Medical MRI image denoising method, device, system, equipment, medium and product

Abstract

This invention discloses a method, apparatus, system, device, medium, and product for denoising medical MRI images, relating to the field of image enhancement technology. The method first converts the medical MRI image signal to the image domain to obtain three-dimensional image data. Then, it selects three-dimensional image sub-blocks and searches for the M most similar three-dimensional image sub-blocks within a predefined neighborhood. These sub-blocks are then vectorized, straightened, and superimposed to obtain the original matrix Y. Next, assuming the original matrix Y = X + N + S, the objective function is minimized to obtain the denoising matrix X. Finally, new sub-blocks corresponding to the aforementioned sub-blocks are reconstructed based on the denoising matrix X. The replacement of the old and new sub-blocks yields new denoised three-dimensional image data. This solves the problem of image blurring after denoising in existing technologies applied to MRI denoising, improving the quality of medical MRI images and facilitating subsequent image analysis.

Description

Technical Field

[0001] This invention belongs to the field of image enhancement technology, specifically relating to a method, apparatus, system, device, medium, and product for denoising medical MRI images. Background Technology

[0002] Magnetic resonance imaging (MRI) plays a vital role in human medicine due to its non-ionizing, radiation-free nature and high contrast for soft tissue imaging. With technological advancements, multi-nuclear MRI has also developed accordingly. In multi-nuclear MRI, commonly used nuclides include… 23 Sodium (Na) is one of the most important electrolytes in human physiology. 23 Sodium (Na+) plays a crucial role in osmotic regulation and cell physiology. Transmembrane transport is maintained through the active transport of sodium-potassium pumps (Na+ / K+-ATPase). 23 The sodium concentration gradient differs by more than 10-fold between intracellular (10-15 mm) and extracellular (140-150 mm) concentrations. Under pathological conditions, impaired cellular metabolism or membrane integrity may disrupt sodium pump function, leading to tissue-specific sodium ionization. 23 Disruptions in sodium homeostasis and abnormal ion concentrations. These changes in tissue sodium concentration (TSC) reflect underlying metabolic dysfunction and have been observed in a variety of neurological diseases, including brain tumors, stroke, multiple sclerosis, and epilepsy. 23 Na magnetic resonance imaging (NMR) can non-invasively measure TSC, providing a valuable tool for studying these pathologies. However, compared with proton ( 1 Compared to H) imaging 23 Na MRI faces greater technical challenges because its inherently low signal-to-noise ratio (SNR) stems from a lower cyclotron ratio, lower in vivo concentration, and shorter transverse relaxation time.

[0003] To further improve the quality of magnetic resonance imaging, especially to improve 23 To improve the quality of Na NMR imaging, various post-processing and reconstruction strategies have been proposed. Broadly speaking, these denoising methods fall into two categories: deep learning-based methods and traditional non-learning techniques.

[0004] Deep learning models leverage the nonlinear mapping and automatic feature extraction capabilities of neural networks to discover latent structures in large datasets. For example, Adlung et al. used a u-net-based architecture to reconstruct highly undersampled data from patients with acute ischemic stroke. 23Na MRI data reduced acquisition time while improving signal-to-noise ratio (SNR) and TSC quantization accuracy. Similarly, Bakeret et al. trained a Convolutional Neural Network (CNN) on synthetically corrupted low SNR data generated by adding Gaussian noise to high-quality 1H k-space in the fastMRI dataset. However, these supervised networks require large amounts of paired, noise-free datasets and often struggle to generalize. While self-supervised denoising methods have been developed that train only on noisy images, they heavily rely on predefined noise models. In multi-channel MRI, phased array coils are commonly used to improve the inherently low SNR; however, inter-channel coupling introduces complex noise patterns, violating the noise assumptions of self-supervised denoising methods. To date, no such methods have been successfully applied to MRI denoising in complex scenarios, especially... 23 Na MRI noise reduction.

[0005] Traditional non-learning-based denoising techniques typically rely on manually crafted priors to formulate noise suppression algorithms. While they perform well in conventional MRI denoising, their denoising capabilities become insufficient when extended to multi-core MRI denoising scenarios. For example, Madelin et al. were among the first to apply compressed sensing (CS) to... 23 One research group working on sodium MRI halved the acquisition time while preserving reconstruction accuracy to some extent. However, this method often resulted in residual blurring and loss of detail, limiting its applicability for quantitative sodium imaging. Lachner et al. and Gnahm et al., through... 1 Incorporating anatomical priors into the CS framework for h MRI addresses this issue, improving fidelity but requiring additional multimodal data acquisition and image registration. Benkhedah et al. utilized an adaptive combined reconstruction method to mitigate correlated noise between coil channels, but this necessitates additional noisy scans and extended total scan time. More recently, Christensen et al. applied Global-Local Higher-Order Singular Value Decomposition (GL-HOSVD) and Tensor Marchenko-Pastur Principal Component Analysis (tMPPCA) to denoise various x-nucleus human data. While effective, both methods introduce subtle artifacts and varying degrees of image smoothing.

[0006] tMPPCA (Tensor Marchenko-Pastur Principal Component Analysis) is an improved denoising method based on High-Order Singular Value Decomposition (HOSVD). By introducing the Marchenko-Pastur distribution to automatically estimate the signal rank, it reduces the need for user-defined parameters and improves the objectivity and robustness of denoising. This method effectively utilizes data redundancy by recursively decomposing the tensor structure of multidimensional data, making it particularly suitable for small data blocks and high-dimensional data (such as multi-echo diffusion MRI). However, tMPPCA still has limitations in MRI denoising: although it can significantly reduce background noise, residual analysis shows that the signal distribution in the brain may be skewed (such as abnormally enhanced signal intensity in the ventricular region), and artifacts are introduced in some slices. In addition, for spatial variability noise caused by parallel imaging reconstruction in clinical data, tMPPCA needs to rely on additional noise level maps or local estimation methods; otherwise, it may not be able to fully adapt to complex noise distributions.

