Computed tomography image processing method, system, equipment and medium
By constructing a training data set of sinusoidal data pairs and training a denoising diffusion model, the problems of excessive output smoothing and difficult GAN training in low-dose CT image optimization are solved, and the effect of efficient denoising and contrast preservation is achieved.
Patent Information
- Application Number
- CN202411998275.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-31
- Publication Date
- 2025-05-06
AI Technical Summary
The prior art uses deep learning techniques to optimize low-dose CT images, which can easily lead to excessive smoothing of outputs, and it is difficult to train and converge with Generative Adversarial Networks (GANs).
The training data set is constructed, including multiple pairs of sinusoidal graph data, each pair of sinusoidal graphs including a low dose CT sinusoidal graph data and its corresponding conventional dose CT sinusoidal graph data. These data are used to train the cold diffusion model to generate a denoising diffusion model. This model uses a degradation operator that maintains the mean to gradually add noise to the conventional dose CT sine graph data to degrade it into the corresponding low dose CT sine graph data to gradually eliminate noise in the low dose CT sine graph data.
While removing noise from low-dose CT sinusoidal data in the projection field, the details in the low-dose CT sinusoidal data are maintained, thereby maintaining the contrast of the DICOM images reconstructed based on the denoised low-dose CT sinusoidal data. This method effectively reduces the number of steps to add noise, speeds up the model's inference, and improves data processing efficiency.
Smart Images
Figure CN119941557A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of medical image processing technology, and in particular to a method, system, device and medium for computer tomography image processing. Background Art
[0002] Computed Tomography (CT) is an imaging modality widely used in clinical diagnosis. Since the ionizing X-ray radiation in CT scans may cause health risks such as hair loss and cancer, radiation exposure should be reduced. However, in low-dose X-ray CT, severe artifacts are often present due to photon starvation, beam hardening, etc., which reduce the reliability of diagnosis. Therefore, obtaining high-quality denoising methods from low-dose X-ray CT data has attracted great interest in the field of medical imaging.
[0003] In recent years, people have devoted their efforts to developing deep learning techniques to optimize low-dose CT images. These methods can be mainly divided into two categories: post-processing in the DICOM image domain and restoration in the projection domain. Although the optimization methods in the DICOM image domain have achieved excellent denoising effects, these methods usually result in over-smoothed outputs. In order to alleviate the problem of over-smoothing, generative adversarial networks (GANs) are usually introduced to preserve more textures and details, as close as possible to the quality of conventional dose CT images. However, due to the adversarial nature of GANs, it is usually difficult to train and converge. Summary of the invention
[0004] In view of the shortcomings of the prior art described above, the purpose of the present application is to provide a computer tomography image processing method, system, device and medium, which are used to solve the technical problems in the prior art of using deep learning technology to optimize the output of low-dose CT images, such as excessive smoothing, and the difficulty in training and convergence of generative adversarial networks.
[0005] To achieve the above objectives and other related objectives, the present application provides a computer tomography image processing method, comprising:
[0006] Constructing a training data set, wherein the training data set includes a plurality of sinogram data pairs, each sinogram data pair including a low-dose CT sinogram data and its corresponding conventional-dose CT sinogram data;
[0007] The cold diffusion model is trained using the training data set to generate a denoised diffusion model, wherein the cold diffusion model uses a mean-preserving degradation operator to gradually add noise to conventional-dose CT sinusoidal data to degenerate it into corresponding low-dose CT sinusoidal data as a diffusion process, and gradually eliminates the noise in the low-dose CT sinusoidal data as an inverse diffusion process;
[0008] The low-dose CT sinogram data to be denoised is input into the denoising diffusion model to obtain denoised CT sinogram data.
