includes:

[0017] Obtain a second sample dataset; the second sample dataset includes sample projection data and iterated projection data corresponding to the sample projection data, the iterated projection data being obtained based on the original system matrix through multiple iterations;

[0018] The data optimization model to be trained is iteratively trained based on the second sample dataset to obtain the trained data optimization model.

[0019] In one embodiment, the preprocessing of the projection data includes:

[0020] Remove random event information and / or scattering event information from the projection data.

[0021] In one embodiment, obtaining the reconstructed image based on the optimized projection data includes:

[0022] The original system matrix is ​​trimmed to obtain the trimmed system matrix;

[0023] The reconstructed image is obtained based on the cropped system matrix and the optimized projection data.

[0024] In one embodiment, the trimming of the original system matrix includes:

[0025] Remove the portion of the original system matrix that is related to the noise of the projected data.

[0026] In one embodiment, obtaining the reconstructed image based on the cropped system matrix and the optimized projection data includes:

[0027] Based on the cropped system matrix, the optimized projection data is iteratively updated to obtain the reconstructed image.

[0028] Secondly, this application also provides an image reconstruction apparatus, comprising:

[0029] The acquisition module is used to acquire projection data obtained from medical imaging scans;

[0030] The processing module is used to preprocess the projection data to obtain preprocessed projection data;

[0031] The optimization module is used to input the preprocessed projection data into the trained data optimization model to obtain optimized projection data;

[0032] The reconstruction module is used to obtain a reconstructed image based on the optimized projection data.

[0033] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0034] Acquire projection data obtained from medical imaging scans;

[0035] The projection data is preprocessed to obtain preprocessed projection data;

[0036] The preprocessed projection data is input into the trained data optimization model to obtain optimized projection data;

[0037] The reconstructed image is obtained based on the optimized projection data.

[0038] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0039] Acquire projection data obtained from medical imaging scans;

[0040] The projection data is preprocessed to obtain preprocessed projection data;

[0041] The preprocessed projection data is input into the trained data optimization model to obtain optimized projection data;

[0042] The reconstructed image is obtained based on the optimized projection data.

[0043] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0044] Acquire projection data obtained from medical imaging scans;

[0045] The projection data is preprocessed to obtain preprocessed projection data;

[0046] The preprocessed projection data is input into the trained data optimization model to obtain optimized projection data;

[0047] The reconstructed image is obtained based on the optimized projection data.

[0048] The aforementioned image reconstruction methods, apparatus, computer devices, computer-readable storage media, and computer program products acquire projection data obtained from medical imaging scans, preprocess the projection data to obtain preprocessed projection data, input the preprocessed projection data into a trained data optimization model to obtain optimized projection data, and obtain a reconstructed image based on the optimized projection data. For both FBP and OSEM methods, the projection data (profile) extracted from the sinogram can be preprocessed and directly input into the trained data optimization model to obtain a reconstructed image based on the optimized projection data. This eliminates the need for frequency domain transformation and its inverse transformation in the FBP method, and also eliminates the need for multiple iterations in the OSEM method, thereby reducing the complexity of image reconstruction and increasing its speed. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating an image reconstruction method in one embodiment;

[0051] Figure 2 This is a flowchart illustrating an example of an FBP-based image reconstruction method.

[0052] Figure 3 This is a flowchart illustrating an OSEM-based image reconstruction method as an example.

[0053] Figure 4 This is a structural block diagram of an image reconstruction apparatus in one embodiment;

[0054] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0056] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0057] In one exemplary embodiment, such as Figure 1 As shown, an image reconstruction method is provided. This embodiment illustrates the method applied to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0058] Step S102: Obtain projection data obtained from medical imaging scans.

[0059] The medical imaging can be, but is not limited to, PET imaging. This application does not impose any restrictions on this, as long as image reconstruction is performed using FBP or OSEM. The projection data can be a sine curve projection at the same scanning angle in a sine graph. The sine graph can be a polar coordinate projection matrix of the raw data acquired by the scanning device. It forms sine curve projection data by recording the angle and radial distance of the Line of Response (LOR). In some embodiments, the rows of the sine graph can correspond to the detector rotation angle (LOR angle), and the columns can correspond to the distance from the annihilation event to the detector rotation center (LOR radial distance). In this case, the projection data can be the row data in the sine graph.

