Method and system for magnetic resonance imaging process

Abstract

According to the present invention, there is provided a method for processing acquired magnetic resonance images, the method comprising the steps of: S1) providing a first magnetic resonance image; S2) providing a second magnetic resonance image; S3) providing a first operator; S4) providing a second operator; and S5) applying the second operator to the first magnetic resonance image and applying the first operator to a magnetic resonance image obtained by applying the second operator to the first magnetic resonance image, thereby deriving a magnetic resonance image instead of the first magnetic resonance image that would be expected if the first magnetic resonance image had been acquired in a first main magnetic field using a magnetic resonance imaging apparatus under shimming conditions for uniformly shimming the first main magnetic field.

Description

[Technical Field]

[0001] The present invention relates to the field of magnetic resonance imaging, and more particularly to the field of processing magnetic resonance images acquired with a magnetic resonance imaging device. [Background technology]

[0002] High-temperature superconductors allow for very rapid ramping of the magnet, making it possible to build magnetic resonance imaging devices with switchable magnetic flux densities in the main magnetic field. This allows for switching the magnetic flux density of the main magnetic field during an examination to other magnetic flux densities in the main magnetic field of the magnetic resonance imaging device. For example, an image is performed with a main magnetic field of 1.5 T. Subsequent switching of the high-temperature superconductor causes a change in magnetic flux density to another main magnetic field of 0.6 T, allowing for the acquisition of another magnetic resonance image at this lower magnetic flux density. However, shimming is still performed so that the main magnetic field at, for example, 1.5 T is uniform. Because no shimming is performed for magnetic flux densities different from the first magnetic flux density, the other main magnetic field will exhibit non-uniformities due to the change in magnetic flux density.

[0003] Shimming is generally the process of providing the greatest possible homogenization of the main magnetic field. A distinction is made between active and passive shimming. Passive shimming is performed by placing and positioning ferromagnetic material inside the magnet that creates the main magnetic field, and helps minimize inhomogeneities caused by the design of the magnetic resonance imaging system. Active shimming can be performed using shim coils provided for this purpose. Active shimming homogenizes inhomogeneities in the main magnetic field caused by patient load. Summary of the Invention [Problem to be solved by the invention]

[0004] Inhomogeneities in the main magnetic field caused by variations in magnetic flux density cause image artifacts in magnetic resonance imaging systems that cannot be sufficiently reduced by active shimming.

[0005] The paper "Feasibility study of novel rapid ramp-down procedure in MgB2 MRI magnet using persistent current switch with high off-resistivity" by Kodama et. al. in Superconductor Science and Technology, 34 (2021) 074003 (13pp) describes a dry magnet with high-temperature superconductors and MgB2, which is equipped with a novel rapid shutdown method that can replace controlled shutdown in an emergency. For this purpose, a power supply is established by a persistent current switch, which is switched off when heated and the power supply is interrupted by a circuit breaker. The energy stored in the solenoid is dissipated in an external resistance.

[0006] Accurate homogenization of the main magnetic field is important in magnetic resonance imaging to reduce inhomogeneity artifacts in the magnetic resonance images. However, with existing rampable magnetic resonance imaging scanners and the ability to vary the main magnetic field, homogenization of the main magnetic field by shimming is not sufficient.

[0007] From US 2020 / 0400764 A1, a magnetic resonance imaging system is known that is configured for imaging near metal implants and in which the magnetic field can be ramped within a clinically acceptable time.

[0008] Summary of the Invention

[0009] It is an object of the present invention to provide a time- and memory-efficient method for reducing the effects of field inhomogeneities in magnetic resonance images. [Means for solving the problem]

[0010] According to the present invention, this object is addressed by the subject matter of the independent claims. Preferred embodiments of the invention are set forth in the dependent claims.