[0009] In an optional embodiment of the present application, constructing a training data set includes:
[0010] Obtain different low-dose CT raw data and corresponding conventional-dose CT raw data;
[0011] Normalizing the different low-dose CT raw data and the corresponding conventional-dose CT raw data to obtain different low-dose CT sinogram data and the corresponding conventional-dose CT sinogram data;
[0012] Sine graph data pairs are constructed using different low-dose CT sinogram data and corresponding conventional-dose CT sinogram data to form the training data set.
[0013] In an optional embodiment of the present application, a pair of sinogram data is constructed using different low-dose CT sinogram data and corresponding conventional-dose CT sinogram data to form the training data set, including:
[0014] Constructing a sinusoidal data pair using different low-dose CT sinusoidal data and corresponding conventional-dose CT sinusoidal data;
[0015] The low-dose CT sinogram data in each sinogram data pair are grouped according to the dose corresponding to the low-dose CT sinogram data, so that the sinogram data pairs with the same dose are divided into the same image group, and the sinogram data pairs with different doses are divided into different image groups, thereby forming the training data set.
[0016] In an optional embodiment of the present application, the mean-preserving degradation operator is defined as
[0017] x t =α t x0+(1-α t )x T
[0018] Among them, x0 is the conventional dose CT sinusoidal data, x T is the low-dose CT sinogram data, T is the total number of diffusion steps, t is one of 1, 2, ..., T, α t is the noise factor corresponding to the tth step, α t <α t-1 .
[0019] In an optional embodiment of the present application, T is greater than or equal to 5 and less than or equal to 20.
[0020] In an optional embodiment of the present application, the low-dose CT sinogram data to be denoised is input into the denoising diffusion model to obtain denoised CT sinogram data, and then the following steps are further included:
[0021] Image reconstruction is performed based on the denoised CT sinogram data to generate a DICOM image.
[0022] In an optional embodiment of the present application, image reconstruction is performed based on the denoised CT sinogram data to generate a DICOM image, including:
[0023] Based on the denoised CT sinogram data, an iterative reconstruction method based on a model is used to perform image reconstruction to generate a DICOM image.
[0024] To achieve the above objectives and other related objectives, the present application also provides a computer tomography image processing system, comprising:
[0025] A training set construction module, used to construct a training data set, wherein the training data set includes a plurality of sinusoidal data pairs, each sinusoidal data pair includes a low-dose CT sinusoidal data and its corresponding conventional-dose CT sinusoidal data;
[0026] A model training module, used for training the cold diffusion model using the training data set to generate a denoised diffusion model, wherein the cold diffusion model uses a mean-preserving degradation operator to gradually add noise to conventional-dose CT sinusoidal data to degenerate it into corresponding low-dose CT sinusoidal data as a diffusion process, and gradually eliminates the noise in the low-dose CT sinusoidal data as an inverse diffusion process;
[0027] The denoising module is used to input the low-dose CT sinogram data to be denoised into the denoising diffusion model to obtain the denoised CT sinogram data.
[0028] To achieve the above objectives and other related objectives, the present application provides an electronic device, the electronic device comprising:
[0029] one or more processors;
[0030] A storage device is used to store one or more programs, and when the one or more programs are executed by the one or more processors, the electronic device implements any of the above-mentioned computer tomography image processing methods.
[0031] To achieve the above-mentioned purpose and other related purposes, the present application provides a computer-readable storage medium, characterized in that a computer program is stored thereon, and when the computer program is executed by a processor of a computer, the computer executes the computer tomography image processing method described in any one of the above claims.
[0032] As described above, the above-mentioned computer tomography image processing method, system, device and medium of the present application have the following
[0033] Beneficial effects:
[0034] By constructing a training data set, the training data set includes a plurality of sinogram data pairs, each sinogram data pair includes a low-dose CT sinogram data and its corresponding conventional-dose CT sinogram data; using the training data set to train a cold diffusion model to generate a denoising diffusion model, wherein the cold diffusion model uses a mean-preserving degradation operator to gradually add noise to the conventional-dose CT sinogram data to degenerate it into the corresponding low-dose CT sinogram data as a diffusion process, and gradually eliminates the noise in the low-dose CT sinogram data as an inverse diffusion process; the low-dose CT sinogram data to be denoised is input into the denoising diffusion model to obtain the denoised CT sinogram data. By using the present application, the noise of the low-dose CT sinogram data can be removed in the projection image domain while maintaining the detail information in the low-dose CT sinogram data, thereby maintaining the contrast of the DICOM image reconstructed based on the denoised low-dose CT sinogram data.