[0007] GL-HOSVD (Global-Local High-Order Singular Value Decomposition) is a hybrid denoising algorithm that combines global pre-filtering with local block processing. It uses pre-denoising in the global HOSVD stage to guide the local HOSVD stage, reducing fringe artifacts generated by local methods at low signal-to-noise ratios. This method performs well in diffusion MRI denoising, but its application in... 23 Na MRI denoising has significant limitations: experimental results show that the brain signal distribution may change after denoising (e.g., abnormal signals in the ventricular region), and brain contour deviations still exist in the residual images, indicating that the denoising process introduces systematic errors. Furthermore, GL-HOSVD relies on user-defined threshold parameters (such as k_global and k_local), which need to be adjusted empirically. In clinical data, noise characteristics may be complicated by preprocessing steps (such as parallel imaging or geometric correction), leading to limited parameter generalization. Meanwhile, in... 23 Artifacts were observed in Na MRI denoising, further limiting its reliability.

[0008] In summary, existing tMPPCA and GL-HOSVD techniques demonstrate excellent performance in conventional MRI denoising; however, they lag behind in multi-core MRI denoising (especially in...). 23 In MRI denoising, varying degrees of image blurring occur, resulting in limited denoising capability in areas with signal intensity, even though noise can be removed. Therefore, it is necessary to improve existing MRI denoising algorithms to enhance their ability to handle not only conventional MRI (e.g., MRI signal-sensitive areas) but also areas with signal intensity. 1 It can effectively denoise H MRI, and can also effectively denoise higher and more complex multi-nucleus MRI. Summary of the Invention

[0009] The purpose of this invention is to provide a medical MRI image denoising method, apparatus, system, computer equipment, computer-readable storage medium, and computer program product to solve the problems of blurred images after denoising and limited denoising ability for signal-containing areas when existing tMPPCA and GL-HOSVD technologies are applied to MRI denoising.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] Firstly, a method for denoising medical MRI images is provided, including:

[0012] Receives medical MRI image signals;

[0013] The medical MRI image signal is converted to the image domain to obtain three-dimensional image data, wherein the three-dimensional image data package contains R layers of two-dimensional images, each layer of two-dimensional images contains P×Q pixels, and R, P and Q represent positive integers respectively;

[0014] A three-dimensional image sub-block is selected from the three-dimensional image data, wherein the three-dimensional image sub-block contains R layers of two-dimensional sub-images, and each layer of two-dimensional sub-images contains... 1 pixel Represents a positive integer less than or equal to P. Represents a positive integer less than or equal to Q;

[0015] From the three-dimensional image data, search for M three-dimensional image similar sub-blocks that are most similar to the three-dimensional image sub-block within a predefined neighborhood of the three-dimensional image sub-block, where M represents a positive integer;

[0016] The three-dimensional image sub-blocks and the M similar three-dimensional image sub-blocks are respectively vectorized and straightened, and the resulting M+1 one-dimensional vectors are superimposed to obtain a result of size M. The original matrix Y;

[0017] Assuming the original matrix Y = X + N + S, the denoised matrix X is obtained by solving the following minimization objective function:

[0018]

[0019] In the formula, N represents Gaussian noise, S represents sparse noise, ||S||1 represents the L1 norm used to constrain the sparse noise S, and ||N|| F This represents the F-norm used to constrain Gaussian noise N. λ1, λ2, and λ3 represent the preset regularization coefficients, W represents the multi-channel weighting matrix in diagonal form, r represents a positive integer less than or equal to R, and δ represents the Schatten p-norm used to constrain the denoising matrix X. r Let I represent the noise standard deviation of the r-th channel, I represent the identity matrix, min() represent the minimum function, i represent a positive integer, and w represent the minimum value. i Indicates assigning σ i The weights, σ i represents the i-th singular value in the denoising matrix X, p represents the parameter value used to determine the norm type and takes a value in the interval (0,1], and w represents the weighting of the nuclear norm;

[0020] Based on the denoising matrix X, a new three-dimensional image sub-block corresponding to the three-dimensional image sub-block and M new three-dimensional image similar blocks corresponding one-to-one with the M three-dimensional image similar sub-blocks are reconstructed;

[0021] In the three-dimensional image data, the three-dimensional image sub-blocks are replaced with new three-dimensional image sub-blocks, and the M three-dimensional image similar sub-blocks are replaced one-to-one with the M new three-dimensional image similar sub-blocks to obtain new three-dimensional image data that has undergone image denoising.

[0022] Based on the above-mentioned invention, a novel MRI image denoising scheme based on multi-channel weighted robust principal component analysis is provided. First, the medical MRI image signal is converted to the image domain to obtain three-dimensional image data. Then, three-dimensional image sub-blocks are selected, and the M most similar three-dimensional image sub-blocks are searched within a predefined neighborhood. These sub-blocks are then vectorized, straightened, and superimposed to obtain the original matrix Y. Next, assuming the original matrix Y = X + N + S, the objective function is minimized to obtain the denoising matrix X. Finally, new sub-blocks corresponding to the aforementioned sub-blocks are reconstructed based on the denoising matrix X. The replacement of the old and new sub-blocks yields new denoised three-dimensional image data. This solves the problems of image blurring after denoising and limited denoising capability for signal-containing regions in existing tMPPCA and GL-HOSVD techniques applied to MRI denoising, improving the quality of medical MRI images, facilitating subsequent image analysis, and enabling practical application and promotion.

[0023] In one possible design, the medical MRI image signal is converted to the image domain to obtain three-dimensional image data, including:

[0024] When the medical MRI image signal is obtained by scanning the subject using a three-dimensional density-adapted radial sequence, the medical MRI image signal is converted to the image domain using a non-uniform fast Fourier transform to obtain the three-dimensional image data.

[0025] In one possible design, the M most similar 3D image sub-blocks located within a predefined neighborhood of the 3D image sub-block and most similar to the 3D image sub-block are searched from the 3D image data, including:

[0026] For each other three-dimensional image sub-block that is located within a predefined neighborhood of the three-dimensional image sub-block in the three-dimensional image data and has the same size as the three-dimensional image sub-block, the Eulerian distance between the three-dimensional image sub-block and the corresponding sub-block is calculated;

[0027] The other three-dimensional image sub-blocks are arranged in order of Eulerian distance from nearest to farthest to obtain a three-dimensional image sub-block sequence;

[0028] The first M three-dimensional image sub-blocks are selected from the sequence of three-dimensional image sub-blocks as the M most similar three-dimensional image sub-blocks to the three-dimensional image sub-blocks, where M represents a positive integer.