[0011] Therefore, according to the present invention, there is provided a method for processing acquired magnetic resonance images, the method comprising the steps of: providing, by a magnetic resonance imaging device, a first magnetic resonance image acquired in a first main magnetic field having a first magnetic flux density under shimming conditions for uniformly shimming a second main magnetic field, the second main magnetic field having a second magnetic flux density different from the first magnetic flux density; providing, by the magnetic resonance imaging device, a second magnetic resonance image acquired in the second main magnetic field under shimming conditions for uniformly shimming the second main magnetic field; and converting the magnetic resonance image acquired by the magnetic resonance imaging device in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field into a magnetic resonance image expected in the second main magnetic field under shimming conditions for uniformly shimming the second main magnetic field. providing a second operator for converting a magnetic resonance image acquired in the first main magnetic field under shimming conditions for uniformly shimming the second main magnetic field by a magnetic resonance imaging device into a magnetic resonance image expected in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field; and deriving a magnetic resonance image instead of the first magnetic resonance image that would be expected if the first magnetic resonance image had been acquired in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field by applying the second operator to the first magnetic resonance image and applying the first operator to the magnetic resonance image obtained by applying the second operator to the first magnetic resonance image.

[0012] An operator is generally a mapping or function that operates on elements of one space to generate elements of another space. The term can be used as a synonym for function or for the application of a machine learning algorithm to elements of a space. The actual elements in this space are the first and second magnetic resonance images by the respective first and second operators.

[0013] The term "uniformly shimming the main magnetic field" means that such shimming is applied to reduce image distortions due to magnetic field inhomogeneities. However, although such shimming can help to significantly reduce magnetic field inhomogeneities, such shimming cannot achieve a perfectly uniform main magnetic field.

[0014] Furthermore, in practice, it may be impossible to create a perfectly functioning operator, i.e., one that can accurately calculate the expected image from a given image. If this were the case, the second operator alone would be sufficient to reconstruct a correctly shimmed image from an imperfectly shimmed image, i.e., an image acquired under different main magnetic field shimming conditions with a higher or lower magnetic flux density. Thus, according to the present invention, the second operator is used in conjunction with the first operator, the first magnetic resonance image, and the second magnetic resonance image to reconstruct a magnetic resonance image in place of the first magnetic resonance image that would be expected if the first magnetic resonance image had been acquired by the magnetic resonance imaging device in the first main magnetic field with the correct shimming conditions for uniformly shimming the first main magnetic field. Because the ground truth formed by the second magnetic resonance image acquired by the magnetic resonance imaging device in the second main magnetic field with the shimming conditions for uniformly shimming the second main magnetic field is provided (i.e., obtained), the operation of the first operator provides access to the consistency of the operation of the second operator and can be corrected or adapted accordingly. The operation of the first operator can further be used as a constraint on the operation of the second operator to improve the consistency of the results of the action of the first operator.

[0015] This maximizes the capabilities of magnetic resonance imaging devices with switchable magnetic flux densities in a time-efficient and resource-saving approach. Transforming a first magnetic resonance image with two operators provides a time-efficient method for examining a patient using a magnetic resonance imaging device that is switchable between two different modes, i.e., two magnetic flux densities, without the need to adjust passive or active shimming between the two magnetic flux densities. In particular, passive shimming, which involves varying the main magnetic field by introducing ferromagnetic materials, is a resource-intensive shimming method for achieving field homogenization. Using operators to transform magnetic resonance images can open new magnetic resonance imaging possibilities, such as reducing susceptibility artifacts caused by acquiring magnetic resonance images at different magnetic flux densities. Furthermore, forming the operators selects a particularly computationally and memory-efficient approach. Once the operator is constructed, training data for the magnetic resonance imaging device on which the first and second magnetic resonance images are generated is no longer required, and therefore the first operator does not necessarily have to be generated locally on the magnetic resonance imaging device in question, thereby achieving a synergistic effect with the first and second operators.

[0016] According to the present invention, various algorithms can be used for the final step in which the operators are applied. However, according to a preferred embodiment of the present invention, the method for performing the final step comprises a trained machine learning algorithm representing the operation of the first operator and the second operator, which, when inputted with the first and second magnetic resonance images, generates a magnetic resonance image in place of the first magnetic resonance image that would be expected if the first magnetic resonance image had been acquired in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field by the magnetic resonance imaging device.