[0035] In addition, in the diffusion process of the cold diffusion model of the present application, conventional-dose CT sinusoidal data is used as the starting point of the diffusion process, and low-dose CT sinusoidal data is used as the end point of the diffusion process. A mean-maintaining degradation operator is used to gradually add noise to the conventional-dose CT sinusoidal data to degenerate it into the corresponding low-dose CT sinusoidal data, thereby effectively reducing the number of steps for adding noise and speeding up the inference speed of the model.
[0036] At the same time, since the amount of sinusoidal data before CT image reconstruction is smaller than that of DICOM images after reconstruction, denoising on the projection image domain with a smaller data volume can process data more efficiently and reduce the consumption of computing resources and processing time. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1 A schematic diagram of a flow chart of a computer tomography image processing method provided in an embodiment of the present application;
[0038] Figure 2 Schematic diagram of the diffusion and denoising process of the denoising diffusion model of this application;
[0039] Figure 3 It is a comparison chart of DICOM images reconstructed based on low-dose CT sinusoidal data, conventional-dose CT sinusoidal data, and low-dose CT sinusoidal data optimized by the present application;
[0040] Figure 4Shown is a structural block diagram of a computer tomography image processing system provided by an embodiment of the present application;
[0041] Figure 5 Shown is a structural schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION
[0042] The following describes the embodiments of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the contents disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.
[0043] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present application, and thus the drawings only show components related to the present application rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed at will, and the component layout may also be more complicated.
[0044] In the following description, a large number of details are discussed to provide a more thorough explanation of the embodiments of the present application. However, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present application difficult to understand.
[0045] The present application proposes a novel denoising method for computed tomography image processing based on a projection domain based on a cold diffusion structure. Compared with the post-processing method in the DICOM image domain, the image denoising method in the projection domain of the present application reduces noise at the source, avoids the propagation of noise in the reconstruction process and the generation of complex artifacts, and the image denoising method in the projection domain of the present application maintains the detail information in the low-dose CT sinusoidal data, so that the DICOM image reconstructed using the optimized sinusoidal data is superior to the optimization method in the DICOM image domain in both quantitative indicators and visual qualitative indicators.
[0046] Figure 1 A computer tomography image processing method in an exemplary embodiment of the present application is shown, including steps S10-S40.
[0047] First, execute step S10 to construct a training data set.
[0048] Specifically, when constructing a training data set, it is necessary to first collect and obtain different low-dose CT raw data and corresponding conventional-dose CT raw data in pairs as the raw data for constructing the training data set. In order to ensure the consistency and comparability of the data and reduce the impact of equipment differences, all raw data are from the same series of CT scanners, such as Siemens Medical's SOMATOM go.Fit. Among them, sinusoidal data can also be called projection data, which is the attenuation measurement value formed after the X-ray passes through different parts of the patient's body.
[0049] As an example, the original data may include four types of dose scanning data, namely 25%, 50%, 75% and 100% dose CT sinusoidal data, wherein 25%, 50%, 75% dose CT sinusoidal data are defined as low-dose CT sinusoidal data, 100% dose CT sinusoidal data are defined as conventional dose CT sinusoidal data, and each low-dose CT sinusoidal data corresponds to a conventional dose CT sinusoidal data.
[0050] Then, the different low-dose CT raw data and the corresponding conventional-dose CT raw data are normalized to obtain different low-dose CT sinogram data and corresponding conventional-dose CT sinogram data. Normalization can eliminate the influence of different scanning conditions and ensure that the cold diffusion model receives consistent and standardized input data, thereby improving training efficiency and model performance.
