MR signal processing device

Abstract

To improve accuracy of MRS signals.SOLUTION: An MR signal processing device comprises a parameter output unit. The parameter output unit applies a learned model to a plurality of MRS signals collected via MR spectroscopy with respect to the same object to output a plurality of parameters for MRS reconfiguration.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The embodiments disclosed in the present specification and drawings relate to an MR signal processing device. [Background technology]

[0002] Data acquisition by magnetic resonance spectroscopy (MRS) is known to be extremely unstable. Therefore, the number of integrations is set to about 50 to 100, and the MRS signals acquired in each integration are summed to obtain an average MRS signal. However, subject movement and static magnetic field disturbances between multiple data acquisitions can cause distortions in the average MRS signal.

[0003] A known technique is to generate MRS signals by least-squares fitting. Approximately 20 artificial spectra of metabolites, etc. are generated using a physical model, and the MRS spectrum of the target object is generated by fitting each artificial spectrum with spectral parameters such as signal intensity values. Another known technique is to use a deep neural network to represent the regression process that obtains spectral parameters for each artificial spectrum from a single MRS spectrum. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Nima Hatami, Michael Sdika, Helene Ratiney, “Magnetic Resonance Spectroscopy Quantification using Deep Learning”, arXiv:1806.07237v1 [cs.CV] 19 Jun 2018. Summary of the Invention [Problem to be solved by the invention]

[0005] One of the problems to be solved by the embodiments disclosed in this specification and the drawings is to improve the accuracy of MRS signals. However, the problems to be solved by the embodiments disclosed in this specification and the drawings are not limited to the above problem. Problems corresponding to the effects of the configurations of the embodiments described below can also be positioned as other problems. [Means for solving the problem]

[0006] The MR signal processing device according to the embodiment includes a parameter output unit that applies a trained model to multiple MRS signals acquired by MR spectroscopy for the same subject and outputs multiple parameters for MRS reconstruction. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a diagram showing an example of the configuration of a magnetic resonance imaging apparatus according to the first embodiment. [Figure 2] FIG. 2 is a diagram schematically illustrating the input / output relationship of the trained model (amount of substance estimation NN) according to the first embodiment. [Figure 3] FIG. 3 is a diagram schematically showing an example of an MRS spectrum. [Figure 4] FIG. 4 is a diagram showing an example of the flow of signal processing by the MR signal processing device according to the first embodiment. [Figure 5] FIG. 5 is a diagram schematically showing a specific example of the input / output relationship of the substance amount estimation NN according to the first embodiment. [Figure 6] FIG. 6 is a diagram showing a schematic diagram of the process of generating a composite MRS spectrum in step SA3 of FIG. [Figure 7] FIG. 7 is a diagram schematically showing a specific example of the input / output relationship of the substance amount estimation NN according to the first modification of the first embodiment. [Figure 8] FIG. 8 is a diagram schematically showing a specific example of the input / output relationship of the substance amount estimation NN according to the second modification of the first embodiment. [Figure 9]FIG. 9 is a diagram schematically showing a specific example of the input / output relationship of the substance amount estimation NN according to the third modification of the first embodiment. [Figure 10] FIG. 10 is a diagram schematically showing a specific example of the input / output relationship of the substance amount estimation NN according to the fourth modification of the first embodiment. [Figure 11] FIG. 11 is a diagram illustrating an example of a network configuration of a first integrated model according to the fifth modification of the first embodiment. [Figure 12] FIG. 12 is a diagram illustrating an example of a network configuration of another first integrated model according to the fifth modification of the first embodiment. [Figure 13] FIG. 13 is a diagram showing an example of the configuration of a magnetic resonance imaging apparatus according to the second embodiment. [Figure 14] FIG. 14 is a diagram schematically illustrating the input / output relationship of the trained model (basis component amount estimation NN) according to the second embodiment. [Figure 15] FIG. 15 is a diagram showing an example of the flow of signal processing by the MR signal processing device according to the second embodiment. [Figure 16] FIG. 16 is a diagram schematically showing the input / output relationship of the basis component amount estimation NN according to the second embodiment. [Figure 17] FIG. 17 is a diagram schematically showing the normality / abnormality determination process (determination result=normal) based on the base component amount information, which is performed in step SB4 of FIG. [Figure 18] FIG. 18 is a diagram schematically showing the normality / abnormality determination process (determination result=abnormality) based on the base component amount information, which is performed in step SB4 of FIG. [Figure 19] FIG. 19 is a diagram showing an example of a display screen of the composite MRS spectrum and the determination results displayed in step SB5 of FIG. [Figure 20] FIG. 20 is a diagram showing an example of the configuration of a magnetic resonance imaging apparatus according to the third embodiment. [Figure 21] FIG. 21 is a diagram schematically illustrating the input / output relationship of the trained model (modulation amount estimation NN) according to the third embodiment. [Figure 22]FIG. 22 is a diagram showing an example of the flow of signal processing by the MR signal processing device according to the third embodiment. [Figure 23] FIG. 23 is a diagram schematically showing a specific example of the input / output relationship of the modulation amount estimation NN according to the third embodiment. [Figure 24] FIG. 24 is a diagram schematically showing the correction process and spectrum generation process in steps SC3 and SC4 of FIG. [Figure 25] FIG. 25 is a diagram showing an example of the configuration of a magnetic resonance imaging apparatus according to the fourth embodiment. [Figure 26] FIG. 26 is a diagram illustrating an example of a network configuration of the second integrated model according to the fourth embodiment. [Figure 27] FIG. 27 is a diagram illustrating an example of a network configuration of another second integrated model according to the fourth embodiment. [Figure 28] FIG. 28 is a diagram schematically illustrating the input / output relationship of the trained model (modulation amount estimation NN) according to the fifth embodiment. [Figure 29] FIG. 29 is a diagram schematically showing a specific example of the input / output relationship of the modulation amount estimation NN according to the fifth embodiment. [Figure 30] FIG. 30 is a diagram schematically showing a further specific example of the input / output relationship of the modulation amount estimation NN shown in FIG. [Figure 31] FIG. 31 is a diagram schematically illustrating the input / output relationship of the trained model NN of the sixth embodiment. [Figure 32] FIG. 32 is a diagram schematically illustrating a processing example according to the seventh embodiment. [Figure 33] FIG. 33 is a diagram schematically illustrating a processing example according to the eighth embodiment. [Figure 34] FIG. 34 is a diagram schematically illustrating a processing example according to the ninth embodiment. [Figure 35] FIG. 35 is a diagram showing an example of the configuration of a magnetic resonance imaging system including an MR signal processing device according to the tenth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] Hereinafter, an embodiment of an MR signal processing device will be described in detail with reference to the drawings.

[0009] The MR signal processing device according to this embodiment is a computer that processes MR signals collected by a magnetic resonance imaging device. The MR signal processing device may be a computer incorporated in the magnetic resonance imaging device, or may be a computer separate from the magnetic resonance imaging device. Hereinafter, the MR signal processing device will be described in several embodiments.

[0010] (First embodiment) Fig. 1 is a diagram showing an example of the configuration of a magnetic resonance imaging apparatus 1 according to the first embodiment. As shown in Fig. 1, the magnetic resonance imaging apparatus 1 includes a gantry 11, a bed 13, a gradient magnetic field power supply 21, a transmission circuit 23, a reception circuit 25, a bed driving device 27, a sequence control circuit 29, and an MR signal processing device (host computer) 50.

[0011] The gantry 11 has a static magnetic field magnet 41 and a gradient magnetic field coil 43. The static magnetic field magnet 41 and the gradient magnetic field coil 43 are housed in a housing of the gantry 11. A hollow bore is formed in the housing of the gantry 11. A transmitting coil 45 and a receiving coil 47 are arranged in the bore of the gantry 11.

[0012] The static magnetic field magnet 41 has a hollow, approximately cylindrical shape and generates a static magnetic field inside the approximately cylinder. For example, a permanent magnet, a superconducting magnet, or a normal-conducting magnet may be used as the static magnetic field magnet 41. Here, the central axis of the static magnetic field magnet 41 is defined as the Z axis, the axis perpendicular to the Z axis is defined as the Y axis, and the axis horizontally perpendicular to the Z axis is defined as the X axis. The X axis, Y axis, and Z axis form an orthogonal three-dimensional coordinate system.

