A method and device for artifact correction based on standardized flow

Abstract

The present application relates to the field of magnetic resonance imaging, and specifically to a method and device for artifact correction based on standardized flow. The method and device learn different artifact types and different levels of flow models for data in the image domain and the K-space domain; use the standardized flow in the flow model to perform dual learning on the data in the image domain and the K-space domain, and construct a bidirectional reversible mapping between artifact data and artifact-free data; and use the bidirectional reversible mapping to perform artifact correction on MRI images. Compared with traditional correction methods, due to the reversibility of the flow method, the invention has better interpretability, and combined with the dual domain constraints, it can also have a better artifact correction effect.

Description

Technical Field

[0001] The present application relates to the field of magnetic resonance imaging, and in particular to a method and device for artifact correction based on standardized flow. Background Art

[0002] Magnetic resonance imaging (MRI) has been widely used in the field of medical diagnosis. However, due to the high requirements for the imaging environment during the imaging process, the patient's breathing and movement will cause artifacts in the image that affect the diagnosis. Therefore, artifacts have also become an important research topic in MRI imaging. Existing research is mainly divided into two directions. One is to remove artifacts based on the image domain, mainly using computer vision methods, and using neural networks to learn the residuals of artifact-free and artifact-free images for artifact removal. The other is to remove artifacts based on K space, treating artifact information as outliers for processing. Both routes have their own shortcomings. Due to the complexity and particularity of MRI images, including the diversity of artifact types, the image domain-based method often makes it difficult for the network to learn the characteristics of the real artifacts, resulting in poor artifact correction results. The K-space-based method is difficult to obtain the initial raw data, resulting in a lack of research. In order to solve this problem, the correction method based on the dual domain is a very good choice. Combining the dual domain information of the image domain and the K-space domain, combined with the physical methods of magnetic resonance imaging, can better ensure the quality of MRI images and has very important scientific significance and application prospects in the fields of auxiliary medical treatment and diagnosis.

[0003] At the same time, nuclear magnetic resonance is a medical imaging technology that has been widely used in recent years. It can reflect the abnormalities and early lesions of human organs from the inside of human molecules, and is superior to X-ray and CT imaging in many aspects. Compared with CT and X-ray, MRI can provide more information on the physiological state of human organs and can perform imaging at any level. Magnetic resonance imaging can also reflect the information about the chemical structure of human tissues by measuring the nuclear relaxation time T1 and T2 of atoms, and is more accurate for quantitative analysis of soft tissues and gray and white matter.

[0004] During the MRI examination, if the patient does not hold his breath, the image information will be disturbed in the K space, thus generating artifacts in the image domain, affecting the speed and accuracy of the doctor's diagnosis. The artifacts of MRI are mostly respiratory artifacts, which are represented as stripes in the image. The blurring degree of the artifacts is positively correlated with the patient's breathing intensity during the examination. Therefore, it is very necessary to correct the artifacts on the MRI images.

[0005] Normalizing flow is a new type of generative model proposed in recent years. Compared with generative adversarial networks (GANs) and variational autoencoders (VAEs), it has better interpretability and traceability. Flow refers to the data flow that passes through a series of bijections (reversible mappings) and is finally mapped to a suitable representation space. Normalizing means that the integral of the variable in the representation space is 1, so that it satisfies the definition of the probability distribution function. This technology can be simply described as: producing a more complex and more expressive probability distribution from a simple probability distribution. Let the initial basic probability distribution change step by step through this series of mapping changes to become the required complex distribution.

