Image reconstruction method and device, electronic device and computer readable storage medium

Abstract

The present application provides an image reconstruction method and device, an electronic device, and a computer-readable storage medium, which relate to the field of medical imaging technology. The method obtains multi-phase actual under-sampled data and corresponding pre-scanned images of moving tissue, and uses a trained deep learning network model to reconstruct the multi-phase actual under-sampled data and the corresponding pre-scanned images to obtain actual moving tissue movie imaging images. It can be seen that the embodiment of the present application can reconstruct the multi-phase k-space actual under-sampled data into a multi-phase k-space full-sampled movie imaging image. The adjacent phase images of the moving tissue movie imaging image contain a large amount of redundant information, and the use of a deep learning network model can extract the common low-frequency information of adjacent phase images and merge the same high-frequency information, thereby reconstructing the final multi-phase image set, which has richer details and a higher signal-to-noise ratio.

Description

Technical Field

[0001] The present application relates to the field of medical imaging technology, and in particular to an image reconstruction method and device, an electronic device, and a computer-readable storage medium. Background Art

[0002] Magnetic Resonance Imaging (MRI), as a multi-parameter, multi-contrast imaging technology, is one of the main imaging methods in modern medical imaging. It can reflect multiple characteristics such as tissue and proton density, and provide information for disease detection and diagnosis. The basic working principle of MRI is to use the magnetic resonance phenomenon, use radio frequency excitation to excite hydrogen protons in the human body, use gradient fields for position encoding, and then use receiving coils to receive electromagnetic signals with position information, and finally use Fourier transform to reconstruct image information.

[0003] Cardiac Magnetic Resonance Cine Imaging (CMRCI), also known as cardiac cine imaging, can quickly image the heart at different stages of a cardiac cycle, obtain multiple images, and play the images of the heart's systolic and diastolic ventricular walls, valves and other structures in the form of movies. It can quantitatively analyze cardiac function parameters and provide more information for diagnosing abnormal cardiac function. Due to special factors such as complex heart beats, disordered blood flow, respiratory movement, and thin cardiac chamber walls, cardiac cine imaging has high requirements for time and space resolution. In order to obtain complete cardiac cycle imaging, it is usually necessary to accelerate the imaging. Currently, commonly used acceleration methods include parallel acquisition and compressed sensing.

[0004] Parallel imaging acquisition technology can use the spatial sensitivity information of the coil, and multiple receiving coils are sampled and encoded independently at the same time, which saves the number of gradient codes required for spatial positioning and saves imaging time exponentially. Parallel imaging acquisition technology speeds up the scanning speed to a certain extent and reduces the impact of motion on the image, but it sacrifices the image signal-to-noise ratio and spatial resolution. Moreover, the larger the acceleration factor, the lower the image signal-to-noise ratio, the more serious the curling artifact, and it is difficult to complete a high-quality cardiac movie scan in one breath hold. Compressed sensing acceleration technology takes the sparsity of the signal as a premise, samples and compresses the signal at the same time, directly obtains the sparse signal, and only uses the highly undersampled k-space to reconstruct the image, reducing the magnetic resonance imaging time. Whether compressed sensing acceleration is successful depends on the sparsity of the image signal, incoherent sampling and nonlinear reconstruction. The sampling template of compressed sensing usually affects the sensitivity of boundary information, such as the contrast between the myocardium and surrounding tissues, and inappropriate reconstruction parameters and iterative reconstruction algorithms can also cause image distortion and blurring.

[0005] It can be seen that the current technical means have a relatively large loss in the signal-to-noise ratio, contrast, sharpness, etc. of the image; in order to obtain a higher spatial resolution, the time of each phase will be correspondingly extended, resulting in the scanning of this phase being affected by the heart beat, and the image is blurred; and the post-processing methods such as interpolation are used to improve the temporal resolution of cardiac movie imaging, which is usually not accurate enough considering the complex movement of the heart; in addition, the compressed sensing algorithm has a large amount of calculation and a long reconstruction time. Therefore, how to ensure the quality of the moving tissue movie imaging image while shortening the acquisition time and improving the acquisition efficiency has become a technical problem that needs to be solved urgently. Summary of the invention

[0006] In view of the above problems, the present application is proposed to provide an image reconstruction method and device, electronic device and computer-readable storage medium that overcome the above problems or at least partially solve the above problems, and can ensure the quality of motion tissue movie imaging images while shortening the acquisition time and improving the acquisition efficiency. The technical solution is as follows:

[0007] In a first aspect, an image reconstruction method is provided, the method comprising:

[0008] Acquire a trained deep learning network model, wherein the trained deep learning network model is obtained by training based on multi-phase data of k-space full sampling of motion tissue and pre-scanned motion tissue images and under-sampling data corresponding to the multi-phase data of k-space full sampling;

[0009] Acquire multi-phase actual under-sampled data and corresponding pre-scanned images of moving tissue, and use the trained deep learning network model to reconstruct the multi-phase actual under-sampled data and the corresponding pre-scanned images to obtain actual moving tissue movie imaging images.

