Method for acquiring magnetic resonance images by using artificial neural network model and method for controlling program therefor

Abstract

The objective of one embodiment of the present disclosure, in comparison to a method for controlling a program for acquiring output magnetic resonance images by using a mathematical model, is to provide a method for controlling a program, performed by a computing device including at least one processor and for acquiring magnetic resonance images by using an artificial neural network model, the method comprising the steps of: using an artificial neural network model so as to set a higher PAT factor applied to input magnetic resonance images or an input magnetic resonance imaging signals to be input for acquisition of output magnetic resonance images, and to set a smaller number of reference lines applied to the input magnetic resonance images or the input magnetic resonance imaging signals; and restoring, through the artificial neural network model, the input magnetic resonance images or the input magnetic resonance imaging signals so as to acquire output magnetic resonance images.

Description

Method for acquiring magnetic resonance images using an artificial neural network model and a method for controlling the program thereof

[0001] The present disclosure relates to a method for acquiring magnetic resonance images using an artificial neural network model and a method for controlling the program.

[0002] X-ray scanners, ultrasound diagnostic devices, computed tomography (CT) scanners, and Magnetic Resonance Imaging (MRI) scanners are utilized to acquire internal body information for observing and diagnosing the inside of a patient's body. Among these, MRI is currently receiving attention for its utility compared to other scanning technologies, particularly because it allows scanning without exposing the patient to radiation or administering contrast agents, and offers high resolution and excellent soft tissue contrast.

[0003] Magnetic Resonance Imaging (MRI) technology faces the problem of requiring a long time to acquire images. Consequently, active research on accelerated scanning technology has been conducted in this field to reduce scanning times. However, MRI images acquired through accelerated scanning technology suffer from low resolution or noise, making accurate analysis difficult. In particular, situations can arise where internal bodily information about the patient is lost from the MRI.

[0004] Ultimately, solving the dual problem of achieving accelerated magnetic resonance imaging (MRI) scans while simultaneously acquiring high-quality images has been a long-standing challenge in this technical field. Artificial intelligence (AI) technology has been proposed as a solution to this. Specifically, this involves using AI models to restore the quality of MRI images acquired through accelerated scanning technology.

[0005] A method for acquiring magnetic resonance images using an artificial neural network and a program control method according to one embodiment of the present disclosure aim to shorten the magnetic resonance image acquisition time while providing magnetic resonance images with excellent resolution, contrast, and SNR.

[0006] The present invention provides a method for controlling a program for acquiring a magnetic resonance image using an artificial neural network model, which is performed by a computing device including at least one processor according to one embodiment of the present disclosure for realizing the aforementioned problem, wherein, compared with a method for controlling a program for acquiring an output magnetic resonance image using a mathematical model, the method comprises: a step of setting a higher PAT factor applied to an input magnetic resonance image or an input magnetic resonance image signal input for acquiring an output magnetic resonance image using an artificial neural network model; and a step of acquiring an output magnetic resonance image by restoring the input magnetic resonance image or the input magnetic resonance signal through the artificial neural network model.

[0007] Alternatively, the present invention provides a method comprising the step of setting the TR (Time to repeat) applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value.

[0008] Alternatively, the present invention provides a method comprising the step of setting the phase resolution applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value.

[0009] Alternatively, the present invention provides a method comprising the step of setting the number of reference lines applied to the input magnetic resonance image or the input magnetic resonance signal to be smaller.

[0010] Alternatively, the present invention provides a method comprising the step of applying a partial Fourier factor in the phase direction to the input magnetic resonance image or the input magnetic resonance signal.

[0011] Alternatively, the present invention provides a method comprising the step of setting the acquisition voxel phase applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value.

[0012] Alternatively, the present invention provides a method comprising the step of setting a turbo factor to be applied to the input magnetic resonance image or the input magnetic resonance signal.

[0013] Alternatively, the present invention provides a method comprising the step of setting the oversampling applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value.

[0014] Alternatively, the present invention provides a method comprising the step of setting the echo spacing applied to the input magnetic resonance image or the input magnetic resonance signal to be shorter.

[0015] The present invention provides a method for controlling a program for acquiring a magnetic resonance image using an artificial neural network model, which is performed by a computing device comprising at least one processor according to another embodiment of the present invention, wherein, compared with a method for controlling a program for acquiring an output magnetic resonance image using a mathematical model, the method comprises the step of setting an average applied to an input magnetic resonance image or an input magnetic resonance image signal input for acquiring an output magnetic resonance image using an artificial neural network model to be lower; and the step of acquiring an output magnetic resonance image by restoring the input magnetic resonance image or the input magnetic resonance signal through the artificial neural network model.

[0016] Alternatively, the present invention provides a method comprising the step of setting the TR (Time to repeat) applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value.

[0017] Alternatively, the present invention provides a method comprising the step of setting the phase resolution applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value.

[0018] Alternatively, the present invention provides a method comprising the step of setting the number of reference lines applied to the input magnetic resonance image or the input magnetic resonance signal to be smaller.

[0019] Alternatively, the present invention provides a method comprising the step of applying a partial Fourier factor in the phase direction to the input magnetic resonance image or the input magnetic resonance signal.

[0020] Alternatively, the present invention provides a method comprising the step of setting the acquisition voxel phase applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value.

[0021] Alternatively, the present invention provides a method comprising the step of setting a turbo factor to be applied to the input magnetic resonance image or the input magnetic resonance signal.

[0022] Alternatively, the present invention provides a method comprising the step of setting the oversampling applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value.

[0023] Alternatively, the present invention provides a method comprising the step of setting the echo spacing applied to the input magnetic resonance image or the input magnetic resonance signal to be shorter.

[0024] The present invention provides a computing device for executing a program to acquire a magnetic resonance image using an artificial neural network model according to another embodiment of the present disclosure, comprising at least one processor; and a memory in which the program is stored, wherein the processor, in accordance with the execution of the program, causes an output magnetic resonance image to be acquired using a mathematical model or an output magnetic resonance image to be acquired using an artificial neural network model, wherein when the output magnetic resonance image is acquired using the artificial neural network model, a fat factor applied to an input magnetic resonance image or an input magnetic resonance signal input for acquiring the output magnetic resonance image is set higher than the fat factor applied when the output magnetic resonance image is acquired using the mathematical model, and the output magnetic resonance image is acquired by restoring the input magnetic resonance image or the input magnetic resonance signal through the artificial neural network model.

