Fast magnetic resonance reconstruction method and system based on multistage wavelet convolutional network

Abstract

The invention provides a fast magnetic resonance reconstruction method and system based on a multistage wavelet convolution network, and particularly relates to the technical field of magnetic resonance imaging and image reconstruction, and the scheme comprises the steps: collecting multi-channel under-sampling K space data; inverse Fourier transform is carried out on the multi-channel under-sampling K space data to obtain an under-sampling magnetic resonance image; and inputting the undersampled magnetic resonance image into a trained multi-stage wavelet convolutional network for reconstruction to obtain a reconstructed magnetic resonance image, the multi-stage wavelet convolutional network being constructed based on multi-stage wavelet transform and a convolutional neural network. According to the scheme, the magnetic resonance K space data is collected in an undersampling mode, the scanning time can be shortened, the multi-stage wavelet convolution network can effectively reserve high-frequency details of the image in the reconstruction process, and therefore the reconstruction quality and efficiency of the multi-channel magnetic resonance image can be remarkably improved on the whole.

Description

Technical Field

[0001] The present invention relates to the technical field of magnetic resonance imaging and image reconstruction, and in particular to a fast magnetic resonance reconstruction method and system based on a multi-level wavelet convolutional network. Background Art

[0002] Magnetic resonance imaging (MRI) is one of the important imaging technologies in modern medicine and has very important clinical value. During the MRI imaging process, static magnetic field, gradient magnetic field and radio frequency (RF) pulse magnetic field of specific frequency are applied to make the imaging object generate magnetic resonance (MR) signals associated with the spatial position.

[0003] In terms of magnetic resonance imaging, the phase encoding step in traditional magnetic resonance imaging must be strictly executed in the order of the imaging sequence through the gradual conversion of the magnetic field gradient, which restricts the data acquisition speed. In conventional scanning, the magnetic resonance scan of a single part can take tens of minutes or even 1-2 hours, which may cause motion artifacts in imaging tasks with high real-time requirements, such as cardiac, perfusion and functional imaging. In addition, the patient remains relatively still in the closed device for a long time, which can easily induce unconscious movement, resulting in low imaging quality.

[0004] Therefore, fast MRI reconstruction has become a hot topic of research. Although deep learning-based fast MRI reconstruction methods such as ADMM-Net and MoDL have achieved certain results, the reconstruction of high-frequency details, local textures, and microstructures of the image is often too blurred or smooth due to the defects in the traditional downsampling structure and reconstruction loss function design, which affects the overall accuracy of the reconstructed image. Therefore, there is an urgent need for a fast MRI reconstruction method that can meet the requirements of clinical high-precision imaging. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the object of the present invention is to provide a fast magnetic resonance reconstruction method and system based on a multi-level wavelet convolutional network, aiming to solve the problems in the prior art that the reconstructed magnetic resonance images are too smooth and the reconstruction accuracy is not high.

[0006] In order to achieve the above object, the first aspect of the present invention provides a fast magnetic resonance reconstruction method based on a multi-level wavelet convolutional network, comprising:

[0007] Collect multi-channel under-sampled K-space data;

[0008] Performing inverse Fourier transform on the multi-channel under-sampled K-space data to obtain an under-sampled magnetic resonance image;

[0009] The under-sampled magnetic resonance image is input into a trained multi-level wavelet convolutional network for reconstruction to obtain a reconstructed magnetic resonance image, wherein the multi-level wavelet convolutional network is constructed based on a multi-level wavelet transform and a convolutional neural network.

[0010] Optionally, the trained multi-level wavelet convolution network includes a multi-level wavelet decomposition network and a multi-level wavelet reconstruction network, and inputting the undersampled magnetic resonance image into the trained multi-level wavelet convolution network for reconstruction to obtain a reconstructed magnetic resonance image includes:

[0011] Using the multi-level wavelet decomposition network to extract and decompose the under-sampled magnetic resonance image to obtain a feature map;

[0012] The feature map is reconstructed using the multi-level wavelet reconstruction network to obtain a reconstructed magnetic resonance image.

[0013] Optionally, the multi-level wavelet decomposition network is composed of a plurality of coding convolutional layers and wavelet decomposition layers stacked alternately, and the multi-level wavelet decomposition network is used to extract and decompose the undersampled magnetic resonance image to obtain a feature map, including:

[0014] Using the coding convolution layer to extract features from the undersampled magnetic resonance image to obtain an initial coding feature map;

[0015] The initial coding feature map is subjected to wavelet decomposition by using the wavelet decomposition layer to obtain an initial coding feature frequency component, and the initial coding feature frequency component is input into the next coding convolution layer and the wavelet decomposition layer for iterative processing in sequence to obtain a feature map.