[0029] In one possible design, the three-dimensional image sub-blocks and the M similar three-dimensional image sub-blocks are vectorized and straightened respectively, and the resulting M+1 one-dimensional vectors are superimposed to obtain a result of size M. The original matrix Y includes:

[0030] For each sub-block in the three-dimensional image sub-block and the M similar three-dimensional image sub-blocks, the pixel value of the pixel located in the p′ row and q′ column in the corresponding r′-th layer two-dimensional sub-image is used as the h′-th element in the corresponding one-dimensional vector to obtain the corresponding one-dimensional vector, where r′ represents a positive integer less than or equal to R, and p′ represents a positive integer less than or equal to R. A positive integer, q′ represents less than or equal to positive integers,

[0031] The m-th one-dimensional vector among the M+1 one-dimensional vectors that correspond one-to-one with the three-dimensional image sub-block and the M similar three-dimensional image sub-blocks will be used as the vector of size M. The original matrix Y is obtained by taking the element in the m-th row of the original matrix Y, where m represents a positive integer less than or equal to M+1;

[0032] Based on the denoising matrix X, a new 3D image sub-block corresponding to the 3D image sub-block and M new 3D image similar blocks corresponding one-to-one with the M 3D image similar sub-blocks are reconstructed, including:

[0033] If the element in the m-th row of the original matrix Y is a one-dimensional vector of the three-dimensional image sub-block, then the element in the m-th row and h′-th column of the denoising matrix X is taken as the pixel value of the pixel in the r′-th layer two-dimensional sub-image in the new three-dimensional image sub-block corresponding to the three-dimensional image sub-block;

[0034] If the element in the m-th row of the original matrix Y is a one-dimensional vector of the m′-th three-dimensional image similarity sub-block among the M three-dimensional image similarity sub-blocks, then the element located in the m-th row and h′-th column of the denoising matrix X is taken as the pixel value of the pixel point located in the p′-th row and q′-th column of the two-dimensional sub-image in the r′-th layer of the new three-dimensional image similarity sub-block corresponding to the m′-th three-dimensional image similarity sub-block.

[0035] In one possible design, the method further includes:

[0036] Different 3D image sub-blocks in the 3D image data are selected by traversal and / or looping. After each selection of a 3D image sub-block, the 3D image data is iteratively updated by sequentially performing a similar sub-block search step, a one-dimensional vector superposition step, a noise reduction matrix solution step, a new sub-block reconstruction step, and an image sub-block replacement step until a preset convergence condition is met.

[0037] In a second aspect, a medical MRI image denoising device is provided, comprising an image signal receiving unit, an image data conversion unit, an image sub-block selection unit, a similar sub-block search unit, a straightening and overlay processing unit, a denoising matrix solving unit, a new sub-block reconstruction unit, and an image data updating unit that are sequentially connected in communication.

[0038] The image signal receiving unit is used to receive medical MRI image signals;

[0039] The image data conversion unit is used to convert the medical MRI image signal into the image domain to obtain three-dimensional image data. The three-dimensional image data package contains R layers of two-dimensional images, and each layer of two-dimensional images contains P×Q pixels, where R, P and Q represent positive integers.

[0040] The image sub-block selection unit is used to select a three-dimensional image sub-block from the three-dimensional image data, wherein the three-dimensional image sub-block contains R layers of two-dimensional sub-images, and each layer of two-dimensional sub-images contains... 1 pixel Represents a positive integer less than or equal to P. Represents a positive integer less than or equal to Q;

[0041] The similar sub-block search unit is used to search for M three-dimensional image similar sub-blocks that are most similar to the three-dimensional image sub-block within a predefined neighborhood of the three-dimensional image sub-block from the three-dimensional image data, where M represents a positive integer;

[0042] The straightening and overlay processing unit is used to perform vectorization straightening processing on the three-dimensional image sub-blocks and the M similar three-dimensional image sub-blocks respectively, and to overlay the resulting M+1 one-dimensional vectors to obtain a vector of size M. The original matrix Y;

[0043] The noise reduction matrix solving unit is used to assume the original matrix Y = X + N + S and solve for the following minimized objective function to obtain the noise reduction matrix X:

[0044]

[0045] In the formula, N represents Gaussian noise, S represents sparse noise, ||S||1 represents the L1 norm used to constrain the sparse noise S, and ||N|| F This represents the F-norm used to constrain Gaussian noise N. λ1, λ2, and λ3 represent the preset regularization coefficients, W represents the multi-channel weighting matrix in diagonal form, r represents a positive integer less than or equal to R, and δ represents the Schatten p-norm used to constrain the denoising matrix X. r Let I represent the noise standard deviation of the r-th channel, I represent the identity matrix, min() represent the minimum function, i represent a positive integer, and w represent the minimum value. i Indicates assigning σ i The weights, σ i represents the i-th singular value in the denoising matrix X, p represents the parameter value used to determine the norm type and takes a value in the interval (0,1], and w represents the weighting of the nuclear norm;

[0046] The new sub-block reconstruction unit is used to reconstruct a new three-dimensional image sub-block corresponding to the three-dimensional image sub-block and M new three-dimensional image similar blocks corresponding one-to-one with the M three-dimensional image similar sub-blocks according to the noise reduction matrix X.

[0047] The image data update unit is used to replace the three-dimensional image sub-blocks with new three-dimensional image sub-blocks in the three-dimensional image data, and to replace the M three-dimensional image similar sub-blocks one by one with the M new three-dimensional image similar sub-blocks, so as to obtain new three-dimensional image data that has been denoised.

[0048] Thirdly, the present invention provides a medical MRI image denoising system, including a magnetic resonance imaging instrument and a host computer connected in communication, wherein the magnetic resonance imaging instrument includes a scanning module and a magnetic resonance receiving coil.

[0049] The scanning module is used to scan the subject by transmitting a sequence suitable for MRI imaging through a radio frequency transmitting coil, so as to excite the hydrogen atoms in the subject to produce a resonance phenomenon.

[0050] The magnetic resonance receiving coil is used to receive and transmit medical MRI image signals to the host computer during the scanning of the subject.

[0051] The host computer is used to execute the medical MRI image denoising method as described in the first aspect or any possible design in the first aspect.

[0052] Fourthly, the present invention provides a computer device comprising a storage module, a processing module, and a transceiver module connected in sequence for communication, wherein the storage module is used to store a computer program, the transceiver module is used to send and receive messages, and the processing module is used to read the computer program and execute the medical MRI image denoising method as described in the first aspect or any possible design in the first aspect.

[0053] Fifthly, the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, perform the medical MRI image denoising method as described in the first aspect or any possible design of the first aspect.