[0025] In a computing device for executing a program to acquire a magnetic resonance image using an artificial neural network model according to another embodiment of the present disclosure, the device comprises at least one processor and a memory in which the program is stored, wherein the processor, in accordance with the execution of the program, causes an output magnetic resonance image to be acquired using a mathematical model or an output magnetic resonance image to be acquired using an artificial neural network model, and when the output magnetic resonance image is acquired using the artificial neural network model, the average applied to the input magnetic resonance image or input magnetic resonance signal input for acquiring the output magnetic resonance image is set lower than the average applied when the output magnetic resonance image is acquired using the mathematical model, and the device acquires an output magnetic resonance image by restoring the input magnetic resonance image or the input magnetic resonance signal through the artificial neural network model.

[0026] A method for acquiring magnetic resonance images using an artificial neural network and a program control method according to one embodiment of the present disclosure can shorten the magnetic resonance image acquisition time while providing magnetic resonance images with excellent resolution, contrast, and SNR.

[0027] FIG. 1 is a block diagram of a computing device for acquiring magnetic resonance images based on a deep learning model according to one embodiment of the present disclosure.

[0028] FIG. 2 is a flowchart showing a method for acquiring magnetic resonance images using an artificial neural network model according to one embodiment of the present disclosure.

[0029] FIG. 3 is a drawing for explaining an artificial neural network model according to one embodiment of the present disclosure.

[0030] FIG. 4 is a schematic diagram illustrating the difference between full sampling and subsampling according to one embodiment of the present invention.

[0031] FIGS. 5a to 5c are drawings showing a comparison of brain magnetic resonance images output according to whether oversampling is applied according to an embodiment of the present invention.

[0032] FIGS. 6a to 6c are drawings showing a comparison of brain magnetic resonance images output by varying the average according to one embodiment of the present disclosure.

[0033] FIGS. 7a to 7c are drawings showing a comparison of brain magnetic resonance images output according to a reference line number setting according to one embodiment of the present disclosure.

[0034] Embodiments of the present disclosure are described below with reference to the attached drawings so that those skilled in the art (hereinafter, those skilled in the art) can easily implement them. The embodiments presented in the present disclosure are provided to enable those skilled in the art to use or implement the contents of the present disclosure. Accordingly, various modifications to the embodiments of the present disclosure will be apparent to those skilled in the art. That is, the present disclosure may be embodied in various different forms and is not limited to the embodiments below.

[0035] Throughout the specification of the present disclosure, identical or similar reference numerals refer to identical or similar components. Additionally, to clearly explain the present disclosure, reference numerals in the drawings that are unrelated to the description of the present disclosure may be omitted.

[0036] The term “or” as used in this disclosure is intended to mean an implicit “or” rather than an exclusive “or.” That is, unless otherwise specified in this disclosure or its meaning is unclear from the context, “X uses A or B” should be understood to mean one of the natural implicit substitutions. For example, unless otherwise specified in this disclosure or its meaning is unclear from the context, “X uses A or B” may be interpreted as X using A, X using B, or X using both A and B.

[0037] The term “and / or” as used in this disclosure should be understood to refer to and include all possible combinations of one or more of the enumerated related concepts.

[0038] The terms “comprising” and / or “comprising” as used in this disclosure should be understood to mean the presence of certain features and / or components. However, the terms “comprising” and / or “comprising” should be understood not to exclude the presence or addition of one or more other features, other components and / or combinations thereof.

[0039] Where not otherwise specified in the present disclosure or where it is not clear from the context that the singular form indicates, the singular should generally be interpreted as including “one or more.”

[0040] The term "the N (N is a natural number)" used in this disclosure may be understood as an expression used to distinguish the components of this disclosure from one another according to certain criteria, such as functional perspectives, structural perspectives, or convenience of explanation. For example, components performing different functional roles in this disclosure may be distinguished as a first component or a second component. However, components that are substantially identical within the technical scope of this disclosure but need to be distinguished for the convenience of explanation may also be distinguished as a first component or a second component.

[0041] The term “acquisition” as used in this disclosure may be understood to mean not only receiving data through a wired or wireless communication network with an external device or system, but also generating data in an on-device form.

[0042] Meanwhile, the terms "module" or "unit" used in this disclosure may be understood as referring to an independent functional unit that processes computing resources, such as a computer-related entity, firmware, software or a part thereof, hardware or a part thereof, or a combination of software and hardware. In this case, "module" or "unit" may be a unit composed of a single element, or a unit expressed as a combination or set of multiple elements. For example, in a narrow sense, "module" or "unit" may refer to a hardware element of a computing device or a set thereof, an application program that performs a specific function of software, a procedure implemented through software execution, or a set of instructions for program execution. Furthermore, in a broad sense, "module" or "unit" may refer to the computing device itself that constitutes the system, or an application executed on the computing device. However, since the above-described concept is merely an example, the concepts of "module" or "part" may be defined in various ways within the scope understandable to those skilled in the art based on the contents of this disclosure.

[0043] As used in this disclosure, the term "model" may be understood as a system implemented using mathematical concepts and language to solve a specific problem, a set of software units to solve a specific problem, or an abstract model regarding a processing process to solve a specific problem. For example, a neural network "model" may refer to an overall system implemented as a neural network that possesses problem-solving capabilities through learning. In this case, the neural network may possess problem-solving capabilities by optimizing parameters connecting nodes or neurons through learning. A neural network "model" may include a single neural network or a set of neural networks composed of multiple neural networks.

[0044] The term “data” as used in this disclosure may include images, signals, etc. The term “image” as used in this disclosure may refer to multidimensional data composed of discrete image elements. In other words, “image” may be understood as a term referring to a digital representation of an object visible to the human eye. For example, “image” may refer to multidimensional data composed of elements corresponding to pixels in a two-dimensional image. “Image” may refer to multidimensional data composed of elements corresponding to voxels in a three-dimensional image.

[0045] For example, the image may include, but is not limited to, medical images acquired by medical imaging devices such as magnetic resonance imaging (MRI) scanners, computed tomography (CT) scanners, ultrasound scanners, or X-ray scanners.

[0046] As used in this disclosure, the term "medical image" is a general concept encompassing all forms of images that include medical knowledge, and may include images acquired through various modalities such as visible light cameras, IR cameras, ultrasound, X-ray, CT, MRI, PET, etc.

[0047] As used in this disclosure, the term "picture archiving and communication system (PACS)" may refer to a system that stores, processes, and transmits medical images in accordance with the Digital Imaging and Communications in Medicine (DICOM) standard. For example, the "picture archiving and communication system" may be linked with digital medical imaging scanning equipment to store medical images, such as magnetic resonance imaging (MRI) and computed tomography (CT) images, in accordance with the Digital Imaging and Communications in Medicine standard. The "picture archiving and communication system" may transmit medical images to terminals inside or outside the hospital via a communication network. At this time, meta information, such as interpretation results and medical records, may be added to the medical images.