[0010] In a possible implementation, a deep learning network model is trained based on multi-phase data of k-space full sampling of motion tissue and pre-scanned motion tissue images and under-sampled data corresponding to the multi-phase data of k-space full sampling to obtain the trained deep learning network model, including:

[0011] The pre-scanned motion tissue image and the under-sampled data corresponding to the multi-phase data fully sampled in k-space are reconstructed by using a preset preliminary reconstruction algorithm to generate a preliminary multi-phase motion tissue image;

[0012] Using the preliminary multi-phase motion tissue image, the pre-scan motion tissue image and the under-sampling data as network inputs of a deep learning network model, and the network output is a reconstructed multi-phase motion tissue image;

[0013] Reconstructing the fully acquired multi-phase data to obtain a fully acquired multi-phase motion tissue image;

[0014] The reconstructed multi-phase motion tissue image is compared with the fully sampled multi-phase motion tissue image, and then the network parameters of the deep learning network model are updated according to the comparison result, so as to obtain the trained deep learning network model.

[0015] In a possible implementation, before reconstructing the pre-scanned motion tissue image and the under-sampled data corresponding to the multi-phase data fully sampled in the k-space using a preset preliminary reconstruction algorithm to generate a preliminary multi-phase motion tissue image, the method further includes:

[0016] Acquire multi-channel and multi-phase data of k-space full sampling of motion tissue movie imaging as label data for deep learning network model training, and simultaneously acquire pre-scanned motion tissue images as antecedent information for the deep learning network model;

[0017] Reconstructing the pre-scanned motion tissue image and the under-sampled data corresponding to the multi-phase data fully sampled in k-space by using a preset preliminary reconstruction algorithm to generate a preliminary multi-phase motion tissue image, including: reconstructing the pre-scanned motion tissue image and the under-sampled data corresponding to the multi-channel multi-phase data fully sampled in k-space by using a preset preliminary reconstruction algorithm to generate a preliminary merged channel multi-phase motion tissue image;

[0018] The preliminary multi-phase motion tissue image, the pre-scan motion tissue image and the under-sampling data are used as network inputs of a deep learning network model, and the network output is a reconstructed multi-phase motion tissue image, including: the preliminary merged channel multi-phase motion tissue image, the pre-scan motion tissue image and the under-sampling data are used as network inputs of a deep learning network model, and the network output is a reconstructed merged channel multi-phase motion tissue image.

[0019] In a possible implementation, the trained deep learning network model is used to reconstruct the multi-phase actual under-sampled data and the corresponding pre-scanned images to obtain actual motion tissue movie imaging images, including:

[0020] Preliminarily reconstructing the multi-phase actual under-sampled data and the corresponding pre-scanned images to generate an actual preliminary multi-phase motion tissue image;

[0021] The actual preliminary multi-phase motion tissue image, the multi-phase actual under-sampled data and the corresponding pre-scan image are input into the trained deep learning network model, and the output obtained is the actual motion tissue movie imaging image.

[0022] In a possible implementation, the trained deep learning network model is used to extract common low-frequency information of adjacent phase images and merge the same high-frequency information.

[0023] In a possible implementation, undersampled data corresponding to the fully sampled multi-phase data of k-space is obtained by the following steps:

[0024] The fully sampled multi-phase data are extracted into under-sampled data in k-space.

[0025] In a possible implementation, extracting the fully sampled multi-phase data into under-sampled data in k-space includes:

[0026] Select the training under-extraction multiple in the preset under-extraction multiple range;

[0027] The fully sampled multi-phase data are extracted in k-space as under-sampled data according to the training under-sampled multiples.

[0028] In a second aspect, an image reconstruction device is provided, the device comprising:

[0029] An acquisition module is used to acquire a trained deep learning network model, wherein the trained deep learning network model is obtained by training based on multi-phase data of k-space full sampling of motion tissue and pre-scanned motion tissue images and under-sampled data corresponding to the multi-phase data of k-space full sampling;

[0030] The reconstruction module is used to obtain multi-phase actual under-sampled data and corresponding pre-scanned images of moving tissue, and use the trained deep learning network model to reconstruct the multi-phase actual under-sampled data and corresponding pre-scanned images to obtain actual moving tissue movie imaging images.

[0031] According to a third aspect, an electronic device is provided, comprising a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute any one of the above-mentioned image reconstruction methods.

[0032] According to a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, wherein the computer program is configured to execute any one of the above-mentioned image reconstruction methods when run.