[0054] In a sixth aspect, the present invention provides a computer program product, including a computer program or instructions, which, when executed by a computer, implement the medical MRI image denoising method as described in the first aspect or any possible design in the first aspect.

[0055] The beneficial effects of the above scheme are:

[0056] (1) This invention creatively provides a new MRI image denoising scheme based on multi-channel weighted robust principal component analysis. First, the medical MRI image signal is converted to the image domain to obtain three-dimensional image data. Then, three-dimensional image sub-blocks are selected and the most similar three-dimensional image similar sub-blocks are searched in a predefined neighborhood. Then, these sub-blocks are vectorized and straightened and superimposed to obtain the original matrix Y. Then, assuming the original matrix Y = X + N + S, the objective function is minimized to obtain the denoising matrix X. Finally, a new sub-block corresponding to the aforementioned sub-block is reconstructed based on the denoising matrix X. The new three-dimensional image data that has been denoised is obtained by replacing the old and new sub-blocks. This can solve the problems of image blurring after denoising and limited denoising ability for signal areas in the application of existing tMPPCA and GL-HOSVD technologies to MRI denoising, improve the quality of medical MRI images, and facilitate subsequent image analysis.

[0057] (2) This scheme extends the Schatten p-norm to a multi-channel weighted form: on the one hand, by assigning different importance to the singular value components, a more accurate low-rank approximation can be achieved; on the other hand, the channel-specific weighting matrix adaptively balances the noise suppression of each coil, and by forcibly enforcing low-rank constraints on all channels, it effectively balances the heterogeneity between channels and the global low-rank structure.

[0058] (3) This scheme also extends robust PCA by combining GRPCA with regularized noise modeling, which can separate low-rank components from noise, thereby improving the fidelity of anatomical structure recovery under complex noise conditions and facilitating practical application and promotion. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a flowchart illustrating the medical MRI image denoising method provided in an embodiment of this application.

[0061] Figure 2 The figure shows a comparison of the denoising effect of the medical MRI image denoising method provided in the embodiments of this application with the tMMPCA method and the GL-HOSVD method.

[0062] Figure 3 This is a schematic diagram of the structure of the medical MRI image denoising device provided in the embodiments of this application.

[0063] Figure 4 This is a schematic diagram of the structure of a medical MRI image denoising system provided in an embodiment of this application.

[0064] Figure 5 A schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0065] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these embodiments without creative effort. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0066] It should be understood that although the terms "first" and "second", etc., may be used herein to describe various objects, these objects should not be limited by these terms. These terms are only used to distinguish one object from another. For example, the first object may be referred to as the second object, and similarly, the second object may be referred to as the first object, without departing from the scope of the exemplary embodiments of the invention.

[0067] It should be understood that the term "and / or" that may appear in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, or A and B exist simultaneously. Another example is A, B and / or C, which can mean that any one of A, B, and C or any combination thereof exists. The term " / and" that may appear in this document describes another relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone or A and B exist simultaneously. In addition, the character " / " that may appear in this document generally indicates that the related objects before and after it are in an "or" relationship.

[0068] Example

[0069] like Figure 1 As shown, the medical MRI image denoising method provided in the first aspect of this embodiment can be executed, but is not limited to, by a computer device with certain computing resources and a communication connection to a magnetic resonance imaging (MRI) scanner. The MRI scanner includes, but is not limited to, a scanning module and a magnetic resonance receiving coil. The scanning module is used to scan a subject (e.g., a patient) by transmitting a sequence suitable for MRI imaging through a radio frequency transmitting coil, so as to excite hydrogen atoms in the subject to resonate. The magnetic resonance receiving coil is used to receive and transmit medical MRI image signals to the computer device during the scanning of the subject. The medical MRI image denoising method includes, but is not limited to, the following steps S1 to S8.

[0070] S1. Receives medical MRI image signals.

[0071] In step S1, the medical MRI image signal originates from the magnetic resonance receiving coil and can be conventionally received using existing wired communication technology. Furthermore, the specific examples of the medical MRI image signal, but not limited to medical... 23 The signal in the Na MRI image.

[0072] S2. The medical MRI image signal is converted to the image domain to obtain three-dimensional image data, wherein the three-dimensional image data package contains R layers of two-dimensional images, each layer of two-dimensional images contains P×Q pixels, and R, P and Q represent positive integers respectively.

[0073] In step S2, the size of the three-dimensional image data can be, for example, 50×80×80, meaning it contains 50 layers of two-dimensional images, each containing 80×80 pixels. Specifically, converting the medical MRI image signal to the image domain to obtain the three-dimensional image data includes, but is not limited to: when the medical MRI image signal is obtained by scanning the subject using a three-dimensional density-adapted radial sequence, using a non-uniform fast Fourier transform to convert the medical MRI image signal to the image domain to obtain the three-dimensional image data. The three-dimensional density-adapted radial sequence (DA-3DPR) is an existing scanning sequence, and the non-uniform fast Fourier transform (NUFFT) is also an existing technique, and will not be elaborated further here.

[0074] S3. Select a three-dimensional image sub-block from the three-dimensional image data, wherein the three-dimensional image sub-block contains R layers of two-dimensional sub-images, and each layer of two-dimensional sub-images contains... 1 pixel Represents a positive integer less than or equal to P. It represents a positive integer less than or equal to Q.

[0075] In step S3, the three-dimensional image sub-block is used as a local reference patch. Based on the example in step S2 above, the size of the three-dimensional image sub-block can be 50×10×10, that is, it contains 50 layers of two-dimensional sub-images, each layer of two-dimensional sub-images containing 10×10 pixels. The specific selection method of the three-dimensional image sub-block can be manual selection, random selection, or traversal selection using a sliding window method (for example, first select a two-dimensional sub-image with a size of 10×10 pixels in the upper left corner of the two-dimensional image, then the initial size of the three-dimensional image sub-block is 50×10×10, and then slide the window to the right by one or more pixels on the two-dimensional image to continue selecting the next three-dimensional image sub-block with a size of 50×10×10).

[0076] S4. Search the three-dimensional image data for M three-dimensional image similar sub-blocks that are most similar to the three-dimensional image sub-block within a predefined neighborhood of the three-dimensional image sub-block, where M represents a positive integer.