[0048] As used in this disclosure, the term "object" refers to a subject of scanning and may include a person, an animal, or a part thereof. For example, an object may include a part of the body (such as an organ or tissue) or a phantom. A phantom refers to a substance having a volume that is very close to the density and effective atomic number of a living organism, and may include a spherical phantom having properties similar to those of the body.

[0049] A Magnetic Resonance Imaging (MRI) system is a system that acquires images of cross-sectional areas of an object by expressing the intensity of a Magnetic Resonance (MR) signal in response to a Radio Frequency (RF) signal generated in a magnetic field of a specific intensity as contrast.

[0050] The MRI system causes the main magnet to form a static magnetic field and aligns the direction of the magnetic dipole moment of specific atomic nuclei of an object located within the static field with the direction of the static field. A gradient field coil applies a gradient signal to the static field to form a gradient field, thereby inducing different resonance frequencies for different parts of the object. An RF coil can irradiate magnetic resonance signals in accordance with the resonance frequency of the area where image acquisition is desired. Additionally, as the gradient field is formed, the RF coil can receive magnetic resonance signals of different resonance frequencies radiated from various parts of the object. The MRI system acquires images by applying image reconstruction techniques to the magnetic resonance signals received through these steps. Furthermore, the MRI system may reconstruct multiple magnetic resonance signals into image data by performing serial or parallel signal processing on multiple magnetic resonance signals received by a multi-channel RF coil.

[0051] The explanation of the foregoing terms is intended to aid in understanding the present disclosure. Accordingly, it should be noted that unless a foregoing term is explicitly stated as a matter limiting the content of the present disclosure, it is not to be used in the sense of limiting the technical concept of the content of the present disclosure.

[0052] FIG. 1 is a block diagram of a computing device for acquiring magnetic resonance images using an artificial neural network model according to one embodiment of the present disclosure.

[0053] Referring to FIG. 1, a computing device (100) for acquiring magnetic resonance images according to one embodiment of the present disclosure may be a hardware device or part of a hardware device that performs comprehensive processing and computation of data, or it may be a software-based computing environment connected to a communication network. For example, the computing device (100) may be a server that performs intensive data processing functions and is an entity that shares resources, or it may be a client that shares resources through interaction with the server. Additionally, the computing device (100) may be a cloud system that enables multiple servers and clients to interact to comprehensively process data. Since the above description is merely one example regarding the type of computing device (100), the type of computing device (100) may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure. As an example, the computing device (100) may include a smartphone, tablet PC, PC, smart TV, micro server, cloud server, etc., that processes magnetic resonance images or performs processing functions. As another example, the computing device (100) may be a magnetic resonance imaging (MRI) device that directly acquires magnetic resonance images.

[0054] The magnetic resonance signal may be K-space data, and the magnetic resonance image may be a two-dimensional or three-dimensional image obtained through an inverse Fourier operation on the magnetic resonance signal. The computing device (100) may obtain pulse sequence data obtained by each of the multiple electronic devices from the multiple electronic devices. Here, the pulse sequence data may include K-space data collected based on a specific pulse sequence used by the other electronic device. The pulse sequence data may include two-dimensional pulse sequence data collected in a two-dimensional space or three-dimensional pulse sequence data collected in a three-dimensional space.

[0055] Here, magnetic resonance imaging and magnetic resonance signals can be included in and transmitted / received within DICOM (Digital Imaging and Communications in Medicine) data. DICOM stands for Digital Imaging and Communications in Medicine, which is a collective term for various standards used for digital image representation and communication in medical devices. DICOM data may primarily contain patient information and media characteristics. For example, the various medical information data included in DICOM data consists of patient-related text information and raw media information collected at the medical site, with no specific restrictions on their format. More specifically, DICOM data may include the patient's vital signs, image information regarding the patient or treatment site generated at the medical site (e.g., magnetic resonance imaging), and information about the device that acquired the images.

[0056] A computing device (100) may include at least one processor (110, hereinafter referred to as a processor), a communication interface (120), and a memory (130). However, since FIG. 1 is merely an example, the computing device (100) may include other configurations for implementing a computing environment. Additionally, only some of the disclosed configurations may be included in the computing device (100).

[0057] A processor (110) according to one embodiment of the present disclosure may be understood as a constituent unit comprising hardware and / or software for performing computing operations. For example, the processor (110) may read a computer program and perform data processing for machine learning. The processor (110) may process computational processes such as processing input data for machine learning, extracting features for machine learning, and calculating errors based on backpropagation. A processor (110) for performing such data processing may include a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), a tensor processing unit (TPU), an application specific integrated circuit (ASIC), or a field programmable gate array (FPGA). Since the above-described type of processor (110) is merely an example, the type of processor (110) may be configured in various ways within a range understandable to a person skilled in the art based on the contents of the present disclosure.

[0058] The processor (110) is electrically connected to other components of the computing device (100) (i.e., communication interface (120), and memory (130)) to control the overall operation of the computing device (100).

[0059] A memory (130) according to one embodiment of the present disclosure may be understood as a unit comprising hardware and / or software for storing and managing data processed in a computing device (100). That is, the memory (130) may store data of any form generated or determined by a processor (110) and data of any form received by a communication interface (120). For example, the memory (130) may include at least one type of storage medium among a flash memory type (130), a hard disk type, a multimedia card micro type, a card type memory, RAM (random access memory), SRAM (static random access memory), ROM (read-only memory), EEPROM (electrically erasable programmable read-only memory), PROM (programmable read-only memory), magnetic memory, a magnetic disk, and an optical disk. Additionally, the memory (130) may include a database system that controls and manages data in a predetermined system. Since the above-described type of memory (130) is merely an example, the type of memory (130) may be configured in various ways within a range understandable to those skilled in the art based on the contents of this disclosure.

[0060] The memory (130) can structure and organize data, combinations of data, and program code executable by the processor (110) that are necessary for the processor (110) to perform calculations. For example, the memory (130) may store a neural network model, a training data set, and a set of setting values ​​for magnetic resonance imaging scans. Additionally, the memory (130) may store program code that operates the neural network model to perform training based on the training data set and the set of setting values ​​for magnetic resonance imaging scans, program code that operates the neural network model to receive magnetic resonance images as input and perform inference according to the purpose of use of the computing device (100), and processed data generated as the program code is executed.

