Magnetic resonance image reconstruction method, apparatus, and program

Abstract

To shorten an imaging time while keeping the quality of a magnetic resonance image.SOLUTION: A magnetic resonance image reconstruction method comprises a step S1 of acquiring a k space signal 21 of an imaging object 99 including a real signal and a virtual signal, a step S2 of reconstructing a pseudo real image 22 based on the k space signal 21, a step S3 of multiplying the k signal 21 by an imaginary unit, a step S4 of reconstructing a pseudo virtual image 24 based on a k space signal 23 obtained by multiplication by the imaginary unit, and a step S5 of composing a magnetic resonance image 25 of the imaging object 99, based on the pseudo real image 22 and the pseudo virtual image 24.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a method, an apparatus and a program for reconstructing a magnetic resonance image. [Background technology]

[0002] Magnetic resonance imaging (MRI) is a technique for capturing magnetic resonance images of an object by applying a radio-frequency magnetic field and a gradient magnetic field to the object placed in a static magnetic field and detecting magnetic resonance signals generated from the object through nuclear magnetic resonance. In a typical MRI device, the magnetic resonance signals detected by the device are converted into k-space signals by passing them through a quadrature phase detector, and then a magnetic resonance image of the object is reconstructed by performing an inverse Fourier transform on the k-space signals. The k-space signals are then stored as imaging data.

[0003] It takes a relatively long time to acquire a magnetic resonance image using MRI. To acquire a magnetic resonance image in a relatively short time, various methods have been proposed.

[0004] As a method for high-speed imaging, for example, there is a method of acquiring complete imaging data (k-space signals) without thinning out the data when detecting magnetic resonance signals, and another method of acquiring imaging data by thinning out the data (also called undersampling). Missing data resulting from the regular thinning out of the imaging data is complemented by assuming that Hermitian symmetry exists in the acquired imaging data.

[0005] Another high-speed imaging technique for acquiring imaging data by thinning it out is a method using compressed sensing, as exemplified in Patent Document 1. In compressed sensing, imaging data is acquired by randomly thinning it out when detecting magnetic resonance signals, thereby estimating and complementing missing data without assuming that Hermitian symmetry exists in the acquired imaging data. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Publication No. 2018-042671 Summary of the Invention [Problem to be solved by the invention]

[0007] In MRI imaging, local phase distortion occurs due to inhomogeneity of the static magnetic field applied to the imaging target (e.g., a subject or biological tissue, hereinafter simply referred to as the target) and differences in the magnetic susceptibility of the target within the imaging region. When local phase distortion exists, the magnetic resonance image reconstructed by simply performing an inverse Fourier transform on the k-space signal is expressed as a complex number rather than a real number. The complex number of the image caused by local phase distortion has a phase that, when expressed as a complex number in terms of magnitude and phase, varies depending on the position on the image.

[0008] According to a method of acquiring complete imaging data without data thinning, the problem of the complex numbering of the magnetic resonance image caused by such local phase distortion is avoided by converting the magnetic resonance image into a real number by taking the absolute value of the complex number. However, the local phase distortion causes various problems in high-speed imaging methods that acquire imaging data by thinning it.

[0009] For example, missing data has been interpolated by assuming that Hermitian symmetry exists in the acquired imaging data. However, Hermitian symmetry does not exist when local phase distortion occurs. As a result, in high-speed imaging techniques that thin out imaging data, the missing data cannot be properly interpolated, resulting in a problem of reduced image quality in magnetic resonance images.

[0010] For example, compressed sensing randomly thins out and acquires imaging data, thereby filling in missing data without assuming Hermitian symmetry in the acquired imaging data. However, compressed sensing does not restore phase information. Therefore, when compressed sensing is used in a situation where local phase distortion is present, attempting to recover phase information lost due to the local phase distortion requires introducing symmetry into the acquired imaging data to restore the phase information. Because compressed sensing is a method for estimating and filling missing data based on the randomness of imaging data, introducing a symmetry-based phase estimation process into compressed sensing reduces the randomness of the imaging data, resulting in a problem of reduced efficiency of compressed sensing.

[0011] It is desired to reduce the imaging time while maintaining the image quality of the magnetic resonance image even under conditions where local phase distortion exists within the imaging region.