[0051] Finally, sinogram data pairs are constructed using different low-dose CT sinogram data and their corresponding conventional-dose CT sinogram data to form the training data set. That is, the training data set includes multiple sinogram data pairs, and each sinogram data pair includes a low-dose CT sinogram data and its corresponding conventional-dose CT sinogram data.
[0052] As an example, for example, a 25% dose CT sinogram data and its corresponding 100% dose CT sinogram data can be paired to form a sinogram data pair; a 50% dose CT sinogram data and its corresponding 100% dose CT sinogram data can be paired to form a sinogram data pair; a 75% dose CT sinogram data and its corresponding 100% dose CT sinogram data can be paired to form a sinogram data pair.
[0053] In a specific embodiment of the present application, when constructing sinusoidal data pairs using different low-dose CT sinusoidal data and corresponding conventional-dose CT sinusoidal data to form the training data set, sinusoidal data pairs can be first constructed using different low-dose CT sinusoidal data and corresponding conventional-dose CT sinusoidal data; and then grouping is performed according to the dose corresponding to the low-dose CT sinusoidal data in each sinusoidal data pair, so that the sinusoidal data pairs with the same dose are divided into the same image group, and the sinusoidal data pairs with different doses are divided into different image groups, thereby forming the training data set.
[0054] As an example, the sinusoidal data pairs can be divided into three groups: 25% dose image group, 50% dose image group, and 75% dose image group according to the dose corresponding to the low-dose CT sinusoidal data in each sinusoidal data pair, wherein the low-dose CT sinusoidal data in all sinusoidal data pairs in the 25% dose image group are 25% dose CT sinusoidal data, the low-dose CT sinusoidal data in all sinusoidal data pairs in the 50% dose image group are 50% dose CT sinusoidal data, and the low-dose CT sinusoidal data in all sinusoidal data pairs in the 75% dose image group are 75% dose CT sinusoidal data.
[0055] Next, step S20 is performed to train a cold diffusion model using the training data set to generate a denoised diffusion model. The cold diffusion model uses a mean-preserving degradation operator to gradually add noise to the conventional dose CT sinogram data to degenerate it into the corresponding low dose CT sinogram data as a diffusion process, and gradually eliminates the noise in the low dose CT sinogram data to generate denoised CT sinogram data as an inverse diffusion process.
[0056] Specifically, Figure 2 As shown, the diffusion process, also known as the forward process, the degradation process, is based on the conventional dose CT sinogram data ( Figure 2 The leftmost figure, Original represents the original image) is used as the starting point of diffusion, and the low-dose CT sinogram data ( Figure 2 The middle figure, Degraded represents the degraded image) is used as the end point of the diffusion process, and the mean-maintaining degradation operator is used to gradually add noise with a mean of zero to the conventional dose CT sinusoidal data to degrade it (Degraded) to the low-dose CT sinusoidal data, thereby effectively reducing the number of steps for adding noise and speeding up the inference speed of the model. The inference time of a single CT sinusoidal data can be reduced by 0.05s-0.1s, for example, 0.05s, 0.075s, and 0.1s. Among them, the mean-maintaining degradation operator is defined as
[0057] x t =α t x0+(1-α t )x T
[0058] Where x0 is the conventional dose CT sinusoidal data; x T is the low-dose CT sinogram data; T is the total number of diffusion steps, T is greater than or equal to 5 and less than or equal to 20, for example, 5, 10, 15, 20; t is one of 1, 2, ..., T, α t is the noise factor corresponding to the tth step, α t <α t-1 , α t Experiments have found that the noise added by the mean-preserving degradation operator of the present application is more consistent with the actual noise distribution of the chord diagram in the original scan data, which is helpful for the simulation and learning of noise.
