Fast MRI image optimization reconstruction method and device based on pre-training diffusion model

Abstract

The invention relates to the technical field of medical images, discloses a rapid MRI image optimization reconstruction method and device based on a pre-training diffusion model, electronic equipment, a computer readable storage medium and a computer program product, and is used for solving the technical problem of insufficient quality of a rapid MRI image obtained through reconstruction in the prior art. Comprising the following steps: acquiring a first conventional MRI image sample and a K space data sample thereof; carrying out undersampling processing on the K space data sample and carrying out image reconstruction to obtain a simulated fast MRI image sample; constructing an initial diffusion model, and pre-training the initial diffusion model based on the first conventional MRI image sample and the simulated fast MRI image sample to obtain a pre-trained diffusion model; acquiring a second conventional MRI image sample and an actual fast MRI image sample of the same acquisition object, and performing fine adjustment on the pre-training diffusion model to obtain an image optimization reconstruction model; and calling an image optimization reconstruction model to carry out image optimization reconstruction on the fast MRI image to obtain a target MRI image.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging technology, and in particular to a fast MRI image optimization and reconstruction method, device, electronic device, computer storage medium, and computer program product based on a pre-trained diffusion model. Background Art

[0002] Magnetic Resonance Imaging (MRI) is a medical imaging technology based on the principle of nuclear magnetic resonance and is widely used in the field of imaging. As an important means in the modern medical diagnosis process, MRI has significant advantages such as non-invasive and high definition. However, the imaging process of MRI usually requires a long scanning time, which not only increases the cost of image acquisition but also increases the possibility of motion artifacts.

[0003] In urgent situations, such as for emergency patients or other situations that require rapid scanning, or when it is difficult for patients to remain stationary for a long time, using fast MRI scanning can significantly shorten the scanning time.

[0004] However, in the prior art, the images obtained by fast MRI scanning often have more noise, poor clarity, and are difficult to meet the diagnostic requirements. Therefore, there is an urgent need for a fast MRI image optimization and reconstruction method that can not only effectively shorten the scanning time but also improve the quality of MRI images, thereby providing more efficient and accurate support for clinical diagnosis. Summary of the Invention

[0005] The main purpose of the present invention is to solve the technical problem that the fast MRI images obtained by optimization and reconstruction in the prior art have more noise, insufficient clarity, and poor image quality.

[0006] The first aspect of the present invention provides a fast MRI image optimization and reconstruction method based on a pre-trained diffusion model, including:

[0007] Obtain a first conventional MRI image sample and a K-space data sample of the first conventional MRI image sample;

[0008] Perform undersampling processing on the K-space data sample and perform image reconstruction to obtain a simulated fast MRI image sample corresponding to the first conventional MRI image sample;

[0009] Construct an initial diffusion model, and pre-train the initial diffusion model based on the first conventional MRI image sample and the simulated fast MRI image sample to obtain a pre-trained diffusion model;

[0010] Obtain a second conventional MRI image sample and an actual fast MRI image sample of the same acquisition object;

[0011] Fine-tune the pre-trained diffusion model based on the second conventional MRI image sample and the actual fast MRI image sample to obtain an image optimization and reconstruction model;

[0012] In response to an image optimization and reconstruction request, call the image optimization and reconstruction model to perform image optimization and reconstruction on the fast MRI image to be optimized and reconstructed, and obtain an optimized and reconstructed target MRI image.

[0013] Optionally, in the first implementation manner of the first aspect of the present invention, the performing undersampling processing on the K-space data sample and performing image reconstruction to obtain the simulated fast MRI image sample corresponding to the first conventional MRI image sample includes:

[0014] Perform undersampling processing on the K-space data sample to obtain an undersampled K-space data sample;

[0015] Call the GRAPPA algorithm to perform data fitting on the undersampled K-space data sample, calculate the data not sampled at the K-space center, and complement the undersampled K-space data sample to obtain a complemented K-space data sample;

[0016] Call the SENSE algorithm, and perform image reconstruction based on the complemented K-space data sample in combination with the estimated sensitivity matrix value to obtain a simulated fast MRI image sample.

[0017] Optionally, in the second implementation manner of the first aspect of the present invention, before performing undersampling processing on the K-space data sample to obtain an undersampled K-space data sample, it further includes:

[0018] Determine the undersampling parameter according to the fast scanning requirement;

[0019] The performing undersampling processing on the K-space data sample to obtain an undersampled K-space data sample includes:

[0020] Discard some sampling rows of the K-space data sample based on the undersampling parameter to obtain an undersampled K-space data sample.

[0021] Optionally, in the third implementation manner of the first aspect of the present invention, after calling the GRAPPA algorithm to perform data fitting on the undersampled K-space data sample, calculating the data not sampled at the K-space center, and complementing the undersampled K-space data sample to obtain a complemented K-space data sample, it further includes:

[0022] Calculate the estimated value of the sensitivity matrix of the scanning coil;

[0023] After performing the above-described SENSE algorithm and reconstructing an image based on the complemented K-space data samples in combination with the estimated sensitivity matrix to obtain a simulated fast MRI image sample, the method further includes:

[0024] Alternately iterating the estimated sensitivity matrix and the simulated fast MRI image sample to optimize the simulated fast MRI image sample.

[0025] Optionally, in the fourth implementation manner of the first aspect of the present invention, before alternately iterating the estimated sensitivity matrix and the simulated fast MRI image sample to optimize the simulated fast MRI image sample, the method further includes:

[0026] Determining the number of iterations of the alternate iteration according to the fast scanning requirement.

[0027] Optionally, in the fifth implementation manner of the first aspect of the present invention, the initial diffusion model is constructed based on a Brownian bridge diffusion model;

[0028] The pre-training of the initial diffusion model based on the first conventional MRI image sample and the simulated fast MRI image sample to obtain a pre-trained diffusion model includes:

[0029] Randomly selecting cross-sectional images of three consecutive layers in the simulated fast MRI image sample as the model input of the initial diffusion model;

[0030] Randomly selecting the number of model steps and randomly generating Gaussian noise to be added to the cross-sectional images of three consecutive layers in the simulated fast MRI image sample;

[0031] Selecting cross-sections of three consecutive layers at corresponding positions in the corresponding first conventional MRI image sample as the model output of the initial diffusion model;

[0032] Invoking an optimizer to pre-train the initial diffusion model until the model converges to obtain a pre-trained diffusion model.