[0012] An object of the present invention is to reduce the imaging time while maintaining the image quality of a magnetic resonance image. [Means for solving the problem]

[0013] The present invention for solving the above problems includes, for example, the following aspects. (Section 1) acquiring k-space signals of an imaged object, the k-space signals including real and imaginary signals; reconstructing a pseudo-real image based on the k-space signals; multiplying the k-space signals by an imaginary unit; reconstructing a pseudo-imaginary image based on the k-space signals multiplied by an imaginary unit; synthesizing a magnetic resonance image of the imaging subject based on the pseudo real image and the pseudo imaginary image; A method for reconstructing a magnetic resonance image, comprising: (Section 2) Item 1. The method according to Item 1, wherein the k-space signal has measurement points randomly thinned out. (Section 3) Item 1. The method according to Item 1, wherein the step of multiplying by an imaginary unit multiplies the k-space signal by −i (where i is an imaginary unit). (Section 4) The step of reconstructing the pseudo-real part image includes estimating the pseudo-real part image from the k-space signal according to a compressed sensing method; Item 1. The method according to Item 1, wherein the step of reconstructing the pseudo-imaginary part image estimates the pseudo-imaginary part image from the k-space signal multiplied by an imaginary unit according to a compressed sensing method. (Section 5) Item 5. The method according to Item 4, wherein the step of reconstructing the pseudo real part image estimates the pseudo real part image by performing an inverse Fourier transform and a wavelet transform on the k-space signal. (Section 6) Item 5. The method according to Item 4, wherein the step of reconstructing the pseudo-imaginary part image estimates the pseudo-imaginary part image by performing an inverse Fourier transform and a wavelet transform on the k-space signal multiplied by an imaginary unit. (Section 7) Item 2. The method according to Item 1, wherein the step of synthesizing the magnetic resonance image of the imaging object includes calculating the L2 norm of the pseudo real image and the pseudo imaginary image. (Section 8) an acquisition unit that acquires k-space signals of an imaging object, including real and imaginary signals; a pseudo real part image reconstruction unit that reconstructs a pseudo real part image based on the k-space signal; an imaginary unit multiplier that multiplies the k-space signal by an imaginary unit; a pseudo-imaginary part image reconstruction unit that reconstructs a pseudo-imaginary part image based on the k-space signal multiplied by an imaginary unit; a magnetic resonance image synthesis unit that synthesizes a magnetic resonance image of the imaging target based on the pseudo real image and the pseudo imaginary image; A magnetic resonance image reconstruction device comprising: (Section 9) A program for causing a computer to execute each step of the method according to any one of items 1 to 7. [Effects of the Invention]

[0014] According to the present invention, it is possible to reduce the imaging time while maintaining the image quality of the magnetic resonance image. [Brief explanation of the drawings]

[0015] [Figure 1] 1 is a diagram for explaining a mode of use of a magnetic resonance image reconstruction device according to an embodiment of the present invention; [Figure 2] 1 is a block diagram for explaining the function of a magnetic resonance image reconstruction device according to an embodiment of the present invention. [Figure 3] 1 is a flowchart for explaining a processing procedure performed by a magnetic resonance image reconstruction device according to an embodiment of the present invention. [Figure 4] 1A and 1B show magnetic resonance images reconstructed by various methods, including a method according to an embodiment of the present invention. (A) is an image according to a comparative example, in which no phase correction has been applied. (B) is an image according to a comparative example, in which conventional phase correction (single phase correction) using phase detection has been applied. (C) is an image according to an embodiment of the present invention. (D) is an image according to a comparative example, in which the image has been reconstructed from complete imaging data in which no measurement points have been reduced. DETAILED DESCRIPTION OF THE INVENTION

[0016] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description and drawings, the same reference numerals will denote the same or similar components, and therefore, redundant descriptions of the same or similar components will be omitted.

[0017] In this specification, the term "signal" does not only mean a signal in its literal sense, but can also be interpreted as a term meaning data. Similarly, the term "data" does not only mean data in its literal sense, but can also be interpreted as a term meaning a signal. [Device Overview] <Usage>

[0018] FIG. 1 is a diagram for explaining a mode of use of a magnetic resonance image reconstruction device according to one embodiment of the present invention.

[0019] A magnetic resonance image reconstruction device 1 (hereinafter also referred to simply as a reconstruction device) according to one embodiment is a device that acquires imaging data (k-space signal 21 (shown in Figure 2)) from an MRI device 90 and reconstructs a magnetic resonance image 25 (shown in Figure 2) of an object 99.

[0020] The MRI device 90 includes a gradient magnetic field coil 91, a main magnet 92 for applying a static magnetic field, a high-frequency receiving coil 93 for detecting magnetic resonance signals, and a quadrature detector (not shown), and images an object 99. In the example shown, the object 99 is a head 99a of a subject. Various known MRI devices can be used for the MRI device 90. The MRI device 90 images the object 99 by randomly thinning out measurement points of the imaging data (k-space signal 21) in advance.