[0006] In 2020, Andreas Lugmayr et al. proposed a super-resolution reconstruction method based on standardized flow (SRFlow) to learn a reversible transformation from low-resolution images to high-resolution images, so that SRFlow can generate multiple high-quality images of different resolutions from the learned complex distribution, showing better representation ability and interpretability. In addition, since SRFlow's transformation is reversible, it can complete the task of downsampling from high-resolution images to low-resolution images. Muhammad Asim et al. proposed a model that uses Flow prior to replace GAN prior to solve the inverse problem:

[0007]

[0008] Where A is the metric matrix, G(·) is the Flow model, and p G is the likelihood function about x guided by G, and λ is a hyperparameter. Asim et al. applied this model to solve the problems of compressed sensing and image denoising, and achieved good performance.

[0009] Traditional artifact correction methods are methods that perform artifact correction based solely on information in the image domain or K-space domain, which have the following disadvantages:

[0010] 1. The amount of information in a single domain (image domain or K-space domain) is relatively small compared to that in dual domains.

[0011] 2. The method is effective for artifacts that have been trained, but is not effective in removing artifact types with unknown distribution.

[0012] 3. The image domain-based method is not effective in correcting heavier artifacts, and some fat tissue will be enhanced during the correction process, resulting in loss of details.

[0013] 4. It is difficult to obtain original data based on K-space methods and there is a lack of sufficient data support. Summary of the invention

[0014] The embodiments of the present application provide a method and device for artifact correction based on standardized flow, so as to at least solve the technical problem that the existing artifact correction methods have poor effects.

[0015] According to an embodiment of the present application, a method for artifact correction based on standardized flow is provided, comprising the following steps:

[0016] Learn different artifact types and different levels of flow models for data in image and K-space domains;

[0017] The standardized flow in the flow model is used to perform dual learning on the data in the image domain and the K-space domain, and a bidirectional reversible mapping between the data with artifacts and the data without artifacts is constructed.

[0018] Artifact correction of MRI images using bidirectional reversible mapping.

[0019] The technical solution adopted by the embodiment of the present application also includes: learning different artifact types and different levels of flow models for data in the image domain and the K-space domain includes:

[0020] Collect corresponding data and use deep learning algorithms to perform paired training on the image domain and K-space domain.

[0021] The technical solution adopted by the embodiment of the present application also includes: collecting corresponding data and using a deep learning algorithm to perform paired training on the image domain and the K-space domain, including:

[0022] Four sets of data are collected, namely, image domain data with artifacts, K-space domain data with artifacts, image domain data without artifacts, and K-space domain data without artifacts. Based on the equipment parameters, the image domain data is inversely Fourier transformed to obtain the corresponding K-space data.

[0023] The technical solution adopted by the embodiment of the present application also includes: learning different artifact types and different levels of flow models for data in the image domain and the K-space domain also includes:

[0024] The artifact data is further subdivided according to the degree, and is divided into light, moderate and severe artifact data;

[0025] Establish datasets of no, mild, moderate, and severe artifacts in image and K-space domains.

[0026] The technical solution adopted by the embodiment of the present application also includes: using the standardized flow in the flow model to perform dual learning on the data in the image domain and the K-space domain, and constructing a bidirectional reversible mapping between the artifact data and the artifact-free data includes:

[0027] Define a differentiable and reversible transformation T: y = f(x), From a basic distribution p xThe sample u in (x) is transformed by T to obtain a random variable:

[0028]

[0029]

[0030] Using a series of reversible mappings Establishing a mapping relationship from original distributed artifact-free data x to new distributed artifact-containing data y;

[0031] The mapping relationship between artifact-free and artifacts of various degrees is established, and a series of reversible mappings transform the original distribution into a new distribution.

[0032] The technical solution adopted by the embodiment of the present application also includes: using bidirectional reversible mapping to correct artifacts of MRI images includes:

[0033] The learned normalized flow model is used to correct artifacts in MRI images.

[0034] The technical solution adopted in the embodiment of the present application also includes: performing artifact correction in the relevant domains on the image domain and K-space domain data respectively, and then fusing the correction results. If only the result of a single domain is available, the final correction result is directly given.