[0017] This means in particular that when the first operators are applied, the intermediate results are not necessarily available to the user of the system, but remain internal to the trained machine learning algorithm, and only the final result is accessible. These operators may therefore be inaccessible execution steps within the trained machine learning algorithm. For example, in a trained artificial neural network, these operators may be the activation functions at the nodes at each level of the trained artificial neural network.

[0018] In principle, the training data set can include a variety of data. However, according to a preferred embodiment of the present invention, the method further includes a training data set for building a first operator in the trained machine learning algorithm, the training data set including magnetic resonance images acquired with a first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field and magnetic resonance images acquired with a second main magnetic field under shimming conditions for uniformly shimming the second main magnetic field.

[0019] Pairs of magnetic resonance images acquired using a first main magnetic field with shimming conditions for uniformly shimming the first main magnetic field and magnetic resonance images acquired using a second main magnetic field with shimming conditions for uniformly shimming the second main magnetic field are input into a machine learning algorithm to construct a training data set. To this end, in situ and in vivo magnetic resonance images of a humanoid subject, an animal subject, or a specialized magnetic resonance imaging phantom intended as an alternative to in vivo magnetic resonance imaging can be acquired at different magnetic flux densities. The training data set can include respective amplitude and phase images.

[0020] In general, the training data set for the first operator can be generated in various ways. However, according to a preferred embodiment of the present invention, the method further comprises: generating a training data set for constructing the first operator in the trained machine learning algorithm using a generative adversarial network formed from two artificial neural networks.

[0021] Magnetic resonance images acquired using a first main magnetic field with a shimming condition for uniformly shimming the first main magnetic field, and magnetic resonance images acquired using a second main magnetic field with a uniformly shimming condition for the second main magnetic field, can be generated by using a generative adversarial network formed from two artificial neural networks to generate training data for a first operator without requiring additional in vivo or in situ acquisition of magnetic resonance images.

[0022] According to a preferred embodiment of the present invention, the method further comprises generating a training data set for constructing the first operator in the trained machine learning algorithm using electromagnetic simulation.

[0023] Using these electromagnetic simulations, for example, Maxwell's equations for a given problem can be solved in a simulation environment and the first operator can be derived accordingly.

[0024] The training data set for the second operator can be generated in various ways. However, according to a preferred embodiment of the present invention, the method further comprises that the training data set for building the second operator in the trained machine learning algorithm comprises magnetic resonance images of a magnetic resonance imaging phantom acquired using the first main magnetic field with shimming conditions for uniformly shimming the second main magnetic field, and magnetic resonance images of a magnetic resonance imaging phantom acquired using the second main magnetic field with shimming conditions for uniformly shimming the second main magnetic field.

[0025] The training data set thus contains the differences between the first and second magnetic resonance images, thereby containing direct measurements of magnetic field inhomogeneities created by changes in magnetic flux density. The machine learning algorithm can be trained to compensate for these inhomogeneities using a second operator. This effect is inherent to the magnetic resonance imaging device on which the first and second magnetic resonance images were generated.

[0026] Preferably, the trained machine learning algorithm for the second operator is a trained artificial neural network.

[0027] In principle, different methods can be provided for approximating a magnetic resonance image alternative to the first magnetic resonance image that would be expected if the first magnetic resonance image had been acquired in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field by the magnetic resonance imaging device. However, according to a preferred embodiment of the present invention, the method further comprises, in a final step, deriving a magnetic resonance image alternative to the first magnetic resonance image that would be expected if the first magnetic resonance image had been acquired in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field by the magnetic resonance imaging device, by minimizing a loss function that uses the first and second magnetic resonance images and the first and second operators.