[0060] Optionally, a sinogram can be obtained through medical imaging scanning. This sinogram is then input into a terminal, which extracts projection data from the sinogram based on the detector rotation angle during the medical imaging process. For example, the terminal can acquire a sinogram obtained from a PET scan and extract the row data from it as projection data (profile).

[0061] Step S104: Preprocess the projection data to obtain preprocessed projection data.

[0062] Optionally, after acquiring the projection data, the terminal can preprocess the projection data to form preprocessed projection data. In practical applications, for FBP reconstruction, preprocessing includes, but is not limited to, noise suppression, artifact correction, and data augmentation; for OSEM reconstruction, preprocessing may involve removing random events and / or scattering events from the projection data.

[0063] Step S106: Input the preprocessed projection data into the trained data optimization model to obtain the optimized projection data.

[0064] The trained data optimization model can be a data processing model for medical image reconstruction, which can be, but is not limited to, a deep learning model. The optimized projection data can be the output data of the trained data optimization model.

[0065] For example, the terminal can pre-train a data optimization model to obtain the trained data optimization model. After preprocessing the projection data extracted from the sine wave, the preprocessed projection data is input into the model to obtain the optimized projection data output by the model. In practical applications, for FBP reconstruction, a filter model can be pre-trained, and the preprocessed projection data can be input into the filter model to obtain the optimized projection data. The filter model can directly filter the preprocessed projection data in the spatial domain without transforming the preprocessed projection data to the frequency domain for filtering, or transforming it back to the spatial domain after filtering, thus reducing processing complexity. For OSEM reconstruction, a system matrix model can be pre-trained, and the preprocessed projection data can be input into the system matrix model to obtain the optimized projection data. Then, the optimized projection data can be iterated once to obtain the reconstructed image, avoiding the slowdown caused by multiple iterations.

[0066] Step S108: Obtain the reconstructed image based on the optimized projection data.

[0067] Among them, the reconstructed image can be an image that restores the distribution of the radioactive tracer in the organism.

[0068] Optionally, the terminal can obtain a reconstructed image of the medical imaging scan based on the optimized projection data. In practical applications, for FBP reconstruction, the optimized projection data output by the filter model can be back-projected to obtain the reconstructed image; for OSEM reconstruction, the optimized projection data output by the system matrix model can be iterated once to obtain the reconstructed image.

[0069] The aforementioned image reconstruction method acquires projection data obtained from medical imaging scans, preprocesses the projection data to obtain preprocessed projection data, inputs the preprocessed projection data into a trained data optimization model to obtain optimized projection data, and obtains a reconstructed image based on the optimized projection data. For both the FBP and OSEM methods, the projection data extracted from the sine wave can be preprocessed and directly input into the trained data optimization model to obtain the reconstructed image based on the optimized projection data. This eliminates the need for frequency domain transformation and its inverse transformation in the FBP method, and also eliminates the need for multiple iterations in the OSEM method, thereby reducing the complexity of image reconstruction and increasing its speed.

[0070] In practical applications, for FBP reconstruction, a filter model can be pre-trained. This filter model can be understood as a model that directly predicts the filtered results of the projected data in the spatial domain, and can be implemented using deep learning. In an exemplary embodiment, the specific training process of the filter model may include: obtaining a first sample dataset; the first sample dataset includes sample projected data and the corresponding filtered projected data, the filtered projected data being obtained based on different filtering parameters; iteratively training the data optimization model to be trained based on the first sample dataset to obtain the trained data optimization model.

[0071] The first sample dataset can be a dataset used to train the filter model. The sample projection data can be projection data used as training samples, for example, projection data from multiple sine waves corresponding to different noise levels and imaging systems at the same scanning angle. The filtered projection data can be the labeled data of the sample projection data, specifically the data obtained by sequentially performing Fourier transform, filtering, and inverse Fourier transform on the sample projection data based on traditional FBP. The filtering parameters can be the parameters used in traditional FBP to filter the Fourier-transformed sample projection data. The data optimization model to be trained can be a filter model whose filtering parameters need to be determined. The trained data optimization model can be a filter model with determined filtering parameters.