[0013] The gradient coil 43 is a hollow, approximately cylindrical coil unit attached to the inside of the static magnetic field magnet 41. The gradient coil 43 generates a gradient magnetic field by receiving a current from the gradient power supply 21. More specifically, the gradient coil 43 has three coils corresponding to the X-axis, Y-axis, and Z-axis, which are orthogonal to each other. The three coils form gradient magnetic fields whose field strength varies along each of the X-axis, Y-axis, and Z-axis. The gradient magnetic fields along the X-axis, Y-axis, and Z-axis are combined to form a slice selection gradient magnetic field Gs, a phase encoding gradient magnetic field Gp, and a frequency encoding gradient magnetic field Gr, which are orthogonal to each other, in desired directions. The slice selection gradient magnetic field Gs is used to arbitrarily determine an imaging plane (slice). The phase encoding gradient magnetic field Gp is used to change the phase of a magnetic resonance signal (hereinafter referred to as an MR signal) according to a spatial position. The frequency encoding gradient magnetic field Gr is used to change the frequency of the MR signal according to a spatial position. In the following description, the gradient direction of the slice selection gradient magnetic field Gs is the Z axis, the gradient direction of the phase encoding gradient magnetic field Gp is the Y axis, and the gradient direction of the frequency encoding gradient magnetic field Gr is the X axis.

[0014] The gradient magnetic field power supply 21 supplies a current to the gradient magnetic field coil 43 in accordance with a sequence control signal from the sequence control circuit 29. The gradient magnetic field power supply 21 supplies a current to the gradient magnetic field coil 43, thereby causing the gradient magnetic field coil 43 to generate gradient magnetic fields along the X-axis, Y-axis, and Z-axis. The gradient magnetic fields are superimposed on the static magnetic field formed by the static magnetic field magnet 41 and applied to the subject P.

[0015] The transmission coil 45 is disposed, for example, inside the gradient magnetic field coil 43, and receives a current from the transmission circuit 23 to generate a radio frequency pulse (hereinafter referred to as an RF pulse).

[0016] The transmission circuitry 23 supplies a current to the transmission coil 45 to apply an RF pulse to the subject P via the transmission coil 45 to excite target protons present in the subject P. The RF pulse oscillates at a resonance frequency specific to the target protons, exciting the target protons. An MR signal is generated from the excited target protons and detected by the reception coil 47. The transmission coil 45 is, for example, a whole-body coil (WB coil). The whole-body coil may be used as a transmission / reception coil.

[0017] The receive coil 47 receives MR signals emitted from target protons present in the subject P in response to the action of RF pulses. The receive coil 47 has multiple receive coil elements capable of receiving MR signals. The received MR signals are supplied to the receiver circuit 25 via wire or wirelessly. Although not shown in FIG. 1 , the receive coil 47 has multiple receive channels implemented in parallel. Each receive channel has receive coil elements that receive MR signals and amplifiers that amplify the MR signals. MR signals are output for each receive channel. The total number of receive channels and the total number of receive coil elements may be the same, or the total number of receive channels may be greater or less than the total number of receive coil elements.

[0018] The receiver circuitry 25 receives MR signals generated from excited target protons via the receiver coil 47. The receiver circuitry 25 processes the received MR signals to generate digital MR signals. The digital MR signals can be expressed in k-space defined by spatial frequencies. Therefore, hereinafter, the digital MR signals will be referred to as k-space data. The k-space data is an example of an MR acquisition signal. The k-space data is supplied to the host computer 50 via a wired or wireless connection.

[0019] The above-described transmitting coil 45 and receiving coil 47 are merely examples. A transmitting / receiving coil having both transmitting and receiving functions may be used instead of the transmitting coil 45 and receiving coil 47. Furthermore, the transmitting coil 45, receiving coil 47, and transmitting / receiving coil may be combined.

[0020] A bed 13 is installed adjacent to the gantry 11. The bed 13 has a top plate 131 and a base 133. A subject P is placed on the top plate 131. The base 133 supports the top plate 131 so that it can slide along the X-axis, Y-axis, and Z-axis. A bed driving device 27 is housed in the base 133. The bed driving device 27 moves the top plate 131 under the control of a sequence control circuit 29. The bed driving device 27 may include any motor, such as a servo motor or a stepping motor.

[0021] The sequence control circuit 29 has, as hardware resources, a processor such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit) and memories such as a ROM (Read Only Memory) and a RAM (Random Access Memory). The sequence control circuit 29 synchronously controls the gradient magnetic field power supply 21, the transmission circuit 23, and the reception circuit 25 based on data acquisition conditions set by a condition setting function 512 of the processing circuit 51A, and performs data acquisition on the subject P according to the data acquisition conditions to acquire k-space data regarding the subject P. The sequence control circuit 29 is an example of a sequence control unit.

[0022] The sequence control circuit 29 according to this embodiment executes data collection for MR spectroscopy, which is a type of chemical shift measurement. Chemical shift measurement is a technique for measuring chemical shifts, which are minute differences in the resonance frequencies of target protons, such as hydrogen nuclei, that occur depending on differences in chemical environment. MR spectroscopy includes a single-voxel method, in which data is collected for a single voxel, and a multi-voxel method, in which data is collected for multiple voxels. This embodiment is applicable to either method. The multi-voxel method is also called chemical shift imaging (CSI) or MRS imaging (MRSI). The voxel to be measured is called the voxel of interest.

[0023] The sequence control circuit 29 executes data acquisition for MR spectroscopy on the subject P. By executing data acquisition for MR spectroscopy, a free induction decay (FID) signal or a spin echo signal is generated from a voxel of interest in the subject P. The receiver circuit 25 receives the FID signal or the spin echo signal via the receiver coil 47 and performs signal processing on the received FID signal or the spin echo signal to acquire k-space data for the voxel of interest. The acquired k-space data is assumed to be digital data that represents the signal intensity value emitted from the voxel of interest as a function of time. The pulse sequence for MR spectroscopy is repeated a number of times equal to the number of excitations (NEX), and k-space data corresponding to the number of excitations is acquired. Hereinafter, the k-space data acquired by MR spectroscopy will be referred to as MRSk data. MRSk data is an example of an MRS signal.

[0024] 1, the MR signal processing device 50 is a computer having a processing circuit 51A, a memory 53, a display 55, an input interface 57, and a communication interface 59. The processing circuit 51A is an example of a processing unit, the memory 53 is an example of a storage unit, the display 55 is an example of a display unit, the input interface 57 is an example of an input unit, and the communication interface 59 is an example of a communication unit.

[0025] The processing circuitry 51A has a processor such as a CPU as a hardware resource. The processing circuitry 51A functions as the core of the magnetic resonance imaging apparatus 1. For example, the processing circuitry 51A executes various programs to realize an acquisition function 511, a condition setting function 512, a signal processing function 513, a substance amount output function 514, a synthesis function 515, a learning function 516, and a display control function 517. The acquisition function 511 is an example of an acquisition unit, the condition setting function 512 is an example of a setting unit, the signal processing function 513 is an example of a spectrum generation unit, the substance amount output function 514 is an example of a parameter output unit, the synthesis function 515 is an example of a synthesis unit, the learning function 516 is an example of a learning unit, and the display control function 517 is an example of a display unit.

[0026] In the acquisition function 511, the processing circuitry 51A acquires various MRS signals. For example, the processing circuitry 51A acquires MRSk data collected by the sequence control circuitry 29. The processing circuitry 51A may acquire the MRSk data directly from the sequence control circuitry 29 or the receiving circuitry 25, or may acquire the MRSk data that has been temporarily stored in the memory 53 from the memory 53.

[0027] In the condition setting function 512, the processing circuitry 51A automatically or manually sets data acquisition conditions. Specifically, in this embodiment, data acquisition conditions related to MR spectroscopy are set. Data acquisition conditions related to MR spectroscopy include, for example, a pulse sequence, a repetition time (TR), an echo time (TE), the number of integrations, a spectral width, the number of samples, a data acquisition method, and a region-selective pulse. Known pulse sequences include, for example, point-resolved spectroscopy (PRESS) and stimulated echo acquisition mode (STEAM). For a long TR, the TR is preferably set to, for example, 5000 ms or more, and for a short TR, it is preferably set to, for example, approximately 1000 to 3000 ms. The longer the TR, the closer the obtained MR signal intensity value is to the true value, but the longer the data acquisition time. For a long TE, the TE is preferably set to, for example, approximately 100 to 300 ms, and for a short TE, it is preferably set to, for example, approximately 20 to 100 ms. A shorter TE increases the number of peaks and improves the accuracy of the MRS spectrum, whereas a longer TE decreases the number of peaks and improves the visibility of the MRS spectrum.