[0028] The minimization of the loss function converges to a magnetic resonance image that would be expected to replace the first magnetic resonance image if the first magnetic resonance image were acquired by the magnetic resonance imaging device in a first main magnetic field under shimming conditions that uniformly shim the first main magnetic field. This is a process that can never be perfect and therefore typically proceeds through iterative asymptotic approximations. Thus, the loss function can stop calculating magnetic resonance images that would be expected to replace the first magnetic resonance image if the first magnetic resonance image were acquired by the magnetic resonance imaging device in a first main magnetic field under shimming conditions that uniformly shim the first main magnetic field once a certain user-definable iteration threshold is reached. Similarly, minimization of the loss function can yield a local minimum instead of a global minimum. However, essentially, artifacts in the expected magnetic resonance image are reduced compared to the first magnetic resonance image.

[0029] Different implementations of minimizing the loss function can be employed, however, according to a preferred embodiment of the present invention, the method further comprises that for the loss function LF the following formula is applied: LF = Operator1(Operator2(Image1)) - Image2, where Operator1 is the first operator, Operator2 is the second operator, Image1 is the first magnetic resonance image, and Image2 is the second magnetic resonance image.

[0030] The loss function iteratively minimizes the distance between the first magnetic resonance image and the second magnetic resonance image to which the first operator and the second operator have been applied, and in minimizing the loss function, the first operator and the second operator are input as initial values ​​and can be changed during the iterative minimization using the loss function.

[0031] Generally, different magnetic resonance imaging devices can be used to acquire the magnetic resonance images of the first operator. However, according to a preferred embodiment of the present invention, the first operator is acquired by including magnetic resonance image characteristics of a magnetic resonance image acquired in a first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field and magnetic resonance image characteristics of a magnetic resonance image acquired in a second main magnetic field under shimming conditions for uniformly shimming the second main magnetic field, wherein the magnetic resonance images acquired in the first main magnetic field and the second main magnetic field, respectively, are not acquired by the magnetic resonance imaging device that generated the first magnetic resonance image and the second magnetic resonance image.

[0032] Thus, the first operator is independent of the magnetic resonance imaging device on which the first and second magnetic resonance images were generated, and there is no need to determine the operator for each particular magnetic resonance imaging device, i.e., this first operator only needs to be determined once.

[0033] According to a preferred embodiment of the present invention, the method further includes the second operator being obtained by acquiring magnetic resonance images from a magnetic resonance imaging phantom simulating a patient load in the first main magnetic field and the second main magnetic field, and the difference between the magnetic resonance images of the magnetic resonance imaging phantom acquired in the first main magnetic field and the second main magnetic field includes an effect of inhomogeneity of the first main magnetic field having a first magnetic flux density under shimming conditions for uniformly shimming the second main magnetic field.

[0034] Further according to the invention there is provided a computer program for magnetic resonance imaging correction comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method described above.

[0035] The present invention further provides a system of a magnetic resonance imaging apparatus and a computing unit, wherein the magnetic resonance imaging apparatus comprises a switch capable of switching main magnetic fields, wherein a first main magnetic field has a first magnetic flux density with shimming conditions for uniformly shimming a second main magnetic field, and wherein the second main magnetic field has a second magnetic flux density with shimming conditions for uniformly shimming the second main magnetic field, the second magnetic flux density being different from the first magnetic flux density, and wherein the computing unit is adapted to perform the above-mentioned method.

[0036] By varying the magnetic flux density to a lower or higher magnetic flux density, artifacts in magnetic resonance images that are dependent on magnetic flux density can be compensated for.

[0037] In principle, different superconducting magnets can be used, however, according to a preferred embodiment of the present invention, superconducting magnets are provided for generating the first and second main magnetic fields, and the superconducting magnets are dry magnets.

[0038] Preferably, the superconducting magnets used to generate the first and second main magnetic fields are high temperature superconductors.

[0039] High temperature superconductors allow the superconducting magnet to rapidly switch between a first magnetic flux density and a second magnetic flux density.