[0033] By means of the above technical solution, the image reconstruction method and device, electronic device and computer-readable storage medium provided in the embodiment of the present application can obtain the multi-phase actual under-sampled data and the corresponding pre-scanned images of the moving tissue, and use the trained deep learning network model to reconstruct the multi-phase actual under-sampled data and the corresponding pre-scanned images to obtain the actual moving tissue movie imaging image, wherein the trained deep learning network model is based on the multi-phase data of the k-space full sampling of the moving tissue and the pre-scanned moving tissue image and the under-sampled data corresponding to the multi-phase data of the k-space full sampling. It can be seen that the embodiment of the present application can reconstruct the multi-phase k-space actual under-sampled data into a multi-phase k-space full sampling movie imaging image, and the adjacent phase images of the moving tissue movie imaging image contain a large amount of redundant information, and the use of the deep learning network model can extract the common low-frequency information of the adjacent phase images and merge the same high-frequency information, thereby reconstructing the final multi-phase image set, which has richer details and higher signal-to-noise ratio, and the same high-frequency information includes completely identical and / or similar information. Furthermore, during actual reconstruction, actual under-sampled data of multiple phases are acquired through motion tissue movie imaging acquisition sequences, which saves acquisition time and improves acquisition efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in describing the embodiments of the present application are briefly introduced below.

[0035] Figure 1 A flowchart of an image reconstruction method provided according to an embodiment of the present application is shown;

[0036] Figure 2 A schematic diagram of cardiac cine imaging acquisition provided according to an embodiment of the present application is shown;

[0037] Figure 3 A flowchart of an image reconstruction method provided according to another embodiment of the present application is shown;

[0038] Figure 4 A schematic diagram showing a comparison between a 6-fold undersampled cardiac cine imaging data reconstructed image and a fully sampled image at different sections according to an embodiment of the present application is shown;

[0039] Figure 5 A structural diagram of an image reconstruction device provided according to an embodiment of the present application is shown;

[0040] Figure 6 A structural diagram of an image reconstruction device provided according to another embodiment of the present application is shown;

[0041] Figure 7 A structural diagram of an electronic device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0042] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0043] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that such use is interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the term "including" and its variants are to be interpreted as open-ended terms meaning "including but not limited to".

[0044] The present application embodiment provides an image reconstruction method, such as Figure 1 As shown, the image reconstruction method may include the following steps S101 to S102:

[0045] Step S101, obtaining a trained deep learning network model, where the trained deep learning network model is obtained by training based on multi-phase data of k-space full sampling of motion tissue and pre-scanned motion tissue images and under-sampled data corresponding to the multi-phase data of k-space full sampling;

[0046] Step S102, obtaining multi-phase actual under-sampled data and corresponding pre-scanned images of moving tissue, and reconstructing the multi-phase actual under-sampled data and corresponding pre-scanned images using the trained deep learning network model to obtain actual moving tissue movie imaging images.

[0047] The image reconstruction method provided by the embodiment of the present application can reconstruct the actual undersampled data of multi-phase k-space into a fully sampled movie imaging image of multi-phase k-space. The adjacent phase images of the motion tissue movie imaging image contain a large amount of redundant information, and the use of a deep learning network model can extract the common low-frequency information of adjacent phase images and merge the same high-frequency information, thereby reconstructing the final multi-phase image set. These multi-phase image sets have richer details and higher signal-to-noise ratios, and the same high-frequency information includes completely identical and / or similar information. In addition, during actual reconstruction, the actual undersampled data of multiple phases is obtained through the motion tissue movie imaging acquisition sequence, which saves acquisition time and improves acquisition efficiency.

[0048] The motor tissue mentioned in the embodiment of the present application may also be referred to as a motor organ, which may specifically be a heart, stomach, lung or other tissue organ, and this embodiment does not impose any limitation on this.

[0049] A possible implementation method is provided in an embodiment of the present application, which can construct a deep learning network model for reconstructing movie images of moving tissue. Specifically, the deep learning network model for reconstructing movie images of moving tissue can be constructed based on one or more combinations of convolutional neural networks, recurrent neural networks, fully connected networks, and generative adversarial networks, etc., but the embodiment of the present application is not limited to this.

[0050] A possible implementation method is provided in an embodiment of the present application. When acquiring multi-phase data of k-space with full sampling of moving tissue, it can be specifically a magnetic resonance moving tissue movie imaging scan triggered by gating of the moving tissue electrical signal. The entire moving tissue electrical signal cycle is divided into multiple phases (which can be equally divided or unequally divided), and each phase collects one or more k-lines in the k-space. The data obtained in a single phase is filled into the k-space corresponding to each phase to obtain fully sampled multi-phase data.

[0051] Taking the moving tissue as the heart as an example, when obtaining multi-phase data with full sampling of the k-space of magnetic resonance cardiac cine imaging, it can be specifically a magnetic resonance cardiac cine imaging scan triggered by ECG gating. The entire ECG cycle is divided into multiple phases (which can be equally divided or unequally divided), and each phase collects one or more k-lines in the k-space. The data obtained from a single phase is filled into the k-space corresponding to each phase to obtain full sampling multi-phase data.