[0021] The reconstruction device 1 acquires imaging data of an object 99 acquired by an MRI device 90, for example, via a network 9. In the example shown, the reconstruction device 1 acquires k-space signals 21 of a head 99a of a subject. The reconstruction device 1 reconstructs a magnetic resonance image 25 of the object 99 based on the acquired k-space signals 21 in accordance with a magnetic resonance image reconstruction method according to an embodiment of the present invention, which will be described later. The reconstruction device 1 can be, for example, a general-purpose computer such as a personal computer.

[0022] The reconstruction device 1 reduces the imaging time while maintaining the image quality of the magnetic resonance image 25 to be reconstructed, mainly by using the following two techniques.

[0023] The first innovation is to randomly thin out the measurement points of the k-space signal 21 used by the reconstruction device 1 when reconstructing a magnetic resonance image 25 of the object 99. The MRI device 90 images the object 99 by randomly thinning out the measurement points of the imaging data (k-space signal 21) in advance. This reduces the imaging time. The reconstruction device 1 acquires the k-space signal 21, the measurement points of which have been randomly thinned in advance, directly from the MRI device 90 or indirectly from another server device (not shown) or the like.

[0024] The second innovation is to use not only the k-space signal 21 but also the k-space signal 23 (shown in FIG. 2 ) multiplied by an imaginary unit to reconstruct the magnetic resonance image 25. This maintains the image quality of the magnetic resonance image 25 obtained by reconstruction. The k-space signal 21, which is a complex signal including a real signal and an imaginary signal, is obtained by transforming (e.g., using a Fourier transform) the product of the signal magnitude, which is a function of position in the imaging region, and the signal phase into the frequency domain. This is a convolution operation of the Fourier transform of the signal magnitude and the Fourier transform of the phase, and the Fourier transform of the phase is a function that depends on the position in the k-space. Here, if the phase does not depend on the position in the imaging region and can be treated as a constant, the Fourier transform of the phase is composed of a delta function, and the result of the convolution operation is the Fourier transform of the signal magnitude corrected for the constant phase. However, since local phase distortion occurs in the imaging region of the MRI device 90, the phase depends on the position in the imaging region, and the k-space signal 21 is obtained by convolution of a function dependent on the position in k-space obtained by Fourier transforming the phase, and is not uniform throughout the k-space signal 21 but is complexly dependent on the position in k-space. Therefore, if a conventional single phase correction that corrects the phase shift between the reference input of phase detection and the input of the magnetic resonance signal by a single phase throughout the k-space is simply applied to the k-space signal 21, an imaginary component remains in the magnetic resonance image and the magnetic resonance image is not converted to a real number.

[0025] The inventors have found that, rather than reconstructing a magnetic resonance image 25 from only the k-space signal 21, the image quality of the magnetic resonance image 25 obtained by reconstruction can be maintained by multiplying the k-space signal 21, which is a complex signal, by an imaginary unit to convert the imaginary part of the signal into a real number and then using the resulting k-space signal 23, which converts the real part of the signal into an imaginary number, in addition to the reconstruction of the magnetic resonance image 25. <Signal Processing>

[0026] The signal processing performed by the magnetic resonance image reconstruction method according to an embodiment of the present invention will be described.

[0027] In one embodiment of the reconstruction method, first, a k-space signal of an object is acquired with measurement points randomly thinned (undersampled). Next, a pseudo-real image and a pseudo-imaginary image are reconstructed from the k-space signal according to a compressed sensing method. Finally, a reconstructed magnetic resonance image of the object is created by combining the pseudo-real image and the pseudo-imaginary image obtained by the reconstruction.

[0028] The k-space signal of the object 99 acquired by the MRI apparatus 90 is expressed as a complex number c+di (i is the imaginary unit). The real part signal c corresponds to the component of the magnetic resonance signal whose phase is 0 degrees, and the imaginary part signal di corresponds to the component of the magnetic resonance signal whose phase is 90 degrees. The magnetic resonance signal of the object detected by the MRI method is separated into the 0-degree component and the 90-degree component by quadrature detection.

[0029] The pseudo real part image is created from the k-space signal c+di. The pseudo real part image is reconstructed by processing the k-space signal c+di according to the compressed sensing method. The pseudo real part image is composed of real components.