[0072] For example, to simplify the FBP reconstruction process, the terminal can acquire multiple sample projection data and the corresponding filtered projection data for each sample projection data. The sample projection data is used as training samples, and the filtered projection data is used as sample labels to form a first sample dataset for model training, resulting in a trained data-optimized model. For instance, the terminal can extract sample projection data from multiple sine waves, perform Fourier transform, filtering, and inverse Fourier transform on the sample projection data sequentially based on the traditional FBP method, obtaining filtered projection data. The sample projection data and the filtered projection data are combined to form the first sample dataset. The terminal can preprocess the sample projection data to obtain preprocessed sample projection data, input the preprocessed sample projection data into the filter model to be trained, and output optimized sample projection data. The optimized sample projection data is compared with the filtered projection data, and the filter parameters of the filter model are continuously adjusted based on the difference between the two until the difference is less than a preset threshold or the change in difference is less than a preset threshold. The filter model is determined based on the final obtained filter parameters and used as the trained data-optimized model.

[0073] In this embodiment, by acquiring a first sample dataset, iteratively training the data optimization model to be trained based on the first sample dataset, a trained data optimization model is obtained. The data optimization model used for FBP reconstruction can be predetermined, and the sine wave projection data can be filtered directly in the spatial domain without frequency domain transformation or transformation back to the spatial domain after filtering, which reduces the complexity of image reconstruction and improves the speed of image reconstruction.

[0074] In an exemplary embodiment, step S108 may specifically include: backprojecting the optimized projection data to obtain a reconstructed image.

[0075] Optionally, after inputting the preprocessed projection data into a trained filter model to obtain optimized projection data, the terminal can backproject the optimized projection data to obtain a reconstructed image of the medical imaging scan. It is understood that for multiple projection data points extracted from the sine wave (corresponding to multiple LORs), each projection data point can be input into a pre-trained filter model to obtain multiple optimized projection data points. Each optimized projection data point can then be backprojected to obtain a reconstructed image.

[0076] In this embodiment, the reconstructed image is obtained by backprojecting the optimized projection data. The reconstructed image can be obtained by directly filtering the sine wave projection data in the spatial domain, which reduces the complexity of image reconstruction and improves the speed of image reconstruction.

[0077] In practical applications, for OSEM reconstruction, a system matrix model can be pre-trained. The system matrix model can be understood as a model that predicts the system matrix of the OSEM method, and can be implemented, but is not limited to, through deep learning. In an exemplary embodiment, the specific training process of the system matrix model may include: obtaining a second sample dataset; the second sample dataset includes sample projection data and iteratively derived projection data corresponding to the sample projection data, the iteratively derived projection data being obtained based on the original system matrix through multiple iterations; and iteratively training the data optimization model to be trained based on the second sample dataset to obtain the trained data optimization model.

[0078] The second sample dataset can be a dataset used to train the system matrix model. The iteratively projected data can be the labeled data of the sample projection data, specifically data obtained through multiple iterations based on traditional OSEM reconstruction. The data optimization model to be trained can be a system matrix model for which the system matrix needs to be determined. The trained data optimization model can be a system matrix model with a determined system matrix.

[0079] In OSEM reconstruction, PET data can be modeled as:

[0080] ;

[0081] in, For sensitivity plots, specifically... , For fuzzy correction matrix, For geometric correction matrix, This is the positron physical effects matrix. Among them, This is the attenuation correction factor; This is the normalization correction factor; For the projection data index, it represents the first digit in the sine curve. The projection data corresponds to the nth projection data. One LOR line; The voxel index represents the first voxel of the detected object. Individual factors; Represents a random event; Indicates a scattering event; Represents the reconstructed image; This represents the projection data.

[0082] Based on the above formula, the original system matrix can be: .

[0083] For example, in order to simplify the OSEM reconstruction process, the terminal can obtain multiple sample projection data and the iterative projection data corresponding to each sample projection data, use the sample projection data as training samples, and use the iterative projection data as sample labels to form a second sample dataset for model training, thereby obtaining a trained data-optimized model.