[0052] Figure 2 A schematic diagram of cardiac movie imaging acquisition provided by an embodiment of the present application is shown, Figure 2 The diagram illustrates the acquisition method of the two cycle phases T1 and T2. Four phases are collected in each T1 and T2 cycle, and each phase collects four k-lines in the k-space (which can be four rows of k-lines in the phase encoding direction). The data obtained from a single phase is filled into the k-space corresponding to each phase, so that fully sampled multi-phase data can be obtained.

[0053] A possible implementation method is provided in an embodiment of the present application. When acquiring a magnetic resonance pre-scan moving tissue image, a magnetic resonance moving tissue movie imaging scan triggered by moving tissue electrical signal gating can be used. The entire moving tissue electrical signal cycle is equally divided into multiple phases, and each phase collects one or more k-lines in the k-space. The data obtained in a single phase is filled into the k-space corresponding to each phase and the respective images are reconstructed. The reconstructed images of all phases constitute the pre-scan moving tissue image.

[0054] Taking the moving tissue as the heart as an example, when obtaining the magnetic resonance pre-scan cardiac image, it can be specifically a magnetic resonance cardiac movie imaging scan triggered by ECG gating. The entire ECG cycle is equally divided into multiple phases, and each phase collects one or more k lines in the k-space. The data obtained in a single phase is filled into the k-space corresponding to each phase and reconstructs each image. The reconstructed images of all phases constitute the pre-scan cardiac image. Schematically, you can continue to refer to Figure 2 , T1 and T2 collected four phases respectively, each phase collected four k lines of k-space, filled the data obtained in a single phase into the k-space corresponding to each phase and reconstructed the respective images a1, a2, a3 and a4, and the reconstructed images of all phases constituted the pre-scan cardiac image. It should be noted that Figure 2 The cardiac cine imaging acquisition shown is only illustrative and does not limit the embodiments of the present invention.

[0055] In an embodiment of the present application, a possible implementation method is provided. When acquiring undersampled data corresponding to multi-phase data fully sampled in k-space, the fully sampled multi-phase data may be extracted as undersampled data in k-space, or a training undersampled multiple may be selected in a preset undersampled multiple interval, and the fully sampled multi-phase data may be extracted as undersampled data in k-space according to the training undersampled multiple. Here, parallel imaging technology or compressed sensing technology may be used to extract the fully sampled multi-phase data in k-space as undersampled data as a training set for a deep learning network model.

[0056] In an embodiment of the present application, a possible implementation method is provided. When a deep learning network model is trained based on multi-phase data of k-space full sampling of motion tissue and pre-scanned motion tissue images and under-sampled data corresponding to the multi-phase data of k-space full sampling, the trained deep learning network model can be obtained by specifically including the following steps A1 to A4:

[0057] Step A1, using a preset preliminary reconstruction algorithm to reconstruct the pre-scanned motion tissue image and the under-sampled data corresponding to the multi-phase data fully sampled in k-space to generate a preliminary multi-phase motion tissue image;

[0058] Step A2, using the preliminary multi-phase motion tissue image, the pre-scan motion tissue image and the under-sampling data as network inputs of the deep learning network model, and the network output is a reconstructed multi-phase motion tissue image;

[0059] Step A3, reconstructing the fully sampled multi-phase data to obtain a fully sampled multi-phase motion tissue image;

[0060] Step A4, comparing the reconstructed multi-phase motion tissue image with the full-sample multi-phase motion tissue image, and then updating the network parameters of the deep learning network model according to the comparison result, so as to obtain the trained deep learning network model.

[0061] In step A1, the preset preliminary reconstruction algorithm may be parallel imaging technology or compressed sensing technology. In step A3, the fully sampled multi-phase data is reconstructed to obtain a fully sampled multi-phase motion tissue image. Specifically, the fully sampled multi-phase data may be reconstructed through Fourier transform to obtain a fully sampled multi-phase motion tissue image.

[0062] In an embodiment of the present application, a possible implementation method is provided. When there are multiple receiving coils for magnetic resonance, the multi-phase data of k-space full sampling of motion tissue movie imaging is specifically multi-channel multi-phase data. Then, in step A1, a preset preliminary reconstruction algorithm is used to reconstruct the pre-scan motion tissue image and the under-sampled data corresponding to the multi-phase data fully sampled in k-space. Before generating the preliminary multi-phase motion tissue image, the multi-channel multi-phase data of k-space full sampling of motion tissue movie imaging can also be obtained as label data for deep learning network model training, and the pre-scan motion tissue image is obtained as the antecedent information of the deep learning network model. At this time, step A1 can specifically be to reconstruct the pre-scan motion tissue image and the under-sampled data corresponding to the multi-channel multi-phase data fully sampled in k-space using a preset preliminary reconstruction algorithm to generate a preliminary merged channel multi-phase motion tissue image. Here, the spatial sensitivity information of the coil can be used to reconstruct the multi-channel multi-phase data into a merged channel multi-phase motion tissue image.