[0030] The pseudo imaginary part image is created from the k-space signal multiplied by the imaginary unit. In a reconstruction method according to one embodiment, the k-space signal is multiplied by a negative imaginary unit, −i, as the imaginary unit. If the k-space signal is expressed as c+di, the k-space signal multiplied by the negative imaginary unit, −i, becomes d-ci. That is, in a reconstruction method according to one embodiment, the pseudo imaginary part image is created from the k-space signal d-ci multiplied by the imaginary unit. The k-space signal d-ci multiplied by the imaginary unit is arithmetically processed according to a compressed sensing method to reconstruct the pseudo imaginary part image. The pseudo imaginary part image is composed of imaginary parts.

[0031] A magnetic resonance image of an object is created from a pseudo-real image and a pseudo-imaginary image. If the pseudo-real image made up of real components is represented as a and the pseudo-imaginary image made up of imaginary components is represented as bi, then:

[0032]

number

[0033] When a k-space signal of an object that has been fully sampled without thinning out measurement points is acquired and a magnetic resonance image of the object is reconstructed under conditions in which local phase distortion exists within the imaging region, the reconstructed magnetic resonance image becomes a complex number a+bi. The square root of the sum of squares shown in the above-mentioned (Equation 1) corrects the local phase distortion at this time. In a reconstruction method according to one embodiment, a and b in the complex number a+bi are pseudo-obtained from the k-space signal of the object whose measurement points have been undersampled. [Device configuration]

[0034] FIG. 2 is a block diagram for explaining the function of a magnetic resonance image reconstruction device according to one embodiment of the present invention.

[0035] A magnetic resonance image reconstruction device 1 according to one embodiment includes a data processing unit 10, an auxiliary storage device 20, an input unit 31, a display unit 32, and a communication interface unit (communication I / F unit) 33. The reconstruction device 1 can be configured using, for example, a general-purpose computer such as a personal computer, a laptop PC, a tablet terminal, or the like.

[0036] In this embodiment, the reconstruction device 1 includes, as hardware components, an auxiliary storage device 20, an input unit 31, a display unit 32, and a communication I / F unit 33. Although not shown, the reconstruction device 1 also includes, as hardware components, a processor such as a CPU that performs data processing, and a memory that the processor uses as a working area for data processing.

[0037] The auxiliary storage device 20 is a non-volatile storage device that stores an operating system (OS), various control programs, data generated by the programs, etc., and is configured, for example, by a flash memory, an eMMC (embedded multi media card), an SSD (solid state drive), etc. In this embodiment, the auxiliary storage device 20 stores a k-space signal 21, a pseudo real part image 22, a k-space signal 23 multiplied by an imaginary unit, a pseudo imaginary part image 24, a magnetic resonance image 25, and an image reconstruction program 29.

[0038] The image reconstruction program 29 is a computer program for realizing the respective units 11 to 16 in the data processing unit 10, which are software-based functional blocks, described below. These functional blocks are realized by installing the image reconstruction program 29 in the auxiliary storage device 20 or memory of the reconstruction device 1 and having a processor execute the image reconstruction program 29. The image reconstruction program 29 may be installed in the reconstruction device 1 via a network 9, such as the Internet, connected via the communication I / F unit 33. Alternatively, the image reconstruction program 29 may be installed in the reconstruction device 1 by having the reconstruction device 1 read a computer-readable, non-transitory, tangible recording medium, such as a memory card, on which the image reconstruction program 29 is recorded. The image reconstruction program 29 may also be an application on, for example, a tablet terminal or the like.

[0039] The input unit 31 can be configured with, for example, a mouse, a keyboard, etc. The display unit 32 can be configured with, for example, a liquid crystal display, an organic EL display, etc. The input unit 31 and the display unit 32 can also be integrated as a touch panel.

[0040] The communication I / F unit 33 transmits and receives data to and from external devices such as the MRI apparatus 90 via a wired or wireless network 9. The communication I / F unit 33 may be connected via various wired or wireless connections such as Ethernet (registered trademark), Bluetooth (registered trademark), Wi-Fi (registered trademark), and ZigBee (registered trademark).

[0041] In this embodiment, the reconstruction device 1 includes, as a software configuration, a data processing unit 10. The data processing unit 10 is a functional block that is realized when a processor executes an image reconstruction program 29.

[0042] The k-space signal acquisition unit 11 acquires k-space signals 21 of the object 99, including real and imaginary part signals. In this embodiment, the k-space signal acquisition unit 11 acquires the k-space signals 21 from the MRI apparatus 90 via the network 9. The acquired k-space signals 21 are recorded in the auxiliary storage device 20.