[0061] A communication interface (120) according to one embodiment of the present disclosure may be understood as a configuration unit that transmits and receives data through any known form of wired or wireless communication system. For example, the communication interface (120) may perform data transmission and reception using wired or wireless communication systems such as a local area network (LAN), wideband code division multiple access (WCDMA), long term evolution (LTE), wireless broadband internet (WiBro), 5th generation mobile communication (5G), ultrawide-band wireless communication, ZigBee, radio frequency (RF) communication, wireless LAN, wireless fidelity (Wi-Fi), near field communication (NFC), or Bluetooth. Since the communication systems described above are merely examples, wired or wireless communication systems for data transmission and reception of the communication interface (120) may be applied in various ways other than those described above.

[0062] The communication interface (120) can receive data necessary for the processor (110) to perform calculations through wired or wireless communication with any system or any client. Additionally, the communication interface (120) can transmit data generated through the calculations of the processor (110) through wired or wireless communication with any system or any client. For example, the communication interface (120) can receive medical data through communication with a database within a hospital environment, a cloud server performing tasks such as standardization of medical data, or a computing device (100). The communication interface (120) can transmit output data of the neural network model (10), and intermediate data and processed data derived during the calculation process of the processor (110), through communication with the aforementioned database, server, or computing device (100).

[0063] According to a method for acquiring a magnetic resonance image according to one embodiment of the present disclosure, a computing device (100) can transmit a magnetic resonance signal or a setting value for acquiring a magnetic resonance image to a magnetic resonance image scanning device using a communication interface (120). The computing device (100) can allow a user to transmit a setting value to the magnetic resonance image scanning device in advance before the magnetic resonance image scanning, and enable the magnetic resonance image shooting to be performed according to the setting value.

[0064] FIG. 2 is a flowchart showing a method for acquiring magnetic resonance images using an artificial neural network model according to one embodiment of the present disclosure.

[0065] Referring to FIG. 2, according to one embodiment of the present disclosure, a processor (110) may input setting value information for acquiring a magnetic resonance signal (S310). Here, the magnetic resonance signal may be acquired through a magnetic field generated by the computing device (100) against an object (e.g., a patient), or it may be acquired from a plurality of other electronic devices that acquire magnetic resonance images through a communication interface (120). For example, the computing device (100) may acquire pulse sequence data acquired by each of the plurality of other electronic devices as a magnetic resonance signal. At this time, the pulse sequence data may include K-space data collected based on a specific pulse sequence used by the other electronic device. The pulse sequence data may include two-dimensional pulse sequence data collected in a two-dimensional space or three-dimensional pulse sequence data collected in a three-dimensional space.

[0066] And the processor (110) can control the acquisition of a magnetic resonance signal to which a set value is applied (S320). The processor (110) may also acquire a magnetic resonance signal by receiving DICOM (Digital Imaging and Communications in Medicine) data from another electronic device. Specifically, the processor (110) may acquire a magnetic resonance signal through a Discrete Fourier Transformation after extracting a magnetic resonance signal included in the DICOM data or extracting a magnetic resonance image. In addition, the processor (110) may transmit various set values ​​applied during scanning to another electronic device, such as a magnetic resonance imaging scan device, or instruct the input of set values, and the MRI scan may proceed in a state set by these set values.

[0067] And the processor (110) can convert the acquired magnetic resonance signal into a magnetic resonance image (S330). For example, the processor (110) can convert the magnetic resonance signal into a magnetic resonance image using an inverse Fourier transform. Alternatively, the processor (110) can convert the magnetic resonance signal into a magnetic resonance image using an artificial neural network model.

[0068] FIG. 3 is a drawing for explaining an artificial neural network model according to one embodiment of the present disclosure.

[0069] Referring to FIG. 3, the artificial neural network model (500) may include at least one neural network. The neural network may include network models such as a Deep Neural Network (DNN), a Recurrent Neural Network (RNN), a Bidirectional Recurrent Deep Neural Network (BRDNN), a Multilayer Perceptron (MLP), and a Convolutional Neural Network (CNN), but is not limited thereto. The artificial neural network model (500) may be, for example, a deep learning model. The artificial neural network model (500) may be trained using a magnetic resonance signal or magnetic resonance image captured with a set value applied in a magnetic resonance image acquisition method according to one embodiment of the present disclosure, and a labeled magnetic resonance image, as a training data set. The artificial neural network model (500) may be constructed using various additional data in addition to subsampled magnetic resonance images and full-sampled magnetic resonance images. For example, as additional data, at least one of k-space data corresponding to the magnetic resonance image, real image data, imaginary image data, magnitude image data, phase image data, sensitivity data of the multi-channel RF coil, and noise pattern image (NP) data may be used.

[0070] And the processor (110) can restore the magnetic resonance image using an artificial neural network model (500) (S340). The pre-trained artificial neural network model (500) receives the input magnetic resonance image (340) captured with the set values ​​applied and can output an output magnetic resonance image (350) with improved resolution, contrast, and SNR (signal-to-noise ratio). Thus, the output magnetic resonance image (350) can be obtained using the artificial neural network model (500) according to one embodiment of the present disclosure.

[0071] A control method for a program to acquire magnetic resonance images using an artificial neural network model according to one embodiment of the present disclosure relates to a control method for a program to obtain an output magnetic resonance image with improved resolution, contrast, and SNR compared to an input magnetic resonance image input to the artificial neural network model, while reducing the magnetic resonance image scan time.

[0072] The processor (110) can shorten the scan time by controlling the program during magnetic resonance image scanning. Additionally, the processor (110) can control various setting values ​​applied to the scanned magnetic resonance image by controlling the program. Alternatively, the initial setting of the program's setting values ​​can be optimized and set by the processor (110) in correspondence with the technology of acquiring magnetic resonance images using an artificial neural network. The user can use the magnetic resonance image acquisition program according to one embodiment of the present disclosure to acquire an input magnetic resonance image in an optimized state for using the technology of inputting into an artificial neural network and outputting the acquired magnetic resonance image.

[0073] Specifically, compared to the control method of a program that acquires output magnetic resonance images using a mathematical model, the method of acquiring magnetic resonance images using an artificial neural network model can reduce the scan time and acquire magnetic resonance images with adjusted contrast, resolution, and SNR.

[0074] Here, acquiring the output magnetic resonance image using a mathematical model does not mean acquiring the output magnetic resonance image using an artificial neural network model, but rather acquiring the magnetic resonance image through a different technique. Here, the image acquisition technique may be, for example, a parallel imaging technique. Here, the parallel imaging technique is a type of image reconstruction technique for acquiring high-accuracy k-space data and / or magnetic resonance images, such as full-sampled k-space data and / or magnetic resonance images from subsampled magnetic resonance signals (310) and / or k-space data.