[0059] The reverse diffusion process, also known as the reverse process or denoising process, uses the cold diffusion model as the denoising diffusion model. The model is trained by a large number of paired sinusoidal data pairs, so that the cold diffusion model learns to extract the denoising information from the low-dose CT sinusoidal data ( Figure 2 middle image) to conventional dose CT sinogram data ( Figure 2 In the rightmost figure, Generated represents the generated image) the reverse diffusion process is used to complete the training of the denoising diffusion model, so that the trained denoising diffusion model can gradually generate optimized CT sinusoidal data from the conventional noise distribution through step-by-step sampling from the low-dose CT sinusoidal image.
[0060] Next, step S30 is performed to input the low-dose CT sinogram data to be denoised into the denoising diffusion model to obtain denoised CT sinogram data.
[0061] After obtaining the trained denoising diffusion model, the denoising diffusion model can be used to denoise the low-dose CT chordogram to be denoised. When using the denoising diffusion model to denoise the low-dose CT chordogram to be denoised, it is necessary to first normalize the low-dose CT chordogram to be denoised, and input the normalized low-dose CT chordogram to be denoised into the denoising diffusion model for denoising to obtain the optimized denoising CT sinogram data.
[0062] Finally, step S40 is executed to perform image reconstruction based on the denoised CT sinogram data to generate a DICOM image. Specifically, when performing image reconstruction based on the denoised CT sinogram data, it is necessary to first perform an anti-normalization operation on the denoised CT sinogram data, and then perform image reconstruction based on the anti-normalized denoised CT sinogram data to generate a DICOM image.
[0063] In the present application, the model-based iterative reconstruction (MBIR) method can be used to reconstruct the image based on the denoised CT sinogram data to generate a DICOM image. The MBIR method has the potential to improve image quality and remove artifacts, further reduce image noise and improve image quality. It is understood that in other embodiments, the filtered back projection (FBP) method can also be used for image reconstruction.
[0064] Figure 3 A comparison diagram of DICOM images reconstructed based on low-dose CT sinogram data, conventional-dose CT sinogram data, and low-dose CT sinogram data optimized by the present application is shown. In the figure, LDCT represents a DICOM image reconstructed based on low-dose CT sinogram data, NDCT represents a DICOM image reconstructed based on conventional-dose CT sinogram data, and Diffusion represents a DICOM image reconstructed based on optimized low-dose CT sinogram data. As can be seen from the figure, the denoising method of the present application can remove the noise of the low-dose CT sinogram data in the projection image domain while maintaining the detail information in the low-dose CT sinogram data, thereby maintaining the contrast of the DICOM image reconstructed based on the denoised low-dose CT sinogram data.
[0065] Based on the same concept, Figure 4 As shown, the present application also provides a computer tomography image processing system 11, which includes a training set construction module 111, a model training module 112, a denoising module 113 and a reconstruction module 114.
[0066] The training set construction module 111 is used to construct a training data set, wherein the training data set includes a plurality of pairs of sinusoidal graph data, each pair of sinusoidal graph data includes a low-dose CT sinusoidal graph data and a corresponding conventional-dose CT sinusoidal graph data;
[0067] The model training module 112 is used to train the cold diffusion model using the training data set to generate a denoised diffusion model, wherein the cold diffusion model uses a mean-preserving degradation operator to gradually add noise to the conventional dose CT sinogram data to degenerate it into corresponding low dose CT sinogram data as a diffusion process, and gradually eliminates the noise in the low dose CT sinogram data as an inverse diffusion process;
[0068] The denoising module 113 is used to input the low-dose CT sinogram data to be denoised into the denoising diffusion model to obtain denoised CT sinogram data;
[0069] The reconstruction module 114 is used to perform image reconstruction based on the denoised CT sinogram data to generate a DICOM image.
