Diffusion model-based joint pet-MRI imaging method and related device

Abstract

Disclosed are a diffusion model-based joint PET-MRI imaging method and a related device. The method comprises: acquiring an MRI image and a PET image which are aligned, and inputting the MRI image and the PET image into a joint diffusion model; on the basis of a noise adder of the joint diffusion model, sequentially adding Gaussian noise to each of the MRI image and the PET image a preset number of times, until the MRI image and the PET image completely become Gaussian noise; on the basis of a sampler of the joint diffusion model, performing data sampling on the Gaussian noise obtained by means of the noise adder, to obtain a denoised PET image and a denoised MRI image; and on the basis of a corrector of the joint diffusion model, correcting the denoised PET image and the denoised MRI image, and performing joint reconstruction, to obtain a high-quality PET image and a high-quality MRI image. The present invention uses under-sampled PET and MRI data to jointly reconstruct PET and MRI data, and can thus reconstruct a high-quality PET image and a high-quality MRI image.

Description

PET-MRI combined imaging method and related equipment based on diffusion model Technical Field

[0001] The present invention relates to the technical field of magnetic resonance imaging and positron emission tomography imaging, and in particular to a PET-MRI combined imaging method, system, terminal and computer-readable storage medium based on a diffusion model. Background Art

[0002] Medical imaging is an important tool for clinical analysis and diagnosis. In the past, single-modality medical imaging modes were widely used for anatomical and functional imaging. For example, PET (Positron Emission Tomography) images are mainly used to obtain functional images, and MRI (Magnetic Resonance Imaging) images are mainly used to obtain structural images. However, PET-MRI images can simultaneously obtain registered PET and MRI images. PET-MRI images combine high-resolution structural and metabolic information. Since MRI image reconstruction requires a long reconstruction time, and PET requires the injection of contrast agents, which are harmful to the body, it is considered whether it is possible to shorten the reconstruction time while injecting as little contrast agent as possible and still obtain high-quality PET-MRI images. This effect can only be achieved by undersampling the data.

[0003] Previous work has primarily reconstructed the two modalities independently, using undersampled PET and MRI data, each using a suitable reconstruction model, and then fusing the two modal images. However, this approach clearly fails to consider the relationship between the two modalities.

[0004] The key idea behind the joint reconstruction task is to effectively utilize the joint probability distribution of the two modalities. In traditional joint reconstruction models, since the functional image largely follows the structural image, the reconstructed images can share edges, as shown in Figure 1. Leveraging the similarity in edges between the functional and structural images, a joint probability distribution is manually designed to facilitate reconstruction.

[0005] The joint reconstruction model of traditional technology does not fully utilize the joint probability distribution information between the two modalities, but only considers the structural similarity between the two modalities. The deep learning method mainly uses GAN generation, and the main problem is mode collapse.

[0006] Therefore, the existing technology still needs to be improved and developed.

[0007] Summary of the Invention

[0008] The main purpose of the present invention is to provide a PET-MRI joint imaging method, system, terminal and computer-readable storage medium based on a diffusion model, aiming to solve the problem in the prior art that the joint reconstruction model does not fully utilize the joint probability distribution information between the two modalities and cannot reconstruct high-quality PET images and high-quality MRI images.

[0009] To achieve the above object, the present invention provides a PET-MRI combined imaging method based on a diffusion model, the PET-MRI combined imaging method based on a diffusion model comprising the following steps:

[0010] Acquire the registered MRI image and PET image, and input the MRI image and the PET image into a joint diffusion model;

[0011] A noise adder based on the joint diffusion model sequentially loads Gaussian noise for a preset number of times on the MRI image and the PET image until the MRI image and the PET image are completely Gaussian noise;

[0012] A sampler based on the joint diffusion model performs data sampling on the Gaussian noise obtained by the noise adder to obtain a denoised PET image and an MRI image;

[0013] The corrector based on the joint diffusion model corrects the denoised PET image and MRI image, and jointly reconstructs the high-quality PET image and MRI image.

[0014] Optionally, in the diffusion model-based PET-MRI combined imaging method, the step of acquiring the registered MRI image and PET image specifically includes:

[0015] Acquire the original MRI image and original PET image of the same target;

[0016] The original MRI image and the original PET image are subjected to data alignment processing to obtain a registered MRI image and PET image.