[0033] Optionally, in the sixth implementation manner of the first aspect of the present invention, the fine-tuning training of the pre-trained diffusion model based on the second conventional MRI image sample and the actual fast MRI image sample to obtain an image optimization and reconstruction model includes:

[0034] Randomly selecting cross-sectional images of three consecutive layers in the actual fast MRI image sample as the model input of the pre-trained diffusion model;

[0035] Randomly selecting the number of model steps and randomly generating Gaussian noise to be added to the cross-sectional images of three consecutive layers in the actual fast MRI image sample;

[0036] Select three consecutive cross-sections at the corresponding positions in the corresponding second conventional MRI image sample as the model output of the pre-trained diffusion model;

[0037] Call an optimizer to fine-tune the pre-trained diffusion model to obtain an image optimization and reconstruction model.

[0038] Optionally, in the seventh implementation manner of the first aspect of the present invention, the calling the image optimization and reconstruction model to perform image optimization and reconstruction on the fast MRI image to be optimized and reconstructed to obtain the optimized and reconstructed target MRI image includes:

[0039] Obtain cross-sectional images of three consecutive layers in the fast MRI image to be optimized for image reconstruction as the model input of the image optimization and reconstruction model, and call the image optimization and reconstruction model to perform image optimization and reconstruction to obtain the prediction result of the target MRI image.

[0040] Optionally, in the eighth implementation manner of the first aspect of the present invention, after obtaining the simulated fast MRI image sample corresponding to the first conventional MRI image sample, it further includes:

[0041] Calculate the image sizes of the first conventional MRI image sample and the simulated fast MRI image sample respectively;

[0042] Based on the image sizes, perform normalization processing on the first conventional MRI image sample and the simulated fast MRI image sample respectively;

[0043] Randomly select the normalized first conventional MRI image sample and the simulated fast MRI image sample, and perform data augmentation operations.

[0044] The second aspect of the present invention provides a fast MRI image optimization and reconstruction device for a pre-trained diffusion model, including:

[0045] A sample simulation module, configured to obtain a first conventional MRI image sample and the K-space data sample of the first conventional MRI image sample; and perform undersampling processing on the K-space data sample and perform image reconstruction to obtain the simulated fast MRI image sample corresponding to the first conventional MRI image sample;

[0046] A model training module, configured to construct an initial diffusion model, pre-train the initial diffusion model based on the first conventional MRI image sample and the simulated fast MRI image sample to obtain a pre-trained diffusion model; and configured to obtain a second conventional MRI image sample and an actual fast MRI image sample of the same acquisition object; fine-tune the pre-trained diffusion model based on the second conventional MRI image sample and the actual fast MRI image sample to obtain an image optimization and reconstruction model;

[0047] An optimization and reconstruction module, configured to, in response to an image optimization and reconstruction request, call the image optimization and reconstruction model to perform image optimization and reconstruction on the fast MRI image to be subjected to image optimization and reconstruction, so as to obtain an optimized and reconstructed target MRI image.

[0048] The third aspect of the present invention provides a fast MRI image optimization and reconstruction device based on a pre-trained diffusion model, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory, so that the fast MRI image optimization and reconstruction device based on the pre-trained diffusion model executes the steps of the above-mentioned fast MRI image optimization and reconstruction method based on the pre-trained diffusion model.

[0049] The fourth aspect of the present invention provides a computer-readable storage medium, in which instructions are stored, and when the instructions are run on a computer, the computer is caused to execute the steps of the above-mentioned fast MRI image optimization and reconstruction method performed by the fast MRI image optimization and reconstruction device based on the pre-trained diffusion model.

[0050] The fifth aspect of the present invention provides a computer program product, including a computer program / instructions, characterized in that when the computer program / instructions are executed by a processor, the steps of the above-mentioned fast MRI image optimization and reconstruction method performed by the fast MRI image optimization and reconstruction device based on the pre-trained diffusion model are implemented.

[0051] In the technical solution provided by the present invention, a first conventional MRI image sample and a K-space data sample of the first conventional MRI image sample are obtained; the K-space data sample is subjected to undersampling processing and image reconstruction to obtain a simulated fast MRI image sample corresponding to the first conventional MRI image sample; an initial diffusion model is constructed, and the initial diffusion model is pre-trained based on the simulated fast MRI image sample and the first conventional MRI image sample to obtain a pre-trained diffusion model; a second conventional MRI image sample and an actual fast MRI image sample of the same acquisition object are obtained; the pre-trained diffusion model is fine-tuned based on the second conventional MRI image sample and the actual fast MRI image sample to obtain an image optimization and reconstruction model; in response to an image optimization and reconstruction request, the image optimization and reconstruction model is called to perform image optimization and reconstruction on the fast MRI image to be optimized and reconstructed, and an optimized and reconstructed target MRI image is obtained. This method can not only effectively shorten the scanning time for obtaining MRI images, but also reduce the noise in the obtained images and improve the quality of fast MRI images, thereby providing more efficient and accurate support for clinical diagnosis. An apparatus, an electronic device, a computer-readable storage medium, and a computer program product provided by the present invention also solve the corresponding technical problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings described herein are used to provide a further understanding of the present application and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0053] Figure 1 It is a schematic flowchart of the first embodiment of the fast MRI image optimization and reconstruction method based on a pre-trained diffusion model in an embodiment of the present invention;

[0054] Figure 2 It is a schematic flowchart of one of the second embodiments of the fast MRI image optimization and reconstruction method based on a pre-trained diffusion model in an embodiment of the present invention;

[0055] Figure 3 It is a schematic flowchart of another of the second embodiments of the fast MRI image optimization and reconstruction method based on a pre-trained diffusion model in an embodiment of the present invention;

[0056] Figure 4 It is a schematic diagram of the image optimization and reconstruction model of the second embodiment in the fast MRI image optimization and reconstruction method based on a pre-trained diffusion model in an embodiment of the present invention;

[0057] Figure 5 It is a schematic flowchart of the third embodiment of the fast MRI image optimization and reconstruction method based on a pre-trained diffusion model in an embodiment of the present invention;

[0058] Figure 6Schematic diagram of the image optimization and reconstruction model of the third embodiment in the fast MRI image optimization and reconstruction method based on the pre-trained diffusion model in the embodiments of the present invention;

[0059] Figure 7 Schematic diagram of an embodiment of the fast MRI image optimization and reconstruction device based on the pre-trained diffusion model in the embodiments of the present invention;

[0060] Figure 8 Schematic diagram of an embodiment of the fast MRI image optimization and reconstruction equipment based on the pre-trained diffusion model in the embodiments of the present invention;

[0061] Figure 9 Schematic diagram of the principle of a computer-readable medium in the embodiments of the present invention. Detailed implementation manners

[0062] Now, the exemplary embodiments of the present invention will be described more comprehensively with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the embodiments described herein. On the contrary, providing these exemplary embodiments enables the present invention to be more comprehensive and complete, and more convenient to fully convey the inventive concept to those skilled in the art. The same reference numerals in the figures denote the same or similar elements, components, or parts, and thus their repeated description will be omitted.