[0084] For example, the above OSEM reconstruction formula can be transformed into:

[0085] ;

[0086] in, This is a trimmed version of the original system matrix; the specific values ​​are known. The unknown part refers to the portion of the original system matrix that needs to be determined. The terminal can extract sample projection data from multiple sine waves. Projecting data for each sample based on traditional OSEM methods Projection data is obtained through multiple iterations. And based on the trimmed version of the original system matrix Obtain the iterative projection data Then, a second sample dataset is constructed. The terminal can also project data from the samples. Remove random events and scattering events The preprocessed sample projection data is obtained. .use express , express Construct the system matrix model to be trained Among them, can be Interpreted as input to a deep learning model, The system matrix model can be understood as the output of a deep learning model. Its training process is based on deep learning, determining the model's output based on the input and output. The process.

[0087] In this embodiment, by obtaining a second sample dataset, the data optimization model to be trained is iteratively trained based on the second sample dataset to obtain a trained data optimization model. This model can be used to construct a data optimization model for OSEM reconstruction, enabling PET reconstruction to be achieved in just one iteration, thus reducing the complexity of image reconstruction and improving the speed of image reconstruction.

[0088] In an exemplary embodiment, step S104 may specifically include: removing random event information and / or scattering event information from the projection data.

[0089] Among them, random event information can be the impact of random events on the projected data. Scattering event information can be the impact of scattering events on the projected data.

[0090] Optionally, after obtaining the projection data from the sine wave, the terminal can remove the influence of random events and / or scattering events from the projection data to obtain preprocessed projection data. For example, let the projection data be... After removing the effects of random and scattering events, the preprocessed projection data is obtained. .

[0091] In this embodiment, by removing random event information and / or scattering event information from the projection data, the influence of random events and scattering events can be removed from the projection data, reducing the complexity of model training and improving the efficiency of model training.

[0092] In an exemplary embodiment, step S108 may specifically include: cropping the original system matrix to obtain a cropped system matrix; and obtaining a reconstructed image based on the cropped system matrix and the optimized projection data.

[0093] The trimmed system matrix can be: .

[0094] Optionally, in determining The specific values ​​are used to obtain the trained system matrix model. Afterwards, the terminal can process the pre-processed projection data. Input the trained system matrix model to obtain optimized projection data. The terminal can access the original system matrix. The system matrix is ​​obtained by pruning. Using the trimmed system matrix For the optimized projection data Correction is performed to obtain the reconstructed image. .

[0095] In this embodiment, the original system matrix is ​​cropped to obtain the cropped system matrix. Based on the cropped system matrix and the optimized projection data, the reconstructed image is obtained. The optimized projection data can be corrected using a geometric correction matrix and a positron physical effect matrix with known specific parameters, directly obtaining the reconstructed image and increasing the accuracy of image reconstruction.

[0096] In an exemplary embodiment, the step of trimming the original system matrix described above may specifically include: removing the portion of the original system matrix that is related to noise in the projected data.

[0097] The part related to projection data noise refers to the terms in the original system matrix that can be considered as noise, such as the fuzzy correction matrix. Attenuation correction factor and normalized correction factor .

[0098] Optionally, the terminal can be in the original system matrix. Remove the portion related to projection data noise. The trimmed system matrix is ​​obtained. .

[0099] In this embodiment, by removing the part of the original system matrix that is related to the noise of the projection data, the optimized projection data can be corrected using the geometric correction matrix and the positron physical effect matrix with known specific parameters, and the reconstructed image can be obtained directly, which increases the accuracy of image reconstruction.

[0100] In an exemplary embodiment, the step of obtaining the reconstructed image based on the cropped system matrix and the optimized projection data may specifically include: performing iterative updates on the optimized projection data based on the cropped system matrix to obtain the reconstructed image.

[0101] Optionally, the terminal can be based on the trimmed system matrix. For the optimized projection data Perform one iteration update to obtain the reconstructed image. .

[0102] In this embodiment, the optimized projection data is iteratively updated based on the cropped system matrix to obtain the reconstructed image. The reconstructed image can be obtained directly through a single iteration of the cropped system matrix, which reduces the complexity of image reconstruction and improves the efficiency of image reconstruction.

[0103] To facilitate a deeper understanding of the embodiments of this application by those skilled in the art, a specific example will be used for illustration below.