[0040] Regarding the details of training (i) the first operator that converts images at a magnetic field / shim setting of one magnetic field strength to images at a magnetic field / shim setting of another magnetic field strength, and (ii) the second operator that converts images at a magnetic field setting with an uncorresponding shim setting to images at the same magnetic field setting and its corresponding shim setting, these first and second operators can be trained based on ground truth annotated training image pairs that are converted into other images by the respective first and second operators, or based on ground truth obtained by computer simulation. These operators can be implemented as trained neural networks to return the desired image from an input image. The second operator can be trained, for example, on a large set of image pairs of a phantom with each magnetic field ramp and associated correct shim setting, or by Bloch simulation. The first operator can be implemented as a trained neural network by training images of a phantom model with one magnetic field ramp and correct / incorrect shim settings. Alternatively, it is sufficient to collect data with incorrect shim settings, and from knowledge of the coding distortion introduced by the "wrong" shim settings, corresponding images with an equal field ramp and correct shim settings can be generated.

[0041] Real data can be used to train the described TH (conversion between contrasts at different magnetic field strengths) and US (unscrambling) networks. Generally, the more data provided for training, the better the results. To improve training without the need to measure more real data, it is possible to artificially enhance the training dataset using a GAN (generative adversarial network). Although the GAN itself needs to be trained, once trained, for example, based on a limited initial training set, it can generate a virtually infinite number of TH and US feature / label pairs that can be used to train the above-described TH and US networks.

[0042] Even more preferably, the superconducting magnets used to generate the first and second main magnetic fields are magnesium diboride magnets.

[0043] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter, but such embodiments do not necessarily represent the full scope of the invention, and reference should therefore be made to the claims and this specification for interpreting the scope of the invention. [Brief explanation of the drawings]

[0044] [Figure 1] 1 shows a schematic scheme of a method according to a preferred embodiment of the present invention. [Figure 2] 1 shows a schematic representation of a system comprising a magnetic resonance imaging device and a computing unit according to a preferred embodiment of the present invention; DETAILED DESCRIPTION OF THE INVENTION

[0045] Figure 1 shows a schematic scheme of a method according to a preferred embodiment of the present invention, which comprises steps S1 to S5.

[0046] S1: First, a magnetic resonance imaging apparatus 1 provides a first magnetic resonance image acquired in a first main magnetic field having a first magnetic flux density under shimming conditions for uniformly shimming a second main magnetic field, the second main magnetic field having a second magnetic flux density different from the first magnetic flux density.

[0047] S2: Secondly, a second magnetic resonance image acquired by the magnetic resonance imaging apparatus 1 in the second main magnetic field under shimming conditions for uniformly shimming the second main magnetic field is provided.

[0048] S3: Third, a first operator is provided for converting a magnetic resonance image acquired in a first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field by the magnetic resonance imaging device 1 into a magnetic resonance image expected in a second main magnetic field under shimming conditions for uniformly shimming the second main magnetic field.

[0049] S4: Fourth, a second operator is provided that converts a magnetic resonance image acquired in the first main magnetic field under shimming conditions for uniformly shimming the second main magnetic field by the magnetic resonance imaging device 1 into a magnetic resonance image expected in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field.

[0050] S5: Fifth, by applying the second operator to the first magnetic resonance image and applying the first operator to the magnetic resonance image obtained by applying the second operator to the first magnetic resonance image, a magnetic resonance image alternative to the first magnetic resonance image is derived that would be expected if the first magnetic resonance image had been acquired in the first main magnetic field using the magnetic co-imaging device 1 under shimming conditions for uniformly shimming the first main magnetic field.

[0051] The first operator is generated by a trained machine learning algorithm using magnetic resonance images acquired in a first main magnetic field with shimming conditions for uniformly shimming the first main magnetic field and magnetic resonance images acquired in a second main magnetic field with shimming conditions for uniformly shimming the second main magnetic field, which images were not acquired by the magnetic resonance imaging device 1 from which the first and second magnetic resonance images were generated. This first operator is inferred only once and is valid for a given set of first and second magnetic flux densities and for the set of predetermined scan parameters for which they were determined. If the scan parameters are changed, for example, if the repetition time (TR) or echo time (TE) is changed and therefore a different image contrast is achieved, a new determination is required.