[0063] On the premise of conforming to the technical concept of the present invention, the features, structures, characteristics, or other details described in a specific embodiment do not exclude being combined in a suitable manner in one or more other embodiments.

[0064] In the description of specific embodiments, the features, structures, characteristics, or other details described in the present invention are for enabling those skilled in the art to fully understand the embodiments. However, it does not exclude that those skilled in the art can practice the technical solutions of the present invention without one or more of the specific features, structures, characteristics, or other details.

[0065] The flowcharts shown in the accompanying drawings are only exemplary illustrations, and do not necessarily include all the contents and operations / steps, nor do they necessarily need to be executed in the described order. For example, some operations / steps can be decomposed, while some operations / steps can be combined or partially combined, so the actual execution order may be changed according to the actual situation.

[0066] The block diagrams shown in the accompanying drawings are only functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0067] The term "and / or" or "and / or" includes all combinations of any one or more of the associated listed items.

[0068] Please refer to Figure 1 , the first embodiment of the fast MRI image optimization and reconstruction method based on a pre-trained diffusion model in the embodiments of the present invention includes:

[0069] S101. Obtain a first conventional MRI image sample and a K-space data sample of the first conventional MRI image sample;

[0070] It can be understood that the execution subject of the present invention can be a fast MRI (Nuclear Magnetic Resonance Imaging) image optimization and reconstruction device based on a pre-trained diffusion model, or a terminal or a server, and specifically is not limited here. In the embodiments of the present invention, the server is taken as an example of the execution subject for illustration.

[0071] In this embodiment, the server can build and train an image optimization model to be used before specifically responding to image optimization and reconstruction. The image optimization model is specifically used to perform optimization and reconstruction on the fast MRI image obtained by fast MRI scanning, remove the image noise brought by fast MRI scanning, and predict the normal MRI image result of the current scanned object based on the fast MRI image, where the normal MRI image refers to a clearer MRI image under normal circumstances with a longer scanning time relative to the fast MRI image.

[0072] Generally, if a large amount of paired data is required to train an image optimization and reconstruction model, and if all the paired data obtained by actual scanning is used as samples, then each time an image sample is obtained, the same scanned object needs to be scanned twice with different parameters. Therefore, it is very difficult to obtain a large amount of paired MRI images as training samples.

[0073] Therefore, in this embodiment, first obtain a conventional MRI image as the first conventional MRI image sample, where the first conventional MRI image sample can be a normal MRI image extracted from the public dataset fast-MRI; at the same time, obtain the K-space data corresponding to the first conventional MRI image sample as the K-space data sample, where the K-space data refers to the intermediate data obtained after encoding the original data obtained during MRI scanning.

[0074] S102. Perform undersampling processing on the K-space data sample and perform image reconstruction to obtain a simulated fast MRI image sample corresponding to the first conventional MRI image sample;

[0075] After obtaining the conventional MRI image as the first conventional MRI image sample, this step constructs the paired data of the first conventional MRI image sample. Specifically, the K-space data sample corresponding to the first conventional MRI image sample is obtained, undersampled, and then image reconstruction is performed based on the undersampled K-space data sample to obtain an MRI image to simulate the fast MRI image of the same scanned object as the first conventional MRI image sample during fast MRI scanning, and this is used as the simulated fast MRI image sample.

[0076] Specifically, the undersampling step can be to reduce the sampling lines of the K-space data sample, which can be achieved by uniformly or randomly discarding some sampling lines in the K-space data sample.

[0077] In a specific implementation, after undersampling the K-space data sample, it further includes a data recovery step for the undersampled data. Specifically, it includes: performing undersampling on the K-space data sample to obtain an undersampled K-space data sample; calling the GRAPPA (GeneRalized Autocalibrating Partial Parallel Acquisition) algorithm to perform data fitting on the undersampled K-space data sample, calculating the unacquired data at the K-space center, and complementing the undersampled K-space data sample to obtain a complemented K-space data sample; calling the SENSE algorithm, and based on the complemented K-space data sample, performing image reconstruction in combination with the estimated sensitivity matrix value to obtain a simulated fast MRI image sample. In this way, the optimization operations that the simulated fast MRI image sample would perform when undersampled are simulated, so that the simulated fast MRI image sample is more similar to the fast MRI image obtained in actual scanning.

[0078] S103. Construct an initial diffusion model, and pre-train the initial diffusion model based on the first conventional MRI image sample and the simulated fast MRI image sample to obtain a pre-trained diffusion model;

[0079] In this embodiment, the initial diffusion model can be constructed before or after obtaining the simulated fast MRI image sample corresponding to the first conventional MRI image sample. Among them, the diffusion model can learn the latent structure of the dataset by modeling the diffusion method of data points in the latent space, and it can be applied to tasks such as image denoising, image inpainting, super-resolution imaging, and image generation. In a specific implementation, the constructed initial diffusion model can be a conditional diffusion model or a diffusion bridge model.

[0080] After constructing the initial diffusion model, pre-train the initial diffusion model with the first conventional MRI image samples and simulated fast MRI image samples obtained in the manner described in the foregoing S101 and S102 to obtain a pre-trained diffusion model. Specifically, during pre-training, randomly select the cross-sectional images of three consecutive layers in the simulated fast MRI image samples as the input of the initial diffusion model, and select the cross-sectional images of three consecutive layers at the corresponding positions in the paired first conventional MRI image samples of the simulated fast MRI image samples as the output of the initial diffusion model, and adjust the parameters of the initial diffusion model to obtain a pre-trained diffusion model.

[0081] S104. Obtain a second conventional MRI image sample and an actual fast MRI image sample of the same acquisition object;

[0082] In this embodiment, it also includes obtaining a second conventional MRI image sample and an actual fast MRI image sample of the same acquisition object, that is, performing two scans with different parameters on the same scanning object to obtain a second conventional MRI image sample and an actual fast MRI image sample respectively.

[0083] S105. Fine-tune and train the pre-trained diffusion model based on the second conventional MRI image sample and the actual fast MRI image sample to obtain an image optimization and reconstruction model;

[0084] Use the second conventional MRI image sample and the actual fast MRI image sample to fine-tune and train the obtained pre-trained diffusion model according to the solution in the foregoing S103 to obtain a diffusion model that can adapt to real data as the image optimization and reconstruction model required in this embodiment.