[0035] According to another embodiment of the present application, there is provided an artifact correction device based on standardized flow, comprising:

[0036] A flow model building unit, used for learning different artifact types and different levels of flow models for data in the image domain and the K-space domain;

[0037] A bidirectional reversible mapping unit, used for performing dual learning on data in the image domain and the K-space domain by using the standardized flow in the flow model, and constructing a bidirectional reversible mapping between data with artifacts and data without artifacts;

[0038] The artifact correction unit is used to perform artifact correction on the MRI image by using bidirectional reversible mapping.

[0039] A storage medium stores a program file capable of implementing any one of the above-mentioned artifact correction methods based on standardized flow.

[0040] A processor is used to run a program, wherein when the program is run, any one of the above-mentioned artifact correction methods based on standardized flow is executed.

[0041] Compared with the prior art, the beneficial effects of the embodiments of the present application are as follows: the artifact correction method and device based on standardized flow in the embodiments of the present application learn different artifact types and different levels of flow models for data in the image domain and the K-space domain; use the standardized flow in the flow model to perform dual learning on the data in the image domain and the K-space domain, and construct a bidirectional reversible mapping between artifact data and artifact-free data; use the bidirectional reversible mapping to perform artifact correction on MRI images. Compared with traditional correction methods, due to the reversibility of the flow method, the invention has better interpretability, and combined with the dual domain constraints, it can also have a better artifact correction effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0043] Figure 1 A schematic diagram of a standardized flow of reversible mapping in an artifact correction method based on a standardized flow according to an embodiment of the present application;

[0044] Figure 2 A schematic diagram of dual learning in the image domain and the K-space domain in the artifact correction method based on the standardized flow in an embodiment of the present application;

[0045] Figure 3 This is a flowchart of the entire artifact correction method based on standardized flow in an embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0047] 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 the data used in this way can be 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 terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0048] Example 1

[0049] According to an embodiment of the present application, a method for artifact correction based on standardized flow is provided, comprising the following steps:

[0050] Learn different artifact types and different levels of flow models for data in image and K-space domains;

[0051] The standardized flow in the flow model is used to perform dual learning on the data in the image domain and the K-space domain, and a bidirectional reversible mapping between the data with artifacts and the data without artifacts is constructed.

[0052] Artifact correction of MRI images using bidirectional reversible mapping.

[0053] The artifact correction method based on standardized flow in the embodiment of the present application learns different artifact types and different levels of flow models for data in the image domain and K-space domain; uses the standardized flow in the flow model to perform dual learning on the data in the image domain and K-space domain, and constructs a bidirectional reversible mapping between artifact data and artifact-free data; and uses the bidirectional reversible mapping to perform artifact correction on MRI images. Compared with traditional correction methods, due to the reversibility of the flow method, the invention has better interpretability, and combined with the dual domain constraints, it can also have a better artifact correction effect.

[0054] Among them, learning different artifact types and different levels of flow models for data in the image domain and K-space domain includes:

[0055] Collect corresponding data and use deep learning algorithms to perform paired training on the image domain and K-space domain.

[0056] Among them, collecting corresponding data and using deep learning algorithms to perform paired training on the image domain and the K-space domain include:

[0057] Four sets of data are collected, namely, image domain data with artifacts, K-space domain data with artifacts, image domain data without artifacts, and K-space domain data without artifacts. Based on the equipment parameters, the image domain data is inversely Fourier transformed to obtain the corresponding K-space data.

[0058] Among them, learning different artifact types and different levels of flow models for data in the image domain and the K-space domain also includes:

[0059] The artifact data is further subdivided according to the degree, and is divided into light, moderate and severe artifact data;

[0060] Establish datasets of no, mild, moderate, and severe artifacts in image and K-space domains.