[0043] The MRI device 90 images an object 99 by randomly thinning out measurement points of the imaging data (k-space signal 21), and the k-space signal 21 acquired from the MRI device 90 by the k-space signal acquisition unit 11 has measurement points that have been randomly thinned out. Because local phase distortion occurs in the imaging region of the MRI device 90, the k-space signal 21 acquired from the MRI device 90 is a complex signal generated by convolution of a function obtained by Fourier transforming the signal magnitude with a function obtained by Fourier transforming the phase that depends on the imaging position. Note that because quadrature detection is performed, the k-space signal 21 is a complex signal even if there is no local phase distortion.

[0044] The pseudo real part image reconstruction unit 12 reconstructs the pseudo real part image 22 based on the acquired k-space signal 21. In this embodiment, the pseudo real part image reconstruction unit 12 reconstructs the pseudo real part image 22 from the k-space signal 21 according to a known compressed sensing method. Preferably, the pseudo real part image reconstruction unit 12 estimates the pseudo real part image 22 by, for example, performing a wavelet transform on the k-space signal 21. More preferably, the pseudo real part image 22 is estimated by performing an inverse Fourier transform on the k-space signal 21 and minimizing the norm after the wavelet transform. The wavelet transform is performed to calculate the norm. The type of norm used to minimize the norm may be either the L1 norm (Manhattan distance) or the L2 norm (Euclidean distance). Preferably, the pseudo real part image reconstruction unit 12 initializes data of measurement points of the k-space signal 21 that have been thinned out and are missing before reconstructing the pseudo real part image 22. The initialization is performed, for example, by padding the data with zeros.

[0045] There are many variations of compressed sensing methods. The term "compressed sensing method" used in this specification encompasses these various variations. As a typical compressed sensing method, in this embodiment, an image is reconstructed from k-space signals by performing, for example, the following six steps 1 to 6.

[0046] Step 1: An image estimated from the k-space signal (estimated image) is obtained by performing an inverse Fourier transform on the k-space signal. Step 2: The estimated image obtained in step 1 is subjected to wavelet transform to calculate the L1 norm of the estimated image. Step 3: Calculate the variance of the estimated image obtained in step 1. Step 4: The degree of data mismatch is calculated between the k-space signal data and the data obtained by inversely transforming (i.e., Fourier transforming) the estimated image obtained in step 1. Step 5: The data of the estimated image is organized so that the sum of the L1 norm, variance, and discrepancy is small. Step 6: Using the estimated image data constructed in step 5, steps 2 to 5 are repeated to repeatedly construct an estimated image. The construction of an estimated image is continued until the L1 norm, variance, and sum of the disparity remain unchanged (for example, until the L1 norm, variance, and sum of the disparity are minimized). The estimated image data when the L1 norm, variance, and sum of the disparity are minimized is used as the image data estimated from the k-space signal by reconstruction (reconstructed image data).

[0047] The calculation in step 2 assumes that the image is sparse, transforms the image into a region that can be represented sparsely, and calculates that the representation of the image in that region after transformation is composed of a small number of elements. Therefore, the value of the L1 norm calculated in step 2 is independent of the consistency of the data. An image being sparse means that the image is composed of a small amount of information (the image is sparse).

[0048] In this specification, the term "Fourier transform" refers not only to the literal Fourier transform but also to the discrete Fourier transform and fast Fourier transform. The same applies to the inverse transform. In one variation of the compressed sensing method, instead of using the wavelet transform in step 2, a transform method other than the wavelet transform that converts to the frequency domain can be used depending on which domain is suitable for sparse representation.

[0049] The imaginary unit multiplication unit 13 multiplies the acquired k-space signal 21 by an imaginary unit. In this embodiment, the imaginary unit multiplication unit 13 multiplies the k-space signal 21, which is a complex signal, by −i (i is an imaginary unit). The k-space signal 23 multiplied by the imaginary unit is recorded in the auxiliary storage device 20.

[0050] The pseudo-imaginary part image reconstruction unit 14 reconstructs the pseudo-imaginary part image 24 based on the k-space signal 23 multiplied by the imaginary unit. In this embodiment, the pseudo-imaginary part image reconstruction unit 14 reconstructs the pseudo-imaginary part image 24 from the k-space signal 23 multiplied by the imaginary unit in accordance with a known compressed sensing method (for example, steps 1 to 6 described above as the procedure of a typical compressed sensing method). Preferably, the pseudo-imaginary part image reconstruction unit 14 estimates the pseudo-imaginary part image 24 by performing, for example, a wavelet transform on the k-space signal 23 multiplied by the imaginary unit. More preferably, the pseudo-imaginary part image 24 is estimated by performing an inverse Fourier transform on the k-space signal 23 multiplied by the imaginary unit and minimizing the norm (for example, either the L1 norm or the L2 norm) after the wavelet transform. Preferably, the pseudo-imaginary part image reconstruction unit 14 initializes the data of the measurement points of the k-space signal 23 multiplied by the imaginary unit, which are missing due to thinning, before reconstructing the pseudo-imaginary part image 24. The initialization is performed, for example, by padding the data with zeros.