[0016] Optionally, the multi-level wavelet reconstruction network is composed of a plurality of decoding convolutional layers and wavelet reconstruction layers stacked alternately, and the multi-level wavelet reconstruction network is used to reconstruct the feature map to obtain a reconstructed magnetic resonance image, including:

[0017] Utilizing the decoding convolutional layer to analyze the feature map, and obtain a number of initial decoding feature frequency components;

[0018] The wavelet reconstruction layer is used to perform wavelet reconstruction on all the initial decoding characteristic frequency components to obtain an initial decoding characteristic map, and the initial decoding characteristic map is input into the next decoding convolution layer and the wavelet reconstruction layer for iterative processing in sequence to obtain a reconstructed magnetic resonance image.

[0019] Optionally, the training process of the multi-level wavelet convolutional network includes:

[0020] Based on the full-sampled K-space data set, a plurality of groups of under-sampled magnetic resonance image samples and corresponding full-sampled magnetic resonance images are obtained;

[0021] Constructing a multi-level wavelet convolution network and initializing network parameters, wherein the multi-level wavelet convolution network is constructed by a multi-level wavelet decomposition network and a multi-level wavelet reconstruction network;

[0022] Inputting the undersampled magnetic resonance image samples into the multi-level wavelet convolutional network to obtain a network reconstructed image;

[0023] Calculating a reconstruction loss based on the network reconstructed image and the corresponding fully sampled magnetic resonance image;

[0024] The parameters of the multi-level wavelet convolutional network are iteratively optimized using a back propagation algorithm until the reconstruction loss reaches a preset loss threshold or the number of iterations reaches a preset iteration number threshold, thereby obtaining a trained multi-level wavelet convolutional network.

[0025] Optionally, the process of calculating the reconstruction loss includes:

[0026] Based on the network reconstructed image and the corresponding fully sampled magnetic resonance image, a reconstruction loss value of each pixel is calculated;

[0027] Based on the reconstruction loss value of each pixel, a unit loss discard probability is calculated;

[0028] Obtaining the total number of pixels of the network reconstructed image, and calculating the loss discard probability of each pixel according to a preset loss discard ratio, the total number of pixels and the unit loss discard probability;

[0029] Calculating a loss weight for each pixel based on the loss discard probability of each pixel;

[0030] Optionally, the calculating the unit loss discard probability based on the reconstruction loss value of each pixel includes:

[0031] The loss value of each pixel is normalized using a preset normalization correction factor to obtain a unit loss discard probability, and the normalization correction factor is used to adjust the difference in the loss discard probability of pixels with different reconstruction difficulties in the network reconstructed image.

[0032] A second aspect of the present invention provides a fast magnetic resonance reconstruction system based on a multi-level wavelet convolutional network, the system comprising:

[0033] An information acquisition module, used for acquiring multi-channel under-sampled K-space data;

[0034] An under-sampled image generation module, used for performing inverse Fourier transform on the multi-channel under-sampled K-space data to obtain an under-sampled magnetic resonance image;

[0035] The image reconstruction module is used to input the undersampled magnetic resonance image into a trained multi-level wavelet convolutional network for reconstruction to obtain a reconstructed magnetic resonance image, wherein the multi-level wavelet convolutional network is constructed based on multi-level wavelet transform and convolutional neural network.

[0036] A third aspect of the present invention provides a magnetic resonance device, comprising a memory for storing executable instructions; and a processor for calling and running the executable instructions in the memory to execute any one of the steps of the above-mentioned fast magnetic resonance reconstruction method based on a multi-level wavelet convolutional network.

[0037] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions. When the program instructions are executed by a processor, any step of the above-mentioned fast magnetic resonance reconstruction method based on a multi-level wavelet convolutional network is implemented.

[0038] Compared with the prior art, the beneficial effects of this solution are as follows:

[0039] The present invention uses undersampling to collect magnetic resonance K-space data, which can shorten the scanning time and help improve the imaging speed; the trained multi-level wavelet convolution network can extract multi-scale features in undersampled magnetic resonance images, while better protecting the high-frequency details of the image, so it can significantly improve the reconstruction quality and efficiency of multi-channel magnetic resonance images. At the same time, the constructed loss function can adaptively adjust the reconstruction loss weights of different regions according to the loss value, prompting the model to focus more attention on the more difficult high-frequency regions, which can significantly improve the network's reconstruction performance of high-frequency details. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0041] Figure 1 It is a flow chart of the fast magnetic resonance reconstruction method based on multi-level wavelet convolution network of the present invention;

[0042] Figure 2 It is a schematic diagram of the process of acquiring multi-channel under-sampled K-space data and under-sampled magnetic resonance images of the present invention;

[0043] Figure 3 It is a schematic diagram of a multi-level wavelet convolutional network module of the present invention;