[0061] Among them, using the standardized flow in the flow model to perform dual learning on the data in the image domain and the K-space domain, and constructing a bidirectional reversible mapping between the artifact data and the artifact-free data includes:

[0062] Define a differentiable and reversible transformation T: y = f(x), From a basic distribution p x The sample u in (x) is transformed by T to obtain a random variable:

[0063]

[0064]

[0065] Using a series of reversible mappings T i Establishing a mapping relationship from original distributed artifact-free data x to new distributed artifact-containing data y;

[0066] The mapping relationship between artifact-free and artifacts of various degrees is established, and a series of reversible mappings transform the original distribution into a new distribution.

[0067] Among them, using bidirectional reversible mapping to correct artifacts in MRI images includes:

[0068] The learned normalized flow model is used to correct artifacts in MRI images.

[0069] Among them, the artifact correction of the relevant domains is performed on the image domain and K-space domain data respectively, and then the correction results are fused. If there is only the result of a single domain, the final correction result is directly given.

[0070] The following is a detailed description of the artifact correction method based on the standardized flow of the present application with a specific embodiment:

[0071] In order to solve the technical problems of MRI imaging image quality, this application will combine the image domain and K-space dual domains to provide more information, and propose a new artifact correction method based on standardized flow, which can effectively correct artifacts in the MRI imaging process, reduce the impact of such unnecessary information on the diagnosis of radiologists, and improve the imaging quality of MRI images. This application uses paired data to construct a reversible mapping from artifact images to artifact-free images during artifact removal, and performs artifact removal based on this mapping. Compared with traditional correction methods, due to the reversibility of the flow method, this invention has better interpretability, and combined with the dual domain constraints, it can also have a better artifact correction effect.

[0072] This application combines existing research. First, the standardized flow is used to perform dual learning on the data in the image domain and the K-space domain to construct a bidirectional reversible mapping between artifact data and artifact-free data. Secondly, different artifact types and different levels of flow models are further learned. Finally, artifact correction of MRI images is achieved. Compared with previous methods, the correction results are better, more interpretable, and more generalizable.

[0073] To achieve the above objectives, the specific implementation method of this application is as follows:

[0074] 1. Dataset

[0075] This application is a deep learning algorithm that performs paired training in the image domain and the K-space domain, and corresponding data needs to be collected.

[0076] First, four sets of data need to be collected: 1) image domain data with artifacts, 2) K-space domain data with artifacts, 3) image domain data without artifacts, and 4) K-space domain data without artifacts. It is not required that the data in 1) and 2) and the data in 3) and 4) must correspond one to one. Paired data exist, and the corresponding K-space data can be obtained from the image domain data through inverse Fourier transform according to the equipment parameters.

[0077] Secondly, the artifact data can be further subdivided and divided into light, moderate and severe artifact data according to their severity.

[0078] Finally, a dataset of no, mild, moderate, and severe artifacts was established in the image domain and K-space domain.

[0079] 2. Establishment of reversible mapping between artifacts and no artifacts

[0080] Step 1: Define a differentiable and reversible transformation T: y = f(x), From a basic distribution p x The sample u in (x) is transformed by T to obtain a random variable:

[0081]

[0082] So

[0083] Step 2: Use a series of reversible mappings Establish a mapping relationship from the original distribution (data without artifacts x) to the new distribution (data with artifacts y), as follows Figure 1 As shown, Figure 1 A schematic diagram of the standardized flow of reversible mapping, from Z0 to Z k The generation process.

[0084] Step 3: Establish the mapping relationship between artifact-free and artifact-free. A series of reversible mappings transform the original distribution into a new distribution. By optimizing this series of distributions, the purpose of transforming a simple Gaussian distribution into a complex true posterior distribution can be achieved. Bidirectional learning strategies based on dual domain data, such as Figure 2 As shown, Figure 2 Schematic diagram of dual learning in image domain and K-space domain.

[0085] Step 4: Use the learned standardized flow model for MRI artifact correction. If there are image domain and K-space domain data, perform artifact correction in the relevant domains respectively, and then fuse the correction results. If there are only results from a single domain, directly give the final correction result. The whole process is as follows: Figure 3 As shown, Figure 3 Flowchart for the entire artifact correction.