[0028] There are no particular restrictions on the number of accumulations, and it should be set to 1 or more. The spectral width number and the number of samples are condition items related to spectral resolution. The spectral width number and the number of samples can be set to any value. As described above, data collection methods include the single-voxel method, which obtains an MRS spectrum for one voxel in one data collection, and the multi-voxel method, which obtains MRS spectra for each of multiple voxels in one data collection. Region-selective pulses include pulses that excite hydrogen nuclei limited to a set region, and pulses that do not excite hydrogen nuclei limited to a set region. Data collection conditions include whether or not to apply a region-selective pulse, frequency information for the selected region, etc.

[0029] In the signal processing function 513, the processing circuitry 51A performs various signal processing on the MRS signals acquired by the acquisition function 511. For example, the processing circuitry 51A generates a spectrum (hereinafter referred to as an MRS spectrum) indicating the signal intensity for each chemical shift based on the MRSk data. The MRS spectrum is an example of an MRS signal.

[0030] In the substance quantity output function 514, the processing circuitry 51A applies the trained model to multiple MRS signals obtained by MR spectroscopy for the same measurement target region of the subject P, and outputs multiple parameters for MRS reconstruction. The multiple parameters according to the first embodiment are spectral parameters for fitting based on multiple spectral models corresponding to the multiple parameters. The multiple spectral models are multiple artificial spectra corresponding to multiple substances contained in the measurement target region of the subject P, respectively. The multiple spectral parameters include substance quantity information for multiple substances related to the measurement target region of the subject P. The substance quantity information refers to information related to the substance quantities of the substances. Hereinafter, the parameters according to the first embodiment will be referred to as substance quantity information. The trained model is a machine learning model trained to input multiple MRS signals and output substance quantity information. A neural network or a deep neural network is used as the machine learning model. Hereinafter, the trained model according to the first embodiment will be referred to as a substance quantity estimation NN.

[0031] In the synthesis function 515, the processing circuitry 51A generates an MRS spectrum for the subject P by fitting based on the plurality of spectral parameters output by the substance quantity output function 514 and the plurality of spectral models. The MRS spectrum generated by the synthesis function 515 is referred to as a synthetic MRS spectrum.

[0032] In the learning function 516, the processing circuit 51A generates a substance amount estimation NN used in the substance amount output function 514 by machine learning based on the learning data.

[0033] In the display control function 517, the processing circuitry 51A displays various information on the display 55. For example, the processing circuitry 51A displays on the display 55 an MRS spectrum, substance amount information, a composite MRS spectrum, a setting screen for data acquisition conditions, and the like.

[0034] The memory 53 is a storage device such as an HDD (Hard Disk Drive), an SSD (Solid State Drive), or an integrated circuit storage device that stores various information. The memory 53 may also be a drive device that reads and writes various information from and to a portable storage medium such as a CD-ROM drive, a DVD drive, or a flash memory. For example, the memory 53 stores a substance amount estimation NN, data acquisition conditions, MRS signals, a control program, and the like.

[0035] The display 55 displays various information using a display control function 517. As the display 55, for example, a CRT display, a liquid crystal display, an organic EL display, an LED display, a plasma display, or any other display known in the art can be used as appropriate.

[0036] The input interface 57 includes input devices that accept various commands from the user. Examples of input devices that can be used include a keyboard, a mouse, various switches, a touch screen, and a touch pad. Note that the input devices are not limited to those equipped with physical operating parts such as a mouse and a keyboard. For example, an example of the input interface 57 also includes an electrical signal processing circuit that receives an electrical signal corresponding to an input operation from an external input device provided separately from the magnetic resonance imaging apparatus 1 and outputs the received electrical signal to various circuits. The input interface 57 may also be a voice recognition device that converts a voice signal collected by a microphone into an instruction signal.

[0037] The communication interface 59 is an interface that connects the magnetic resonance imaging apparatus 1 to a workstation, a PACS (Picture Archiving and Communication System), an HIS (Hospital Information System), a RIS (Radiology Information System), etc. via a LAN (Local Area Network), etc. The network IF transmits and receives various types of information to and from the connected workstation, PACS, HIS, and RIS.

[0038] The above configuration is merely an example and is not limiting. For example, the sequence control circuit 29 may be incorporated into the MR signal processing device 50. Furthermore, the sequence control circuit 29 and the processing circuit 51A may be mounted on the same board. Furthermore, the condition setting function 512 does not necessarily have to be mounted in the processing circuit 51A of the magnetic resonance imaging apparatus 1. For example, the condition setting function 512 may be mounted in a computer for setting data acquisition conditions that is separate from the magnetic resonance imaging apparatus 1. In this case, the data acquisition conditions generated by the computer are supplied to the magnetic resonance imaging apparatus 1 via a network, a portable recording medium, or the like. Furthermore, the storage area for the data acquisition conditions in the memory 53 does not have to be mounted in the magnetic resonance imaging apparatus 1, and may be mounted, for example, in a storage device connected to the magnetic resonance imaging apparatus 1 via a network.

[0039] An example of the operation of the MR signal processing device 50 according to the first embodiment will be described in detail below.

[0040] As described above, the substance amount output function 514 allows the processing circuitry 51A to apply the substance amount estimation NN to a plurality of MRS signals and output substance amount information of each substance contained in the measurement target region of the plurality of MRS signals.

[0041] FIG. 2 is a diagram schematically illustrating the input / output relationship of a trained model (amount of substance estimation NN) according to the first embodiment. As shown in FIG. 2, the amount of substance estimation NN is a machine learning model in which parameters have been trained so as to receive a first MRS signal and a second MRS signal as input and output amount of substance information regarding each substance contained in the measurement target region of the first MRS signal and the second MRS signal. The first MRS signal and the second MRS signal are MRS signals for the same subject P. The data acquisition conditions for the first MRS signal and the second MRS signal may be the same or different. The MRS signal input to the amount of substance estimation NN can be MRSk data or an MRS spectrum. Although the number of MRS signals input to the amount of substance estimation NN shown in FIG. 2 is two, this is an example, and any number greater than two may be used.

[0042] The substance quantity information is information regarding the substance quantity of a substance to be discriminated by MR spectroscopy. The substance to be discriminated is not particularly limited, and any substance that may be contained in the measurement target region can be discriminated. The substance to be discriminated varies depending on the measurement target region, but various molecules such as propylene glycol, ethanol, acetate, and acetone are known. The substance quantity may be a numerical value indicating signal intensity, a class according to signal intensity, or any analytical value calculated based on signal intensity. The substance quantity may be a discrete value including 0 or a continuous value. The substance quantity corresponds to a spectral parameter assigned to an artificial spectrum for fitting the artificial spectrum, which is a spectral model corresponding to the substance.

[0043] FIG. 3 is a diagram schematically illustrating an example of an MRS spectrum. The MRS spectrum shown in FIG. 3 is an example of an MRS spectrum of the head. As shown in FIG. 3, the vertical axis of the MRS spectrum is defined as the MR signal intensity value [AU (Arbitrary Unit)], and the horizontal axis is defined as the difference from the reference frequency, i.e., the chemical shift [ppm (parts per million)]. The reference frequency is set to the frequency of an arbitrarily selected reference material. The reference material is not particularly limited, but is set to, for example, tetramethylsilane (TMS). The MRS spectrum makes it possible to visualize the amounts of various substances, such as propylene glycol, ethanol, acetate, and acetone, present in the voxel of interest.

[0044] The MRS spectrum according to this embodiment may be waveform data of the MRS spectrum, or may be numerical data representing a combination of an identifier (name or symbol of the substance) of a substance corresponding to a peak and an amount of the substance, or may be numerical data representing a combination of a frequency difference corresponding to a peak and an amount of the substance. Furthermore, the amount-of-substance information may include reliability information regarding the amount of the substance in addition to the amount of the substance. The reliability information corresponds to the half-width of the peak included in the MRS spectrum. For example, the full width at half maximum or the half width at half maximum may be used as the half-width.