[0085] S106. In response to an image optimization and reconstruction request, call the image optimization and reconstruction model to perform image optimization and reconstruction on the fast MRI image to be image-optimized and reconstructed, and obtain an optimized and reconstructed target MRI image.

[0086] After obtaining the image optimization and reconstruction model, specific image optimization and reconstruction requests can be executed. Specifically, in this embodiment, an image optimization and reconstruction request can be received and responded to, and the fast MRI image to be image-optimized and reconstructed included in the image optimization and reconstruction request can be obtained. Input the fast MRI image to be image-optimized and reconstructed into the image optimization and reconstruction model for image optimization and reconstruction, and predict the target MRI image with better clarity that the scanning object corresponding to the fast MRI image included in the image optimization and reconstruction request may obtain during a scan at a conventional speed.

[0087] The method provided in the embodiment of the present invention can not only effectively shorten the scanning time for obtaining MRI images, but also reduce the noise in the obtained images and improve the quality of fast MRI images, thereby providing more efficient and accurate support for clinical diagnosis.

[0088] Please refer to Figures 2-4 , the second embodiment of the fast MRI image optimization and reconstruction method based on the pre-trained diffusion model in the embodiments of the present invention includes:

[0089] S201. Obtain the first conventional MRI image sample and the K-space data sample of the first conventional MRI image sample;

[0090] The content in step S201 in this embodiment is basically the same as the content in step S101 in the foregoing embodiment, so it will not be elaborated here.

[0091] S202. Perform undersampling processing on the K-space data sample to obtain an undersampled K-space data sample;

[0092] Subsequently, simulate the fast MRI image sample based on the K-space data sample. In a specific implementation, uniformly or randomly discard some sampling rows of the K-space data sample based on the undersampling parameter to obtain an undersampled K-space data sample.

[0093] In a specific implementation, before performing the specific undersampling operation, the undersampling parameter is determined according to the fast scanning requirements required when simulating the fast MRI image sample.

[0094] S203. Invoke the GRAPPA algorithm to perform data fitting on the undersampled K-space data sample, calculate the data not sampled at the K-space center, and complement the undersampled K-space data sample to obtain a complemented K-space data sample;

[0095] After obtaining the undersampled K-space data sample, invoke the GRAPPA algorithm for data fitting to calculate more data not sampled at the K-space center. The specific calculation expression is:

[0096]

[0097] where I j (k u +mΔk y ,k x 0 is the magnetic resonance signal of the jth coil of the MRI at 9k y +mΔk y ,k x ), w is the weight coefficient, R is the acceleration factor, j represents the target coil, l represents the coil coefficient, b and h are the data blocks in the phase encoding k y and frequency encoding direction k x directions of the reconstruction area.

[0098] Based on the above fitting steps, an estimated value of the sensitivity matrix of the coil is calculated.

[0099] S204, calling the SENSE algorithm, combining the sensitivity matrix estimation value and performing image reconstruction based on the completed K-space data samples, to obtain simulated fast MRI image samples;

[0100] Next, reconstruction based on SENSE is performed. The specific reconstruction calculation expression is:

[0101] I=Eρ

[0102] Where I represents the simulated fast MRI image sample; ρ represents the image domain coordinates generated according to the K-space data; E is the system encoding matrix, and its specific expression is:

[0103]

[0104] Among them, r p refers to the pth image pixel, k K is the kth K-space value, C j (r p ) is the jth coil located at r p The coil sensitivity function at the position; in this step, the conjugate gradient iterative reconstruction method is introduced to solve it in order to reduce the calculation time and improve the image accuracy.

[0105] S205, alternately iterating the sensitivity matrix estimation value and the simulated fast MRI image sample to optimize the simulated fast MRI image sample;

[0106] In this step, the estimated values ​​of the sensitivity matrix obtained in the first two steps and the reconstructed image are alternately iterated to perform further optimization.

[0107] In one embodiment, the coil sensitivity function is represented by a polynomial fitting method, and the specific expression is:

[0108]

[0109] Among them, C l (r) represents the sensitivity function, (x, y) = r represents the pixel position, a l,i,j represents the polynomial fitting coefficient, and N represents the order of the fitting polynomial;

[0110] The specific iterative process is as follows: the sensitivity obtained in the previous step is used to obtain the reconstructed image f. The specific expression is:

[0111]

[0112] f is obtained by the conjugate gradient iteration method and used to iteratively obtain the sensitivity coefficient a. The specific expression is:

[0113]

[0114] Among them, the sensitivity coefficient a is a parameter used to describe the sensitivity matrix;

[0115] Perform alternating iteration until the objective function stops decreasing, and the reconstructed image obtained is the final reconstructed image result. The specific expression of the objective function is:

[0116]

[0117] Among them, E(a) is the encoding matrix, and the reconstruction result is the image obtained in the last step.

[0118] Using the above process, by controlling the parameters of undersampling and the number of iterations, simulated fast MRI image samples simulating the sampling time and noise level of different devices can be obtained, so as to obtain paired data of simulated conventional MRI images - fast MRI image samples.

[0119] In a specific implementation manner, please refer to Figure 3 , and it also includes the step of data preprocessing. Calculate the image sizes of the first conventional MRI image sample and the simulated fast MRI image sample respectively; based on the image sizes, perform normalization processing on the first conventional MRI image sample and the simulated fast MRI image sample respectively; so as to achieve the same resolution for data in the same batch and meet the requirements of model input, and enable the network to perform stable training. Subsequently, randomly select the normalized first conventional MRI image sample and the simulated fast MRI image sample, and perform data augmentation operations, including randomly translating, rotating, flipping the image, etc., to enhance the robustness of the model.

[0120] S206. Construct an initial diffusion model, and pre-train the initial diffusion model based on the first conventional MRI image sample and the simulated fast MRI image sample to obtain a pre-trained diffusion model;

[0121] Before or after obtaining the simulated fast MRI image sample corresponding to the first conventional MRI image sample, construct an initial diffusion model. The initial diffusion model can be a bridge diffusion model or a conditional diffusion model. Taking the conditional diffusion model as an example, its process schematic diagram can be as Figure 4 shown.

[0122] In one embodiment, three consecutive cross-sectional images of the simulated fast MRI image samples are randomly selected as the model input of the initial diffusion model; the number of model steps is randomly selected, and Gaussian noise is randomly generated and added to the three consecutive cross-sectional images of the simulated fast MRI image samples; three consecutive cross-sections at the corresponding positions in the corresponding first conventional MRI image samples are selected as the model output of the initial diffusion model; an optimizer is called to pre-train the initial diffusion model until the model converges, obtaining a pre-trained diffusion model.