[0063] Further, in step A2, the preliminary multi-phase motion tissue image, the pre-scanned motion tissue image and the under-sampled data are used as the network input of the deep learning network model, and the network output is the reconstructed multi-phase motion tissue image. Specifically, the preliminary merged channel multi-phase motion tissue image, the pre-scanned motion tissue image and the under-sampled data are used as the network input of the deep learning network model, and the network output is the reconstructed merged channel multi-phase motion tissue image. And, in step A3, the fully sampled multi-phase data is reconstructed to obtain the fully sampled multi-phase motion tissue image. Specifically, the fully sampled multi-channel multi-phase data is reconstructed to obtain the fully sampled merged channel multi-phase motion tissue image. In step A4, the reconstructed multi-phase motion tissue image is compared with the fully sampled multi-phase motion tissue image, and then the network parameters of the deep learning network model are updated according to the comparison result, so as to obtain the trained deep learning network model. Specifically, the reconstructed merged channel multi-phase motion tissue image is compared with the fully sampled merged channel multi-phase motion tissue image, and then the network parameters of the deep learning network model are updated according to the comparison result, so as to obtain the trained deep learning network model.

[0064] Furthermore, in step S102 above, the trained deep learning network model is used to reconstruct the multi-phase actual under-sampled data and the corresponding pre-scanned images to obtain the actual motion tissue movie imaging image, which may specifically include the following steps B1 to B2:

[0065] Step B1, preliminarily reconstructing the multi-phase actual under-sampled data and the corresponding pre-scanned images to generate an actual preliminary multi-phase motion tissue image;

[0066] Step B2, inputting the actual preliminary multi-phase motion tissue image, the multi-phase actual under-sampled data and the corresponding pre-scanned image into the trained deep learning network model, and the output obtained is the actual motion tissue movie imaging image.

[0067] In the above embodiments, by combining traditional imaging technology with deep learning network models, the image reconstruction effect is improved, and the unexplainability and instability caused by the algorithm black box problem of the deep learning network model are reduced. Compared with traditional algorithms, deep learning network models can ensure that the reconstructed images are richer in details and have a higher signal-to-noise ratio by extracting redundant information of images in multiple phases. Compared with traditional reconstruction algorithms, deep learning network models have short reconstruction time and occupy fewer resources.

[0068] The above introduces Figure 1 The various implementation methods of each link in the illustrated embodiment are described below. The image reconstruction method provided by the embodiment of the present invention is further described through specific embodiments.

[0069] Figure 3 FIG. 2 is a flowchart of an image reconstruction method according to another embodiment of the present application. In this embodiment, taking the moving tissue as the heart as an example, there are multiple magnetic resonance receiving coils, so that multi-channel and multi-phase data can be collected. Figure 3 As shown, the image reconstruction method may include the following steps S301 to S309.

[0070] Step S301, obtaining multi-channel and multi-phase data of full k-space sampling of magnetic resonance cardiac movie imaging as label data for deep learning network model training, and simultaneously obtaining magnetic resonance pre-scan cardiac images as antecedent information of the deep learning network model.

[0071] Step S302: By adopting parallel imaging technology or compressed sensing technology, the fully sampled multi-channel and multi-phase data are extracted into under-sampled data in k-space as a training set for the deep learning network model.

[0072] Step S303: Reconstruct the pre-scanned cardiac image and the under-sampled data using parallel imaging technology or compressed sensing technology to generate a preliminary merged channel multi-phase cardiac image.

[0073] Step S304, the preliminary merged channel multi-phase cardiac image, the pre-scan cardiac image and the under-sampled data are used as the network input of the deep learning network model, and the network output is the reconstructed merged channel multi-phase cardiac image. Here, the deep learning network model for reconstructing cardiac movie images can be constructed based on one or more combinations of convolutional neural networks, recurrent neural networks, fully connected networks and adversarial generative networks, etc., and the embodiment of the present invention is not limited to this.

[0074] Step S305 , reconstructing the fully sampled multi-channel and multi-phase data to obtain a fully sampled merged channel and multi-phase cardiac image.

[0075] Step S306, comparing the reconstructed merged channel multi-phase cardiac image with the fully sampled merged channel multi-phase cardiac image, and then updating the network parameters of the deep learning network model according to the comparison result, so as to obtain a trained deep learning network model.

[0076] In this step, the deep learning network model structure includes a data fidelity module. During the network reconstruction process, by controlling the gap between the reconstructed image and the full-sample merged channel multi-phase cardiac image, it is ensured that the data reconstruction process will not deviate significantly from the full-sample multi-phase cardiac image.

[0077] Step S307, acquiring multi-channel multi-phase actual under-sampling data and corresponding pre-scan images through a cardiac cine imaging acquisition sequence.

[0078] Step S308, using parallel imaging technology or compressed sensing technology to preliminarily reconstruct the multi-channel multi-phase actual under-sampled data and the corresponding pre-scanned images to generate an actual preliminary merged channel multi-phase cardiac image.