[0017] Optionally, the PET-MRI joint imaging method based on the diffusion model, wherein the denoiser based on the joint diffusion model sequentially loads Gaussian noise to the MRI image and the PET image for a preset number of times until the MRI image and the PET image are completely Gaussian noise, specifically includes:

[0018] The MRI image and the PET image are loaded with Gaussian noise one by one through the noise adder of the joint diffusion model:

[0019] Among them, (u i , vi ) is the PET image and MRI image with noise added for the first time; (u i-1 , v i-1 ) are the PET and MRI images with noise added for the i-1th time; σ i is a fixed parameter value, z i-1 is the i-1th Gaussian noise; N represents the preset number of times the Gaussian noise is loaded;

[0020] After each addition of Gaussian noise, the deep neural network of encoding and decoding is used to learn and obtain the derivative s of the logarithm of the joint probability distribution of the noisy PET and MRI. θ (u i+1 , v i+1 , t), t represents the value of (1, 1000).

[0021] Optionally, the PET-MRI joint imaging method based on the diffusion model, wherein the sampler based on the joint diffusion model performs data sampling on the Gaussian noise obtained by the noise adder to obtain a denoised PET image and MRI image, specifically includes:

[0022] The Gaussian noise obtained by the noise adder is sampled by the sampler of the joint diffusion model:

[0023] Among them, (u i-1 , v i-1 ) are the PET and MRI images with noise added for the i+1th time; represents Radon transform; represents the unit matrix; f represents PET data; g represents MRI data; represents Fourier transform;

[0024] After data sampling, denoised PET images and MRI images are obtained.

[0025] Optionally, the PET-MRI joint imaging method based on the diffusion model, wherein the corrector based on the joint diffusion model corrects the denoised PET image and MRI image to jointly reconstruct high-quality PET and MRI images, specifically includes:

[0026] The PET image and MRI image obtained by data sampling by the sampler are corrected by the corrector of the joint diffusion model:

[0027] Among them, η i is a parameter.

[0028] High-quality PET images and MRI images are obtained by performing joint reconstruction through the sampler and the corrector.

[0029] Optionally, in the diffusion model-based PET-MRI combined imaging method, the original MRI image and the original PET image are undersampled data.

[0030] Optionally, in the diffusion model-based PET-MRI combined imaging method, the preset number of times is in the range of 800-1000 times.

[0031] To achieve the above object, the present invention provides a PET-MRI combined imaging system based on a diffusion model, wherein the PET-MRI combined imaging system based on a diffusion model comprises:

[0032] an image acquisition module, configured to acquire a registered MRI image and a PET image, and input the MRI image and the PET image into a joint diffusion model;

[0033] an image diffusion module, which sequentially loads a preset number of Gaussian noises on the MRI image and the PET image based on a noise adder of the joint diffusion model until the MRI image and the PET image are completely Gaussian noises;

[0034] An image inverse diffusion module, which performs data sampling on the Gaussian noise obtained by the noise adder based on the sampler of the joint diffusion model to obtain denoised PET images and MRI images;

[0035] The image correction module corrects the denoised PET image and MRI image based on the corrector of the joint diffusion model, and jointly reconstructs the PET image and MRI image to obtain high-quality PET image and MRI image.

[0036] In addition, to achieve the above-mentioned purpose, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a diffusion model-based PET-MRI combined imaging program stored in the memory and executable on the processor, wherein the diffusion model-based PET-MRI combined imaging program, when executed by the processor, implements the steps of the diffusion model-based PET-MRI combined imaging method described above.

[0037] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a PET-MRI combined imaging program based on a diffusion model, and when the PET-MRI combined imaging program based on a diffusion model is executed by a processor, the steps of the PET-MRI combined imaging method based on a diffusion model as described above are implemented.