[0045] The substance amount estimation NN is generated by the learning function 516 of the processing circuit 51A. The processing circuit 51A trains a machine learning model based on a plurality of training samples to generate the substance amount estimation NN. The training samples are a combination of first and second MRS signals, which are input data, and substance amount information, which is correct data (hereinafter referred to as correct substance amount information). The first and second MRS signals, which are input data, are generated by the magnetic resonance imaging apparatus 1 or another magnetic resonance imaging apparatus. The correct substance amount information is obtained by signal analysis based on the input data. The processing circuit 51A applies the machine learning model to the first and second MRS signals to perform forward propagation processing and outputs substance amount information (hereinafter referred to as estimated substance amount information). Next, the processing circuit 51A applies the difference (error) between the estimated substance amount information and the correct substance amount information to the machine learning model to perform back propagation processing and calculates a gradient vector, which is a differential coefficient of an error function, which is a function of parameters. Next, the processing circuit 51A updates the parameters of the machine learning model based on the gradient vector. These forward propagation, back propagation, and parameter update processes are repeated while changing the training sample, and the parameters that minimize the error function are determined according to a predetermined optimization method. This generates a substance amount estimation NN.

[0046] Next, signal processing by the MR signal processing apparatus 50 according to the first embodiment will be described with reference to Fig. 4. In the following description, it is assumed that the MRS signal input to the substance amount estimation NN is an MRS spectrum.

[0047] FIG. 4 is a diagram showing an example of the flow of signal processing by the MR signal processing device 50 according to the first embodiment.

[0048] As shown in FIG. 4, the processing circuitry 51A acquires a first MRS spectrum and a second MRS spectrum by implementing the acquisition function 511 (step SA1). The first MRS spectrum and the second MRS spectrum are assumed to be MRS spectra for the same measurement target region acquired under the same data acquisition conditions, for example, by the single-voxel method. In MR spectroscopy, the number of accumulations is set to, for example, about 50 to 150, as a data acquisition condition, and MRS spectra corresponding to the number of accumulations are generated by the processing circuitry 51A. The processing circuitry 51A performs addition processing, such as addition or averaging, on the MRS spectra corresponding to the number of accumulations to generate one MRS spectrum. In step SA1, the processing circuitry 51A selects and acquires any two MRS spectra from the MRS spectra corresponding to the number of accumulations before the addition processing as the first MRS spectrum and the second MRS spectrum. The two MRS spectra are not particularly limited. They may be two MRS spectra acquired at consecutive data acquisition times or two MRS spectra separated in time. The two MRS spectra may be selected manually by the user according to an instruction via the input interface 57 or automatically according to a predetermined algorithm.

[0049] The addition process may be performed on the MRSk data corresponding to the number of integration times. In this case, the processing circuitry 51A may select two MRSk data from the MRSk data corresponding to the number of integration times before the addition process, generate two MRS spectra from the selected two MRSk data, and acquire the generated two MRS spectra as the first MRS spectrum and the second MRS spectrum.

[0050] When step SA1 is performed, the processing circuitry 51A, by implementing the substance amount output function 514, applies the substance amount estimation NN to the first MRS spectrum and the second MRS spectrum acquired in step SA1 and outputs substance amount information (step SA2).

[0051] FIG. 5 is a diagram schematically illustrating a specific example of the input / output relationship of the substance amount estimation NN. As shown in FIG. 5, the substance amount estimation NN is a machine learning model in which parameters are trained so as to input a first MRS spectrum and a second MRS spectrum and output substance amount information regarding substances contained in the measurement target region of the first MRS spectrum and the second MRS spectrum. Specific examples of the substance amount information include the substance amount M1 and half-width W1 of propylene glycol, the substance amount M2 and half-width W2 of ethanol, the substance amount M3 and half-width W3 of acetate, and the substance amount M4 and half-width W4 of acetone. Note that if multiple peaks exist for one substance, the substance amount and half-width are output for each peak. Peaks may be identified by frequency differences.

[0052] The processing circuitry 51A applies the amount-of-substance estimation NN to the first and second MRS spectra acquired in step SA1, and outputs the amount-of-substance M1 and half-width W1 of propylene glycol, the amount-of-substance M2 and half-width W2 of ethanol, the amount-of-substance M3 and half-width W3 of acetate, and the amount-of-substance M4 and half-width W4 of acetone. According to the first embodiment, amount-of-substance information can be estimated from a plurality of MRS spectra related to the same measurement target site, and therefore amount-of-substance information of substances contained in the measurement target site can be estimated with high accuracy.

[0053] When step SA2 is performed, the processing circuitry 51A generates a composite MRS spectrum based on the substance amount information output in step SA2 by implementing the synthesis function 515 (step SA3).

[0054] FIG. 6 is a diagram schematically illustrating the synthetic MRS spectrum generation process in step SA3. As shown in FIG. 6, it is assumed that in step SA2, substance amount information IM1 for propylene glycol, substance amount information IM2 for ethanol, substance amount information IM3 for acetate, and substance amount information IM4 for acetone are output. Meanwhile, artificial spectra corresponding to each substance to be discriminated by the substance amount estimation NN are generated in advance and stored in memory 53 or the like. The artificial spectra are theoretical MRS spectra for the substances and are generated by predictive calculations using physical models or the like. The artificial spectra can be considered ideal MRS spectra free of noise resulting from instability in MR spectroscopy data collection, body movement of the subject P, disturbances in the static magnetic field, and the like. In the example of FIG. 6, an artificial spectrum ES1 for propylene glycol, an artificial spectrum ES2 for ethanol, an artificial spectrum ES3 for acetate, and an artificial spectrum ES4 for acetone are generated in advance.

[0055] In step SA3, the processing circuitry 51A generates a composite MRS spectrum based on the amount of substance and half-width of each of the plurality of substances and the plurality of artificial spectra corresponding to the plurality of substances. More specifically, the processing circuitry 51A generates a plurality of corrected artificial spectra by assigning the amount of substance and half-width of each of the plurality of substances to the artificial spectra of the plurality of substances, and generates a composite MRS spectrum by fitting based on the generated plurality of corrected artificial spectra.

[0056] Specifically, the processing circuitry 51A applies the mass of substance M1 and half-width W1 of propylene glycol to the artificial spectrum ES1 to generate a corrected artificial spectrum of propylene glycol (step SA31). Similarly, corrected artificial spectra are generated for ethanol, acetate, and acetone (step SA31). Next, the processing circuitry 51A fits the corrected artificial spectrum of propylene glycol, the corrected artificial spectrum of ethanol, the corrected artificial spectrum of acetate, and the corrected artificial spectrum of acetone to generate a composite MRS spectrum FS1 (step SA32). Any fitting method, such as least-squares fitting, may be used. Since the composite MRS spectrum FS1 is generated based on ideal artificial spectra ES1 to ES4, noise due to instability in MR spectroscopy data collection, body movement of the subject P, disturbances in the static magnetic field, etc., is reduced compared to the input first and second MRS spectra. In other words, the substance amount estimation by the substance amount estimation NN (step SA2) and the generation of a synthetic MRS spectrum based on the substance amount information (step SA3) can be regarded as the same as the denoising process of the first MRS spectrum and the second MRS spectrum.

[0057] After step SA3 is performed, the processing circuitry 51A, by implementing the display control function 517, displays the composite MRS spectrum generated in step SA3 (step SA4). The composite MRS spectrum is displayed on the display 55. This allows the user to observe the composite MRS spectrum, which is a denoised MRS spectrum.

[0058] With the above, the signal processing by the MR signal processing device 50 according to the first embodiment is completed.

[0059] The first embodiment can be modified in various ways.

[0060] (Variation 1) 7 is a diagram schematically illustrating a specific example of the input / output relationship of a substance amount estimation NN according to Modification 1 of the first embodiment. The substance amount estimation NN according to Modification 1 is a machine learning model in which parameters are trained so as to receive a first MRS spectrum and a second MRS spectrum as input and output substance amount information related to substances contained in the measurement target region of the first MRS spectrum and the second MRS spectrum. The first MRS spectrum is an MRS spectrum collected by MR spectroscopy using a pulse sequence without application of a region-selective pulse. The second MRS spectrum is an MRS spectrum collected by MR spectroscopy using a pulse sequence with application of a region-selective pulse.