[0104] To address the time-consuming nature of image reconstruction, this application proposes an AI-based signal processing method based on noise reduction. This method performs one-dimensional signal processing on the sinogram for both FBP and OSEM methods. For FBP, an optimized filter design is proposed. For OSEM, the system matrix used for modeling is cropped, divided into a sinogram and an image space. The relevant terms in the sinogram are treated as noise factors, and deep learning is used to learn the overall mapping factor for this part. The remaining part serves as the cropped system matrix for the sinogram and image space. Compared to other image-to-image end-to-end reconstruction methods, this method preserves signal integrity to the greatest extent, improves image reconstruction speed, and avoids the influence of iterative methods on initial conditions and the iteration process.

[0105] For FBP in image reconstruction, since FBP is a profile that has undergone a one-dimensional Fourier transform and is filtered in the frequency domain, it is equivalent to performing a convolution operation on the profile in the spatial domain. Therefore, during deep learning training, the Fourier transform operation can be avoided, and the filter parameters can be learned directly from the profile.

[0106] For OSEM, PET data can be modeled as follows:

[0107] ;

[0108] in, It is a measured value (projected data), and it is the first... Reconstructed image values ​​of individual pixels. It is a system matrix. and They are the first Random events and scattering events on the LOR line. Random estimation is obtained by individually counting the time-delay channels in the system hardware design. Scattering estimation is obtained through Monte Carlo simulation of the actual scanned object. In this application, both random events and scattering events need to be saved for the samples used in deep learning.

[0109] It is a sensitivity map, which models physical processes such as geometry projection, inter-crystal scattering, crystal penetration, non-linearity, and positron range. ,in It acts on the sine space, and Acting in image space, These are the attenuation correction factor and the normalization correction factor, respectively, which also act on the sine graph space. It models a blur kernel to perform filtering in the image space. It depends on the type of nuclide used in the current acquisition and generates blur kernels in real time for convolution operations. The lookup table is obtained through system simulation. It is a unified manifold approximation and projection (UMAP) obtained by transforming CT images of real-time scanned objects. This also needs to be calculated in real time during the reconstruction process.

[0110] In this embodiment, Disassemble, and believe as well as These are the related terms in the sinusoidal graph space, which can be considered as noise in the profile. Using deep learning methods, we can find the correlation between the original profiles and the collected profiles. The overall parameter mapping relationship is obtained, thus eliminating the need for real-time calculation of the above parameters and accelerating the reconstruction speed.

[0111] The specific process of image reconstruction is as follows:

[0112] Step S301, Data preparation.

[0113] To accommodate data diversity, the training data includes FBP-reconstructed images and OSEM-reconstructed images with different noise levels and imaging systems, as well as sine wave data acquired by the system during the reconstruction of these images. First, the projection data of the sine waves acquired at different angles are obtained as the training input data for FBP.

[0114] For OSEM, the acquired LOR (Local Area Reconstruction) needs to be processed. First, random and scattering events need to be subtracted. Then, the OSEM-reconstructed image is used as the target, and a cropped version of the system matrix is ​​applied. Propagation is performed to obtain profiles. This establishes a mapping relationship between the input and output profiles.

[0115] The training objective can be to find the parameter mapping relationship between the original profiles and the projected profiles at each angle. Model design can include using deep learning models such as DeepFilterNet for noise suppression as training models for parameter tuning. The training output can be a .pt parameter file.

[0116] Step S302, Engineering Deployment.

[0117] First, configure the filter parameter file to the system.

[0118] Secondly, the parameters of the FBP and OSEM methods mentioned above are allocated and stored in the system. When the actual scanning is performed, the stored parameters are read directly from the noisy sine wave projection data to perform a deblurring operation, and the filtered profiles are obtained.

[0119] Then, for this processed profile, if it is FBP, it is directly back-projected; if it is OSEM, it is iteratively reconstructed in one iteration.

[0120] When selecting FBP as the data reconstruction algorithm, parameters are pre-loaded for filtering. Specifically, checkboxes are provided during deployment for users to select between the default non-AI filter and the filter designed in this system. Users can simultaneously evaluate the generated images during image analysis, and the evaluation results can be used as feedback to further optimize the filter during training.