[0044] Figure 4It is a schematic diagram of the module of the fast magnetic resonance reconstruction system based on the multi-level wavelet convolution network of the present invention;

[0045] Figure 5 It is a schematic diagram of the structure of the magnetic resonance device of the present invention. DETAILED DESCRIPTION

[0046] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0047] It should be understood that when used in this specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0048] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0049] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0050] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0052] Magnetic resonance imaging is one of the important imaging techniques in modern medicine. Rapid MRI reconstruction technology can significantly improve the speed and quality of MRI and has very important clinical value. In recent years, the development of deep learning and its wide application in image processing have provided new ideas for rapid MRI reconstruction. On the one hand, deep learning networks allow network layers to extract more abstract and representative features without manual design, so compared with traditional MRI reconstruction algorithms, more stable and higher quality results can be obtained. On the other hand, the highly parallel computing and parallel processing capabilities of deep learning networks make real-time MRI possible, thus providing more timely and effective image information for clinical diagnosis. This is especially important for application scenarios such as heart and perfusion that require rapid dynamic imaging. Therefore, the research on rapid MRI reconstruction based on deep learning has important clinical significance and brings new opportunities and prospects for the development of medical imaging.

[0053] The fast magnetic resonance reconstruction algorithm based on deep learning performs well in capturing the global features of images. However, affected by the traditional downsampling structure and the design defects of the reconstruction loss function, the current mainstream deep learning reconstruction methods, such as ADMM-Net and MoDL, are still not ideal for reconstructing high-frequency details of images. There are still certain limitations in the reconstruction of local textures and tiny structures, which may cause the reconstructed image to be blurred and too smooth.

[0054] Based on this, the present invention provides a fast magnetic resonance reconstruction method based on a multi-level wavelet convolutional network. The method first combines a multi-level wavelet transform with a convolutional neural network to construct a multi-level wavelet convolutional network; then uses paired input data and reference data to determine all parameters in the network structure through a back propagation algorithm; finally, the test data (i.e., undersampled magnetic resonance images) are reconstructed through a forward propagation process. In the training phase, paired undersampled multi-channel K-space data and corresponding fully sampled K-space data are input into the multi-level wavelet convolutional network to train the network to establish an "end-to-end" mapping relationship, that is, the training network can automatically and accurately map the input data to the reference data. In the testing phase, the network only needs to be provided with test data as input, and the trained multi-level wavelet convolutional network can be used to perform the image reconstruction process. Since the present scheme adopts undersampling to collect magnetic resonance K-space data, the scanning time can be shortened, which is beneficial to improving the imaging speed; since the wavelet transform is reversible, it can better avoid the problem of loss of detail information compared with traditional downsampling operations (such as maximum pooling), so the present scheme can better protect the high-frequency details of the image and improve the reconstruction quality of the image; at the same time, the present invention proposes a new focal reconstruction loss function (Focal Reconstruction Loss, FRL), which can adaptively adjust the reconstruction loss weights of different regions according to the loss value, prompting the model to focus more attention on high-frequency regions with higher reconstruction difficulty, thereby effectively improving the reconstruction performance of the multi-level wavelet convolutional network for high-frequency details, and then improving the overall reconstruction quality of the image.

[0055] The embodiment of the present invention provides a fast magnetic resonance reconstruction method based on a multi-level wavelet convolutional network, which is deployed on electronic devices such as computers and servers and is applied to the scene of reconstructing magnetic resonance images, especially for the case of reconstructing under-sampled magnetic resonance images. Specifically, Figure 1-Figure 3 As shown, the steps of the method in this embodiment include:

[0056] Step S100: Acquiring multi-channel under-sampled K-space data;

[0057] Specifically, undersampling is a sparse sampling method that reduces the sampling frequency when the bandwidth capacity of the test equipment is insufficient or in order to shorten the scanning time. In magnetic resonance imaging, undersampling is used to simultaneously collect partial data points in the K space using multiple receiving coils. Since different coils correspond to different receiving channels, it is beneficial to improve the imaging speed and image quality. Therefore, in the process of data acquisition based on a multi-channel magnetic resonance system in this embodiment, the layer selection gradient field and radio frequency excitation are first used to realize signal layer selection, determine the imaging layer and position, and control the layer thickness and layer position by adjusting the intensity and direction of the layer selection gradient field, as well as the bandwidth and center frequency of the radio frequency field. Then, data is collected simultaneously through each receiving coil, and the collected data is filled into the K space in the order of phase encoding and frequency encoding to form a two-dimensional complex matrix, that is, the multi-channel under-sampled K space data K u , where the multi-channel under-sampled K-space data K u The undersampling method used in this embodiment can be a random undersampling mode or an evenly spaced undersampling mode, such as Figure 2 The figure shows multi-channel under-sampled K-space data constructed by the interval under-sampling mode. The data size can be expressed as Nx*Ny*Nc, where Nx represents the number of rows of acquired data, Ny represents the number of columns of data, and Nc represents the number of receiving channels.