[0061] Application of a region-selective pulse excites hydrogen nuclei of a substance belonging to a specific frequency band, generating a second MRS spectrum showing the chemical shift distribution of signal intensities from the excited hydrogen nuclei. Alternatively, application of a region-selective pulse does not excite hydrogen nuclei of a substance belonging to a specific frequency band, generating a second MRS spectrum showing the chemical shift distribution of signal intensities from hydrogen nuclei of substances belonging to other frequency bands. On the other hand, when a region-selective pulse is not applied, hydrogen nuclei of substances belonging to all frequency bands are excited, generating a first MRS spectrum showing the chemical shift distribution of signal intensities from hydrogen nuclei of substances belonging to all frequency bands. The region-selective pulse is used to selectively excite or de-excite a substance belonging to the specific frequency band among multiple substances in order to separate multiple peaks on the MRS spectrum that are caused by different substances but have overlapping chemical shifts (frequency differences).

[0062] By inputting the first MRS spectrum and the second MRS spectrum into the amount-of-substance estimation NN, the amount-of-substance estimation NN can recognize the specific substance belonging to the specific frequency band separately from other substances, thereby improving the estimation accuracy of the amount-of-substance information of the specific substance.

[0063] (Variation 2) 8 is a diagram schematically illustrating a specific example of the input / output relationship of a substance amount estimation NN according to Modification 2 of the first embodiment. The substance amount estimation NN according to Modification 2 is a machine learning model in which parameters are trained so as to receive a first MRS spectrum, a second MRS spectrum, and selection information as inputs, and output quantity information related to substances contained in measurement regions of the first MRS spectrum and the second MRS spectrum. The first MRS spectrum is an MRS spectrum acquired by MR spectroscopy using a pulse sequence without application of a region-selective pulse. The second MRS spectrum is an MRS spectrum acquired by MR spectroscopy using a pulse sequence with application of a region-selective pulse. The selection information is information for identifying a region selected by the region-selective pulse, and includes, for example, frequency information of a frequency band that selectively excites or de-excites.

[0064] By inputting the first MRS spectrum, the second MRS spectrum, and the selection information into the amount-of-substance estimation NN, it is possible to easily separate specific substances belonging to specific frequency bands using the amount-of-substance estimation NN, and it is expected that the estimation accuracy of the amount-of-substance information of specific substances will be further improved.

[0065] (Variation 3) 9 is a diagram schematically illustrating a specific example of the input / output relationship of a substance amount estimation NN according to Modification 3 of the first embodiment. The substance amount estimation NN according to Modification 3 is a machine learning model in which parameters are trained so as to receive a first MRS spectrum and a second MRS spectrum as inputs and output substance amount information related to substances contained in the measurement target region of the first MRS spectrum and the second MRS spectrum. The first MRS spectrum and the second MRS spectrum are MRS spectra with different combinations of TR and TE, which are data acquisition conditions. TR and TE significantly affect the SNR (Signal-to-Noise Ratio) of MR spectroscopy. The longer the TR, the closer the obtained signal intensity value is to the true value, but the longer the data acquisition time. The shorter the TE, the greater the number of observed peaks, improving the accuracy of the MRS spectrum. The longer the TE, the fewer the observed peaks, improving the visibility of the MRS spectrum.

[0066] The first MRS spectrum is collected under a first combination of TR and TE, and the second MRS spectrum is collected under a second combination of TR and TE. The first and second MRS spectra may have the same TR but different TE, or may have different TRs but the same TE, or may have different TRs and different TEs. Note that although an example in which two types of MRS spectra are input has been shown, it is also possible to collect MRS spectra approximately 3 to 1000 times while changing the TE or TR each time, and use some or all of the obtained MRS spectra as input to the quantity of substance estimation NN.

[0067] By inputting the first and second MRS spectra, which have different combinations of TR and TE, into the substance amount estimation NN, the peak recognition performance of the substance amount estimation NN is improved, and therefore, the estimation accuracy of substance amount information is expected to improve.

[0068] (Variation 4) 10 is a diagram schematically showing a specific example of the input / output relationship of the amount-of-substance estimation NN according to Modification 4 of the first embodiment. The amount-of-substance estimation NN according to Modification 4 is a machine learning model in which parameters are trained so as to receive an MRS spectrum and a composite MRS spectrum as inputs and output amount-of-substance information relating to substances contained in the measurement target region of the MRS spectrum and the composite MRS spectrum. The MRS spectrum is an MRS spectrum generated by the signal processing function 513 based on MRSk data. The composite MRS spectrum is an MRS spectrum generated by the synthesis function 515. The composite MRS spectrum may or may not be generated based on the other input MRS spectrum.

[0069] By inputting the MRS spectrum and the composite MRS spectrum into the substance amount estimation NN, the ability of the substance amount estimation NN to distinguish between peaks in the MRS spectrum and noise can be improved, and it is expected that the estimation accuracy of substance amount information will be improved.

[0070] (Variation 5) In the above embodiment, only the amount of substance estimation NN by the amount of substance output function 514 is built in by a machine learning model. However, the amount of substance output function 514 and the synthesis function 515 may be built in by a single machine learning model (hereinafter referred to as a first integrated model).

[0071] FIG. 11 is a diagram showing an example of a network configuration of a first integrated model NN10 according to Modification 5 of the first embodiment. As shown in FIG. 11, the first integrated model NN10 is a deep neural network having a substance amount information output layer NN11, a synthesis layer NN12, a substance amount information output layer NN13, and a synthesis layer NN14. The substance amount information output layer NN11 and the substance amount information output layer NN13 are neural network layers corresponding to the substance amount estimation NNs described in some of the above examples. The substance amount information output layer NN11 and the substance amount information output layer NN13 are examples of a parameter output unit. The synthesis layer NN12 and the synthesis layer NN14 are neural network layers that perform processing using a synthesis function 515 to generate a synthesis MRS spectrum based on substance amount information. The synthesis layer NN12 and the synthesis layer NN14 are examples of a synthesis unit.

[0072] Specifically, as shown in FIG. 11 , the substance amount information output layer NN11 receives the first MRS spectrum and the second MRS spectrum as input, and outputs the substance amount information of propylene glycol, ethanol, acetate, and acetone. The synthesis layer NN12 receives the substance amount information of propylene glycol, ethanol, acetate, and acetone output from the substance amount information output layer NN11 as input, and outputs a synthesized MRS spectrum. The substance amount information output layer NN13 receives the first MRS spectrum and the synthesized MRS spectrum output from the synthesis layer NN12 as input, and outputs the substance amount information of propylene glycol, ethanol, acetate, and acetone. The synthesis layer NN14 receives the substance amount information of propylene glycol, ethanol, acetate, and acetone output from the substance amount information output layer NN13 as input, and outputs a synthesized MRS spectrum.

[0073] The first integrated model NN10 shown in Fig. 11 has two unit layers, each of which has a substance amount information output layer and a synthesis layer connected in series. However, the present invention is not limited to this, and the integrated model NN10 may have only one unit layer, or may have three or more unit layers. Furthermore, although the substance amount information output layer has two input channels, the present invention is not limited to this, and the number may be one or three.

[0074] Fig. 12 is a diagram showing an example of a network configuration of another first integrated model NN20 according to Modification 5 of the first embodiment. As shown in Fig. 12, the first integrated model NN20 is a deep neural network having a substance amount information output layer NN21, a synthesis layer NN22, a substance amount information output layer NN23, and a synthesis layer NN24. The substance amount information output layer NN21 and the substance amount information output layer NN23 are examples of a parameter output unit. The synthesis layer NN22 and the synthesis layer NN24 are examples of a synthesis unit.

[0075] The substance amount information output layer NN21 receives the first MRS spectrum and the second MRS spectrum as input, and outputs the substance amount information of propylene glycol, ethanol, acetate, and acetone. The synthesis layer NN22 receives the substance amount information of propylene glycol, ethanol, acetate, and acetone output from the substance amount information output layer NN21 as input, and outputs a synthesized MRS spectrum. The substance amount information output layer NN23 receives the first MRS spectrum, the second MRS spectrum, and the synthesized MRS spectrum output from the synthesis layer NN12 as input, and outputs the substance amount information of propylene glycol, ethanol, acetate, and acetone. The synthesis layer NN24 receives the substance amount information of propylene glycol, ethanol, acetate, and acetone output from the substance amount information output layer NN23 as input, and outputs a synthesized MRS spectrum.