[0075] FIG. 4 is a schematic diagram illustrating the difference between full sampling and subsampling according to one embodiment of the present invention.

[0076] Referring to FIG. 4, the subsampled magnetic resonance signal may be a magnetic resonance signal sampled at a sampling rate lower than the Nyquist sampling rate. Additionally, the subsampled magnetic resonance image is an image obtained by sampling the magnetic resonance signal at a sampling rate lower than the Nyquist sampling rate. The subsampled magnetic resonance image may be an image containing aliasing artifacts. Aliasing artifacts may be artificial images that occur in the magnetic resonance image when the scanned object is larger than the field of view (FOV). Meanwhile, the full-sampled magnetic resonance image may be an image obtained by sampling k-space data at a sampling rate greater than or equal to the Nyquist sampling rate.

[0077] In performing image reconstruction according to parallel imaging techniques, any known parallel imaging techniques such as SPACE RIP (Sensitivity Profiles From an Array of Coils for Encoding and Reconstruction in Parallel), SMASH (Simultaneous acquisition of spatial harmonics), PILS (Partially Parallel Imaging With Localized Sensitivity), GRAPPA (Generalized Autocalibrating Partially Parallel Acquisitions), and SPIRiT (iterative Self-consistent Parallel Imaging Reconstruction) may be applied without restriction. A reconstructed magnetic resonance image can be generated by performing an inverse Fourier transform on complete k-space image data generated according to parallel imaging techniques. Meanwhile, the mathematical model is not limited to the parallel imaging techniques described above and may include a method for acquiring enhanced magnetic resonance signal data or acquiring a magnetic resonance image by reconstructing insufficient magnetic resonance signal data without using an artificial neural network model.

[0078] Meanwhile, acquiring magnetic resonance images using an artificial neural network model may involve obtaining k-space image data through scanning, generating a DICOM magnetic resonance image by performing an inverse Fourier transform, and using this as input to reconstruct the output magnetic resonance image through an artificial neural network model. Alternatively, it may involve obtaining k-space image data through scanning and inputting it into an artificial neural network model to acquire a DICOM magnetic resonance image. In this case, the artificial neural network model can convert the k-space magnetic resonance image into a DICOM magnetic resonance image.

[0079] Here, a processor (110) according to one embodiment of the present disclosure can control the scan setting value of a magnetic resonance image of a magnetic resonance image acquisition program.

[0080] For example, the processor (110) can control settings related to reducing the magnetic resonance imaging scan time, settings related to the resolution of the magnetic resonance imaging, or settings related to the contrast of the magnetic resonance imaging. In addition, it can control the FOV of the magnetic resonance imaging. These settings may be set differently for each series of magnetic resonance imaging scans. Here, a magnetic resonance imaging scan series may refer to a combination of various magnetic resonance images scanned according to the body part, lesion, scanning method, scanning device, etc. that are the target of the scan.

[0081] For example, the series applied to brain scans will be explained based on [Table 1] below.

[0082] Key Settings Control Group Experimental Group Pulse sequence Spin Echo Spin Echo Average 11 TE (ms), TR (ms) 10, 700 10, 450 Acquisition Matrix (Frequency, Phase) 256 x 177 Scan time 4 : 12 2 : 43

[0083] Referring to Table 1, the processor (110) can set the TR (Time to repeat) applied to the magnetic resonance image input to the artificial neural network model lower in the method of acquiring magnetic resonance images through an artificial neural network model (experimental group) than in the method of acquiring magnetic resonance images through a mathematical model (control group). TR refers to the time interval between the occurrence of continuous RF pulses (Radiofrequency pulses), and the unit is milliseconds (ms). TR can refer to the time from sending one RF pulse until sending the next RF pulse. The shorter the TR, the more sensitive it becomes to the T1 signal, and the contrast between tissues becomes distinct due to the difference in the T1 signal, making it suitable for obtaining T1-weighted imaging. Reducing the TR can shorten the scan time.

[0084] Echo time (TE) refers to the time taken to collect the echo signal after sending an RF pulse, and its unit is milliseconds (ms). It can be defined as the time from when magnetization is tilted by the RF pulse until the resulting signal (echo) is detected. If the TE is too long, the signal may be excessively attenuated, potentially degrading image quality. A long TE can also increase scan time. Reducing the TE can decrease the Transistor (TR) and shorten scan time. A short TE results in a strong signal but may lead to lower contrast between tissues. Short TR and a short TE may be suitable for T1-weighted imaging. Long TR and a long TE may be suitable for T2-weighted imaging.

[0085] Key Settings Control Group Experimental Group Pulse sequence Turbo Spin Echo Turbo Spin Echo Average 11 TE (ms), TR (ms) 10⁹, 3900 Acquisition Matrix (Frequency, Phase) 384 x 384 Scan time 0 : 4 30 : 24

[0086] Referring to Table 2, the processor (110) can set the phase resolution applied to the magnetic resonance image input to the artificial neural network model to be lower. Phase resolution refers to the density of sampling points in the phase encoding direction. Phase resolution indicates how fine information can be obtained in the phase direction of the image. Phase resolution may refer to the ratio of the exquisition matrix size (phase) to the exquisition matrix size (frequency). Also, for example, a low phase resolution may mean that the exquisition matrix size (phase) is small. Also, a low phase resolution may mean that the exquisition voxel size (phase) is large. Phase resolution can be adjusted by changing the data points (number of sampling lines) in the phase encoding direction. As the phase resolution increases, the number of acquired sampling lines increases, and the scan time may increase. Increasing the phase resolution increases the resolution but may decrease the SNR. Lowering the phase resolution may decrease the scan time. The processor (110) can set the acquisition matrix size (phase) applied to the input magnetic resonance image to a lower value. The acquisition matrix size (phase) refers to the size of the matrix set in the phase direction during the magnetic resonance image data acquisition process. The acquisition matrix size (phase) and the scan time may be proportional. This can have a significant impact on the resolution and scan time of the magnetic resonance image scan. Reducing the acquisition matrix size (phase) reduces the scan time, reduces the resolution, and reduces the SNR. The acquisition matrix size (frequency) refers to the size of the matrix set in the frequency direction during the magnetic resonance image data acquisition process. If the exquisite matrix size (frequency) is reduced, the resolution decreases and the SNR may increase.However, in this case, the reduced resolution and SNR can be compensated for using super-resolution. In the 3D T1 TFE SAG series, when an artificial neural network model is applied, the exquisite matrix size (phases) is set higher to increase scan time while improving resolution and SNR. If improving resolution is more important than reducing scan time, the exquisite matrix size (phases) can be increased.