[0070] It should be noted that the computer tomography image processing system 11 provided in the above embodiment and the computer tomography image processing method provided in the above embodiment belong to the same concept, wherein the specific manner in which each module and unit performs the operation has been described in detail in the method embodiment, and will not be repeated here. In actual application, the computer tomography image processing system 11 provided in the above embodiment can allocate the above functions to different functional modules as needed, that is, divide the internal structure of the device into different functional modules to complete all or part of the functions described above, and this is not limited here.
[0071] Figure 5 A structural schematic diagram of an electronic device for implementing the computer tomography image processing method of the present application is shown.
[0072] The electronic device 1 may include a memory 12, a processor 13 and a bus, and may also include a computer program stored in the memory 12 and executable on the processor 13, such as a computer tomography image processing program.
[0073] Among them, the memory 12 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 12 can also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. Further, the memory 12 can also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 12 can not only be used to store application software and various types of data installed in the electronic device 1, such as codes for computer tomography image processing, etc., but can also be used to temporarily store data that has been output or is to be output.
[0074] In some embodiments, the processor 13 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 13 is the control core (Control Unit) of the electronic device 1, and uses various interfaces and lines to connect various components of the entire electronic device 1, and executes or executes programs or modules (such as computer tomography image processing programs, etc.) stored in the memory 12, and calls data stored in the memory 12 to execute various functions of the electronic device 1 and process data.
[0075] The processor 13 executes the operating system and various installed application programs of the electronic device 1. The processor 13 executes the application programs to implement the steps in the above-mentioned computer tomography image processing method.
[0076] Exemplarily, the computer program may be divided into one or more modules, which are stored in the memory 12 and executed by the processor 13 to complete the present application. The one or more modules may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the electronic device 1. For example, the computer program may be divided into a training set construction module 111, a model training module 112, a denoising module 113, and a reconstruction module 114.
[0077] The above-mentioned integrated unit implemented in the form of a software function module can be stored in a computer-readable storage medium, and the computer-readable storage medium can be non-volatile or volatile. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a computer device, or a network device, etc.) or a processor to perform part of the functions of the computer tomography image processing method described in each embodiment of the present application.
[0078] In summary, the present application discloses a method, system, device and medium for processing computed tomography images. The method constructs a training data set, wherein the training data set includes multiple pairs of sinusoidal data, each pair of sinusoidal data includes a low-dose CT sinusoidal data and a corresponding conventional-dose CT sinusoidal data; the training data set is used to train a cold diffusion model to generate a denoising diffusion model, wherein the cold diffusion model uses a mean-preserving degradation operator to gradually add noise to the conventional-dose CT sinusoidal data to degenerate it into the corresponding low-dose CT sinusoidal data as a diffusion process, and gradually eliminates the noise in the low-dose CT sinusoidal data as an inverse diffusion process; the low-dose CT sinusoidal data to be denoised is input into the denoising diffusion model to obtain the denoised CT sinusoidal data. By using the present application, the noise of the low-dose CT sinusoidal data can be removed in the projection image domain while maintaining the detail information in the low-dose CT sinusoidal data, thereby maintaining the contrast of the DICOM image reconstructed based on the denoised low-dose CT sinusoidal data. In addition, in the diffusion process of the cold diffusion model, the present application uses conventional-dose CT sinogram data as the starting point of the diffusion process and low-dose CT sinogram data as the end point of the diffusion process, and uses a mean-maintaining degradation operator to gradually add noise to the conventional-dose CT sinogram data to degenerate it into the corresponding low-dose CT sinogram data, thereby effectively reducing the number of steps for adding noise and speeding up the reasoning speed of the model. At the same time, since the amount of sinogram data before CT image reconstruction is smaller than that of the reconstructed DICOM image, denoising is performed on the projection image domain with a smaller amount of data, which can process data more efficiently and reduce the consumption of computing resources and processing time. Therefore, the present application effectively overcomes the various shortcomings of the prior art and has a high industrial utilization value.
[0079] The above embodiments are merely illustrative of the principles and effects of the present application and are not intended to limit the present application. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by a person of ordinary skill in the art without departing from the spirit and technical ideas disclosed in the present application shall still be covered by the claims of the present application.