[0077] In step S4, based on the example in step S3 above, the size of the three-dimensional image similar sub-block is also 50×10×10. The size of the predefined neighborhood can be specified by the user or conventionally determined based on the results of multiple limited experiments. Specifically, searching for M three-dimensional image similar sub-blocks that are most similar to the three-dimensional image sub-block within the predefined neighborhood of the three-dimensional image sub-block from the three-dimensional image data includes, but is not limited to, the following steps S41 to S43.

[0078] S41. For each other three-dimensional image sub-block that is located within a predefined neighborhood of the three-dimensional image sub-block in the three-dimensional image data and has the same size as the three-dimensional image sub-block, calculate the Eulerian distance between the three-dimensional image sub-block and the corresponding sub-block.

[0079] In step S41, since the three-dimensional image sub-block and the other three-dimensional image sub-blocks can all be regarded as three-dimensional vectors, the Euler distance between them can be calculated based on the existing Euler distance formula.

[0080] S42. Arrange the other three-dimensional image sub-blocks in order of Eulerian distance from nearest to farthest to obtain a three-dimensional image sub-block sequence.

[0081] S43. Select the first M three-dimensional image sub-blocks from the three-dimensional image sub-block sequence as the M three-dimensional image similar sub-blocks most similar to the three-dimensional image sub-blocks, where M represents a positive integer.

[0082] S5. Perform vectorization and straightening processing on the three-dimensional image sub-blocks and the M similar three-dimensional image sub-blocks respectively, and superimpose the resulting M+1 one-dimensional vectors to obtain a result of size... The original matrix Y.

[0083] In step S5, the vectorization straightening process refers to converting the current 3D vector of the corresponding sub-block into a 1D vector. Specifically, the 3D image sub-block and the M similar 3D image sub-blocks are vectorized and straightened respectively, and the resulting M+1 1-dimensional vectors are superimposed to obtain a vector of size... The original matrix Y includes, but is not limited to, the following steps S51 to S52.

[0084] S51. For each sub-block in the three-dimensional image sub-block and the M similar three-dimensional image sub-blocks, the pixel value of the pixel located in the p′ row and q′ column in the corresponding r′ layer two-dimensional sub-image is taken as the h′ element in the corresponding one-dimensional vector to obtain the corresponding one-dimensional vector, where r′ represents a positive integer less than or equal to R, and p′ represents a positive integer less than or equal to R. A positive integer, q′ represents less than or equal to positive integers,

[0085] S52. Take the m-th one-dimensional vector from the M+1 one-dimensional vectors that correspond one-to-one with the three-dimensional image sub-block and the M similar three-dimensional image sub-blocks, and use it as a vector of size... The original matrix Y is obtained by taking the element in the m-th row of the original matrix Y, where m represents a positive integer less than or equal to M+1.

[0086] S6. Assume the original matrix Y = X + N + S, and solve the following minimization objective function to obtain the denoised matrix X:

[0087]

[0088] In the formula, N represents Gaussian noise, S represents sparse noise, ||S||1 represents the L1 norm used to constrain the sparse noise S, and ||N|| F This represents the F-norm used to constrain Gaussian noise N. λ1, λ2, and λ3 represent the preset regularization coefficients, W represents the multi-channel weighting matrix in diagonal form, r represents a positive integer less than or equal to R, and δ represents the Schatten p-norm used to constrain the denoising matrix X. r Let I represent the noise standard deviation of the r-th channel, I represent the identity matrix, min() represent the minimum function, i represent a positive integer, and w represent the minimum value. i Indicates assigning σ i The weights, σ i Let represent the i-th singular value in the denoising matrix X, p represent the parameter value used to determine the norm type and take values ​​in the interval (0,1], and w represent the weighted nuclear norm.

[0089] In step S6, the denoising matrix X represents the potential clean image (i.e., a clean, noise-free image that meets the requirements). The L1 norm, F norm, and Schatten p-norm are all existing norms. Furthermore, the specific solution process for minimizing the objective function can be implemented, but is not limited to, using the existing Alternating Direction Method of Multipliers (ADMM) algorithm.

[0090] S7. Based on the denoising matrix X, reconstruct a new three-dimensional image sub-block corresponding to the three-dimensional image sub-block and M new three-dimensional image similar sub-blocks corresponding one-to-one with the M three-dimensional image similar sub-blocks.

[0091] In step S7, the reconstruction process of the new 3D image sub-block and the M new 3D image similar sub-blocks is the reverse process of the aforementioned step S5. Specifically, based on the denoising matrix X, a new 3D image sub-block corresponding to the 3D image sub-block and M new 3D image similar sub-blocks corresponding one-to-one with the M 3D image similar sub-blocks are reconstructed, including but not limited to: if the element in the m-th row of the original matrix Y is a one-dimensional vector of the 3D image sub-block, then the element located in the m-th row and h′-th column of the denoising matrix X is taken as... The pixel value of the pixel located in the p′ row and q′ column of the two-dimensional sub-image at the r′ layer in the new three-dimensional image sub-block corresponding to the three-dimensional image sub-block; if the element in the m-th row of the original matrix Y is a one-dimensional vector of the m′-th three-dimensional image similar sub-block among the M three-dimensional image similar sub-blocks, then the element located in the m-th row and h′ column of the denoising matrix X is taken as the pixel value of the pixel located in the p′ row and q′ column of the two-dimensional sub-image at the r′ layer in the new three-dimensional image similar sub-block corresponding to the m′-th three-dimensional image similar sub-block.

[0092] S8. In the three-dimensional image data, the three-dimensional image sub-blocks are replaced with new three-dimensional image sub-blocks, and the M three-dimensional image similar sub-blocks are replaced one-to-one with the M new three-dimensional image similar sub-blocks to obtain new three-dimensional image data that has been denoised.

[0093] Based on the above steps S1 to S8, this embodiment also performed the following tests: using medical... 23 Na MRI images were obtained, and the noise from the actual acquisitions was added to the simulation data to obtain, as shown below. Figure 2 The noise reduction effect comparison chart and the technical indicator comparison table shown in Table 1 below are as follows:

[0094] Table 1. Comparison of denoising performance of this embodiment with tMMPCA and GL-HOSVD methods.

[0095]

[0096] Therefore, it can be seen that the denoising effect of the method described in this embodiment is better than that of the tMMPCA and GL-HOSVD methods. This is evident from the PSNR (Peak Signal-to-Noise Ratio), SSIM (Structural Similarity Index), and MSE (Mean Squared Error) indices, as well as the difference plot. The denoising quality of the method described in this embodiment is better and closer to the real data, demonstrating its effectiveness in removing medical data. 23 Noise in NaMRI images.