[0058] Step S200: performing inverse Fourier transform on the under-sampled scan data to obtain an under-sampled magnetic resonance image;

[0059] Specifically, the collected multi-channel under-sampled K-space data K u Perform a two-dimensional Fourier transform (2D FFT) to convert it into an undersampled magnetic resonance image I u .

[0060] Step S300: inputting the undersampled magnetic resonance image into a trained multi-level wavelet convolutional network for reconstruction to obtain a reconstructed magnetic resonance image, wherein the multi-level wavelet convolutional network is constructed based on a convolutional neural network and a multi-level wavelet transform.

[0061] In the training phase, a large number of high-quality fully sampled MRI images and corresponding under-sampled MRI images are used as training data. The network parameters are optimized through the back-propagation algorithm so that the multi-level wavelet convolution network can learn to restore high-quality images from under-sampled MRI images. In the inference phase, the under-sampled MRI images are directly used as the input of the trained multi-level wavelet convolution network, and the reconstructed high-quality MRI images can be obtained through network prediction.

[0062] In this embodiment, the undersampling method is used to collect magnetic resonance K-space data, which can shorten the scanning time and help improve the imaging speed; the trained multi-level wavelet convolution network can reconstruct the under-sampled magnetic resonance images into high-quality magnetic resonance images, while effectively retaining the high-frequency details of the images, thereby significantly improving the reconstruction efficiency and quality of multi-channel magnetic resonance images.

[0063] In a preferred embodiment, the trained multi-level wavelet convolution network includes a multi-level wavelet decomposition network and a multi-level wavelet reconstruction network. The step S300 of inputting the undersampled magnetic resonance image into the trained multi-level wavelet convolution network for reconstruction to obtain the reconstructed magnetic resonance image includes:

[0064] Step S310: extracting and decomposing features of the undersampled magnetic resonance image using the multi-level wavelet decomposition network to obtain a feature map;

[0065] Step S320: reconstruct the feature map using the multi-level wavelet reconstruction network to obtain a reconstructed magnetic resonance image.

[0066] Specifically, the multi-level wavelet convolution network trained in this embodiment includes a multi-level wavelet decomposition network and a multi-level wavelet reconstruction network. The multi-level wavelet decomposition network uses multi-level wavelet transform and convolution operation to gradually decompose the input undersampled magnetic resonance image into feature maps of different frequencies and resolutions. These feature maps contain both the spatial information of the original image and the frequency information of the image at different scales, such as high-frequency components (edges, textures) and low-frequency components (smooth areas), which can provide rich feature information for the subsequent reconstruction process. The multi-level wavelet reconstruction network gradually restores the resolution and details of the image through multi-level wavelet inverse transform and convolution operation, so as to reconstruct the feature map extracted by the multi-level wavelet decomposition network into a high-quality magnetic resonance image. In practical applications, the parameters and structures of the multi-level wavelet decomposition network and the multi-level wavelet reconstruction network can be designed and adjusted according to specific magnetic resonance imaging tasks and data.

[0067] In a preferred embodiment, the multi-level wavelet decomposition network is composed of a plurality of coding convolutional layers and wavelet decomposition layers stacked alternately, and the step S310 of extracting and decomposing features of the undersampled magnetic resonance image using the multi-level wavelet decomposition network to obtain a feature map includes:

[0068] Step S311: extracting features of the undersampled magnetic resonance image using the coded convolutional layer to obtain an initial coded feature map;

[0069] Step S312: using the wavelet decomposition layer to perform wavelet decomposition transformation on the initial coding feature map to obtain initial coding feature frequency components, and inputting the initial coding feature frequency components into the next coding convolution layer and the wavelet decomposition layer for iterative processing in sequence to obtain a feature map.

[0070] Specifically, Figure 3 As shown, the multi-level wavelet decomposition network in this embodiment is formed by alternately stacking multiple coded convolution layers and discrete wavelet transform (Discrete Wavelet Transform, DWT) layers. The process of extracting and decomposing the under-sampled magnetic resonance image using the multi-level wavelet decomposition network includes: first, using the first coded convolution layer to perform preliminary feature extraction on the under-sampled magnetic resonance image, specifically, sliding a series of learnable convolution kernels (or filters) on the image, calculating the local dot product between the input image and the convolution kernel, and then obtaining the initial features in the image through normalization and nonlinear activation, and expressing them in the form of a feature map, which contains low-level features of the image, such as edge, texture and color information. Then, the initial features are subjected to discrete wavelet transform using the discrete wavelet transform layer to obtain different frequency components of the initial feature map. According to the above steps, iterative processing is performed, and optimization is performed layer by layer until the last coded convolution layer outputs a feature map containing multi-scale features and rich detail information. It should be stated that in practical applications, the parameters and structure of the multi-level wavelet decomposition network can be adjusted and optimized according to specific tasks and data to obtain better performance.