[0121] In one exemplary embodiment, such as Figure 2 The diagram shows a flowchart of an FBP-based image reconstruction method, including the following steps:

[0122] Step S401: Obtain the first sample dataset; the first sample dataset includes sample projection data and the filtered projection data corresponding to the sample projection data, and the filtered projection data is obtained based on different filtering parameters;

[0123] Step S402: Iteratively train the data optimization model to be trained based on the first sample dataset to obtain the trained data optimization model;

[0124] Step S403: Obtain projection data obtained from medical imaging scans;

[0125] Step S404: Preprocess the projection data to obtain preprocessed projection data;

[0126] Step S405: Input the preprocessed projection data into the trained data optimization model to obtain the optimized projection data;

[0127] Step S406: Backproject the optimized projection data to obtain the reconstructed image.

[0128] Optionally, during the model training phase, the terminal can extract sample projection data from multiple sine waves and obtain filtered projection data corresponding to each sample projection data based on the traditional FBP method. The sample projection data and the filtered projection data are combined to form a first sample dataset. A data optimization model is then trained based on this first sample dataset to obtain the trained data optimization model. During the model application phase, the terminal can extract projection data from the sine waves obtained from PET scans, preprocess the projection data, and input the preprocessed projection data into the trained data optimization model to obtain optimized projection data. By backprojecting the optimized projection data, a reconstructed image can be obtained.

[0129] The aforementioned FBP-based image reconstruction method can preprocess the projection data extracted from the sine wave and directly input it into the trained data optimization model. Then, the reconstructed image is obtained through backprojection. This eliminates the need for frequency domain transformation and its inverse transformation in the FBP method, thereby reducing the complexity of image reconstruction and increasing its speed.

[0130] In one exemplary embodiment, such as Figure 3 The diagram shows a flowchart of an OSEM-based image reconstruction method, including the following steps:

[0131] Step S501: Obtain the second sample dataset; the second sample dataset includes sample projection data and the iterative projection data corresponding to the sample projection data. The iterative projection data is obtained based on the original system matrix through multiple iterations.

[0132] Step S502: Iteratively train the data optimization model to be trained based on the second sample dataset to obtain the trained data optimization model;

[0133] Step S503: Obtain projection data obtained from medical imaging scan;

[0134] Step S504: Remove random event information and / or scattering event information from the projection data to obtain preprocessed projection data;

[0135] Step S505: Input the preprocessed projection data into the trained data optimization model to obtain the optimized projection data;

[0136] Step S506: Remove the part of the original system matrix that is related to the noise of the projection data to obtain the cropped system matrix. Based on the cropped system matrix, perform iterative updates on the optimized projection data to obtain the reconstructed image.

[0137] Optionally, during the model training phase, the terminal can extract sample projection data from multiple sine waves and obtain iterated projection data corresponding to each sample projection data based on the traditional OSEM method. The sample projection data and the iterated projection data are combined to form a second sample dataset. The data optimization model is trained based on the second sample dataset to obtain the trained data optimization model. During the model application phase, the terminal can extract projection data from the sine wave obtained from the PET scan. ,from After removing random event information and / or scattering event information, the preprocessed projection data is obtained. ,Will enter , thus obtaining optimized projection data The cropped system matrix is ​​obtained by removing the part related to the noise of the projected data from the original system matrix. ,right Perform one iteration update to obtain the reconstructed image. .

[0138] The aforementioned OSEM-based image reconstruction method can directly obtain the reconstructed image by preprocessing the projection data extracted from the sine wave and performing a single iteration based on the cropped system matrix, without needing to perform multiple iterations in the OSEM method. This reduces the complexity of image reconstruction and improves its speed.

[0139] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0140] Based on the same inventive concept, this application also provides an image reconstruction apparatus for implementing the image reconstruction method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more image reconstruction apparatus embodiments provided below can be found in the limitations of the image reconstruction method described above, and will not be repeated here.

[0141] In one exemplary embodiment, such as Figure 4 As shown, an image reconstruction apparatus is provided, including: an acquisition module 602, a processing module 604, an optimization module 606, and a reconstruction module 608, wherein:

[0142] The acquisition module 602 is used to acquire projection data obtained from medical imaging scans;

[0143] Processing module 604 is used to preprocess the projection data to obtain preprocessed projection data;

[0144] Optimization module 606 is used to input the preprocessed projection data into a trained data optimization model to obtain optimized projection data;

[0145] The reconstruction module 608 is used to obtain a reconstructed image based on the optimized projection data.