[0097] Therefore, based on the medical MRI image denoising method described in steps S1 to S8 above, a new MRI image denoising scheme based on multi-channel weighted robust principal component analysis is provided. First, the medical MRI image signal is converted to the image domain to obtain three-dimensional image data. Then, three-dimensional image sub-blocks are selected, and the M most similar three-dimensional image sub-blocks are searched within a predefined neighborhood. These sub-blocks are then vectorized, straightened, and superimposed to obtain the original matrix Y. Next, assuming the original matrix Y = X + N + S, the objective function is minimized to obtain the denoising matrix X. Finally, new sub-blocks corresponding to the aforementioned sub-blocks are reconstructed based on the denoising matrix X. The replacement of the old and new sub-blocks yields new denoised three-dimensional image data. This solves the problems of image blurring after denoising and limited denoising capability for signal-containing regions in existing tMPPCA and GL-HOSVD techniques applied to MRI denoising, improving the quality of medical MRI images. 23 The results were validated on NaMRI images, which facilitates subsequent image analysis and promotes practical application and widespread adoption.

[0098] Based on the aforementioned first aspect of the technical solution, this embodiment also provides a possible design for iteratively performing medical MRI image denoising. That is, the method further includes, but is not limited to: selecting different three-dimensional image sub-blocks in the three-dimensional image data in a traversal manner and / or a loop manner, and after each selection of a three-dimensional image sub-block, iteratively updating the three-dimensional image data in sequence through the similar sub-block search step (i.e., step S4), the one-dimensional vector superposition step (i.e., step S5), the denoising matrix solution step (i.e., step S6), the new sub-block reconstruction step (i.e., step S7), and the image sub-block replacement step (i.e., step S8) in the medical MRI image denoising method described in the first aspect, until a preset convergence condition is met. The aforementioned convergence conditions have the following two cases: (1) Set the maximum number of denoising loops to K in advance (which needs to be set according to experience based on different data and noise intensity), traverse the entire image to select reference patch blocks (i.e., the three-dimensional image sub-blocks), and then select similar blocks in a predefined neighborhood for each reference patch block to denoise; denoising all similar blocks once completes one denoising loop; if the current number of denoising loops exceeds K, stop; (2) When ||X| is satisfied at the same time k -Z k || F ≤Tol、||X k+1 -X k || F ≤Tol and ||Z k+1 -Z k || F When ≤Tol>0, convergence is confirmed, where Tol represents a preset and small common difference, and Tol>0, k represents a positive integer, and Xk This represents the denoised matrix X, Z obtained by the k-th alternation when solving using the alternating direction multiplier algorithm. k This represents the auxiliary variable matrix for the k-th alternation when solving using the alternating direction multiplier algorithm, || || F This represents the F-norm.

[0099] like Figure 3 As shown, the second aspect of this embodiment provides a virtual device for implementing the medical MRI image denoising method described in the first aspect or possibly the first design, including an image signal receiving unit, an image data conversion unit, an image sub-block selection unit, a similar sub-block search unit, a straightening and overlay processing unit, a denoising matrix solving unit, a new sub-block reconstruction unit, and an image data updating unit that are sequentially connected in communication.

[0100] The image signal receiving unit is used to receive medical MRI image signals;

[0101] The image data conversion unit is used to convert the medical MRI image signal into the image domain to obtain three-dimensional image data. The three-dimensional image data package contains R layers of two-dimensional images, and each layer of two-dimensional images contains P×Q pixels, where R, P and Q represent positive integers.

[0102] The image sub-block selection unit is used to select a three-dimensional image sub-block from the three-dimensional image data, wherein the three-dimensional image sub-block contains R layers of two-dimensional sub-images, and each layer of two-dimensional sub-images contains... 1 pixel Represents a positive integer less than or equal to P. Represents a positive integer less than or equal to Q;

[0103] The similar sub-block search unit is used to search for M three-dimensional image similar sub-blocks that are most similar to the three-dimensional image sub-block within a predefined neighborhood of the three-dimensional image sub-block from the three-dimensional image data, where M represents a positive integer;

[0104] The straightening and overlay processing unit is used to perform vectorization straightening processing on the three-dimensional image sub-blocks and the M similar three-dimensional image sub-blocks respectively, and to overlay the resulting M+1 one-dimensional vectors to obtain a vector of size M. The original matrix Y;

[0105] The noise reduction matrix solving unit is used to assume the original matrix Y = X + N + S and solve for the following minimized objective function to obtain the noise reduction matrix X:

[0106]

[0107] In the formula, N represents Gaussian noise, S represents sparse noise, ||S||1 represents the L1 norm used to constrain the sparse noise S, and ||N|| F This represents the F-norm used to constrain Gaussian noise N. λ1, λ2, and λ3 represent the preset regularization coefficients, W represents the multi-channel weighting matrix in diagonal form, r represents a positive integer less than or equal to R, and δ represents the Schatten p-norm used to constrain the denoising matrix X. r Let I represent the noise standard deviation of the r-th channel, I represent the identity matrix, min() represent the minimum function, i represent a positive integer, and w represent the minimum value. i Indicates assigning σ i The weights, σ i represents the i-th singular value in the denoising matrix X, p represents the parameter value used to determine the norm type and takes a value in the interval (0,1], and w represents the weighting of the nuclear norm;

[0108] The new sub-block reconstruction unit is used to reconstruct a new three-dimensional image sub-block corresponding to the three-dimensional image sub-block and M new three-dimensional image similar blocks corresponding one-to-one with the M three-dimensional image similar sub-blocks according to the noise reduction matrix X.

[0109] The image data update unit is used to replace the three-dimensional image sub-blocks with new three-dimensional image sub-blocks in the three-dimensional image data, and to replace the M three-dimensional image similar sub-blocks one by one with the M new three-dimensional image similar sub-blocks, so as to obtain new three-dimensional image data that has been denoised.

[0110] The working process, working details and technical effects of the aforementioned device provided in the second aspect of this embodiment can be found in the medical MRI image denoising method described in the first aspect or possible design, and will not be repeated here.