[0086] The key points and intended protection points of the present application method are at least:

[0087] 1. Based on the standardized flow, a reversible mapping relationship between artifact-free, mild, moderate, and severe artifact data is established to achieve free conversion between various data.

[0088] 2. Use the information of the dual domain of image domain and K-space domain to correct artifacts and enhance the quality of the final corrected image.

[0089] 3. The present application can be considered as an artifact correction method based on the reversibility of the same mapping, and can also be considered as an artifact generation method based on the reversibility of the same mapping.

[0090] The advantages of the present application method and the existing respiratory artifact correction method are as follows:

[0091] First of all, this application optimizes the artifact correction method based on dual domain constraints. Compared with the traditional single-domain-based respiratory artifact correction method, it has better effect, can effectively restore the image, and has better traceability and interpretability.

[0092] Secondly, based on the reversibility of the standardized flow, compared with other methods that simply perform artifact correction, the present application can simultaneously generate more realistic artifacts compared with other artifact generation methods, and can further expand the data set and improve the generalization of the model.

[0093] Finally, the dual domain of the image domain and the K-space domain is optimized simultaneously, which can provide richer detail information than a single domain and reduce the detail loss of the image after artifact correction.

[0094] The changes or alternatives to this application method are:

[0095] In addition to being used for MRI artifact correction, this method can also be used for CT image artifact correction after appropriate transformation.

[0096] This method is further optimized and can be applied to the joint task of MRI and CT reconstruction and artifact correction, using flow priors to build a data-driven joint model of image reconstruction and artifact correction.

[0097] Based on the reversibility of the flow model, this method can be used to generate MR image artifacts.

[0098] Example 2

[0099] According to another embodiment of the present application, there is provided an artifact correction device based on standardized flow, comprising:

[0100] A flow model building unit, used for learning different artifact types and different levels of flow models for data in the image domain and the K-space domain;

[0101] A bidirectional reversible mapping unit, used for performing dual learning on data in the image domain and the K-space domain by using the standardized flow in the flow model, and constructing a bidirectional reversible mapping between data with artifacts and data without artifacts;

[0102] The artifact correction unit is used to perform artifact correction on the MRI image by using bidirectional reversible mapping.

[0103] The artifact correction device based on standardized flow in the embodiment of the present application learns different artifact types and different levels of flow models for data in the image domain and K-space domain; uses the standardized flow in the flow model to perform dual learning on the data in the image domain and K-space domain, and constructs a bidirectional reversible mapping between artifact data and artifact-free data; and uses the bidirectional reversible mapping to perform artifact correction on MRI images. Compared with traditional correction methods, due to the reversibility of the flow method, the invention has better interpretability, and combined with the dual domain constraints, it can also have a better artifact correction effect.

[0104] The following is a detailed description of the artifact correction device based on standardized flow of the present application with a specific embodiment:

[0105] In order to solve the technical problems of MRI imaging image quality, this application will combine the image domain and K-space dual domains to provide more information, and propose a new artifact correction device based on standardized flow to effectively correct artifacts in the MRI imaging process, reduce the impact of such unnecessary information on the diagnosis of radiologists, and improve the imaging quality of MRI images. This application uses paired data to construct a reversible mapping from artifact images to artifact-free images during artifact removal, and performs artifact removal based on this mapping. Compared with traditional correction methods, due to the reversibility of the flow method, this invention has better interpretability. At the same time, combined with the dual domain constraints, it can also have a better artifact correction effect.

[0106] This application combines existing research. First, the standardized flow is used to perform dual learning on the data in the image domain and the K-space domain to construct a bidirectional reversible mapping between artifact data and artifact-free data. Secondly, different artifact types and different levels of flow models are further learned. Finally, artifact correction of MRI images is achieved. Compared with previous methods, the correction results are better, more interpretable, and more generalizable.