[0123] S207. Obtain a second conventional MRI image sample and an actual fast MRI image sample of the same acquisition object;

[0124] S208. Fine-tune the pre-trained diffusion model based on the second conventional MRI image sample and the actual fast MRI image sample to obtain an image optimization and reconstruction model;

[0125] Three consecutive cross-sectional images of the actual fast MRI image sample are randomly selected as the model input of the pre-trained diffusion model; the number of model steps is randomly selected, and Gaussian noise is randomly generated and added to the three consecutive cross-sectional images of the actual fast MRI image sample; three consecutive cross-sections at the corresponding positions in the corresponding second conventional MRI image sample are selected as the model output of the pre-trained diffusion model; an optimizer is called to fine-tune the pre-trained diffusion model to obtain an image optimization and reconstruction model.

[0126] In a specific embodiment, please continue to refer to Figure 3 , after obtaining the second conventional MRI image sample and the actual fast MRI image sample, it further includes a data preprocessing step, which is specifically basically similar to the preprocessing step in step S205 of this embodiment, including the following steps: calculating the image sizes of the second conventional MRI image sample and the actual fast MRI image sample respectively; based on the image sizes, normalizing the second conventional MRI image sample and the actual fast MRI image sample respectively; to achieve the same resolution for the same batch of data and meet the requirements of the model input, and enable the network to perform stable training. Subsequently, it further includes randomly selecting the normalized second conventional MRI image sample and the actual fast MRI image sample, and performing data augmentation operations, including randomly translating, rotating, flipping the images, etc., to enhance the robustness of the model.

[0127] S209. In response to an image optimization and reconstruction request, call the image optimization and reconstruction model to perform image optimization and reconstruction on the fast MRI image to be optimized and reconstructed, obtaining an optimized and reconstructed target MRI image.

[0128] Input the fast MRI image to be optimized and reconstructed into the image optimization and reconstruction model for image optimization and reconstruction, and predict the target MRI image with better clarity that the scanned object corresponding to the fast MRI image included in the image optimization and reconstruction request may obtain when scanned at a conventional speed.

[0129] In a specific implementation manner, when performing image optimization and reconstruction, specifically, the target image is obtained based on the reverse diffusion process of the image optimization and reconstruction model.

[0130] The method provided in the embodiments of the present invention can not only effectively shorten the scanning time for obtaining MRI images, but also reduce the noise in the obtained images and improve the quality of fast MRI images, thereby providing more efficient and accurate support for clinical diagnosis.

[0131] Please refer to Figures 5-6 , the third embodiment of the fast MRI image optimization and reconstruction method based on the pre-trained diffusion model in the embodiments of the present invention includes:

[0132] S301. Obtain the first conventional MRI image sample and the K-space data sample of the first conventional MRI image sample;

[0133] S302. Perform undersampling processing on the K-space data sample and perform image reconstruction to obtain the simulated fast MRI image sample corresponding to the first conventional MRI image sample;

[0134] The basic content in steps S301 - S302 in this embodiment is basically the same as the content in S201 - S205 in the foregoing embodiment, so it will not be elaborated here.

[0135] In a specific implementation manner, after obtaining the first conventional MRI image sample and the simulated fast MRI image sample, it further includes the step of data preprocessing, calculating the image sizes of the first conventional MRI image sample and the simulated fast MRI image sample respectively; based on the image sizes, normalizing the first conventional MRI image sample and the simulated fast MRI image sample respectively; so as to achieve the same resolution for the same batch of data and meet the requirements of model input, and enable the network to perform stable training. Subsequently, it further includes randomly selecting the normalized first conventional MRI image sample and the simulated fast MRI image sample and performing data augmentation operations, including randomly translating, rotating, flipping the image, etc., to enhance the robustness of the model.

[0136] S303. Construct an initial diffusion model based on the Brownian bridge diffusion model;

[0137] In this embodiment, an initial diffusion model is constructed based on the Brownian bridge diffusion model. Please refer to Figure 6, which shows the forward diffusion process and the reverse diffusion process of the Brownian bridge diffusion model. Specifically, in this embodiment, the expression of the forward diffusion of the initial diffusion model constructed based on the Brownian bridge diffusion model is:

[0138] q BB (x t |(x 0 ,y) = N(x t ; (1 - m t )x 0 + m t y, δ t I))

[0139] Where x 0 = x, t is the time step, T is the total number of steps of the diffusion model, and δ t represents the noise intensity, and the expression is: Where ∈ is randomly sampled Gaussian noise, and ∈ θ is the noise prediction network in the diffusion model.

[0140] S304. Pre-train the initial diffusion model based on the first conventional MRI image sample and the simulated fast MRI image sample to obtain a pre-trained diffusion model;

[0141] In this embodiment, first, the initial diffusion model is pre-trained using the simulated first conventional MRI image sample and the simulated fast MRI image sample.

[0142] In a specific implementation manner, randomly select the cross-sectional images of three consecutive layers in the simulated fast MRI image sample as the model input of the initial diffusion model; secondly, randomly select the number of model steps and randomly generate Gaussian noise and add it to the cross-sectional images of three consecutive layers in the simulated fast MRI image sample; thirdly, select the cross-sectional images of three consecutive layers at the corresponding positions in the corresponding first conventional MRI image sample as the model output of the initial diffusion model; in addition, call the optimizer to pre-train the initial diffusion model until the model converges to obtain a pre-trained diffusion model.

[0143] In a specific implementation manner, when the initial diffusion model is constructed based on the Brownian bridge diffusion model, the pre-training method specifically includes: randomly select the cross-sectional images of three consecutive layers of the simulated fast MRI as the input y, and select the cross-sectional images of three consecutive layers at the corresponding positions of the paired normal MRI as x 0 , randomly select the time step t from 1 to T, and randomly generate Gaussian noise ∈ to simulate the noise degradation process. Then calculate the image x t at the t-th forward step as the target output, and the specific expression is:

[0144]

[0145] The loss function formula is as follows:

[0146]

[0147] Subsequently, the Adam optimizer is called to train the model until convergence, thereby obtaining a pre-trained diffusion model.

[0148] S305. Obtain the second conventional MRI image sample and the actual fast MRI image sample of the same acquisition object;

[0149] In this embodiment, the content of step S305 is basically the same as that of step S104 in the foregoing embodiment, so it will not be elaborated here.

[0150] In a specific implementation manner, it further includes performing preprocessing operations on the second conventional MRI image sample and the actual fast MRI image sample that are similar to those on the first conventional MRI image sample and the simulated fast MRI image sample.