[0087] Additionally, the acquired voxel size refers to the size of the voxel from which data is acquired. It can be input in three directions: M: measurement (frequency encoding) direction, P: phase encoding direction, and S: slice thickness (voxel size in the case of 3D). Increasing the voxel size can increase the SNR.

[0088] Main settingsControl groupExperimental groupPulse sequenceTurbo Spin EchoTurbo Spin EchoAverage11TE (ms),TR (ms)94, 406094, 4060Acqusition Matrix (Frequency, Phase)320 x 320320 x 320Phase Oversampling50%0%Scan time0: 570: 41

[0089] Additionally, the processor (110) can set the oversampling applied to the input magnetic resonance image to a lower value. Oversampling is a technique primarily used to improve signal acquisition in the phase direction or slice direction (in the case of 3D). Oversampling is a method of sampling more data than is actually required in the phase direction or slice direction. This increases the SNR and can reduce artifacts. Setting the oversampling to a lower value can reduce the scan time. Slice oversampling can prevent backfolding in the slice direction in 3D scans. Oversampling can be used to prevent the occurrence of wrapping artifacts or to increase the SNR. FIGS. 5a to 5c are drawings showing a comparison of brain MRI images output according to whether oversampling is applied according to an embodiment of the present invention. Fig. 5a is a control image taken with phase oversampling set to 50%, Fig. 5b is an image restored by an artificial neural network with phase oversampling set to 0%, and Fig. 5c is an image taken with phase oversampling set to 0%.

[0090] Comparing Fig. 5a and Fig. 5c, it can be seen that the image in Fig. 5c contains a lot of noise. Looking at the image in Fig. 5b, it can be visually confirmed that the noise has been removed and an SNR similar to that of the image in Fig. 5a has been secured.

[0091] Key Settings Control Group Experimental Group Pulse sequence Turbo Spin Echo Turbo Spin Echo Average 11 TE (ms), TR (ms) 93 , 40 60 Acquisition Matrix (Frequency, Phase) 320 x 320 PAT Factor 1 Phase oversampling 00 Scan time 1 : 1 30 : 41

[0092] Additionally, referring to Table 4, the processor (110) can set a higher PAT factor applied to the input magnetic resonance image. The PAT factor is an acceleration factor indicating how much data sampling is reduced. Increasing the PAT factor can shorten the scan time. Increasing the PAT factor may reduce the SNR or cause aliasing artifacts to appear. Increasing the PAT factor in the phase direction and slice direction can shorten the scan time. The reconstruction matrix can determine the apparent resolution of the output image. If the selected reconstruction matrix size is larger than the scan matrix size, an interpolated image may be generated. The reconstruction matrix is ​​related to the resolution of the output image, and if the size of the reconstruction matrix is ​​larger than the size of the scan matrix, interpolated images may occur. The setting value of the reconstruction matrix must be at least equal to the scan matrix value. The reason for increasing the setting value of the recon matrix is ​​to create a room where the super-resolution algorithm can be applied.

[0093] Key Settings Control Group Experimental Group Pulse sequence Turbo Spin Echo Average 11 TE (ms), TR (ms) 108, 60 2 0 105, 588 Acquisition Matrix (Frequency, Phase) 512 x 256 256 x 256 Scan time 1 : 30 1 : 28

[0094] Referring to Table 5, it can be seen that the exquisite matrix size (frequency) in the experimental group was set to half the value compared to the control group. An increase in the exquisite matrix size (frequency) can affect the increase in resolution. Resolution can be determined by adjusting the exquisite matrix size or the voxel size.

[0095] Key Settings Control Group Experimental Group Pulse sequence Turbo Spin Echo Average 11TE (ms),TR (ms) 109, 3900 111, 4340 Acquisition Matrix (Frequency, Phase) 384 x 384 512 x 297 Scan time 0 : 430 : 39

[0096] Referring to Table 6, it can be seen that compared to the control group, the exquisite matrix size (frequency) of the experimental group increased and the exquisite matrix size (phase) decreased. This allows for a reduction in scan time while minimizing the degradation of resolution. Although increasing the exquisite matrix (frequency) does not increase the resolution in the phase direction, the resolution in the phase direction can be increased due to the influence of the resolution in the frequency direction.

[0097] Key Settings Control Group Experimental Group Pulse sequence Gradient Echo Gradient Echo Average 11 TE (ms), TR (ms) 2.5 , 250 Acquisition Matrix (Frequency, Phase) 320 x 256 320 x 256 Partial Fourier Factor None 6 / 8 Scan time 0 : 350 : 27

[0098] Referring to Table 7, it can be seen that a Partial Fourier Factor has been applied in the phase direction. Scan time can be reduced depending on the ratio of the applied Partial Fourier Factor. The Partial Fourier Factor refers to the ratio used in the Partial Fourier Technique. This technique samples only a portion of k-space and mathematically compensates for the remainder. Through this, imaging speed can be increased, scan time reduced, and SNR optimized.

[0099] Main setting valuesControl groupExperimental group 1Experimental group 2Pulse sequenceTurbo Spin EchoTurbo Spin EchoTurbo Spin EchoTE (ms),TR (ms)77, 500075 , 500075 , 5000Acqusition Matrix (Frequency, Phase)320 x 256320 x 256320 x 256Echo Spacing / ETL / ETD15.3 / 10 / 1539.4 / 16 / 1508.1 / 19 / 153Bandwidth100250250RF Mode, Gradient ModeNormal , NormalNormal , NormalFast, FastScan time1:050:400:35

[0100] Referring to Table 8, it can be seen that among the key settings, the Pulse Sequence is set to Turbo Spin Echo. Generally, if the Echo Train Duration is increased, the SNR of the magnetic resonance image decreases, and blurring may occur. Among the key settings, ETL can be referred to as the Turbo Factor and Echo Train Length. As the Turbo Factor increases, the scan time may decrease. Meanwhile, ETD is the product of Echo Spacing and ETL; if ETD becomes infinitely long, blurring may occur in the image and the SNR may decrease. Therefore, to reduce scan time, it is desirable to increase ETL while decreasing Echo Spacing. Below, we will explain the characteristic aspects of the settings applied to different series in magnetic resonance imaging scans of the brain.