[0071] In this embodiment, the advantages of discrete wavelet transform and convolutional neural network are combined to gradually extract and decompose image features through iterative processing, which can not only effectively extract the multi-scale features of the image, but also better protect the high-frequency details, local texture and micro-structure information of the image, thereby improving the quality and accuracy of image reconstruction.

[0072] In a preferred embodiment, the multi-level wavelet reconstruction network is composed of a plurality of decoding convolutional layers and wavelet reconstruction layers stacked alternately, and the step S320 of reconstructing the feature map using the multi-level wavelet reconstruction network to obtain a reconstructed magnetic resonance image includes:

[0073] Step S321: using the decoding convolution layer to analyze the feature map to obtain a number of initial decoding feature frequency components;

[0074] Step S322: using the wavelet reconstruction layer to perform wavelet reconstruction on all the initial decoding feature frequency components to obtain initial decoding features, and inputting the initial decoding features into the next decoding convolution layer and the wavelet reconstruction layer for iterative processing in sequence to obtain a reconstructed magnetic resonance image.

[0075] Specifically, Figure 3 As shown, the multi-level wavelet reconstruction network in this embodiment is formed by alternately stacking decoding convolution layers and wavelet reconstruction layers, wherein the wavelet reconstruction layer refers to the inverse wavelet transform (IWT) layer. The process of reconstructing the feature map using the multi-level wavelet reconstruction network includes: first, using the decoding convolution layer to parse the feature map output by the multi-level wavelet decomposition network to obtain the initial characteristic frequency components. Then, the wavelet reconstruction layer reconstructs these initial characteristic frequency components using the inverse wavelet transform to obtain the reconstructed initial decoding feature map. According to the above steps, iterative processing is performed and layer by layer optimization is performed until the last decoding convolution layer outputs the final reconstructed high-quality magnetic resonance image.

[0076] In this embodiment, the reconstruction capability of the inverse wavelet transform and the image restoration advantage of the convolutional neural network are fully combined, and the image features are gradually analyzed and the image details are restored through multiple iterations to achieve high-precision image reconstruction.

[0077] In a preferred embodiment, the training process of the multi-level wavelet convolutional network includes:

[0078] Step M100: Based on the full-sampled K-space data set, a plurality of groups of under-sampled magnetic resonance image samples and corresponding full-sampled magnetic resonance images are obtained;

[0079] Specifically, based on the multi-channel fully sampled K-space data, undersampled K-space data is generated by retrospective undersampling to obtain a number of paired undersampled K-space data and corresponding fully sampled K-space data. Then, inverse Fourier transform is used to generate undersampled magnetic resonance image samples corresponding to the undersampled K-space data and fully sampled magnetic resonance images corresponding to the corresponding fully sampled K-space data.

[0080] Step M200: construct a multi-level wavelet convolution network and initialize network parameters, wherein the multi-level wavelet convolution network is constructed by a multi-level wavelet decomposition network and a multi-level wavelet reconstruction network;

[0081] Specifically, the network consists of a multi-level wavelet decomposition network and a multi-level wavelet reconstruction network. The multi-level wavelet decomposition network is composed of alternating coded convolutional layers (CNN) and wavelet decomposition layers (DWT). Among them, the coded convolutional layer contains calculation units such as convolution, normalization, and nonlinear activation, and its function is to extract image features; the wavelet decomposition layer refers to discrete wavelet transform, and its function is to perform wavelet decomposition on image features to obtain feature components of different frequencies or scales, and pass them to the next level of convolutional layer. In the present invention, the type of wavelet can be arbitrary. This embodiment takes Haar wavelet (HaarWavelet) as an example to illustrate the calculation process of the wavelet decomposition layer. Assume that the feature output by the convolutional layer is F, where N, C, H, and W are the batch size, number of channels, length, and width of the feature, respectively. The Haar wavelet contains a low-pass filter f LL and three high-pass filters f LH 、f HL and f HH , the expression is as follows:

[0082]

[0083] The above four filters are used to filter the feature F batch by batch and channel by channel to obtain four frequency components X LL , X LH , X HL and X HH , splicing in the channel direction to obtain new features After being processed by the multi-level wavelet decomposition network, the output feature map will be further input into the multi-level wavelet reconstruction network.