[0151] S306. Fine-tune and train the pre-trained diffusion model based on the second conventional MRI image sample and the actual fast MRI image sample to obtain an image optimization and reconstruction model;

[0152] First, randomly select the cross-sectional images of three consecutive layers in the actual fast MRI image sample as the model input of the pre-trained diffusion model; secondly, randomly select the model steps and randomly generate Gaussian noise and add it to the cross-sectional images of three consecutive layers in the actual fast MRI image sample; thirdly, select the cross-sectional images of three consecutive layers at the corresponding positions in the corresponding second conventional MRI image sample as the model output of the pre-trained diffusion model; in addition, call the optimizer to fine-tune and train the pre-trained diffusion model to obtain an image optimization and reconstruction model.

[0153] S307. Obtain the cross-sectional images of three consecutive layers in the fast MRI image to be optimized for image reconstruction as the model input of the image optimization and reconstruction model, and call the image optimization and reconstruction model to perform image optimization and reconstruction to obtain the prediction result of the target MRI image.

[0154] Select the cross-sectional images of three consecutive layers of the fast MRI as the input y = x T , sequentially select the time step t in T,..., 1, and calculate the intermediate step x of the reverse diffusion according to the expression t-1 , and the specific expression is:

[0155]

[0156] Among them, Iterate until t = 1 to obtain the predicted x 0 , which is the prediction result of the target MRI image.

[0157] In the embodiments of the present invention, the method is provided to avoid the poor model effect caused by too little training data. Normal MRI images and their corresponding k-space data in the public dataset are used to construct simulated fast MRI image samples. The model is trained with the simulated image samples, which greatly expands the number of paired data available for training and increases the diversity of fast MRI, thereby enhancing the robustness and generalization of the model. At the same time, the model is constructed based on the diffusion model structure, which can improve the details of the generated images. The method in the embodiments of the present invention can not only effectively shorten the scanning time for obtaining MRI images, but also reduce the noise in the obtained images, improve the quality of fast MRI images, and thus provide more efficient and accurate support for clinical diagnosis.

[0158] The above describes the fast MRI image optimization and reconstruction method based on the pre-trained diffusion model in the embodiments of the present invention. Next, the fast MRI image optimization and reconstruction device based on the pre-trained diffusion model in the embodiments of the present invention will be described. Please refer to Figure 7 , an embodiment of the fast MRI image optimization and reconstruction device based on the pre-trained diffusion model in the embodiments of the present invention includes:

[0159] A sample simulation module 701, configured to obtain a first conventional MRI image sample and a k-space data sample of the first conventional MRI image sample; and perform undersampling processing on the k-space data sample and perform image reconstruction to obtain a simulated fast MRI image sample corresponding to the first conventional MRI image sample;

[0160] A model training module 702, configured to construct an initial diffusion model, pre-train the initial diffusion model based on the first conventional MRI image sample and the simulated fast MRI image sample to obtain a pre-trained diffusion model; and configured to obtain a second conventional MRI image sample and an actual fast MRI image sample of the same acquisition object; fine-tune and train the pre-trained diffusion model based on the second conventional MRI image sample and the actual fast MRI image sample to obtain an image optimization and reconstruction model;

[0161] An optimization and reconstruction module 703, configured to respond to an image optimization and reconstruction request, call the image optimization and reconstruction model to perform image optimization and reconstruction on the fast MRI image to be optimized and reconstructed, and obtain an optimized and reconstructed target MRI image.

[0162] The device provided in the embodiments of the present invention can not only effectively shorten the scanning time for obtaining MRI images, but also reduce the noise in the obtained images, improve the quality of fast MRI images, and thus provide more efficient and accurate support for clinical diagnosis.

[0163] In another embodiment of the present application, the sample simulation module 701 includes:

[0164] An undersampling unit for performing undersampling processing on the K-space data samples to obtain undersampled K-space data samples;

[0165] A completion unit for calling the GRAPPA algorithm to perform data fitting on the undersampled K-space data samples, calculating the unacquired data at the K-space center, and completing the undersampled K-space data samples to obtain completed K-space data samples;

[0166] A reconstruction unit for calling the SENSE algorithm and performing image reconstruction based on the completed K-space data samples in combination with the estimated sensitivity matrix value to obtain simulated fast MRI image samples.

[0167] In another embodiment of the present application, the fast MRI image optimization and reconstruction device based on the pre-trained diffusion model further includes a requirement acquisition module, and the requirement acquisition module is specifically used to determine the undersampling parameters according to the fast scanning requirements;

[0168] The undersampling unit is specifically used to: discard some sampling rows of the K-space data samples based on the undersampling parameters to obtain undersampled K-space data samples.

[0169] In another embodiment of the present application, the completion unit is further used to: calculate the estimated value of the sensitivity matrix of the scanning coil;

[0170] The reconstruction unit is further used to alternately iterate the estimated sensitivity matrix value and the simulated fast MRI image samples to optimize the simulated fast MRI image samples.

[0171] In another embodiment of the present application, the requirement acquisition module is further used to: determine the number of iterations of the alternate iteration according to the fast scanning requirements.

[0172] In another embodiment of the present application, the initial diffusion model is constructed based on the Brownian bridge diffusion model; the model training module 702 further includes a pre-training unit, and the pre-training unit is specifically used to:

[0173] Randomly select the cross-sectional images of three consecutive layers in the simulated fast MRI image samples as the model input of the initial diffusion model;

[0174] Randomly select the number of model steps and randomly generate Gaussian noise and add it to the cross-sectional images of three consecutive layers in the simulated fast MRI image samples;

[0175] Select the cross-sections of three consecutive layers at the corresponding positions in the corresponding first conventional MRI image samples as the model output of the initial diffusion model;

[0176] Call the optimizer to pre-train the initial diffusion model until the model converges to obtain a pre-trained diffusion model.

[0177] In another embodiment of the present application, the model training module 702 further includes a fine-tuning training unit, and the fine-tuning training unit is specifically configured to:

[0178] Randomly select cross-sectional images of three consecutive layers in the actual fast MRI image samples as the model input of the pre-trained diffusion model;

[0179] Randomly select the number of model steps, and randomly generate Gaussian noise and add it to the cross-sectional images of three consecutive layers in the actual fast MRI image samples;

[0180] Select cross-sections of three consecutive layers at the corresponding positions in the corresponding second conventional MRI image samples as the model output of the pre-trained diffusion model;

[0181] Call the optimizer to perform fine-tuning training on the pre-trained diffusion model to obtain an image optimization and reconstruction model.

[0182] In another embodiment of the present application, the optimization and reconstruction module 703 is specifically configured to: Obtain cross-sectional images of three consecutive layers in the fast MRI image to be optimized and reconstructed as the model input of the image optimization and reconstruction model, and call the image optimization and reconstruction model to perform image optimization and reconstruction to obtain the prediction result of the target MRI image.