[0079] Step S309, inputting the actual preliminary merged channel multi-phase cardiac image, the multi-channel multi-phase actual under-sampled data and the corresponding pre-scanned image into the trained deep learning network model, and the output obtained is the actual cardiac movie imaging image.

[0080] The present embodiment provides a method based on a deep learning network combined with a traditional reconstruction algorithm to improve the problem of low temporal resolution and spatial resolution of cardiac movie imaging. Compared with the traditional algorithm, the reconstruction method of the present embodiment can improve the image quality and increase the under-sampling multiple of the acquisition, that is, the acquisition time is shorter under the same image quality. Therefore, more phases or higher resolution images can be acquired at the acquisition end, thereby achieving the effect of improving the temporal resolution and spatial resolution of cardiac movie imaging. In addition, by extracting redundant information of the image in multiple dimensions such as multiple channels and multiple phases, it can be ensured that the reconstructed image has richer details and a higher signal-to-noise ratio. On the other hand, the use of a method based on a deep learning network combined with a traditional reconstruction algorithm can protect the original data from being modified to the greatest extent, improve the stability of the algorithm, and reduce the reconstruction time and corresponding computing resources.

[0081] Figure 4 A schematic diagram is shown of a comparison between a 6-fold undersampled cardiac movie imaging data reconstructed image and a fully sampled image at different sections according to an embodiment of the present application. By comparison, it can be found that the image reconstruction method provided in the embodiment of the present application can completely reconstruct the 6-fold undersampled image, and the image contrast and detail information are consistent with the fully sampled label image, that is, P1, P2, and P3 are fully sampled label images, Q1, Q2, and Q3 are completely reconstructed images of the 6-fold undersampled image, P1 is consistent with Q1, P2 is consistent with Q2, and P3 is consistent with Q3.

[0082] It should be noted that, in practical applications, all possible implementation methods described above can be arbitrarily combined to form possible embodiments of the present application, which will not be described one by one here.

[0083] Based on the image reconstruction methods provided in the above embodiments and based on the same inventive concept, an embodiment of the present application further provides an image reconstruction device.

[0084] Figure 5 FIG. 4 shows a structural diagram of an image reconstruction device provided according to an embodiment of the present application. Figure 5 As shown, the image reconstruction device may include an acquisition module 510 and a reconstruction module 520 .

[0085] An acquisition module 510 is used to acquire a trained deep learning network model, where the trained deep learning network model is obtained by training based on multi-phase data of k-space full sampling of motion tissue and pre-scanned motion tissue images and under-sampling data corresponding to the multi-phase data of k-space full sampling;

[0086] The reconstruction module 520 is used to obtain multi-phase actual under-sampled data and corresponding pre-scanned images of moving tissue, and use the trained deep learning network model to reconstruct the multi-phase actual under-sampled data and the corresponding pre-scanned images to obtain actual moving tissue movie imaging images.

[0087] A possible implementation method is provided in the embodiment of the present application, such as Figure 6 As shown above Figure 5 The image reconstruction apparatus may further include a training module 610, which is further configured to:

[0088] The pre-scanned motion tissue image and the under-sampled data corresponding to the multi-phase data fully sampled in k-space are reconstructed by using a preset preliminary reconstruction algorithm to generate a preliminary multi-phase motion tissue image;

[0089] The preliminary multi-phase motion tissue image, the pre-scan motion tissue image and the under-sampling data are used as the network input of the deep learning network model, and the network output is the reconstructed multi-phase motion tissue image;

[0090] Reconstruct the fully acquired multi-phase data to obtain a fully acquired multi-phase motion tissue image;

[0091] The reconstructed multi-phase motion tissue image is compared with the full-sample multi-phase motion tissue image, and then the network parameters of the deep learning network model are updated according to the comparison results to obtain the trained deep learning network model.

[0092] A possible implementation method is provided in the embodiment of the present application, and the training module 610 is further used to:

[0093] Acquire multi-channel and multi-phase data of k-space full sampling of motion tissue movie imaging as label data for deep learning network model training, and simultaneously acquire pre-scanned motion tissue images as antecedent information for the deep learning network model;

[0094] The pre-scanned motion tissue image and the under-sampled data corresponding to the multi-channel multi-phase data fully sampled in k-space are reconstructed using a preset preliminary reconstruction algorithm to generate a preliminary merged channel multi-phase motion tissue image;

[0095] The preliminary merged channel multi-phase motion tissue image, the pre-scanned motion tissue image and the under-sampling data are used as the network input of the deep learning network model, and the network output is the reconstructed merged channel multi-phase motion tissue image.