[0084] The multi-level wavelet reconstruction network is composed of a decoding convolutional layer (CNN) and a wavelet reconstruction layer (IWT) stacked alternately. The decoding convolutional layer contains convolution, normalization, nonlinear activation and other computing units, which is used to analyze the features and obtain four different characteristic frequency components X LL , X LH , X HL and X HH ; The wavelet reconstruction layer refers to the inverse wavelet transform, which is used to perform wavelet reconstruction on the four characteristic frequency components and continuously restore and reconstruct the high-resolution feature X. For the Haar wavelet, the inverse wavelet transform calculation process is as follows:

[0085]

[0086] Step M300: inputting the undersampled magnetic resonance image sample into the multi-level wavelet convolution network to obtain a network reconstructed image;

[0087] Specifically, the undersampled image is input into a multi-level wavelet convolutional network to capture the multi-scale features in the image, and the resolution and detail information of the image are gradually restored through multiple convolutional layers, and finally a reconstructed magnetic resonance image is output, that is, the network reconstructed image. This image should have fewer curling artifacts and noise than the input undersampled image, while retaining the high-frequency details, local texture and micro-structure information in the original image, improving the quality and accuracy of image reconstruction.

[0088] Step M400: calculating the reconstruction loss based on the network reconstructed image and the corresponding full-sampled magnetic resonance image;

[0089] Specifically, when calculating the reconstruction loss, a preset reconstruction loss function is used to calculate the difference between the network reconstructed image and the corresponding full-sampled magnetic resonance image in terms of pixel value accuracy, structural similarity, texture details and other parameters.

[0090] Step M500: Iteratively optimize the parameters of the multi-level wavelet convolutional network using a back propagation algorithm until the reconstruction loss reaches a preset loss threshold or the number of iterations reaches a preset iteration threshold, thereby obtaining a trained multi-level wavelet convolutional network.

[0091] In a preferred embodiment, the process of calculating the reconstruction loss in step M400 includes:

[0092] Step M410: calculating a reconstruction loss value of each pixel based on the network reconstructed image and the corresponding fully sampled magnetic resonance image;

[0093] Step M420: Calculate the unit loss discard probability based on the reconstruction loss value of each pixel;

[0094] Step M430: obtaining the total number of pixels of the network reconstructed image, and calculating the loss discard probability of each pixel according to the preset loss discard ratio, the total number of pixels and the unit loss discard probability;

[0095] Step M440: Calculate the total reconstruction loss based on the loss weight of each pixel and the reconstruction loss value of each pixel.

[0096] Specifically, for the magnetic resonance image reconstruction task, this embodiment proposes a new focal reconstruction loss function (Focal Reconstruction Loss, FRL), which can be effectively combined with common pixel-wise types of losses such as L1 loss, L2 loss and SSIM loss to improve the reconstruction performance of the model. The principle of FRL is mainly: due to the difference in the difficulty of reconstructing different areas on the image by the neural network, the high-frequency area is theoretically more difficult than the low-frequency area, and the existing reconstruction loss function generally ignores this difference, resulting in different areas having the same loss weight, which causes the low-frequency area with a larger proportion to contribute most of the gradient, which is not conducive to the network's reconstruction of high-frequency details. To address this problem, FRL can adaptively adjust the reconstruction loss weights of different areas according to the size of the loss value, prompting the model to focus more attention on the more difficult high-frequency areas, thereby improving the network's reconstruction performance of high-frequency details. In the embodiment, the weight value is set to only have two states: 0 or 1. When the weight is 0, it means that the loss of the pixel is discarded and does not participate in the loss calculation. When the weight is 1, the loss is calculated normally. Assume that the network reconstruction result is P, and the corresponding fully sampled magnetic resonance image is t, where M represents the number of image pixels. For any Pixel-wiseN type reconstruction loss function (such as L1 loss, L2 loss, and SSIM loss), the calculation process of FRL is as follows:

[0097] First, based on the network reconstructed image and the corresponding fully sampled magnetic resonance image, the loss value g of each pixel is calculated i =Q(P[i],t[i]), and then perform the following normalization:

[0098]

[0099] Among them, h i It is called the unit loss discard probability, which means the probability of discarding the loss of pixel i when discarding the loss of one pixel, and β>0 is a normalized correction factor, which is used to adjust the difference in the loss discard probability of pixels with different reconstruction difficulties in the reconstructed magnetic resonance image. In this embodiment, for the convenience of description, the pixels in the reconstructed magnetic resonance image are divided into easy-to-reconstruct pixels and difficult-to-reconstruct pixels according to the reconstruction difficulty, wherein the difficult-to-reconstruct pixels include pixel areas such as high-frequency details, local textures and microstructures in the reconstructed magnetic resonance image; and the easy-to-reconstruct pixels refer to the pixel areas in the reconstructed magnetic resonance image other than the difficult-to-reconstruct pixels.