[0183] In another embodiment of the present application, the sample simulation module 701 is specifically further configured to:

[0184] Calculate the image sizes of the first conventional MRI image samples and the simulated fast MRI image samples respectively;

[0185] Based on the image sizes, perform normalization processing on the first conventional MRI image samples and the simulated fast MRI image samples respectively;

[0186] Randomly select the normalized first conventional MRI image samples and the simulated fast MRI image samples, and perform data augmentation operations.

[0187] In order to avoid the poor model effect caused by too little training data, the device provided in the embodiment of the present invention uses normal MRI images in the public data set and their corresponding K-space data to construct simulated fast MRI image samples. The model is trained by the simulated image samples, which greatly expands the number of paired data that can be used for training and increases the diversity of fast MRI, which improves the robustness and generalization of the model. At the same time, the model is constructed based on the diffusion model structure, which can improve the details of the generated image. The device in the embodiment of the present invention can not only effectively shorten the scanning time for obtaining MRI images, but also reduce the noise in the obtained images, and improve the quality of fast MRI images, thereby providing more efficient and accurate support for clinical diagnosis.

[0188] Based on the same inventive concept, an embodiment of this specification also provides an electronic device for fast MRI image optimization reconstruction based on a pre-trained diffusion model. The electronic device for fast MRI image optimization reconstruction based on a pre-trained diffusion model in an embodiment of the present invention is described in detail below from the perspective of hardware processing.

[0189] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this specification. Figure 8 The electronic device 800 according to this embodiment of the present invention is described. Figure 8 The electronic device 800 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0190] like Figure 8 As shown, the electronic device 800 is in the form of a general computing device. The components of the electronic device 800 may include, but are not limited to: at least one processing unit 810, at least one storage unit 820, a bus 830 connecting different system components (including the storage unit 820 and the processing unit 810), a display unit 840, etc.

[0191] The storage unit stores program codes, which can be executed by the processing unit 810, so that the processing unit 810 performs the steps according to various exemplary embodiments of the present invention described in the above processing method section of this specification. For example, the processing unit 810 can perform the following steps: Figure 1 , Figure 2 or Figure 5 The steps of the method are shown.

[0192] The storage unit 820 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 8201 and / or a cache memory unit 8202 , and may further include a read-only memory unit (ROM) 8203 .

[0193] The storage unit 820 may further include a program / utility 8204 having a set (at least one) of program modules 8205. Such program modules 8205 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment.

[0194] The bus 830 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus structures.

[0195] The electronic device 800 may also communicate with one or more external devices 100 (such as a keyboard, a pointing device, a Bluetooth device, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 800, and / or may communicate with any device that enables the electronic device 800 to communicate with one or more other computing devices (such as a router, a modem, etc.). Such communication may be through the input / output (I / O) interface 850. And, the electronic device 800 may also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 860. The network adapter 860 may communicate with other modules of the electronic device 800 through the bus 830. It should be understood that although Figure 8 not shown, other hardware and / or software modules may be used in conjunction with the electronic device 800, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0196] Through the description of the above embodiments, those skilled in the art can easily understand that the exemplary embodiments described in the present invention can be implemented by software, or can be implemented by the way of software combined with necessary hardware. Therefore, the technical solutions according to the embodiments of the present invention can be embodied in the form of a software product, and the software product can be stored in a computer-readable storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, or a network device, etc.) to execute the above method according to the present invention. When the computer program is executed by a data processing device, the computer-readable medium can implement the above method of the present invention, that is: as Figure 1 、 Figure 2 or Figure 5 shown in the method.

[0197] Figure 9 Schematic diagram of the principle of a computer-readable medium provided by the embodiments of this specification.

[0198] Implement Figure 1 、 Figure 2 Or Figure 5 The computer program for implementing the method shown can be stored on one or more computer-readable media. The computer-readable media can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0199] The computer-readable storage medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries the readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable storage medium can also be any readable medium other than the readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium can be transmitted using any appropriate medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination of the above.

[0200] The program code for performing the operations of the present invention can be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and also including conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, executed as a stand-alone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, by using an Internet service provider to connect through the Internet).

[0201] In summary, the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that general-purpose data processing devices such as microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components in the embodiments of the present invention. The present invention can also be implemented as a device or device program (e.g., a computer program and a computer program product) for executing some or all of the methods described herein. Such a program implementing the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, or provided on a carrier signal, or in any other form.

[0202] In addition, the present invention also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implements the fast MRI image optimization and reconstruction method based on a pre-trained diffusion model described in any of the above embodiments.

[0203] The specific embodiments described above further elaborate on the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the present invention is not inherently related to any specific computer, virtual device, or electronic device, and various general-purpose devices can also implement the present invention. The above are only specific embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

[0204] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments.

[0205] If the technical solution of this application involves personal information, before the product applying the technical solution of this application processes personal information, it has clearly informed the personal information processing rules and obtained the individual's independent consent. If the technical solution of this application involves sensitive personal information, before the product applying the technical solution of this application processes sensitive personal information, it has obtained the individual's separate consent and at the same time meets the requirements of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform that the personal information collection scope has been entered and personal information will be collected. If an individual voluntarily enters the collection scope, it is regarded as consenting to the collection of their personal information; or on the device for personal information processing, when the personal information processing rules are informed by obvious signs / information, personal authorization is obtained through pop-up messages or by asking the individual to upload their personal information by themselves, etc.; among them, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0206] The above are only the embodiments of this application and are not used to limit this application. For those skilled in the art, various changes and modifications can be made to this application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of this application shall be included within the scope of the claims of this application.

Claims

1. A fast MRI image optimization reconstruction method based on a pre-trained diffusion model, characterized in that: include: Acquire a first conventional MRI image sample and a K-space data sample of the first conventional MRI image sample; Performing undersampling processing on the K-space data samples and performing image reconstruction to obtain simulated fast MRI image samples corresponding to the first conventional MRI image samples; constructing an initial diffusion model, and pre-training the initial diffusion model based on the first conventional MRI image sample and the simulated fast MRI image sample to obtain a pre-trained diffusion model; Acquire a second conventional MRI image sample and an actual rapid MRI image sample of the same acquisition object; Fine-tune the pre-trained diffusion model based on the second conventional MRI image sample and the actual fast MRI image sample to obtain an image optimization reconstruction model; In response to an image optimization reconstruction request, the image optimization reconstruction model is called to perform image optimization reconstruction on the fast MRI image to be subjected to image optimization reconstruction, so as to obtain an optimized reconstructed target MRI image.