[0051] The magnetic resonance image synthesis unit 15 synthesizes a magnetic resonance image 25 of the object 99 based on the pseudo real image 22 and the pseudo imaginary image 24. Preferably, the synthesis of the magnetic resonance image 25 is performed by calculating the L2 norm of the pseudo real image 22 and the pseudo imaginary image 24. The synthesized magnetic resonance image 25 is recorded in, for example, the auxiliary storage device 20. The magnetic resonance image display unit 16 displays the synthesized magnetic resonance image 25 on, for example, the display unit 32. [Processing Procedure]

[0052] FIG. 3 is a flowchart for explaining the procedure of processing performed by the magnetic resonance image reconstruction device according to one embodiment of the present invention.

[0053] In step S1, k-space signals 21 of an object 99, including real and imaginary part signals, are acquired. In step S2, a pseudo-real part image 22 is reconstructed based on the acquired k-space signals 21. In step S3, the acquired k-space signals 21 are multiplied by an imaginary unit. In step S4, a pseudo-imaginary part image 24 is reconstructed based on the k-space signals 23 multiplied by the imaginary unit. In step S5, a magnetic resonance image 25 of the object 99 is synthesized based on the pseudo-real part image 22 and the pseudo-imaginary part image 24. [effect]

[0054] As described above, the magnetic resonance image reconstruction device 1 and reconstruction method according to one embodiment of the present invention can reduce the imaging time while maintaining the image quality of the magnetic resonance image 25.

[0055] In magnetic resonance imaging using MRI, the magnitude of local phase distortion present within the imaging region varies depending on the magnitude of the static magnetic field applied to the object 99. In MRI with a relatively low static magnetic field (e.g., approximately 1.5 T or less), the magnitude of phase distortion is small and its effect can be largely ignored. However, in recent years, in order to acquire magnetic resonance image signals with high intensity and improve image resolution, the magnitude of the static magnetic field applied to the object 99 has become relatively high (e.g., approximately 3 T or more), and efforts are being made to increase the magnetic field strength of MRI.

[0056] The magnetic resonance image reconstruction device 1 according to one embodiment of the present invention solves the problem of local phase distortion, which is particularly noticeable in a high magnetic field (for example, about 3 T or higher), and can shorten the imaging time while maintaining the image quality of the magnetic resonance image 25, even in a relatively high magnetic field MRI. This makes it possible to achieve both improved diagnostic accuracy due to the improved image resolution achieved by the higher magnetic field of MRI, and reduced burden on the object 99 (for example, a subject) due to the shortened imaging time.

[0057] Furthermore, the magnetic resonance image reconstruction device 1 according to one embodiment of the present invention is realized by calculating imaging data (k-space signals 21) acquired from the MRI device 90, and there is no need to make any changes to the imaging procedure on the MRI device 90 side. As a result, the magnetic resonance image reconstruction device 1 according to one embodiment of the present invention does not require additional connection of new hardware to an existing MRI device, and can be implemented on any existing MRI device. Preparation on the MRI device 90 side simply requires randomly thinning out the measurement points and capturing an image of the object 99, thereby randomly thinning out the imaging data (k-space signals 21) to be provided to the reconstruction device 1 in advance. [Other forms]

[0058] Although the present invention has been described above with reference to specific embodiments, the present invention is not limited to the above-described embodiments.

[0059] In the above embodiment, the reconstruction device 1 is realized as an integrated device, but the reconstruction device 1 does not need to be an integrated device, and the processor, memory, auxiliary storage device 20, etc. may be located in different places and connected to each other via a network. The input unit 31 and the display unit 32 do not necessarily need to be located in the same place, and may be located in different places and connected to each other via the network 9 so that they can communicate with each other.

[0060] In the above embodiment, each functional block of the reconstruction device 1 is executed by a single processor, but each functional block does not necessarily have to be executed by a single processor and may be processed in a distributed manner by multiple processors. Furthermore, instead of a processor, a field programmable gate array (FPGA) may perform the processing, or, for example, a graphics processing unit (GPU) may be used as an accelerator to assist the parallel arithmetic processing performed by the processor. In other words, the processing performed by the processor includes processing performed by the processor or FPGA using an accelerator such as a GPU.