[0100] Assume that the unit loss discarding probability of easy-to-reconstruct pixels and difficult-to-reconstruct pixels is h m and h n , then the difference in the loss discard probability between the two is:

[0101]

[0102] In the above formula, g m <g n , so the larger β is, the greater the difference in the discarding probability between easy-to-reconstruct pixels and difficult-to-reconstruct pixels is, that is, easy-to-reconstruct pixels are more likely to be discarded than difficult-to-reconstruct pixels. Assuming that the total number of pixels in the image is M, and it is hoped that a proportion of γ pixels will be discarded in each training, then the probability of each pixel being discarded s i for:

[0103] s i =γMh i (5)

[0104] In the above formula, γ∈[0,1] is set manually. When γ=0, it means that all pixels are retained; when γ=1, it means that all pixels are discarded. Next, the random discard step is performed. If a pixel is discarded, its loss weight is reset to 0 so that it does not participate in the loss calculation. Otherwise, it is set to 1, that is:

[0105]

[0106] Among them, x i ∈U(0,1) is a random number sampled from a standard uniform distribution, w i is the loss weight of pixel i, and the final focus reconstruction loss is:

[0107]

[0108] The focused reconstruction loss function FRL constructed in this embodiment can adaptively adjust the reconstruction loss weights of different regions according to the loss value, prompting the model to focus more attention on pixel areas where reconstruction is more difficult, and can significantly improve the network's reconstruction performance for high-frequency details. At the same time, the focused reconstruction loss function FRL can be combined with various types of existing loss functions to meet different actual calculation accuracy, calculation amount and other requirements, thereby improving the flexibility of calculating the loss value of the reconstructed image.

[0109] like Figure 4 As shown, corresponding to the above-mentioned fast magnetic resonance reconstruction method based on the multi-level wavelet convolution network, the embodiment of the present invention further provides a fast magnetic resonance reconstruction system based on the multi-level wavelet convolution network, and the above-mentioned fast magnetic resonance reconstruction system based on the multi-level wavelet convolution network includes:

[0110] An information acquisition module 410 is used to acquire multi-channel under-sampled K-space data;

[0111] An under-sampled image generation module 420 is used to perform inverse Fourier transform on the multi-channel under-sampled K-space data to obtain an under-sampled magnetic resonance image;

[0112] The image reconstruction module 430 is used to input the undersampled magnetic resonance image into a trained multi-level wavelet convolutional network for reconstruction to obtain a reconstructed magnetic resonance image. The multi-level wavelet convolutional network is constructed based on multi-level wavelet transform and convolutional neural network.

[0113] Specifically, in this embodiment, the specific functions of the above-mentioned fast magnetic resonance reconstruction system based on the multi-level wavelet convolution network can also refer to the corresponding description in the above-mentioned fast magnetic resonance reconstruction method based on the multi-level wavelet convolution network, and will not be repeated here.

[0114] Based on the above embodiments, the present invention further provides a magnetic resonance device, whose principle block diagram can be shown as follows: Figure 5 As shown. The magnetic resonance device can be used to execute the fast magnetic resonance reconstruction method based on the multi-level wavelet convolution network provided in the above embodiment, which will not be described here for the sake of brevity. The magnetic resonance device includes: a processor, the processor is coupled to a memory, the memory is used to store a computer program or instruction, and the processor is used to execute the computer program or instruction stored in the memory, so that the method in the above method embodiment is executed.

[0115] The present invention also provides a computer-readable storage medium on which computer instructions for implementing the method in the above method embodiment are stored.

[0116] For example, when the computer program is executed by a computer, the computer can implement the method in the above method embodiment.

[0117] The embodiment of the present application also provides a computer program product including instructions, which, when executed by a computer, enables the computer to implement the method in the above method embodiment.

[0118] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0119] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0120] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0121] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0122] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0123] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

Claims

1. A fast magnetic resonance reconstruction method based on a multi-level wavelet convolutional network, characterized in that: The following steps are involved: Collect multi-channel under-sampled K-space data; Performing inverse Fourier transform on the multi-channel under-sampled K-space data to obtain an under-sampled magnetic resonance image; The under-sampled magnetic resonance image is input into a trained multi-level wavelet convolutional network for reconstruction to obtain a reconstructed magnetic resonance image, wherein the multi-level wavelet convolutional network is constructed based on a multi-level wavelet transform and a convolutional neural network.

2. The fast magnetic resonance reconstruction method based on a multi-level wavelet convolutional network according to claim 1, characterized in that: The trained multi-level wavelet convolution network includes a multi-level wavelet decomposition network and a multi-level wavelet reconstruction network, and the under-sampled magnetic resonance image is input into the trained multi-level wavelet convolution network for reconstruction to obtain a reconstructed magnetic resonance image, including: Using the multi-level wavelet decomposition network to extract and decompose the under-sampled magnetic resonance image to obtain a feature map; The feature map is reconstructed using the multi-level wavelet reconstruction network to obtain a reconstructed magnetic resonance image.