[0101] Key Settings Control Group Experimental Group Pulse sequence Turbo Spin Echo Turbo Spin Echo Average 21TE (ms),TR (ms) 94 , 40 60 Acquisition Matrix (Frequency, Phase) 320 x 320 PAT Factor 22Phase oversampling 00 Scan time 1 : 1 70 : 41

[0102] Referring to Table 9, it can be seen that the average of the experimental group is set to half the value of the control group. Average is a setting value representing the number of times signals are repeatedly collected and averaged. Average can be used to improve the image's SNR and enhance image quality. Consequently, the scan time for magnetic resonance imaging acquisition using artificial neural networks can be reduced. Increasing the average increases the SNR, extends the scan time, and can reduce motion artifacts and flow artifacts. Average may be referred to as "Average" by Siemens, "NEX" (Number of excitations) by GE, and "NSA" (Number of signal averages) by Philips. Decreasing the average from 2 to 1 can shorten the scan time by half. Changes in average do not affect resolution but may be proportional to the SNR. However, reducing the average may increase motion artifacts. FIGS. 6a to 6c are drawings showing a comparison of brain magnetic resonance images output by varying the average according to an embodiment of the present disclosure. FIG. 6a is a control image scanned with the average set to 2, FIG. 6b is an image restored using an artificial neural network model from an image scanned with the average set to 1, and FIG. 6c is an image scanned with the average set to 1. When comparing FIG. 6a and FIG. 6c, it can be seen that there is a lot of noise in the image of FIG. 6a. When comparing the image of FIG. 6b with the image of FIG. 6a, it can be seen that the noise in the image of FIG. 6b has been removed to create an image similar in level to the image of FIG. 6c.

[0103] Key Settings Control Group Experimental Group Pulse sequence Gradient Echo Gradient Echo Average 11 TE (ms), TR (ms) 2.46 , 160 Acquisition Matrix (Frequency, Phase) 320 x 256 PAT Factor 22 Number of Reference Line 80 Scan time 0 : 270 : 21

[0104] Referring to Table 10, the number of reference lines in the experimental group can be set to be smaller than that of the control group. When using a graphite-based PAT, data is acquired by skipping N lines in the phase direction in K-space; in this case, the central part of K-space is acquired without omission, and the number of phase encoding lines acquired at this time can be referred to as reference lines. Reference lines can largely serve as data for fitting the kernel, or they can fill K-space regardless of PAT to increase the SNR. The more reference lines there are, the higher the relative SNR (the relative magnitude of the SNR between two images) can be. If the number of reference lines increases, the number of phase encoding lines that need to be acquired increases, which may increase the scan time. Figures 7a and 7b are drawings comparing magnetic resonance images output according to the setting of the number of reference lines in an embodiment of the present disclosure. Figure 7a is a magnetic resonance image with the reference line set to 80, 7b is a magnetic resonance image restored by an artificial neural network model with the reference line set to 8, and 7c is a magnetic resonance image with the reference line set to 8. By referring to this, it can be seen that the SNR of the 7c image is lower than that of the 7a image, but the SNR of the 7b image is higher than that of the 7a image.

[0105] Meanwhile, in methods for acquiring magnetic resonance imaging using artificial neural networks, the imaging sequence applied to the input magnetic resonance image can be set to FFE. In MRI, the Fast Field Echo (FFE) sequence is one of the methods for rapidly acquiring images and is a Gradient Echo-based sequence. FFE is used specifically to reduce scan time and emphasize specific physical effects. FFE sequences can reduce scan time by maintaining a low flip angle and utilizing Gradient Echo. Due to its short TR and rapid acquisition, the FFE sequence is suitable for 3D and high-precision imaging. The flip angle is the angle at which the magnetization rotates due to the RF pulse. A small flip angle enables fast sequences and can be frequently used in Gradient Echo-based sequences. In MRI, a sequence can refer to a combination of a series of RF pulses and magnetic field gradients that manipulate magnetization to generate an image. Each sequence generates and manipulates signals for a specific purpose, and consequently, can provide different types of contrast and image characteristics.

[0106] The processor (110) can set at least one of the Recon matrix size (frequency) or Recon matrix size (phase) applied to the input magnetic resonance image to a lower value. The Recon matrix size (frequency) and the Recon matrix size (phase) are factors that define the size and resolution of the matrix in the frequency direction and the phase direction, respectively. If the Recon matrix size (frequency) and the Recon matrix size (phase) are reduced, the Recon matrix size increases and super-resolution can be applied. Meanwhile, regarding the Recon matrix size (phase), if the number of samplings in the phase direction increases, the scan time may be extended. Super-resolution is a technique that improves the spatial resolution of an image. The processor (110) can use super-resolution to process images scanned at a low resolution due to limited data acquisition time and SNR issues in the magnetic resonance image, thereby restoring a more detailed and high-resolution image.

[0107] The reconstruction matrix size determines the size of the matrix in which data is reconstructed and must be smaller than or equal to the size of the smallest exquisite matrix. Only in the case of 3D scanning can both the voxel size acquired in the slice direction and the reconstructed voxel size be determined. Reducing the reconstructed voxel size in the slice direction does not affect scan time, but it can increase the number of reconstructed slices and create overlap between slices.

[0108] 주요 설정값대조군실험군Pulse sequenceTurbo Spin EchoTurbo Spin EchoTE (ms),TR (ms)77 , 565075 , 5000Acqusition Matrix (Frequency, Phase)320 x 256320 x 256Echo Spacing / ETL / ETD15.3 / 16 / 2409.4 / 16 / 150Bandwidth100250RF Mode, Gradient Mode320 x 256Normal , NormalScan time0 : 450 : 40

[0109] Referring to Table 11, the processor (110) can control the echo spacing applied to the input magnetic resonance image to be set to be shorter. Echo spacing refers to the time interval between the collection of one echo signal and the collection of the next echo signal. Short echo spacing allows for the rapid collection of more echoes, thereby reducing scan time. Echo spacing can be set to be short to reduce TR. Long echo spacing may increase scan time. It is preferable to keep echo spacing relatively short to avoid blurring. Additionally, the TSE factor can be viewed as a numerical value of scan time acceleration. Profile order is intended to define which part of k-space to collect data from and in what order to fill the data. The order may vary depending on the purpose of data acquisition (e.g., rapid scan, dynamic imaging). Regarding the characteristic settings applied to a series of MRI scans for the spine, the Ultra Short function can be applied to the input MRI images. Ultra Short is an MRI sequence that reduces echo time to the level of several hundred microseconds (μs). While the Echo Time (TE) in standard MRI is several milliseconds (ms) or longer, Ultra Short sequences can reduce this to less than 1ms, enabling very fast signal detection. Ultra Short can be described as a function that shortens echo spacing. In summary, the Fat Factor can be increased from 1 to 2 to prioritize scan time reduction. Generally, increasing the Fat Factor to 2 minimizes artifact occurrence and reduces SNR degradation, making it a preferred setting. However, raising the Fat Factor from 2 to 3 drastically increases the likelihood of aliasing artifacts and decreases the scan time reduction rate, so caution is required when applying this setting.