[0076] As described above, according to Modification 5, it is possible to implement the substance amount output function 514 and the synthesis function 515 using a deep neural network in which a plurality of unit layers (combinations of substance amount information output layers and synthesis layers) are connected in series, as in the first integrated models 10 and 20. In this way, the process of generating a synthetic MRS spectrum from a plurality of MRS spectra can be implemented using a deep neural network, making it possible to generate a synthetic MRS spectrum quickly and with high accuracy.

[0077] As in the above embodiment, the MR signal processing device 50 according to the first embodiment includes a processing circuitry 51A. The processing circuitry 51A applies a trained model to multiple MRS signals acquired by MR spectroscopy for the same subject, and outputs physical quantity information of multiple substances for MRS reconstruction. The processing circuitry 51A generates an MRS spectrum for the measurement target region by fitting based on the physical quantity information of the multiple substances and multiple artificial spectra.

[0078] According to the above configuration, since physical quantity information is obtained based on a plurality of MRS signals, the accuracy of the physical quantity information can be improved. Also, since an MRS spectrum is generated based on such physical quantity information, the accuracy of the MRS spectrum can be improved.

[0079] (Second embodiment) Next, an MR signal processing device according to a second embodiment will be described. In the following description, components having substantially the same functions as those in the first embodiment will be given the same reference numerals and will be described only when necessary.

[0080] Fig. 13 is a diagram showing an example of the configuration of a magnetic resonance imaging apparatus 1 according to the second embodiment. As shown in Fig. 13, the magnetic resonance imaging apparatus 1 has a gantry 11, a bed 13, a gradient magnetic field power supply 21, a transmission circuit 23, a reception circuit 25, a bed driving device 27, a sequence control circuit 29, and an MR signal processing device (host computer) 50. The MR signal processing device 50 is a computer having a processing circuit 51B, a memory 53, a display 55, an input interface 57, and a communication interface 59.

[0081] The processing circuitry 51B has a processor such as a CPU as a hardware resource. The processing circuitry 51B functions as the core of the magnetic resonance imaging apparatus 1. For example, the processing circuitry 51B executes various programs to realize an acquisition function 521, a condition setting function 522, a signal processing function 523, a component amount output function 524, a synthesis function 525, a determination function 526, a learning function 527, and a display control function 528. The acquisition function 521 is an example of an acquisition unit, the condition setting function 522 is an example of a setting unit, the signal processing function 523 is an example of a spectrum generation unit, the substance amount output function 524 is an example of a parameter output unit, the synthesis function 525 is an example of a synthesis unit, the determination function 526 is an example of a determination unit, the learning function 527 is an example of a learning unit, and the display control function 528 is an example of a display unit.

[0082] The acquisition function 521, the condition setting function 522, and the signal processing function 523 are functions that are approximately identical to the acquisition function 511, the condition setting function 512, and the signal processing function 513 according to the first embodiment, respectively.

[0083] In the component quantity output function 524, the processing circuitry 51B applies the trained model to multiple MRS signals collected by MR spectroscopy for the same target P, and outputs multiple parameters related to MRS reconstruction. The multiple parameters according to the second embodiment are spectral parameters for fitting based on multiple spectral models corresponding to the multiple parameters. The multiple spectral models are multiple basis spectra corresponding to multiple bases related to the measurement target region. The multiple bases are obtained by performing data conversion on the MRS spectra related to the measurement target region. The multiple spectral parameters include component quantity information for the multiple bases. The component quantity information refers to information related to the component quantities of the bases. Hereinafter, the parameters according to the second embodiment will be referred to as basis component quantity information. Note that the bases include, for example, a first basis based on MRS spectra related to a healthy subject (hereinafter referred to as a healthy subject basis) and a second basis based on MRS spectra related to a non-healthy subject (hereinafter referred to as a non-healthy subject basis). The trained model is a machine learning model trained to output component quantity information using multiple MRS signals as input. The machine learning model may be a neural network or a deep neural network. Hereinafter, the trained model according to the second embodiment will be referred to as a basis component amount estimation NN.

[0084] In the synthesis function 525, the processing circuitry 51B generates a synthesized MRS signal for the target P in which noise is reduced compared to the multiple MRS signals, based on the basis component amount information output by the component amount output function 524.

[0085] In the determination function 526, the processing circuitry 51B determines whether the subject P is normal or abnormal based on the basis component amount information regarding the healthy subject's basis and the basis component amount information regarding the unhealthy subject's basis.

[0086] In the learning function 527, the processing circuit 51B generates a basis component amount estimation NN used in the component amount output function 524 by machine learning based on the learning data.

[0087] In the display control function 528, the processing circuitry 51B displays various information on the display 55. For example, the processing circuitry 51B displays on the display 55 MRS signals, basis component amount information, composite MRS signals, normal / abnormal determination results, a setting screen for data acquisition conditions, etc.

[0088] An example of the operation of the MR signal processing device 50 according to the second embodiment will be described in detail below.

[0089] As described above, the component amount output function 524 allows the processing circuitry 51B to apply the basis component amount estimation NN to a plurality of MRS signals and output basis component amount information of the basis.

[0090] FIG. 14 is a diagram schematically illustrating the input / output relationship of a trained model (basis component amount estimation NN) according to the second embodiment. As shown in FIG. 14, the basis component amount estimation NN is a machine learning model in which parameters are trained so as to receive a first MRS signal and a second MRS signal as input and output basis component amount information of a basis. The first MRS signal and the second MRS signal are MRS signals related to the same measurement target region of the subject P. The data acquisition conditions for the first MRS signal and the second MRS signal may be the same or different. The MRS signal input to the basis component amount estimation NN can be MRSk data or an MRS spectrum. Although the number of MRS signals input to the basis component amount estimation NN shown in FIG. 14 is two, this is an example, and any number greater than two may be used.

[0091] The basis component quantity information is information about the component quantity of each basis when the MRS spectra representing the measurement target region of the first MRS signal and the second MRS signal are mathematically transformed into multiple bases. The bases can be obtained by performing data compression on the MRS spectra representing the measurement target region. Examples of data compression methods that can be used include low-rank approximation, principal component analysis, singular value decomposition, and autoencoder. The MRS spectrum representing the measurement target region may be an MRS spectrum of the measurement target region of the subject P measured in advance, or an MRS spectrum of the same measurement target region of a person other than the subject P. The obtained multiple bases are used in the basis component quantity estimation neural network. Alternatively, multiple MRS spectra of the same measurement target region may be converted into multiple bases by data compression, and a basis to be used in the basis component quantity estimation neural network may be selected from these multiple bases. For example, the multiple bases may be divided into multiple clusters by clustering or the like, and a basis to be used in the basis component quantity estimation neural network may be selected from each cluster.

[0092] The basis component amount estimation NN is generated by the learning function 527 of the processing circuit 51B. The processing circuit 51B trains a machine learning model based on multiple learning samples to generate the basis component amount estimation NN. The learning samples are a combination of first and second MRS signals, which are input data, and basis component amount information, which is correct data (hereinafter referred to as correct basis component amount information). The first and second MRS signals, which are input data, are generated by the magnetic resonance imaging apparatus 1 or another magnetic resonance imaging apparatus. The correct basis component amount information can be obtained as component amounts of multiple bases when an MRS spectrum corresponding to the input data is converted into multiple bases obtained in advance. The processing circuit 51B applies the machine learning model to the first and second MRS signals to perform forward propagation processing and outputs basis component amount information (hereinafter referred to as estimated basis component amount information). Next, the processing circuit 51B applies the difference (error) between the estimated basis component amount information and the correct basis component amount information to the machine learning model to perform backpropagation processing, and calculates a gradient vector, which is the differential coefficient of the error function, which is a function of the parameters.The processing circuit 51B then updates the parameters of the machine learning model based on the gradient vector.The forward propagation processing, backpropagation processing, and parameter update processing are repeated while changing the learning sample, and the parameters that minimize the error function are determined according to a predetermined optimization method.This generates a basis component amount estimation NN.

[0093] Next, signal processing by the MR signal processing device 50 according to the second embodiment will be described with reference to Fig. 15. In the following description, it is assumed that the MRS signal input to the basis component amount estimation NN is an MRS spectrum.

[0094] FIG. 15 is a diagram showing an example of the flow of signal processing by the MR signal processing device 50 according to the second embodiment.

[0095] 15, the processing circuitry 51B acquires a first MRS spectrum and a second MRS spectrum (step SB1) by implementing the acquisition function 521. Step SB1 is performed in the same manner as step SA1.