[0111] like Figure 4 As shown, the third aspect of this embodiment provides a physical system for implementing the medical MRI image denoising method described in the first aspect or possibly the design, including a magnetic resonance imaging instrument and a host computer that are communicatively connected, wherein the magnetic resonance imaging instrument includes a scanning module and a magnetic resonance receiving coil.

[0112] The scanning module is used to scan the subject by transmitting a sequence suitable for MRI imaging through a radio frequency transmitting coil, so as to excite the hydrogen atoms in the subject to produce a resonance phenomenon.

[0113] The magnetic resonance receiving coil is used to receive and transmit medical MRI image signals to the host computer during the scanning of the subject.

[0114] The host computer is used to execute the medical MRI image denoising method as described in the first aspect or possibly Design 1.

[0115] The working process, working details and technical effects of the aforementioned system provided in the third aspect of this embodiment can be found in the medical MRI image denoising method described in the first aspect or possible design, and will not be repeated here.

[0116] like Figure 5 As shown, the fourth aspect of this embodiment provides a computer device for performing the medical MRI image denoising method as described in the first aspect or a possible design, including a storage module, a processing module, and a transceiver module connected in sequence. The storage module stores a computer program, the transceiver module sends and receives messages, and the processing module reads the computer program and performs the medical MRI image denoising method as described in the first aspect or a possible design. Specifically, the storage module may include, but is not limited to, random-access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out (FIFO) memory, and / or first-in-last-out (FILO) memory, etc.; the processing module may, but is not limited to, use a microprocessor of the STM32F105 series. Furthermore, the computer device may also include, but is not limited to, a power supply module, a display screen, and other necessary components.

[0117] The working process, working details and technical effects of the aforementioned computer device provided in the fourth aspect of this embodiment can be found in the medical MRI image denoising method described in the first aspect or possible design, and will not be repeated here.

[0118] This fifth aspect of the embodiment provides a computer-readable storage medium storing instructions comprising a medical MRI image denoising method as described in the first aspect or possible design one. Specifically, the computer-readable storage medium stores instructions that, when executed on a computer, perform the medical MRI image denoising method as described in the first aspect or possible design one. The computer-readable storage medium refers to a data storage medium, which may include, but is not limited to, floppy disks, optical disks, hard disks, flash memory, USB flash drives, and / or Memory Sticks. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.

[0119] The working process, working details and technical effects of the aforementioned computer-readable storage medium provided in the fifth aspect of this embodiment can be found in the medical MRI image denoising method described in the first aspect or possible design, and will not be repeated here.

[0120] The sixth aspect of this embodiment provides a computer program product, including a computer program or instructions, which, when executed by a computer, implement the medical MRI image denoising method as described in the first aspect or possible design. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.

[0121] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for denoising medical MRI images, characterized in that, include: Receives medical MRI image signals; The medical MRI image signal is converted to the image domain to obtain three-dimensional image data, wherein the three-dimensional image data package contains R layers of two-dimensional images, each layer of two-dimensional images contains P×Q pixels, and R, P and Q represent positive integers respectively; A three-dimensional image sub-block is selected from the three-dimensional image data, wherein the three-dimensional image sub-block contains R layers of two-dimensional sub-images, and each layer of two-dimensional sub-images contains... 1 pixel Represents a positive integer less than or equal to P. Represents a positive integer less than or equal to Q; From the three-dimensional image data, search for M three-dimensional image similar sub-blocks that are most similar to the three-dimensional image sub-block within a predefined neighborhood of the three-dimensional image sub-block, where M represents a positive integer; The three-dimensional image sub-blocks and the M similar three-dimensional image sub-blocks are respectively vectorized and straightened, and the resulting M+1 one-dimensional vectors are superimposed to obtain a result of size M. The original matrix Y; Assuming the original matrix Y = X + N + S, the denoised matrix X is obtained by solving the following minimization objective function: In the formula, N represents Gaussian noise, S represents sparse noise, ||S||1 represents the L1 norm used to constrain the sparse noise S, and ||N|| F This represents the F-norm used to constrain Gaussian noise N. λ1, λ2, and λ3 represent the preset regularization coefficients, W represents the multi-channel weighting matrix in diagonal form, r represents a positive integer less than or equal to R, and δ represents the Schatten p-norm used to constrain the denoising matrix X. r Let I represent the noise standard deviation of the r-th channel, I represent the identity matrix, min() represent the minimum function, i represent a positive integer, and w represent the minimum value. i Indicates assigning σ i The weights, σ i represents the i-th singular value in the denoising matrix X, p represents the parameter value used to determine the norm type and takes a value in the interval (0,1], and w represents the weighting of the nuclear norm; Based on the denoising matrix X, a new three-dimensional image sub-block corresponding to the three-dimensional image sub-block and M new three-dimensional image similar blocks corresponding one-to-one with the M three-dimensional image similar sub-blocks are reconstructed; In the three-dimensional image data, the three-dimensional image sub-blocks are replaced with new three-dimensional image sub-blocks, and the M three-dimensional image similar sub-blocks are replaced one-to-one with the M new three-dimensional image similar sub-blocks to obtain new three-dimensional image data that has undergone image denoising.

2. The medical MRI image denoising method according to claim 1, characterized in that, The medical MRI image signal is converted to the image domain to obtain three-dimensional image data, including: When the medical MRI image signal is obtained by scanning the subject using a three-dimensional density-adapted radial sequence, the medical MRI image signal is converted to the image domain using a non-uniform fast Fourier transform to obtain the three-dimensional image data.

3. The medical MRI image denoising method according to claim 1, characterized in that, From the 3D image data, search for M most similar 3D image sub-blocks located within a predefined neighborhood of the 3D image sub-block and including: For each other three-dimensional image sub-block that is located within a predefined neighborhood of the three-dimensional image sub-block in the three-dimensional image data and has the same size as the three-dimensional image sub-block, the Eulerian distance between the three-dimensional image sub-block and the corresponding sub-block is calculated; The other three-dimensional image sub-blocks are arranged in order of Eulerian distance from nearest to farthest to obtain a three-dimensional image sub-block sequence; The first M three-dimensional image sub-blocks are selected from the sequence of three-dimensional image sub-blocks as the M most similar three-dimensional image sub-blocks to the three-dimensional image sub-blocks, where M represents a positive integer.