[0061] In the above embodiment, the functional blocks 11 to 16 constituting the data processing unit 10 are realized by software, but some or all of these functional blocks 11 to 16 may be realized as hardware. The processing of each of the functional blocks 11 to 16 constituting the data processing unit 10 does not need to be performed by a single processor, but may be distributed among multiple processors. Some or all of the functional blocks 11 to 16 constituting the data processing unit 10 and the data items 21 to 25 in the auxiliary storage device 20 may be cloud-based in another server device (not shown) connected via the communication I / F unit 33.

[0062] Examples of the present invention will be described below to clarify the features of the present invention. [Example]

[0063] In Example 1, magnetic resonance images were reconstructed and created using various methods, including the method according to the present invention, and the image quality of the created magnetic resonance images was evaluated. The image quality was evaluated visually and using quantitative indicators.

[0064] The subject of MRI imaging was a Japanese rice fish (medaka). Medaka is a small fish with tissues of different magnetic susceptibility densely concentrated in a small area, making it suitable for use in evaluating the image quality of magnetic resonance images. When the local magnetic field changes in a small area, phase distortion due to the local magnetic field also occurs in the small area, and the rate of phase change with location increases, resulting in a significant impact on image quality due to phase distortion. When acquiring MRI imaging data, the static magnetic field strength applied to the subject (medaka in this example) was approximately 14 T. <Visual evaluation>

[0065] FIG. 4 shows magnetic resonance images reconstructed by various methods, including a method according to an embodiment of the present invention. (A) is an image according to a comparative example, in which no phase correction has been applied. (B) is an image according to a comparative example, in which conventional phase correction (single phase correction) using phase detection has been applied. (C) is an image according to an embodiment of the present invention. (D) is an image according to a comparative example, in which the image has been reconstructed from complete imaging data in which no measurement points have been reduced. In the following discussion, image (D) is considered to be the correct image.

[0066] Images (A) to (C) were reconstructed according to the following procedure. First, imaging data with decimated measurement points was prepared, which was used in common for reconstructing images (A) to (C), by randomly decimating the imaging data from the complete imaging data with no measurement points reduced to 50%. Next, images (A) to (C) were reconstructed by applying the compressed sensing method to the prepared imaging data, according to the respective methods described for images (A) to (C).

[0067] The image quality of images (A) to (D) shown in Figure 4 will be considered. In Figure 4, reference numeral 41 indicates the head of the medaka, reference numeral 42 the telencephalon, reference numeral 43 the midbrain, reference numeral 44 the pharyngeal tissue, reference numeral 45 the kidney, reference numeral 46 the swim bladder, and reference numeral 47 the tail.

[0068] Focus on the left and right kidneys 45 in the correct image (D). In the image (C) of the example, the presence of the left and right kidneys 45 could be confirmed in the same way as in the correct image (D). In contrast, in the images (A) and (B) of the comparative example, the left and right kidneys 45 had disappeared.

[0069] Focus on the telencephalon 42 and midbrain 43 in the correct image (D). In the correct image (D), it was possible to confirm that the telencephalon 42 and midbrain 43 existed in a line, in contact with each other. In the image (C) of the example, it was possible to confirm the presence and contact state of the telencephalon 42 and midbrain 43, similar to the correct image (D). In contrast, in the images (A) and (B) of the comparative example, it was not possible to confirm the presence and contact state of the telencephalon 42 and midbrain 43. In the image (A) of the comparative example, the telencephalon 42 and midbrain 43 were in a disconnected state. In the image (B) of the comparative example, the telencephalon 42 had disappeared.

[0070] Focus on the pharyngeal tissue 44 in the correct image (D). In the correct image (D), the presence of the pharyngeal tissue 44 could be confirmed as tissue having the same extent as the midbrain 43. In the image (C) of the example, it could also be confirmed that the pharyngeal tissue 44 was tissue having the same extent as the midbrain 43. In contrast, in both of the images (A) and (B) of the comparative example, the peripheral portion of the pharyngeal tissue 44 was missing, and it was not depicted as a circular tissue.