[0096] When step SB1 is performed, the processing circuitry 51B, by implementing the component amount output function 524, applies the basis component amount estimation NN to the first MRS spectrum and the second MRS spectrum acquired in step SB1 and outputs basis component amount information (step SB2).

[0097] Fig. 16 is a diagram schematically showing the input / output relationship of the basis component amount estimation NN. As shown in Fig. 16, the basis component amount estimation NN is a machine learning model in which parameters are trained so as to receive a first MRS spectrum and a second MRS spectrum as input and output basis component amount information. By converting the first MRS spectrum and the second MRS spectrum into basis component amount information, the degradation process of the first MRS spectrum and the second MRS spectrum can be incorporated into the basis.

[0098] The basis component amount information includes component amount information and reliability information for each basis. In FIG. 16, specific examples are shown, including a component amount C1 and a half-width W1 for healthy subject basis #1, a component amount C2 and a half-width W2 for healthy subject basis #2, a component amount C3 and a half-width W3 for unhealthy subject basis #3, and a component amount C4 and a half-width W4 for unhealthy subject basis #4. Healthy subject basis #1 and #2 are bases obtained by applying data compression to the MRS spectra of healthy subjects. A healthy subject is a person who is evaluated as having no abnormality with respect to the measurement target regions of the first and second MRS spectra. Unhealthy subject basis #3 and #4 are bases obtained by applying data compression to the MRS spectra of unhealthy subjects. An unhealthy subject is a person who is evaluated as having an abnormality with respect to the measurement target regions of the first and second MRS spectra. If multiple peaks exist for one base, the component amount and half-width are output for each peak. Peaks can be identified by frequency difference.

[0099] The processing circuitry 51B applies the basis component amount estimation NN to the first MRS spectrum and the second MRS spectrum acquired in step SB1, and outputs the component amount C1 and half width W1 of the healthy subject basis #1, the component amount C2 and half width W2 of the healthy subject basis #2, the component amount C3 and half width W3 of the unhealthy subject basis #3, and the component amount C4 and half width W4 of the unhealthy subject basis #4. According to the second embodiment, it is possible to estimate basis component amount information from a plurality of MRS spectra related to the same measurement target region, and therefore it is possible to accurately estimate basis component amount information of the bases included in the measurement target region.

[0100] When step SB2 is performed, the processing circuitry 51B, by implementing the synthesis function 515, generates a composite MRS spectrum based on the basis component amount information output in step SB2 (step SB3). Step SB3 is performed in the same manner as step SA3. That is, a plurality of basis spectra corresponding to a plurality of bases are stored in advance. In step SB3, the processing circuitry 51B applies the component amount and half-width of each of the plurality of bases to the basis spectrum to generate a plurality of corrected basis spectra, and fits the generated plurality of corrected basis spectra to generate a composite MRS spectrum. Since the composite MRS spectrum is generated by fitting based on the basis spectra, noise due to instability in MR spectroscopy data acquisition, body movement of the subject P, disturbance of the static magnetic field, etc. is reduced compared to the input first MRS spectrum and second MRS spectrum. In other words, the basis component amount estimation by the basis component amount estimation NN (step SB2) and the generation of the composite MRS spectrum based on the basis component amount information (step SB3) can be regarded as the same as the denoising process of the first MRS spectrum and the second MRS spectrum.

[0101] When step SB3 is performed, the processing circuit 51B determines whether the signal is normal or abnormal based on the base component amount information output in step SB2 by implementing the determination function 526 (step SB4).

[0102] 17 and 18 are diagrams schematically illustrating a process for determining whether a signal is normal or abnormal based on basis component amount information, with FIG. 17 relating to normality determination and FIG. 18 relating to abnormality determination. As shown in FIGS. 17 and 18, in step SB2, component amounts C11 and C12 of healthy subject basis #1, component amounts C21 and C22 of healthy subject basis #2, component amounts C31 and C32 of unhealthy subject basis #3, and component amounts C41 and C42 of unhealthy subject basis #4 are assumed to be output as basis component amount information. The processing circuit 51B determines whether a signal is normal or abnormal based on the ratio of the component amounts C11 and C12 and component amounts C21 and C22 of healthy subject basis #1 and #2 to the component amounts C31, C32 and component amounts C41 and C42 of unhealthy subject basis #3 and #4. The processing circuit 51B determines that the healthy subject bases #1 and #2 are normal when the component amounts C11 and C21 are greater than the component amounts C31 and C41 of the unhealthy subject bases #3 and #4, as shown in Fig. 17. On the other hand, the processing circuit 51B determines that the healthy subject bases #1 and #2 are abnormal when the component amounts C32 and C42 of the unhealthy subject bases #3 and #4 are greater than the component amounts C12 and C22 of the healthy subject bases #1 and #2, as shown in Fig. 18.

[0103] After step SB4 is performed, the processing circuitry 51B displays the composite MRS spectrum generated in step SB3 and the determination result obtained in step SB4 (step SB5) by implementing the display control function 528. The composite MRS spectrum and the determination result are displayed on the display 55 in a predetermined layout.

[0104] FIG. 19 is a diagram showing an example of a display screen I1 displaying a composite MRS spectrum I12 and a determination result I13. As shown in FIG. 19, a voxel of interest setting image I11 is displayed on the display screen I1. The voxel of interest setting image I11 is an MR image in which a voxel of interest V1 of a first MRS spectrum and a second MRS spectrum is set. For example, a brain tumor region R1 is depicted in the voxel of interest setting image I11, and the voxel of interest V1 is depicted in the brain tumor region R1. When the voxels of interest of the first MRS spectrum and the second MRS spectrum are set at different positions, the voxel of interest of the first MRS spectrum and the voxel of interest of the second MRS spectrum are depicted at different positions. The voxel of interest setting image I11 may be an MR image acquired by any imaging method.

[0105] 19, the composite MRS spectrum I12 generated in step SB3 is displayed on the display screen I1, allowing the user to observe the composite MRS spectrum, which is a denoised MRS spectrum.

[0106] As shown in Fig. 19, the display screen I1 displays the normal or abnormal determination result I13 obtained in step SB4. For example, if an abnormality is determined in step SB4, the determination result is displayed as "Suspected abnormality." Information accompanying the determination result, such as "Detailed examination required," may also be displayed.

[0107] With the above, the signal processing by the MR signal processing device 50 according to the second embodiment is completed.

[0108] The second embodiment can be modified in various ways. For example, in Fig. 15, the process of generating a composite MRS spectrum is performed after the process of determining whether the spectrum is normal or abnormal. However, the process of generating a composite MRS spectrum may be performed after the process of determining whether the spectrum is normal or abnormal.

[0109] In the above embodiment, a normal or abnormal determination is made. However, the presence or absence of a specific disease, such as a brain tumor, leukoencephalopathy, stroke, dementia, or trauma, may also be determined. In this case, a basis vector specific to each disease (hereinafter referred to as a disease basis vector) is prepared as the unhealthy subject basis. The disease basis vector can be obtained by applying data compression technology to the MRS spectrum of an unhealthy subject diagnosed with the disease. The processing circuitry 51B compares the amount of basis components of the disease basis vector with a threshold value. If the amount of basis components exceeds the threshold value, the processing circuitry 51B determines that the subject has the disease. If the amount of basis components does not exceed the threshold value, the processing circuitry 51B determines that the subject does not have the disease.

[0110] As in the above embodiment, the MR signal processing device 50 according to the second embodiment includes a processing circuitry 51B. The processing circuitry 51B applies a trained model to multiple MRS signals acquired by MR spectroscopy for the same subject, and outputs basis component amount information for multiple bases for MRS reconstruction. The processing circuitry 51B generates an MRS spectrum for the measurement target region by fitting based on the basis component amount information for the multiple bases and multiple basis spectra.

[0111] According to the above configuration, since basis component amount information is obtained based on a plurality of MRS signals, the accuracy of the basis component amount information can be improved. Also, since an MRS spectrum is generated based on such basis component amount information, the accuracy of the MRS spectrum can be improved.

[0112] (Third embodiment) Next, an MR signal processing device according to a third embodiment will be described. In the following description, components having substantially the same functions as those in the first and second embodiments will be denoted by the same reference numerals, and will be described only when necessary.