[0038] In the present invention, a registered MRI image and PET image are acquired and input into a joint diffusion model. A denoiser based on the joint diffusion model sequentially loads a preset number of Gaussian noises onto the MRI image and the PET image until the MRI image and the PET image are completely Gaussian. A sampler based on the joint diffusion model samples the Gaussian noise obtained by the denoiser to obtain denoised PET and MRI images. A corrector based on the joint diffusion model corrects the denoised PET and MRI images, and jointly reconstructs high-quality PET and MRI images. The present invention utilizes undersampled PET and MRI data to jointly reconstruct the PET and MRI data, thereby reconstructing high-quality PET and MRI images and improving the resolution of positron emission tomography and nuclear magnetic resonance imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] FIG1 is a schematic diagram of the shared edge between PET and MRI;

[0040] FIG2 is a flow chart of a preferred embodiment of the PET-MRI combined imaging method based on the diffusion model of the present invention;

[0041] FIG3 is a schematic diagram showing the addition of noise of the same scale to an MRI image and a PET image in multiple steps in a preferred embodiment of the PET-MRI combined imaging method based on a diffusion model of the present invention;

[0042] FIG4 is a schematic diagram of a joint reconstruction process of PET and MRI data in a preferred embodiment of the PET-MRI joint imaging method based on a diffusion model of the present invention;

[0043] FIG5 is a schematic diagram showing a comparison of the results of joint reconstruction using a diffusion model, a linear level set joint reconstruction, and an iterative shrinkage threshold network joint reconstruction in a preferred embodiment of the PET-MRI joint imaging method based on a diffusion model of the present invention;

[0044] FIG6 is a schematic diagram showing the principle of a preferred embodiment of a PET-MRI combined imaging system based on a diffusion model of the present invention;

[0045] FIG7 is a schematic diagram of an operating environment of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the present invention more clear and distinct, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0047] The PET-MRI combined imaging method based on the diffusion model according to the preferred embodiment of the present invention is shown in FIG2 . The PET-MRI combined imaging method based on the diffusion model comprises the following steps:

[0048] Step S10: Acquire the registered MRI image and PET image, and input the MRI image and the PET image into a joint diffusion model.

[0049] Specifically, an original MRI image and an original PET image of the same target are acquired, wherein the original MRI image and the original PET image are undersampled data. Since full-sampled data imaging requires a long time, and full-sampled PET data requires sufficient contrast agent to be injected into the body, the present invention aims to accelerate MRI imaging and reduce contrast agent injection. To achieve the above-mentioned goals, undersampled MRI (less data) and undersampled PET (injecting a small amount of contrast agent) can be used; the original MRI image and the original PET image are aligned to obtain registered MRI and PET images.

[0050] Step S20 : ​​The noise adder based on the joint diffusion model sequentially loads a preset number of Gaussian noises on the MRI image and the PET image respectively, until the MRI image and the PET image are completely Gaussian noises.

[0051] Specifically, as shown in FIG3 and FIG4 , the MRI image and the PET image (i.e., the MRI data and the PET data) are successively loaded with Gaussian noise by the noise adder of the joint diffusion model (diffusion process, which refers to the direction of the arrow from left to right in FIG3 ):

[0052] Among them, (u i , v i ) is the PET image and MRI image with noise added for the i-th time; (u i-1 , v i-1 ) are the PET and MRI images with noise added for the i-1th time; σ i is a fixed parameter value, z i-1 is the i-1th Gaussian noise; N represents the preset number of times the Gaussian noise is loaded (usually 800-1000 times, preferably 1000 times).

[0053] After each addition of Gaussian noise, the deep neural network of encoding and decoding is used to learn and obtain the derivative s of the logarithm of the joint probability distribution of the noisy PET and MRI. θ (u i+1 , v i+1 , t), t represents the value of (1, 1000).

[0054] Gaussian noise is loaded on the MRI image and the PET image respectively for a preset number of times until the MRI image and the PET image are completely Gaussian noise, which is a process of gradually turning the image into pure noise. The information content of the image is gradually reduced by gradually increasing the noise, so that the variance of the conditional distribution is gradually reduced in each step.

[0055] Step S30: The sampler based on the joint diffusion model performs data sampling on the Gaussian noise obtained by the noise adder to obtain a denoised PET image and an MRI image.

[0056] Specifically, as shown in FIG4 , the Gaussian noise obtained by the noise adder is sampled by the sampler of the joint diffusion model (inverse diffusion process, the inverse diffusion process refers to the direction of the arrow from right to left in the lower part of FIG3 ):

[0057] Among them, (u i+1 , v i+1 ) are the PET and MRI images with noise added for the i+1th time; represents Radon transform; represents the unit matrix; f represents PET data; g represents MRI data; represents Fourier transform;

[0058] After data sampling, denoised PET images and MRI images are obtained.