[0110] Regarding the average, the processor (110) can generally set the average to 1, although it varies depending on the body part being scanned. If the existing average is for the purpose of improving SNR rather than preventing motion artifacts, the average can be shortened to reduce the scan time.

[0111] Increasing ETL while reducing echo spacing can reduce scan time and prevent the occurrence of artifacts. Alternatively, maintaining ETL while reducing echo spacing can shorten ETD and reducing minTR, thereby reducing scan time. Resolution and SNR can be improved by acquiring magnetic resonance images using an artificial neural network model.

[0112] The size of the exquisite voxel is inversely proportional to the SNR, and the size of the exquisite voxel can be increased when further reduction in scan time is required, even though the SNR is low as a result of shortening the scan time by reducing the SNR. The SNR can be increased by increasing the exquisite voxel size, and the scan time can be reduced by decreasing the exquisite matrix size.

[0113] In addition, scan time can be reduced by increasing the exquisite matrix size (frequency) and decreasing the exquisite matrix size (phase). Scan time can be reduced by applying oversampling.

[0114] A method for acquiring magnetic resonance images using an artificial neural network according to one embodiment of the present disclosure described above and a program control method thereof can acquire magnetic resonance images having excellent resolution, contrast, and SNR while reducing the magnetic resonance image acquisition time.

[0115] The various embodiments of the present disclosure described above may be combined with additional embodiments and modified to the extent understandable to those skilled in the art in light of the detailed description above. The embodiments of the present disclosure are illustrative in all respects and should be understood as not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form. Accordingly, all modifications or variations derived from the meaning, scope, and equivalents of the claims of the present disclosure should be interpreted as being included within the scope of the present disclosure.

Claims

1. A method for controlling a program for acquiring magnetic resonance images using an artificial neural network model, performed by a computing device comprising at least one processor, wherein Compared to a control method of a program for acquiring an output magnetic resonance image using a mathematical model, a step of setting a higher PAT factor applied to an input magnetic resonance image or an input magnetic resonance image signal for acquiring an output magnetic resonance image using an artificial neural network model, and setting a lower number of reference lines applied to the input magnetic resonance image or an input magnetic resonance image signal; and A method comprising the step of obtaining an output magnetic resonance image by restoring the input magnetic resonance image or the input magnetic resonance signal through the artificial neural network model. method.

2. In Paragraph 1, A step comprising setting the TR (Time to repeat) applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value. method.

3. In Paragraph 1, A step comprising setting the phase resolution applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value. method.

4. In Paragraph 1, A step comprising applying a partial Fourier factor in the phase direction to the input magnetic resonance image or the input magnetic resonance signal. method.

5. In Paragraph 1, A step comprising setting the exquisite voxel size applied to the input magnetic resonance image or the input magnetic resonance signal to be larger, method.

6. In Paragraph 1, A step comprising setting a turbo factor to be applied to the input magnetic resonance image or the input magnetic resonance signal. method.

7. In Paragraph 1, A step comprising setting the oversampling applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value. method.

8. In Paragraph 1, A step comprising setting the echo spacing applied to the input magnetic resonance image or the input magnetic resonance signal to be shorter. method.

9. A method for controlling a program for acquiring magnetic resonance images using an artificial neural network model, performed by a computing device comprising at least one processor, wherein Compared to a control method of a program for acquiring an output magnetic resonance image using a mathematical model, a step of setting the Average and TR (Time to repeat) applied to an input magnetic resonance image or an input magnetic resonance image signal input for acquiring an output magnetic resonance image using an artificial neural network model to be lower, and setting the Echo Spacing applied to the input magnetic resonance image or the input magnetic resonance signal to be shorter; and A step of obtaining an output magnetic resonance image by restoring the input magnetic resonance image or the input magnetic resonance signal through the artificial neural network model; method.

10. In Paragraph 9, A step comprising setting the phase resolution applied to the input magnetic resonance image or the input magnetic resonance signal to a lower value. method.

11. In Paragraph 9, A step comprising applying a partial Fourier factor in the phase direction to the input magnetic resonance image or the input magnetic resonance signal. method.

12. In Paragraph 9, A step comprising setting the exquisite voxel size applied to the input magnetic resonance image or the input magnetic resonance signal to be larger, method.

13. In Paragraph 9, A step comprising setting a turbo factor to be applied to the input magnetic resonance image or the input magnetic resonance signal. method.

14. In a computing device in which a program for acquiring magnetic resonance images using an artificial neural network model is executed, At least one processor; and The above program includes memory where it is stored, and The above processor, according to the execution of the above program, The output magnetic resonance image is acquired using a mathematical model or an artificial neural network model, and When acquiring the output magnetic resonance image using the artificial neural network model, the Fat Factor applied to the input magnetic resonance image or input magnetic resonance signal for acquiring the output magnetic resonance image is set higher than the Fat Factor applied when acquiring the output magnetic resonance image using the mathematical model, and the Number of Reference Lines applied to the input magnetic resonance image or input magnetic resonance image signal for acquiring the output magnetic resonance image using the artificial neural network model is set lower than the Number of Reference Lines applied when acquiring the output magnetic resonance image using the mathematical model, and the output magnetic resonance image is acquired by restoring the input magnetic resonance image or the input magnetic resonance signal through the artificial neural network model. Computing device.

15. In a computing device in which a program for acquiring magnetic resonance images using an artificial neural network model is executed, At least one processor; and The above program includes memory where it is stored, and The above processor, according to the execution of the above program, The output magnetic resonance image is acquired using a mathematical model or an artificial neural network model, and When acquiring the output magnetic resonance image using the artificial neural network model, the average and TR (Time to repeat) applied to the input magnetic resonance image or input magnetic resonance signal input for acquiring the output magnetic resonance image are set lower than the average and TR applied when acquiring the output magnetic resonance image using the mathematical model, and the echo spacing applied to the input magnetic resonance image or input magnetic resonance image signal input for acquiring the output magnetic resonance image using the artificial neural network model is set shorter, and the output magnetic resonance image is acquired by restoring the input magnetic resonance image or the input magnetic resonance signal through the artificial neural network model. Computing device.

Images