[0096] A possible implementation method is provided in the embodiment of the present application, and the training module 610 is further used to:

[0097] Reconstruct the fully acquired multi-channel and multi-phase data to obtain a fully acquired merged channel multi-phase motion tissue image;

[0098] The reconstructed merged channel multi-phase motion tissue image is compared with the full-sampled merged channel multi-phase motion tissue image, and then the network parameters of the deep learning network model are updated according to the comparison results to obtain the trained deep learning network model

[0099] A possible implementation is provided in an embodiment of the present application, where the reconstruction module 520 is further configured to:

[0100] Preliminarily reconstructing the multi-phase actual under-sampled data and the corresponding pre-scanned images to generate an actual preliminary multi-phase motion tissue image;

[0101] The actual preliminary multi-phase motion tissue images, multi-phase actual under-sampled data and corresponding pre-scan images are input into the trained deep learning network model, and the output obtained is the actual motion tissue movie imaging image.

[0102] A possible implementation method is provided in an embodiment of the present application, in which a trained deep learning network model is used to extract common low-frequency information of adjacent phase images and merge the same high-frequency information.

[0103] A possible implementation method is provided in the embodiment of the present application, and the training module 610 is further used to:

[0104] The fully sampled multi-phase data are extracted into under-sampled data in k-space.

[0105] A possible implementation method is provided in the embodiment of the present application, and the training module 610 is further used to:

[0106] Select the training under-extraction multiple in the preset under-extraction multiple range;

[0107] The fully sampled multi-phase data are extracted as under-sampled data in k-space according to the training under-sampled multiples.

[0108] Based on the same inventive concept, an embodiment of the present application further provides an electronic device, including a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the image reconstruction method of any one of the above embodiments.

[0109] In an exemplary embodiment, an electronic device is provided, such as Figure 7 As shown, Figure 7 The electronic device 700 shown includes: a processor 701 and a memory 703. The processor 701 and the memory 703 are connected, such as through a bus 702. Optionally, the electronic device 700 may also include a transceiver 704. It should be noted that in actual applications, the transceiver 704 is not limited to one, and the structure of the electronic device 700 does not constitute a limitation on the embodiments of the present application.

[0110] Processor 701 may be a CPU (Central Processing Unit), a general purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 701 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0111] The bus 702 may include a path to transmit information between the above components. The bus 702 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 702 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0112] The memory 703 can be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0113] The memory 703 is used to store the application code for executing the solution of the present application, and the execution is controlled by the processor 701. The processor 701 is used to execute the application code stored in the memory 703 to implement the contents shown in the above method embodiment.

[0114] Among them, electronic devices include but are not limited to: mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 7 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0115] Based on the same inventive concept, an embodiment of the present application further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the image reconstruction method of any one of the above embodiments when executed.

[0116] Those skilled in the art can clearly understand that the specific working processes of the systems, devices, and modules described above can refer to the corresponding processes in the aforementioned method embodiments, and for the sake of brevity, they are not further described here.

[0117] Those skilled in the art can understand that the technical solution of the present application can be essentially or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, which includes a number of program instructions, so that an electronic device (such as a personal computer, a server, or a network device, etc.) executes all or part of the steps of the method described in each embodiment of the present application when running the program instructions. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0118] Alternatively, all or part of the steps of implementing the aforementioned method embodiments may be accomplished by hardware associated with program instructions (such as electronic devices such as personal computers, servers, or network devices), and the program instructions may be stored in a computer-readable storage medium. When the program instructions are executed by a processor of an electronic device, the electronic device executes all or part of the steps of the methods described in the embodiments of the present application.

[0119] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that within the spirit and principles of the present application, they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not deviate from the protection scope of the present application.

Claims

1. An image reconstruction method, characterized in that: include: Obtaining a trained deep learning network model, wherein the trained deep learning network model is obtained by training based on multi-phase data of k-space full sampling of moving tissue and pre-scanned moving tissue images and under-sampling data corresponding to the multi-phase data of k-space full sampling; wherein, a magnetic resonance moving tissue movie imaging scan triggered by gated moving tissue electrical signal is adopted, the entire moving tissue electrical signal cycle is divided into multiple phases, each phase acquires one or more k-lines of k-space, the data obtained in a single phase is filled into the k-space corresponding to each phase and the respective images are reconstructed, and the reconstructed images of all phases constitute the pre-scanned moving tissue image; Acquire multi-phase actual under-sampled data and corresponding pre-scanned images of moving tissue, and reconstruct the multi-phase actual under-sampled data and corresponding pre-scanned images using the trained deep learning network model to obtain actual moving tissue movie imaging images; The deep learning network model is trained based on the multi-phase data of k-space full sampling of the motion tissue and the pre-scanned motion tissue image and the under-sampling data corresponding to the multi-phase data of k-space full sampling to obtain the trained deep learning network model, including: The pre-scanned motion tissue image and the under-sampled data corresponding to the multi-phase data fully sampled in k-space are reconstructed by using a preset preliminary reconstruction algorithm to generate a preliminary multi-phase motion tissue image; Using the preliminary multi-phase motion tissue image, the pre-scan motion tissue image and the under-sampling data as network inputs of a deep learning network model, and the network output is a reconstructed multi-phase motion tissue image; Reconstructing the fully acquired multi-phase data to obtain a fully acquired multi-phase motion tissue image; The reconstructed multi-phase motion tissue image is compared with the fully sampled multi-phase motion tissue image, and then the network parameters of the deep learning network model are updated according to the comparison result, so as to obtain the trained deep learning network model.