[0059] Step S40: Correcting the denoised PET image and MRI image using a corrector based on the joint diffusion model, and jointly reconstructing the PET image and MRI image to obtain high-quality PET image and MRI image.

[0060] Specifically, the PET image and the MRI image obtained by the sampler through data sampling are corrected by the corrector of the joint diffusion model:

[0061] Among them, η i is a parameter.

[0062] High-quality PET images and MRI images are obtained by performing joint reconstruction through the sampler and the corrector.

[0063] As shown in Figure 5, the ground truth refers to the original MRI and PET images. The results of the joint reconstruction diffusion model of the present invention (Ours in Figure 5) are compared with those of the linear level set joint reconstruction (LPLS in Figure 5) and the iterative shrinkage threshold network joint reconstruction (ISTA_Net in Figure 5). As shown in Figure 5, simulation experiments demonstrate that the joint reconstruction diffusion model produces higher-quality dual-modal images. Simulations and experiments have proven the feasibility of the present invention. Compared with traditional joint reconstruction models and iterative shrinkage threshold network algorithms, the present model has a higher peak signal-to-noise ratio and structural similarity, as shown in Figure 5.

[0064] This invention applies the joint diffusion model to joint PET-MRI reconstruction for the first time. Undersampled data is reconstructed using the learned logarithm derivative of the joint probability distribution. This logarithm derivative is learned using the diffusion model, resulting in high-quality PET and MRI images, improving the resolution of positron emission tomography and nuclear magnetic resonance imaging. This invention fully utilizes the joint probability distribution of the two modalities and effectively exploits shared information between the modalities, resulting in higher-quality bimodal reconstructed images. By learning the gradient of the logarithm of the joint probability distribution between the two modalities using the joint diffusion model, the invention fully utilizes the joint probability distribution to produce jointly reconstructed bimodal images.

[0065] Furthermore, as shown in FIG6 , based on the above-mentioned diffusion model-based PET-MRI combined imaging method, the present invention also provides a diffusion model-based PET-MRI combined imaging system, wherein the diffusion model-based PET-MRI combined imaging system includes:

[0066] An image acquisition module 51 is configured to acquire a registered MRI image and a registered PET image, and input the MRI image and the PET image into a joint diffusion model;

[0067] an image diffusion module 52 configured to sequentially load a preset number of Gaussian noises onto the MRI image and the PET image respectively based on a noise adder of the joint diffusion model until the MRI image and the PET image are completely filled with Gaussian noises;

[0068] An image inverse diffusion module 53 is configured to perform data sampling on the Gaussian noise obtained by the noise adder based on the sampler of the joint diffusion model to obtain a denoised PET image and an MRI image;

[0069] The image correction module 54 is configured to correct the denoised PET image and MRI image based on the corrector of the joint diffusion model, and jointly reconstruct the PET image and MRI image to obtain high-quality PET image and MRI image.

[0070] Furthermore, as shown in FIG7 , based on the above-described diffusion model-based PET-MRI combined imaging method and system, the present invention also provides a terminal, which includes a processor 10, a memory 20, and a display 30. FIG7 only shows some components of the terminal, but it should be understood that implementation of all shown components is not required, and more or fewer components may be implemented instead.

[0071] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal. Furthermore, the memory 20 may also include both an internal storage unit of the terminal and an external storage device. The memory 20 is used to store application software and various types of data installed on the terminal, such as program codes of the installation terminal. The memory 20 may also be used to temporarily store data that has been output or is about to be output. In one embodiment, a PET-MRI combined imaging program 40 based on a diffusion model is stored on the memory 20, and the PET-MRI combined imaging program 40 based on a diffusion model can be executed by the processor 10, thereby realizing the PET-MRI combined imaging method based on a diffusion model in the present application.

[0072] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, configured to execute program codes or process data stored in the memory 20, such as executing the diffusion model-based PET-MRI combined imaging method.

[0073] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The components 10-30 of the terminal communicate with each other via a system bus.

[0074] In one embodiment, when the processor 10 executes the diffusion model-based PET-MRI combined imaging program 40 in the memory 20, the following steps are implemented:

[0075] Acquire the registered MRI image and PET image, and input the MRI image and the PET image into a joint diffusion model;

[0076] A noise adder based on the joint diffusion model sequentially loads Gaussian noise for a preset number of times on the MRI image and the PET image until the MRI image and the PET image are completely Gaussian noise;

[0077] A sampler based on the joint diffusion model performs data sampling on the Gaussian noise obtained by the noise adder to obtain a denoised PET image and an MRI image;

[0078] The corrector based on the joint diffusion model corrects the denoised PET image and MRI image, and jointly reconstructs the high-quality PET image and MRI image.