[0107] To achieve the above objectives, the specific implementation method of this application is as follows:

[0108] 1. Flow model building unit: learning the data in the image domain and the K-space domain to obtain different artifact types and different levels of flow models.

[0109] This application is a deep learning algorithm that performs paired training in the image domain and the K-space domain, and requires the collection of corresponding data.

[0110] First, four sets of data need to be collected: 1) image domain data with artifacts, 2) K-space domain data with artifacts, 3) image domain data without artifacts, and 4) K-space domain data without artifacts. It is not required that the data in 1) and 2) and the data in 3) and 4) must correspond one to one. Paired data exist, and the corresponding K-space data can be obtained from the image domain data through inverse Fourier transform according to the equipment parameters.

[0111] Secondly, the artifact data can be further subdivided and divided into light, moderate and severe artifact data according to their severity.

[0112] Finally, a dataset of no, mild, moderate, and severe artifacts was established in the image domain and K-space domain.

[0113] 2. Bidirectional reversible mapping unit: establishment of reversible mapping between artifacts and no artifacts

[0114] Step 1: Define a differentiable and reversible transformation T: y = f(x), From a basic distribution p x The sample u in (x) is transformed by T to obtain a random variable:

[0115]

[0116] So

[0117] Step 2: Use a series of reversible mappings Establish a mapping relationship from the original distribution (data without artifacts x) to the new distribution (data with artifacts y), as follows Figure 1 As shown, Figure 1 A schematic diagram of the standardized flow of reversible mapping, from Z0 to Z k The generation process.

[0118] Step 3: Establish the mapping relationship between artifact-free and artifact-free. A series of reversible mappings transform the original distribution into a new distribution. By optimizing this series of distributions, the purpose of transforming a simple Gaussian distribution into a complex true posterior distribution can be achieved. Bidirectional learning strategies based on dual domain data, such as Figure 2 As shown, Figure 2 Schematic diagram of dual domain learning between image domain and K-space domain.

[0119] Step 4, artifact correction unit: Use the learned standardized flow model for MRI artifact correction. If there is image domain and K-space domain data, perform artifact correction in the relevant domains respectively, and then fuse the correction results. If there is only the result of a single domain, directly give the final correction result. The whole process is as follows: Figure 3 As shown, Figure 3 Flowchart for the entire artifact correction.

[0120] The key points and protection points of the device of the present application are at least:

[0121] 1. Based on the standardized flow, a reversible mapping relationship between artifact-free, mild, moderate, and severe artifact data is established to achieve free conversion between various data.

[0122] 2. Use the information of the dual domain of image domain and K-space domain to correct artifacts and enhance the quality of the final corrected image.

[0123] 3. The present application can be considered as an artifact correction method based on the reversibility of the same mapping, and can also be considered as an artifact generation method based on the reversibility of the same mapping.

[0124] The advantages of the present application device and the existing respiratory artifact correction method are as follows:

[0125] First of all, this application optimizes the artifact correction method based on dual domain constraints. Compared with the traditional single-domain-based respiratory artifact correction method, it has better effect, can effectively restore the image, and has better traceability and interpretability.

[0126] Secondly, based on the reversibility of the standardized flow, compared with other methods that simply perform artifact correction, the present application can simultaneously generate more realistic artifacts compared with other artifact generation methods, and can further expand the data set and improve the generalization of the model.

[0127] Finally, the dual domain of the image domain and the K-space domain is optimized simultaneously, which can provide richer detail information than a single domain and reduce the detail loss of the image after artifact correction.

[0128] The modified design or alternative scheme of the device of this application is:

[0129] In addition to being used for MRI artifact correction, the device can also be used for CT image artifact correction after appropriate transformation.

[0130] The device is further optimized to be applied to the joint task of MRI and CT reconstruction and artifact correction, using flow priors to build a data-driven joint model of image reconstruction and artifact correction.