[0146] In an exemplary embodiment, the image reconstruction apparatus further includes a training module for acquiring a first sample dataset; the first sample dataset includes sample projection data and filtered projection data corresponding to the sample projection data, wherein the filtered projection data is obtained based on different filtering parameters; and the data optimization model to be trained is iteratively trained according to the first sample dataset to obtain the trained data optimization model.

[0147] In an exemplary embodiment, the reconstruction module 608 is further configured to backproject the optimized projection data to obtain the reconstructed image.

[0148] In an exemplary embodiment, the training module is further configured to obtain a second sample dataset; the second sample dataset includes sample projection data and iterated projection data corresponding to the sample projection data, the iterated projection data being obtained through multiple iterations based on the original system matrix; and iteratively training the data optimization model to be trained according to the second sample dataset to obtain the trained data optimization model.

[0149] In an exemplary embodiment, the processing module 604 is further configured to remove random event information and / or scattering event information from the projection data.

[0150] In an exemplary embodiment, the reconstruction module 608 is further configured to crop the original system matrix to obtain a cropped system matrix; and to obtain the reconstructed image based on the cropped system matrix and the optimized projection data.

[0151] In an exemplary embodiment, the reconstruction module 608 described above is further configured to remove the portion of the original system matrix that is related to noise in the projected data.

[0152] In an exemplary embodiment, the reconstruction module 608 is further configured to perform iterative updates on the optimized projection data based on the cropped system matrix to obtain the reconstructed image.

[0153] Each module in the aforementioned image reconstruction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0154] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements an image reconstruction method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0155] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0156] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0157] In one exemplary embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above-described method embodiments.

[0158] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0159] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0160] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0161] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0162] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An image reconstruction method, characterized in that, The method includes: Acquire projection data obtained from medical imaging scans; The projection data is preprocessed to obtain preprocessed projection data; The preprocessed projection data is input into the trained data optimization model to obtain optimized projection data; The reconstructed image is obtained based on the optimized projection data.

2. The method according to claim 1, characterized in that, The method further includes: Obtain a first sample dataset; the first sample dataset includes sample projection data and filtered projection data corresponding to the sample projection data, wherein the filtered projection data is obtained based on different filtering parameters; The data optimization model to be trained is iteratively trained based on the first sample dataset to obtain the trained data optimization model.

3. The method according to claim 2, characterized in that, The step of obtaining the reconstructed image based on the optimized projection data includes: The optimized projection data is back-projected to obtain the reconstructed image.

4. The method according to claim 1, characterized in that, The method further includes: Obtain a second sample dataset; the second sample dataset includes sample projection data and iterated projection data corresponding to the sample projection data, the iterated projection data being obtained based on the original system matrix through multiple iterations; The data optimization model to be trained is iteratively trained based on the second sample dataset to obtain the trained data optimization model.

5. The method according to claim 4, characterized in that, The preprocessing of the projection data includes: Remove random event information and / or scattering event information from the projection data.

6. The method according to claim 4, characterized in that, The step of obtaining the reconstructed image based on the optimized projection data includes: The original system matrix is ​​trimmed to obtain the trimmed system matrix; The reconstructed image is obtained based on the cropped system matrix and the optimized projection data.

7. The method according to claim 6, characterized in that, The trimming of the original system matrix includes: Remove the portion of the original system matrix that is related to the noise of the projected data.

8. The method according to claim 6, characterized in that, The step of obtaining the reconstructed image based on the cropped system matrix and the optimized projection data includes: Based on the cropped system matrix, the optimized projection data is iteratively updated to obtain the reconstructed image.

9. An image reconstruction apparatus, characterized in that, The device includes: The acquisition module is used to acquire projection data obtained from medical imaging scans; The processing module is used to preprocess the projection data to obtain preprocessed projection data; The optimization module is used to input the preprocessed projection data into the trained data optimization model to obtain optimized projection data; The reconstruction module is used to obtain a reconstructed image based on the optimized projection data.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

Images