[0079] The step of obtaining the registered MRI image and PET image specifically includes:

[0080] Acquire the original MRI image and original PET image of the same target;

[0081] The original MRI image and the original PET image are subjected to data alignment processing to obtain a registered MRI image and PET image.

[0082] The denoiser based on the joint diffusion model sequentially loads Gaussian noise for a preset number of times on the MRI image and the PET image until the MRI image and the PET image are completely Gaussian noise, specifically comprising:

[0083] The MRI image and the PET image are loaded with Gaussian noise one by one through the noise adder of the joint diffusion model:

[0084] Among them, (u i , v i ) is the PET image and MRI image with noise added for the i-th time; (u i-1 , v i-1 ) are the PET and MRI images with noise added for the i-1th time; σ i is a fixed parameter value, z i -1 is the i-1th Gaussian noise; N represents the preset number of times the Gaussian noise is loaded;

[0085] After each addition of Gaussian noise, the deep neural network of encoding and decoding is used to learn and obtain the derivative s of the logarithm of the joint probability distribution of the noisy PET and MRI. θ (u i+1 , v i+1 , t), t represents the value of (1, 1000).

[0086] The sampler based on the joint diffusion model performs data sampling on the Gaussian noise obtained by the noise adder to obtain the denoised PET image and MRI image, specifically including:

[0087] The Gaussian noise obtained by the noise adder is sampled by the sampler of the joint diffusion model:

[0088] Among them, (u i+1 , v i+1 ) are the PET and MRI images with noise added for the i+1th time; represents Radon transform; represents the unit matrix; f represents PET data; g represents MRI data; represents Fourier transform;

[0089] After data sampling, denoised PET images and MRI images are obtained.

[0090] The corrector based on the joint diffusion model corrects the denoised PET image and MRI image, and jointly reconstructs high-quality PET image and MRI image, specifically including:

[0091] The PET image and MRI image obtained by data sampling by the sampler are corrected by the corrector of the joint diffusion model:

[0092] Among them, η i is a parameter.

[0093] High-quality PET images and MRI images are obtained by performing joint reconstruction through the sampler and the corrector.

[0094] The original MRI image and the original PET image are undersampled data.

[0095] The preset number of times is in the range of 800-1000 times.

[0096] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a PET-MRI combined imaging program based on a diffusion model, and when the PET-MRI combined imaging program based on a diffusion model is executed by a processor, the steps of the PET-MRI combined imaging method based on a diffusion model as described above are implemented.

[0097] In summary, the present invention provides a diffusion model-based PET-MRI joint imaging method and related equipment. The method comprises: acquiring an aligned MRI image and PET image, and inputting the MRI image and the PET image into a joint diffusion model; a denoiser based on the joint diffusion model sequentially loading a preset number of Gaussian noises onto the MRI image and the PET image until the MRI image and the PET image are completely Gaussian; a sampler based on the joint diffusion model performs data sampling on the Gaussian noise obtained by the denoiser to obtain denoised PET and MRI images; and a corrector based on the joint diffusion model corrects the denoised PET and MRI images, and jointly reconstructs high-quality PET and MRI images. The present invention utilizes undersampled PET and MRI data to jointly reconstruct the PET and MRI data, thereby reconstructing high-quality PET and MRI images, thereby improving the resolution of positron emission tomography and nuclear magnetic resonance imaging.

[0098] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or terminal comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or terminal comprising the element.

[0099] Of course, those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware (such as a processor, controller, etc.) through a computer program. The program can be stored in a computer-readable storage medium that can be read by a computer. When the program is executed, it can include the processes in the above-described method embodiments. The computer-readable storage medium can be a memory, a magnetic disk, an optical disk, etc.