[0052] The second operator is also generated by a machine learning algorithm trained using training data, the training data including magnetic resonance images of a magnetic resonance imaging phantom 5 for simulating a patient load acquired in a first main magnetic field with shimming conditions for uniformly shimming the second main magnetic field, and magnetic resonance images of the magnetic resonance imaging phantom 5 acquired in the second main magnetic field with shimming conditions for uniformly shimming the second main magnetic field for each set of predetermined scan parameters.

[0053] Next, the first and second operators are implemented in a loss function, and by minimizing this loss function using the first and second magnetic resonance images and the first and second operators, a magnetic resonance image that replaces the first magnetic resonance image is derived, which would be expected if the first magnetic resonance image were acquired by the magnetic resonance imaging device 1 in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field.

[0054] 2 shows a schematic diagram of a system including a magnetic resonance imaging apparatus 1 and a computing unit 2 according to a preferred embodiment of the present invention. A main magnetic field in the bore of the magnetic resonance imaging apparatus 1 is generated by a superconducting magnet 6, which is a high-temperature superconductor. The magnetic flux density of the main magnetic field generated by the superconducting magnet 6 is switchable by a switch 7 between a first main magnetic field having a first magnetic flux density in a shimming condition for uniformly shimming the second main magnetic field, and a second main magnetic field having a second magnetic flux density different from the first magnetic flux density in a shimming condition for uniformly shimming the second main magnetic field.

[0055] The magnetic resonance imaging phantom 5 is placed on the treatment couch 4 of the patient positioning system 3. The switch 7 is connected to the calculation unit 2 and can be controlled by the calculation unit 2. Furthermore, the calculation unit 2 is configured to perform the method of steps S1 to S5 to acquire a magnetic resonance image replacing the first magnetic resonance image that would be expected if the first magnetic resonance image were acquired in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field by the magnetic resonance imaging apparatus, by applying a second operator to the first magnetic resonance image and applying a first operator to the magnetic resonance image obtained by applying the second operator to the first magnetic resonance image, and then to calculate a magnetic resonance image by applying the first operator to the first magnetic resonance image and the second operator to the second magnetic resonance image.

[0056] While the invention has been illustrated and described in detail in the drawings and the foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive. The invention is not limited to the disclosed embodiments. Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. Any reference signs in the claims are not to be construed as limiting the scope. Moreover, for the sake of clarity, not all elements in the drawings have been labeled with reference signs. [Explanation of symbols]

[0057] 1. Magnetic resonance imaging device 2 Computational Units 3 Patient Positioning Systems 4 Treatment table 5. Magnetic Resonance Imaging Phantom 6 Superconducting magnets 7 Switch

Claims

1. 1. A method for processing acquired magnetic resonance images, comprising: S1) providing a first magnetic resonance image acquired by a magnetic resonance imaging device with a first main magnetic field having a first magnetic flux density under shimming conditions for uniformly shimming a second main magnetic field, the second main magnetic field having a second magnetic flux density different from the first magnetic flux density; S2) providing a second magnetic resonance image acquired by the magnetic resonance imaging device in the second main magnetic field under shimming conditions for uniformly shimming the second main magnetic field; S3) providing a first operator for transforming a magnetic resonance image acquired by the magnetic resonance imaging device in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field into a magnetic resonance image expected in the second magnetic field under shimming conditions for uniformly shimming the second main magnetic field; S4) providing a second operator for transforming a magnetic resonance image acquired by the magnetic resonance imaging device in the first main magnetic field under shimming conditions for uniformly shimming the second main magnetic field into a magnetic resonance image expected in the first magnetic field under shimming conditions for uniformly shimming the first main magnetic field; S5) applying the second operator to the first magnetic resonance image and applying the first operator to the magnetic resonance image obtained by applying the second operator to the first magnetic resonance image, thereby deriving a magnetic resonance image that replaces the first magnetic resonance image, the magnetic resonance image being expected if the first magnetic resonance image had been acquired by the magnetic resonance imaging apparatus under shimming conditions for uniformly shimming the first main magnetic field with the first main magnetic field; A method having the following.