Claims
1. A method for processing a computer tomography image, characterized in that: include: Constructing a training data set, wherein the training data set includes a plurality of sinogram data pairs, each sinogram data pair including a low-dose CT sinogram data and its corresponding conventional-dose CT sinogram data; The cold diffusion model is trained using the training data set to generate a denoised diffusion model, wherein the cold diffusion model uses a mean-preserving degradation operator to gradually add noise to conventional-dose CT sinusoidal data to degenerate it into corresponding low-dose CT sinusoidal data as a diffusion process, and gradually eliminates the noise in the low-dose CT sinusoidal data as an inverse diffusion process; The low-dose CT sinogram data to be denoised is input into the denoising diffusion model to obtain denoised CT sinogram data.
2. The computer tomography image processing method according to claim 1, characterized in that: Construct a training dataset, including: Obtain different low-dose CT raw data and corresponding conventional-dose CT raw data; Normalizing the different low-dose CT raw data and the corresponding conventional-dose CT raw data to obtain different low-dose CT sinogram data and the corresponding conventional-dose CT sinogram data; Sine graph data pairs are constructed using different low-dose CT sinogram data and corresponding conventional-dose CT sinogram data to form the training data set.
3. The computer tomography image processing method according to claim 2, characterized in that: Constructing a pair of sinogram data using different low-dose CT sinogram data and corresponding conventional-dose CT sinogram data to form the training data set includes: Constructing a sinusoidal data pair using different low-dose CT sinusoidal data and corresponding conventional-dose CT sinusoidal data; The low-dose CT sinogram data in each sinogram data pair are grouped according to the dose corresponding to the low-dose CT sinogram data, so that the sinogram data pairs with the same dose are divided into the same image group, and the sinogram data pairs with different doses are divided into different image groups, thereby forming the training data set.
4. The computer tomography image processing method according to claim 1, characterized in that: The mean-preserving degradation operator is defined as x t =a t x0+(1-a t )x T Among them, x0 is the conventional dose CT sinusoidal data, x T is the low-dose CT sinogram data, T is the total number of diffusion steps, t is one of 1, 2, ..., T, a t is the noise factor corresponding to the tth step, α t <α t-1 .
5. The computer tomography image processing method according to claim 4, characterized in that: T is greater than or equal to 5 and less than or equal to 20.
6. The computer tomography image processing method according to claim 1, characterized in that: The low-dose CT sinogram data to be denoised is input into the denoising diffusion model to obtain denoised CT sinogram data, and then the following steps are further included: Image reconstruction is performed based on the denoised CT sinogram data to generate a DICOM image.
7. The computer tomography image processing method according to claim 6, characterized in that: Performing image reconstruction based on the denoised CT sinogram data to generate a DICOM image includes: Based on the denoised CT sinogram data, an iterative reconstruction method based on a model is used to perform image reconstruction to generate a DICOM image.
8. A computer tomography image processing system, characterized in that: include: A training set construction module, used to construct a training data set, wherein the training data set includes a plurality of sinusoidal data pairs, each sinusoidal data pair includes a low-dose CT sinusoidal data and its corresponding conventional-dose CT sinusoidal data; A model training module, for training a cold diffusion model using the training data set to generate a denoised diffusion model, wherein the cold diffusion model uses a mean-preserving degradation operator to gradually add noise to conventional-dose CT sinusoidal data to degenerate it into corresponding low-dose CT sinusoidal data as a diffusion process, and gradually eliminates the noise in the low-dose CT sinusoidal data as an inverse diffusion process; The denoising module is used to input the low-dose CT sinogram data to be denoised into the denoising diffusion model to obtain the denoised CT sinogram data.
9. An electronic device, characterized in that: The electronic device comprises: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the electronic device to implement the computer tomography image processing method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor of a computer, the computer is caused to execute the computer tomography image processing method according to any one of claims 1 to 7.