[0100] It should be understood that the application of the present invention is not limited to the above examples. For those skilled in the art, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A PET-MRI combined imaging method based on a diffusion model, characterized in that: The PET-MRI combined imaging method based on the diffusion model includes: Acquire the registered MRI image and PET image, and input the MRI image and the PET image into a joint diffusion model; A noise adder based on the joint diffusion model sequentially loads Gaussian noise for a preset number of times on the MRI image and the PET image until the MRI image and the PET image are completely Gaussian noise; A sampler based on the joint diffusion model performs data sampling on the Gaussian noise obtained by the noise adder to obtain a denoised PET image and an MRI image; The corrector based on the joint diffusion model corrects the denoised PET image and MRI image, and jointly reconstructs the high-quality PET image and MRI image.

2. The PET-MRI combined imaging method based on the diffusion model according to claim 1, characterized in that: The obtaining of the registered MRI image and PET image specifically includes: Acquire the original MRI image and original PET image of the same target; The original MRI image and the original PET image are subjected to data alignment processing to obtain a registered MRI image and PET image.

3. The PET-MRI combined imaging method based on the diffusion model according to claim 1, characterized in that: The noise adder based on the joint diffusion model sequentially loads Gaussian noise for a preset number of times on the MRI image and the PET image until the MRI image and the PET image are completely Gaussian noise, specifically comprising: The MRI image and the PET image are loaded with Gaussian noise one by one through the noise adder of the joint diffusion model: Among them, (u i , v i ) is the PET image and MRI image with noise added for the i-th time; (u i-1 , v i-1 ) are the PET and MRI images with noise added for the i-1th time; σ i is a fixed parameter value, z i-1 is the i-1th Gaussian noise; N represents the preset number of times the Gaussian noise is loaded; After each addition of Gaussian noise, the deep neural network of encoding and decoding is used to learn and obtain the derivative s of the logarithm of the joint probability distribution of the noisy PET and the noisy MRI. θ (u i+1 , v i+1 , t), t represents the value of (1, 1000).

4. The PET-MRI combined imaging method based on the diffusion model according to claim 3, characterized in that: The sampler based on the joint diffusion model performs data sampling on the Gaussian noise obtained by the noise adder to obtain a denoised PET image and an MRI image, specifically comprising: The Gaussian noise obtained by the noise adder is sampled by the sampler of the joint diffusion model: Among them, (u i+1 , v i+1 ) are the PET and MRI images with noise added for the i+1th time; represents Radon transform; represents the unit matrix; f represents PET data; g represents MRI data; represents Fourier transform; After data sampling, denoised PET images and MRI images are obtained.

5. The PET-MRI combined imaging method based on the diffusion model according to claim 4, characterized in that: The corrector based on the joint diffusion model corrects the denoised PET image and MRI image, and jointly reconstructs high-quality PET images and MRI images, specifically including: The PET image and MRI image obtained by data sampling by the sampler are corrected by the corrector of the joint diffusion model: Among them, η i is a parameter. High-quality PET images and MRI images are obtained by performing joint reconstruction through the sampler and the corrector.

6. The PET-MRI combined imaging method based on the diffusion model according to claim 2, characterized in that: The original MRI image and the original PET image are undersampled data.

7. The PET-MRI combined imaging method based on the diffusion model according to claim 1 or 3, characterized in that: The preset number of times ranges from 800 to 1000 times.

8. A PET-MRI combined imaging system based on a diffusion model, characterized in that: The PET-MRI combined imaging system based on the diffusion model includes: an image acquisition module, configured to acquire a registered MRI image and a PET image, and input the MRI image and the PET image into a joint diffusion model; an image diffusion module, configured to sequentially load a preset number of Gaussian noises onto the MRI image and the PET image respectively based on a noise adder of the joint diffusion model until the MRI image and the PET image are completely filled with Gaussian noises; An image inverse diffusion module, configured to perform data sampling on the Gaussian noise obtained by the noise adder based on a sampler of the joint diffusion model, to obtain denoised PET images and MRI images; The image correction module is used to correct the denoised PET image and MRI image based on the corrector of the joint diffusion model, and jointly reconstruct the PET image and MRI image to obtain high-quality PET image and MRI image.

9. A terminal, characterized in that: The terminal includes: a memory, a processor, and a diffusion model-based PET-MRI combined imaging program stored in the memory and executable on the processor. When the diffusion model-based PET-MRI combined imaging program is executed by the processor, the steps of the diffusion model-based PET-MRI combined imaging method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a diffusion model-based PET-MRI combined imaging program, which, when executed by a processor, implements the steps of the diffusion model-based PET-MRI combined imaging method according to any one of claims 1 to 7.

Images