2. The image reconstruction method according to claim 1, characterized in that: Before reconstructing the pre-scanned motion tissue image and the under-sampled data corresponding to the multi-phase data fully sampled in the k-space by using a preset preliminary reconstruction algorithm to generate a preliminary multi-phase motion tissue image, the method further includes: Acquire multi-channel and multi-phase data of k-space full sampling of motion tissue movie imaging as label data for deep learning network model training, and simultaneously acquire pre-scanned motion tissue images as antecedent information for the deep learning network model; Reconstructing the pre-scanned motion tissue image and the under-sampled data corresponding to the multi-phase data fully sampled in k-space by using a preset preliminary reconstruction algorithm to generate a preliminary multi-phase motion tissue image, including: reconstructing the pre-scanned motion tissue image and the under-sampled data corresponding to the multi-channel multi-phase data fully sampled in k-space by using a preset preliminary reconstruction algorithm to generate a preliminary merged channel multi-phase motion tissue image; The preliminary multi-phase motion tissue image, the pre-scan motion tissue image and the under-sampling data are used as network inputs of a deep learning network model, and the network output is a reconstructed multi-phase motion tissue image, including: the preliminary merged channel multi-phase motion tissue image, the pre-scan motion tissue image and the under-sampling data are used as network inputs of a deep learning network model, and the network output is a reconstructed merged channel multi-phase motion tissue image.

3. The image reconstruction method according to claim 1, characterized in that: The trained deep learning network model is used to reconstruct the multi-phase actual under-sampled data and the corresponding pre-scanned images to obtain actual motion tissue movie imaging images, including: Preliminarily reconstructing the multi-phase actual under-sampled data and the corresponding pre-scanned images to generate an actual preliminary multi-phase motion tissue image; The actual preliminary multi-phase motion tissue image, the multi-phase actual under-sampled data and the corresponding pre-scan image are input into the trained deep learning network model, and the output obtained is the actual motion tissue movie imaging image.

4. The image reconstruction method according to claim 1, characterized in that: The trained deep learning network model is used to extract common low-frequency information of adjacent phase images and merge the same high-frequency information.

5. The image reconstruction method according to any one of claims 1 to 4, characterized in that: The undersampled data corresponding to the multi-phase data fully sampled in k-space are obtained by the following steps: The fully sampled multi-phase data are extracted into under-sampled data in k-space.

6. The image reconstruction method according to claim 5, characterized in that: The step of extracting the fully sampled multi-phase data into under-sampled data in k-space includes: Select the training under-extraction multiple in the preset under-extraction multiple range; The fully sampled multi-phase data are extracted in k-space as undersampled data according to the training undersampled multiples.

7. An image reconstruction device, characterized in that: include: An acquisition module is used to acquire a trained deep learning network model, wherein the trained deep learning network model is obtained by training based on multi-phase data of k-space full sampling of motion tissue and pre-scanned motion tissue images and under-sampling data corresponding to the multi-phase data of k-space full sampling; wherein, a magnetic resonance motion tissue movie imaging scan triggered by gated motion tissue electrical signal is adopted, the entire motion tissue electrical signal cycle is divided into multiple phases, each phase acquires one or more k-lines of k-space, the data obtained in a single phase is filled into the k-space corresponding to each phase and the respective images are reconstructed, and the reconstructed images of all phases constitute the pre-scanned motion tissue image; A reconstruction module is used to obtain multi-phase actual under-sampled data of motion tissue and corresponding pre-scanned images, and use the trained deep learning network model to reconstruct the multi-phase actual under-sampled data and the corresponding pre-scanned images to obtain actual motion tissue movie imaging images; The device further comprises a training module, which is used for: The pre-scanned motion tissue image and the under-sampled data corresponding to the multi-phase data fully sampled in k-space are reconstructed by using a preset preliminary reconstruction algorithm to generate a preliminary multi-phase motion tissue image; The preliminary multi-phase motion tissue image, the pre-scan motion tissue image and the under-sampling data are used as the network input of the deep learning network model, and the network output is the reconstructed multi-phase motion tissue image; Reconstruct the fully acquired multi-phase data to obtain a fully acquired multi-phase motion tissue image; The reconstructed multi-phase motion tissue image is compared with the full-sample multi-phase motion tissue image, and then the network parameters of the deep learning network model are updated according to the comparison results to obtain the trained deep learning network model.

8. An electronic device, characterized in that: The invention comprises a processor and a memory, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the image reconstruction method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the image reconstruction method according to any one of claims 1 to 6 when executed.

Images