4. The medical MRI image denoising method according to claim 1, characterized in that, The three-dimensional image sub-blocks and the M similar three-dimensional image sub-blocks are respectively vectorized and straightened, and the resulting M+1 one-dimensional vectors are superimposed to obtain a result of size M. The original matrix Y includes: For each sub-block in the three-dimensional image sub-block and the M similar three-dimensional image sub-blocks, the pixel value of the pixel located in the p′ row and q′ column in the corresponding r′-th layer two-dimensional sub-image is used as the h′-th element in the corresponding one-dimensional vector to obtain the corresponding one-dimensional vector, where r′ represents a positive integer less than or equal to R, and p′ represents a positive integer less than or equal to R. A positive integer, q′ represents less than or equal to positive integers, The m-th one-dimensional vector among the M+1 one-dimensional vectors that correspond one-to-one with the three-dimensional image sub-block and the M similar three-dimensional image sub-blocks will be used as the vector of size M. The original matrix Y is obtained by taking the element in the m-th row of the original matrix Y, where m represents a positive integer less than or equal to M+1; Based on the denoising matrix X, a new 3D image sub-block corresponding to the 3D image sub-block and M new 3D image similar blocks corresponding one-to-one with the M 3D image similar sub-blocks are reconstructed, including: If the element in the m-th row of the original matrix Y is a one-dimensional vector of the three-dimensional image sub-block, then the element in the m-th row and h′-th column of the denoising matrix X is taken as the pixel value of the pixel in the r′-th layer two-dimensional sub-image in the new three-dimensional image sub-block corresponding to the three-dimensional image sub-block; If the element in the m-th row of the original matrix Y is a one-dimensional vector of the m′-th three-dimensional image similarity sub-block among the M three-dimensional image similarity sub-blocks, then the element located in the m-th row and h′-th column of the denoising matrix X is taken as the pixel value of the pixel point located in the p′-th row and q′-th column of the two-dimensional sub-image in the r′-th layer of the new three-dimensional image similarity sub-block corresponding to the m′-th three-dimensional image similarity sub-block.

5. The medical MRI image denoising method according to claim 1, characterized in that, The method further includes: Different three-dimensional image sub-blocks in the three-dimensional image data are selected by traversal and / or looping. After each selection of a three-dimensional image sub-block, the three-dimensional image data is iteratively updated by sequentially performing the similar sub-block search step, the one-dimensional vector superposition step, the noise reduction matrix solution step, the new sub-block reconstruction step, and the image sub-block replacement step in the medical MRI image denoising method as described in claim 1, until the preset convergence condition is met.

6. A medical MRI image denoising device, characterized in that, It includes an image signal receiving unit, an image data conversion unit, an image sub-block selection unit, a similar sub-block search unit, a straightening and overlay processing unit, a noise reduction matrix solving unit, a new sub-block reconstruction unit, and an image data updating unit, which are connected in sequence. The image signal receiving unit is used to receive medical MRI image signals; The image data conversion unit is used to convert the medical MRI image signal into the image domain to obtain three-dimensional image data. The three-dimensional image data package contains R layers of two-dimensional images, and each layer of two-dimensional images contains P×Q pixels, where R, P and Q represent positive integers. The image sub-block selection unit is used to select a three-dimensional image sub-block from the three-dimensional image data, wherein the three-dimensional image sub-block contains R layers of two-dimensional sub-images, and each layer of two-dimensional sub-images contains... 1 pixel Represents a positive integer less than or equal to P. Represents a positive integer less than or equal to Q; The similar sub-block search unit is used to search for M three-dimensional image similar sub-blocks that are most similar to the three-dimensional image sub-block within a predefined neighborhood of the three-dimensional image sub-block from the three-dimensional image data, where M represents a positive integer; The straightening and overlay processing unit is used to perform vectorization straightening processing on the three-dimensional image sub-blocks and the M similar three-dimensional image sub-blocks respectively, and to overlay the resulting M+1 one-dimensional vectors to obtain a vector of size M. The original matrix Y; The noise reduction matrix solving unit is used to assume the original matrix Y = X + N + S and solve for the following minimized objective function to obtain the noise reduction matrix X: In the formula, N represents Gaussian noise, S represents sparse noise, ||S||1 represents the L1 norm used to constrain the sparse noise S, and ||N|| F This represents the F-norm used to constrain Gaussian noise N. λ1, λ2, and λ3 represent the preset regularization coefficients, W represents the multi-channel weighting matrix in diagonal form, r represents a positive integer less than or equal to R, and δ represents the Schatten p-norm used to constrain the denoising matrix X. r Let I represent the noise standard deviation of the r-th channel, I represent the identity matrix, min() represent the minimum function, i represent a positive integer, and w represent the minimum value. i Indicates assigning σ i The weights, σ i represents the i-th singular value in the denoising matrix X, p represents the parameter value used to determine the norm type and takes a value in the interval (0,1], and w represents the weighting of the nuclear norm; The new sub-block reconstruction unit is used to reconstruct a new three-dimensional image sub-block corresponding to the three-dimensional image sub-block and M new three-dimensional image similar blocks corresponding one-to-one with the M three-dimensional image similar sub-blocks according to the noise reduction matrix X. The image data update unit is used to replace the three-dimensional image sub-blocks with new three-dimensional image sub-blocks in the three-dimensional image data, and to replace the M three-dimensional image similar sub-blocks one by one with the M new three-dimensional image similar sub-blocks, so as to obtain new three-dimensional image data that has been denoised.

7. A medical MRI image denoising system, characterized in that, It includes a magnetic resonance imaging device and a host computer that are connected by communication, wherein the magnetic resonance imaging device includes a scanning module and a magnetic resonance receiving coil; The scanning module is used to scan the subject by transmitting a sequence suitable for MRI imaging through a radio frequency transmitting coil, so as to excite the hydrogen atoms in the subject to produce a resonance phenomenon. The magnetic resonance receiving coil is used to receive and transmit medical MRI image signals to the host computer during the scanning of the subject. The host computer is used to execute the medical MRI image denoising method as described in any one of claims 1 to 5.

8. A computer device, characterized in that, The device includes a storage module, a processing module, and a transceiver module that are sequentially connected in communication. The storage module is used to store a computer program, the transceiver module is used to send and receive messages, and the processing module is used to read the computer program and execute the medical MRI image denoising method as described in any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that... The computer-readable storage medium stores instructions that, when executed on a computer, perform the medical MRI image denoising method as described in any one of claims 1 to 5.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or the instructions are executed by the computer, they implement the medical MRI image denoising method as described in any one of claims 1 to 5.

Images