[0131] Based on the reversibility of the flow model, the device can be used to generate MR image artifacts.

[0132] Example 3

[0133] A storage medium stores a program file capable of implementing any one of the above-mentioned artifact correction methods based on standardized flow.

[0134] Example 4

[0135] A processor is used to run a program, wherein when the program is run, any one of the above-mentioned artifact correction methods based on standardized flow is executed.

[0136] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0137] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0138] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the system embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, 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 units or modules, which can be electrical or other forms.

[0139] 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 over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0140] 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. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0141] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of each embodiment method of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc. Various media that can store program codes.

[0142] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for artifact correction based on standardized flow, characterized in that: The following steps are involved: Learn different artifact types and different levels of flow models for data in image and K-space domains; The standardized flow in the flow model is used to perform dual learning on the data in the image domain and the K-space domain, and a bidirectional reversible mapping between the data with artifacts and the data without artifacts is constructed. Artifact correction of MRI images using bidirectional reversible mapping.

2. The artifact correction method based on standardized flow according to claim 1, characterized in that: The method of learning different artifact types and different levels of flow models for data in the image domain and the K-space domain includes: Collect corresponding data and use deep learning algorithms to perform paired training on the image domain and K-space domain.

3. The artifact correction method based on standardized flow according to claim 2, characterized in that: The collecting of corresponding data and using a deep learning algorithm to perform paired training on the image domain and the K-space domain includes: Four sets of data are collected, namely, image domain data with artifacts, K-space domain data with artifacts, image domain data without artifacts, and K-space domain data without artifacts. Based on the equipment parameters, the image domain data is inversely Fourier transformed to obtain the corresponding K-space data.

4. The artifact correction method based on standardized flow according to claim 2, characterized in that: The learning of different artifact types and different levels of flow models for data in the image domain and the K-space domain also includes: The artifact data is further subdivided according to the degree, and is divided into light, moderate and severe artifact data; Establish artifact-free, mild, moderate, and severe datasets in image and K-space domains.

5. The artifact correction method based on standardized flow according to claim 1, characterized in that: The method of using the standardized flow in the flow model to perform dual learning on the data in the image domain and the K-space domain to construct a bidirectional reversible mapping between the data with artifacts and the data without artifacts includes: Define a differentiable and reversible transformation T: y = f(x), From a basic distribution p x The sample u in (x) is transformed by T to obtain a random variable: Using a series of reversible mappings Establishing a mapping relationship from original distributed artifact-free data x to new distributed artifact-containing data y; The mapping relationship between artifact-free and artifacts of various degrees is established, and a series of reversible mappings transform the original distribution into a new distribution.

6. The artifact correction method based on standardized flow according to claim 1, characterized in that: The artifact correction of the MRI image using bidirectional reversible mapping comprises: The learned normalized flow model is used to correct artifacts in MRI images.

7. The artifact correction method based on standardized flow according to claim 6, characterized in that: The artifact correction of the relevant domains is performed on the image domain and K-space domain data respectively, and then the correction results are fused. If there is only the result of a single domain, the final correction result is directly given.

8. An artifact correction device based on standardized flow, characterized in that: include: A flow model building unit, used for learning different artifact types and different levels of flow models for data in the image domain and the K-space domain; A bidirectional reversible mapping unit, used for performing dual learning on data in the image domain and the K-space domain by using the standardized flow in the flow model, and constructing a bidirectional reversible mapping between data with artifacts and data without artifacts; The artifact correction unit is used to perform artifact correction on the MRI image by using bidirectional reversible mapping.

9. A storage medium, characterized in that: The storage medium stores a program file capable of implementing the artifact correction method based on standardized flow as described in any one of claims 1 to 7.

10. A processor, characterized in that: The processor is used to run a program, wherein the program, when running, executes the artifact correction method based on standardized flow as described in any one of claims 1 to 7.

Images