2. The fast MRI image optimization reconstruction method based on the pre-trained diffusion model according to claim 1, characterized in that: The performing undersampling processing on the K-space data samples and image reconstruction to obtain simulated fast MRI image samples corresponding to the first conventional MRI image samples comprises: Performing undersampling processing on the K-space data samples to obtain undersampled K-space data samples; Calling the GRAPPA algorithm to perform data fitting on the under-sampled K-space data samples, calculating the unsampled data at the center of the K-space, and completing the under-sampled K-space data samples to obtain completed K-space data samples; The SENSE algorithm is called, and image reconstruction is performed based on the completed K-space data samples in combination with the sensitivity matrix estimation value to obtain simulated fast MRI image samples.

3. The fast MRI image optimization reconstruction method based on the pre-trained diffusion model according to claim 2, characterized in that: Before performing undersampling processing on the K-space data samples to obtain undersampled K-space data samples, the method further includes: Determine undersampling parameters based on fast scanning requirements; The performing undersampling processing on the K-space data samples to obtain undersampling K-space data samples comprises: Part of the sampling rows of the K-space data samples are discarded based on the under-sampling parameters to obtain under-sampled K-space data samples.

4. The fast MRI image optimization reconstruction method based on the pre-trained diffusion model according to claim 2, characterized in that: After calling the GRAPPA algorithm to perform data fitting on the under-sampled K-space data samples, calculating the unsampled data at the center of the K-space, and completing the under-sampled K-space data samples to obtain completed K-space data samples, the method further includes: Calculating an estimate of the sensitivity matrix of the scanning coil; After the SENSE algorithm is called and image reconstruction is performed based on the completed K-space data samples in combination with the sensitivity matrix estimation value to obtain simulated fast MRI image samples, the method further includes: The sensitivity matrix estimation value and the simulated fast MRI image samples are iterated alternately to optimize the simulated fast MRI image samples.

5. The fast MRI image optimization reconstruction method based on the pre-trained diffusion model according to claim 4, characterized in that: Before the step of alternately iterating the sensitivity matrix estimation value and simulating fast MRI image samples and optimizing the simulated fast MRI image samples, the method further includes: The number of iterations of the alternating iterations is determined based on the fast scanning requirements.

6. The fast MRI image optimization reconstruction method based on the pre-trained diffusion model according to claim 1, characterized in that: The initial diffusion model is constructed based on the Brownian bridge diffusion model; The pre-training of the initial diffusion model based on the first conventional MRI image sample and the simulated fast MRI image sample to obtain the pre-trained diffusion model comprises: Randomly selecting three consecutive cross-sectional images from the simulated fast MRI image sample as model inputs of the initial diffusion model; Randomly selecting the model step number, and randomly generating Gaussian noise to add to the cross-sectional images of three consecutive layers in the simulated fast MRI image sample; Selecting three consecutive cross-sections at corresponding positions in the corresponding first conventional MRI image sample as the model output of the initial diffusion model; The optimizer is called to pre-train the initial diffusion model until the model converges to obtain a pre-trained diffusion model.

7. The fast MRI image optimization reconstruction method based on the pre-trained diffusion model according to claim 6, characterized in that: The fine-tuning training of the pre-trained diffusion model based on the second conventional MRI image sample and the actual fast MRI image sample to obtain the image optimization reconstruction model comprises: Randomly selecting three consecutive cross-sectional images from the actual rapid MRI image sample as model inputs of the pre-trained diffusion model; Randomly selecting the model step number, and randomly generating Gaussian noise to add to the cross-sectional images of three consecutive layers in the actual fast MRI image sample; Selecting three consecutive cross-sections at corresponding positions in the corresponding second conventional MRI image sample as the model output of the pre-trained diffusion model; The optimizer is called to perform fine-tuning training on the pre-trained diffusion model to obtain an image optimization reconstruction model.

8. The fast MRI image optimization reconstruction method based on the pre-trained diffusion model according to claim 7, characterized in that: Calling the image optimization reconstruction model to perform image optimization reconstruction on the fast MRI image to be subjected to image optimization reconstruction to obtain the optimized reconstructed target MRI image comprises: The cross-sectional images of three consecutive layers in the fast MRI image to be optimized for image reconstruction are obtained as the model input of the image optimization reconstruction model, and the image optimization reconstruction model is called to perform image optimization reconstruction to obtain the prediction result of the target MRI image.

9. The fast MRI image optimization reconstruction method based on a pre-trained diffusion model according to any one of claims 1 to 8, characterized in that: After obtaining the simulated fast MRI image sample corresponding to the first conventional MRI image sample, the method further includes: respectively calculating the image sizes of the first conventional MRI image sample and the simulated fast MRI image sample; Based on the image size, respectively normalizing the first conventional MRI image sample and the simulated fast MRI image sample; The first conventional MRI image samples and the simulated fast MRI image samples after normalization are randomly selected to perform a data enhancement operation.

10. A fast MRI image optimization reconstruction device based on a pre-trained diffusion model, characterized in that: The fast MRI image optimization reconstruction device based on the pre-trained diffusion model includes: a sample simulation module, configured to obtain a first conventional MRI image sample and a K-space data sample of the first conventional MRI image sample; and to perform undersampling processing on the K-space data sample and image reconstruction to obtain a simulated fast MRI image sample corresponding to the first conventional MRI image sample; A model training module is used to construct an initial diffusion model, pre-train the initial diffusion model based on the first conventional MRI image sample and the simulated fast MRI image sample to obtain a pre-trained diffusion model; and to obtain a second conventional MRI image sample and an actual fast MRI image sample of the same acquisition object; and fine-tune the pre-trained diffusion model based on the second conventional MRI image sample and the actual fast MRI image sample to obtain an image optimization reconstruction model; The optimization reconstruction module is used to respond to the image optimization reconstruction request, call the image optimization reconstruction model to perform image optimization reconstruction on the fast MRI image to be image optimized and reconstructed, and obtain the target MRI image after optimization and reconstruction.

11. A fast MRI image optimization reconstruction device based on a pre-trained diffusion model, characterized in that: The fast MRI image optimization reconstruction device based on the pre-trained diffusion model includes: a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instructions in the memory to enable the fast MRI image optimization and reconstruction device based on the pre-trained diffusion model to perform the steps of the fast MRI image optimization and reconstruction method based on the pre-trained diffusion model as described in any one of claims 1-9.

12. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the program / instructions are executed by a processor, the steps of the fast MRI image optimization reconstruction method based on a pre-trained diffusion model as described in any one of claims 1 to 9 are implemented.

13. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the fast MRI image optimization reconstruction method based on a pre-trained diffusion model as described in any one of claims 1 to 9 are implemented.

Images