2. 2. The method of claim 1, wherein, to perform step S5, a trained machine learning algorithm representing the operations of the first operator and the second operator is provided, and the trained machine learning algorithm, upon input of the first magnetic resonance image and the second magnetic resonance image, generates the magnetic resonance image replacing the first magnetic resonance image that would be expected if the first magnetic resonance image were acquired in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field by the magnetic resonance imaging device.

3. 3. The method of claim 2, wherein a training data set for constructing the first operator in the trained machine learning algorithm comprises the magnetic resonance images acquired in the first main magnetic field with a shimming condition for uniformly shimming the first main magnetic field and the magnetic resonance images acquired using the second main magnetic field with a shimming condition for uniformly shimming the second main magnetic field.

4. 4. The method of claim 3, wherein the training data set for constructing the first operator in the trained machine learning algorithm is generated using a generative adversarial network formed from two artificial neural networks.

5. The method of any one of claims 2 to 4, wherein a training data set for constructing the first operator in the trained machine learning algorithm is generated using electromagnetic simulation.

6. 6. The method of claim 2, wherein a training data set for constructing the second operator in the trained machine learning algorithm comprises magnetic resonance images of a magnetic resonance imaging phantom for simulating a patient load acquired in the first main magnetic field with a shimming condition for uniformly shimming the second main magnetic field, and magnetic resonance images of the magnetic resonance imaging phantom acquired in the second main magnetic field with a shimming condition for uniformly shimming the second main magnetic field.

7. 7. The method according to claim 1, wherein in step S5, the magnetic resonance image replacing the first magnetic resonance image that would be expected if the first magnetic resonance image were acquired in the first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field by the magnetic resonance imaging device is derived by minimizing a loss function that uses the first magnetic resonance image, the second magnetic resonance image, a first operator, and a second operator.

8. Regarding the loss function (LF), LF = Operator1(Operator2(Image1)) - Image2 8. The method of claim 7, wherein Operator1 is a first operator, Operator2 is a second operator, Image1 is a first magnetic resonance image, and Image2 is a second magnetic resonance image.

9. 8. The method according to claim 1, wherein the first operator is obtained by including magnetic resonance image characteristics of a magnetic resonance image acquired in a first main magnetic field under shimming conditions for uniformly shimming the first main magnetic field and magnetic resonance image characteristics of a magnetic resonance image acquired in a second main magnetic field under shimming conditions for uniformly shimming the second main magnetic field, and wherein the magnetic resonance images acquired in each of the first main magnetic field and the second main magnetic field are not acquired by the magnetic resonance imaging device by which the first magnetic resonance image and the second magnetic resonance image are generated.

10. 10. The method of claim 1, wherein the second operator is obtained by acquiring magnetic resonance images from a magnetic resonance imaging phantom simulating a patient load in the first main magnetic field and the second main magnetic field, and a difference between the acquired magnetic resonance images of the magnetic resonance imaging phantom in the first main magnetic field and the second main magnetic field includes an effect of inhomogeneity of the first main magnetic field having the first magnetic flux density at a shimming condition for uniformly shimming the second main magnetic field.

11. A computer program for magnetic resonance imaging correction, said computer program comprising instructions which, when said computer program is executed by a computer, cause said computer to carry out the method according to any one of claims 1 to 10.

12. A system comprising a magnetic resonance imaging device and a computing unit, the magnetic resonance imaging apparatus includes a switch capable of switching a main magnetic field, the first main magnetic field having a first magnetic flux density under a shimming condition for uniformly shimming the second main magnetic field, the second main magnetic field having a second magnetic flux density under a shimming condition for uniformly shimming the second main magnetic field, the second magnetic flux density being different from the first magnetic flux density; A system, wherein the computing unit is configured to perform the method according to any one of claims 1 to 11.

13. 13. The system of claim 12, wherein a superconducting magnet is provided to generate the first main magnetic field and the second main magnetic field, and the superconducting magnet is a dry magnet.

Images