3. The fast magnetic resonance reconstruction method based on a multi-level wavelet convolutional network according to claim 2, characterized in that: The multi-level wavelet decomposition network is composed of a plurality of coding convolution layers and wavelet decomposition layers stacked alternately, and the multi-level wavelet decomposition network is used to extract and decompose the under-sampled magnetic resonance image to obtain a feature map, including: Using the coding convolution layer to extract features from the undersampled magnetic resonance image to obtain an initial coding feature map; The initial coding feature map is subjected to wavelet decomposition by using the wavelet decomposition layer to obtain an initial coding feature frequency component, and the initial coding feature frequency component is input into the next coding convolution layer and the wavelet decomposition layer for iterative processing in sequence to obtain a feature map.

4. The fast magnetic resonance reconstruction method based on multi-level wavelet convolutional network according to claim 2, characterized in that: The multi-level wavelet reconstruction network is composed of a plurality of decoding convolutional layers and wavelet reconstruction layers stacked alternately, and the multi-level wavelet reconstruction network is used to reconstruct the feature map to obtain a reconstructed magnetic resonance image, including: Utilizing the decoding convolutional layer to analyze the feature map, and obtain a number of initial decoding feature frequency components; The wavelet reconstruction layer is used to perform wavelet reconstruction on all the initial decoding characteristic frequency components to obtain an initial decoding characteristic map, and the initial decoding characteristic map is input into the next decoding convolution layer and the wavelet reconstruction layer for iterative processing in sequence to obtain a reconstructed magnetic resonance image.

5. The fast magnetic resonance reconstruction method based on multi-level wavelet convolutional network according to claim 1, characterized in that: The training process of the multi-level wavelet convolutional network includes: Based on the full-sampled K-space data set, a plurality of groups of under-sampled magnetic resonance image samples and corresponding full-sampled magnetic resonance images are obtained; Constructing a multi-level wavelet convolution network and initializing network parameters, wherein the multi-level wavelet convolution network is constructed by a multi-level wavelet decomposition network and a multi-level wavelet reconstruction network; Inputting the undersampled magnetic resonance image samples into the multi-level wavelet convolutional network to obtain a network reconstructed image; Calculating a reconstruction loss based on the network reconstructed image and the corresponding fully sampled magnetic resonance image; The parameters of the multi-level wavelet convolutional network are iteratively optimized using a back propagation algorithm until the reconstruction loss reaches a preset loss threshold or the number of iterations reaches a preset iteration number threshold, thereby obtaining a trained multi-level wavelet convolutional network.

6. The fast magnetic resonance reconstruction method based on multi-level wavelet convolutional network according to claim 5, characterized in that: The process of calculating the reconstruction loss includes: Based on the network reconstructed image and the corresponding fully sampled magnetic resonance image, a reconstruction loss value of each pixel is calculated; Based on the reconstruction loss value of each pixel, a unit loss discard probability is calculated; Obtaining the total number of pixels of the network reconstructed image, and calculating the loss discard probability of each pixel according to a preset loss discard ratio, the total number of pixels and the unit loss discard probability; Calculating a loss weight for each pixel based on the loss discard probability of each pixel; Based on the loss weight of each pixel and the reconstruction loss value of each pixel, a total reconstruction loss is calculated.

7. The fast magnetic resonance reconstruction method based on multi-level wavelet convolutional network according to claim 6, characterized in that: The calculating the unit loss discard probability based on the reconstruction loss value of each pixel includes: The loss value of each pixel is normalized using a preset normalization correction factor to obtain a unit loss discard probability, and the normalization correction factor is used to adjust the difference in the loss discard probability of pixels with different reconstruction difficulties in the network reconstructed image.

8. A fast magnetic resonance reconstruction system based on a multi-level wavelet convolutional network, characterized in that: The system comprises: An information acquisition module, used for acquiring multi-channel under-sampled K-space data; An under-sampled image generation module, used for performing inverse Fourier transform on the multi-channel under-sampled K-space data to obtain an under-sampled magnetic resonance image; The image reconstruction module is used to input the undersampled magnetic resonance image into a trained multi-level wavelet convolutional network for reconstruction to obtain a reconstructed magnetic resonance image, wherein the multi-level wavelet convolutional network is constructed based on multi-level wavelet transform and convolutional neural network.

9. A magnetic resonance device, characterized in that: include: A memory for storing executable instructions; A processor is used to call and run the executable instructions in the memory to execute the steps of the fast magnetic resonance reconstruction method based on a multi-level wavelet convolutional network as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are executed by the processor, the fast magnetic resonance reconstruction method based on the multi-level wavelet convolutional network as described in any one of claims 1 to 7 is implemented.

Images