[0071] From the above, it was confirmed that the image quality of the image (C) according to the embodiment of the present invention was maintained to a level that allowed differences in tissue in a living body to be distinguished to the same extent as the correct image (D). <Evaluation based on quantitative indicators>

[0072] For each of images (A) to (C), we calculated SSIM (structural similarity) and PSNR (Peak-Signal to Noise Ratio), which are quantitative indices for evaluating image quality. We then compared the image quality of images (A) to (C) based on the calculated indices. The results are shown in Table 1. Note that the percentages shown in parentheses for the SSIM and PSNR indices in Table 1 represent the reduction rate of measurement points before applying the compressed sensing method to reconstruct the image for each of images (A) to (C). For example, a reduction rate of measurement points of 30% means that the number of measurement points has been reduced by 30%.

[0073] [Table 1]

[0074] Consider the SSIM index. Images (A) and (B) from the comparative example all had SSIM values ​​of approximately 0.75 or less. In contrast, image (C) from the embodiment of the present invention confirmed that, under the condition of a reduction rate of 50%, the SSIM value was approximately 0.9 or more, which is considered to be the target for practical use.

[0075] Consider the PSNR index. Images (A) and (B) from the comparative example all had PSNR values ​​less than approximately 20 dB. In contrast, image (C) from the example of the present invention had a PSNR value that was significantly improved to approximately 25 dB to approximately 27 dB, exceeding the approximately 20 dB to approximately 25 dB range that can be judged as being practical, and approaching 30 dB, the lower limit at which it can be judged by visual inspection that there is no degradation in image quality.

[0076] From the above, it was confirmed that the image quality of the image (C) obtained in the embodiment of the present invention is of a level that allows practical use. [Explanation of symbols]

[0077] 1 Reconfiguration device 9 Network 10 Data Processing Unit 11 k space signal acquisition section 12 Pseudo real part image reconstruction unit 13 Imaginary unit multiplication 14 Pseudo-imaginary part image reconstruction unit 15 Magnetic Resonance Image Synthesis Unit 16 Magnetic resonance image display unit 20 Auxiliary storage 21 k-space signals 22 Pseudo Real Image 23 k-space signal (multiplied by the imaginary unit) 24 Pseudo-imaginary image 25 Magnetic Resonance Imaging 29 Image Reconstruction Program 31 Input section 32 Display section 33 Communication interface section (communication I / F section) 41 Head 42 Telencephalon 43 Midbrain 44 Pharyngeal tissue 45 Kidneys 46 Swim Bladder 47 Tail 90 MRI machine 91 Gradient magnetic field coil 92 Main magnet 93 High frequency receiving coil 99 Imaged object

Claims

1. acquiring k-space signals of an imaged object, the k-space signals including real and imaginary signals; reconstructing a pseudo-real image based on the k-space signals; multiplying the k-space signals by an imaginary unit; reconstructing a pseudo-imaginary image based on the k-space signals multiplied by an imaginary unit; synthesizing a magnetic resonance image of the imaging subject based on the pseudo real image and the pseudo imaginary image; A method for reconstructing a magnetic resonance image, comprising:

2. The method of claim 1 , wherein the k-space signal has measurement points randomly thinned out.

3. 2. The method of claim 1, wherein the step of multiplying by an imaginary unit multiplies the k-space signal by −i, where i is the imaginary unit.

4. the step of reconstructing the pseudo-real part image includes estimating the pseudo-real part image from the k-space signal according to a compressed sensing method; The method of claim 1 , wherein the step of reconstructing the pseudo-imaginary image estimates the pseudo-imaginary image from the k-space signals multiplied by an imaginary unit according to a compressed sensing method.

5. The method of claim 4 , wherein the step of reconstructing the pseudo-real part image estimates the pseudo-real part image by performing an inverse Fourier transform and a wavelet transform on the k-space signal.

6. The method of claim 4 , wherein the step of reconstructing the pseudo-imaginary image estimates the pseudo-imaginary image by performing an inverse Fourier transform and a wavelet transform on the k-space signals multiplied by an imaginary unit.

7. The method of claim 1 , wherein the step of synthesizing the magnetic resonance images of the imaging subject comprises calculating an L2 norm of the pseudo-real image and the pseudo-imaginary image.

8. an acquisition unit that acquires k-space signals of an imaging object, including real and imaginary signals; a pseudo-real part image reconstruction unit that reconstructs a pseudo-real part image based on the k-space signal; an imaginary unit multiplier that multiplies the k-space signal by an imaginary unit; a pseudo-imaginary part image reconstruction unit that reconstructs a pseudo-imaginary part image based on the k-space signal multiplied by an imaginary unit; a magnetic resonance image synthesis unit that synthesizes a magnetic resonance image of the imaging target based on the pseudo real image and the pseudo imaginary image; A magnetic resonance image reconstruction device comprising:

9. A program for causing a computer to execute each step of the method according to any one of claims 1 to 7.

Images