[0113] Fig. 20 is a diagram showing an example of the configuration of a magnetic resonance imaging apparatus 1 according to the third embodiment. As shown in Fig. 20, the magnetic resonance imaging apparatus 1 according to the third embodiment includes a gantry 11, a bed 13, a gradient magnetic field power supply 21, a transmission circuit 23, a reception circuit 25, a bed driving device 27, a sequence control circuit 29, and an MR signal processing device (host computer) 50. The MR signal processing device 50 is a computer including a processing circuit 51C, a memory 53, a display 55, an input interface 57, and a communication interface 59.

[0114] The processing circuitry 51C has a processor such as a CPU as a hardware resource. The processing circuitry 51C functions as the core of the magnetic resonance imaging apparatus 1. For example, the processing circuitry 51C executes various programs to realize an acquisition function 531, a condition setting function 532, a signal processing function 533, a modulation amount output function 534, a correction function 535, a learning function 536, and a display control function 537. The acquisition function 531 is an example of an acquisition unit, the condition setting function 532 is an example of a setting unit, the signal processing function 533 is an example of a spectrum generation unit, the modulation amount output function 534 is an example of a parameter output unit, the correction function 535 is an example of a correction unit, the learning function 536 is an example of a learning unit, and the display control function 537 is an example of a display unit.

[0115] The acquisition function 531, the condition setting function 532, and the signal processing function 533 are functions that are approximately identical to the acquisition function 511, the condition setting function 512, and the signal processing function 513 according to the first embodiment, respectively.

[0116] In the modulation amount output function 534, the processing circuitry 51C applies the trained model to multiple MRS signals collected by MR spectroscopy for the subject P, and outputs multiple parameters for MRS reconstruction. The multiple parameters according to the third embodiment include information about k-space modulation caused by magnetic field modulation between multiple MRS signals. Hereinafter, the parameters according to the third embodiment will be referred to as k-space modulation amount information. The trained model is a machine learning model trained to input multiple MRS signals and output k-space modulation amount information. A neural network or a deep neural network is used as the machine learning model. Hereinafter, the trained model according to the third embodiment will be referred to as a modulation amount estimation NN.

[0117] In the correction function 535, the processing circuitry 51C corrects the MRS signal based on the k-space modulation amount information output in the modulation amount output function 534. Hereinafter, the corrected MRS signal will be referred to as a corrected MRS signal.

[0118] In the learning function 536, the processing circuit 51C generates a modulation amount estimation NN used in the modulation amount output function 534 by machine learning based on the learning data.

[0119] In the display control function 537, the processing circuitry 51C displays various information on the display 55. For example, the processing circuitry 51C displays on the display 55 MRS signals, modulation amount information, corrected MRS signals, a setting screen for data acquisition conditions, and the like.

[0120] An example of the operation of the MR signal processing device 50 according to the third embodiment will be described in detail below.

[0121] As described above, the modulation amount output function 534 causes the processing circuit 51C to apply the modulation amount estimation NN to the multiple MRS signals and output k-space modulation amount information regarding the k-space modulation caused by the magnetic field modulation between data acquisitions for the multiple MRS signals.

[0122] FIG. 21 is a diagram schematically illustrating the input / output relationship of a trained model (modulation amount estimation NN) according to the third embodiment. As shown in FIG. 21, the modulation amount estimation NN is a machine learning model in which parameters are trained so as to receive a first MRS signal and a second MRS signal as input and output k-space modulation amount information related to k-space modulation resulting from magnetic field modulation between data acquisitions of the first MRS signal and the second MRS signal. The first MRS signal and the second MRS signal are MRS signals for the same subject P. The data acquisition conditions for the first MRS signal and the second MRS signal may be the same or different. The MRS signal input to the modulation amount estimation NN can be MRSk data or an MRS spectrum. Although the number of MRS signals input to the modulation amount estimation NN shown in FIG. 21 is two, this is an example, and any number greater than or equal to two may be used.

[0123] The k-space modulation amount information is information regarding the modulation amount of k-space modulation caused by magnetic field modulation between data acquisitions for the first MRS signal and the second MRS signal. Various factors can cause magnetic field modulation, such as static magnetic field modulation, gradient magnetic field modulation, and noise. Factors causing static magnetic field modulation include static magnetic field inhomogeneity and static magnetic field offset (f0 shift). Factors causing gradient magnetic field modulation include transient response (phase shift) of the gradient magnetic field. Magnetic field modulation appears as a deviation in k-space between actual MRSk data and ideal MRSk data without magnetic field modulation. If magnetic field modulation occurs between data acquisitions for the first MRS signal and the second MRS signal, a deviation will occur in k-space between the first MRSk data and the second MRSk data.

[0124] The modulation amount estimation NN is generated by the learning function 536 of the processing circuit 51C. The processing circuit 51C trains a machine learning model based on a plurality of learning samples to generate the modulation amount estimation NN. The learning samples are a combination of first and second MRS signals as input data and k-space modulation amount information as correct data (hereinafter referred to as correct k-space modulation amount information). The first and second MRS signals as input data are generated by the magnetic resonance imaging apparatus 1 or another magnetic resonance imaging apparatus. The correct k-space modulation amount information is obtained by measuring the deviation in k-space of the second MRSk data relative to the first MRSk data. The processing circuit 51C applies the machine learning model to the first and second MRS signals to perform forward propagation processing and outputs k-space modulation amount information (hereinafter referred to as estimated k-space modulation amount information). Next, the processing circuit 51C applies the difference (error) between the estimated k-space modulation amount information and the correct k-space modulation amount information to the machine learning model to perform backpropagation processing, and calculates a gradient vector, which is the differential coefficient of the error function, which is a function of the parameters. Next, the processing circuit 51C updates the parameters of the machine learning model based on the gradient vector. The forward propagation processing, backpropagation processing, and parameter update processing are repeated while changing the learning sample, and the parameters that minimize the error function are determined according to a predetermined optimization method. This generates a modulation amount estimation NN.

[0125] Next, signal processing by the MR signal processing device 50 according to the third embodiment will be described with reference to Fig. 22. In the following description, it is assumed that the MRS signal input to the modulation amount estimation NN is MRSk data.

[0126] FIG. 22 is a diagram showing an example of the flow of signal processing by the MR signal processing device 50 according to the third embodiment.

[0127] 22, the processing circuitry 51C acquires first MRSk data and second MRSk data by implementing the acquisition function 531 (step SC1). The first MRSk data and second MRSk data are assumed to be MRSk data related to the same measurement target region acquired under the same data acquisition conditions by, for example, the single-voxel method. The processing circuitry 51C selects and acquires any two MRSk data from the MRSk data for the number of integrations before the addition process as the first MRSk data and the second MRSk data. The two MRSk data may be selected manually by a user in accordance with an instruction via the input interface 57 or automatically in accordance with a predetermined algorithm.

[0128] When step SC1 is performed, the processing circuit 51C, by implementing the modulation amount output function 534, applies the modulation amount estimation NN to the first MRSk data and the second MRSk data acquired in step SC1 and outputs k-space modulation amount information (step SC2).

[0129] 23 is a diagram schematically illustrating a specific example of the input / output relationship of the modulation amount estimation NN. As shown in FIG. 23, the modulation amount estimation NN is a machine learning model in which parameters are trained so as to receive first MRSk data and second MRSk data as input and output k-space modulation amount information between data acquisitions of the first MRSk data and the second MRSk data. Specific examples of the k-space modulation amount information include an f0 shift and a phase shift. The k-space modulation amount information related to the f0 shift is the amount of deviation of the MRSk data in the k-space caused by static magnetic field modulation, and the phase shift is the amount of deviation of the MRSk data in the k-space caused by gradient magnetic field modulation. The k-space modulation amount information may be a numerical value indicating the amount of deviation of the k-space modulation, or may be parameters such as a time constant, gain, or time origin of an impulse response representing the k-space modulation.

[0130] The processing circuitry 51C applies the modulation amount estimation NN to the first MRSk data and the second MRSk data acquired in step SC1, and outputs k-space modulation amount information between data acquisitions of the first MRSk data and the second MRSk data. According to the third embodiment, it is possible to estimate the k-space modulation amount information during data acquisition of the MRSk data. Furthermore, since it is possible to estimate the k-space modulation amount information from a plurality of MRSk data, it is possible to estimate the k-space modulation amount information with high accuracy.

[0131] When step SC2 is performed, the processing circuitry 51C corrects the MRSk data based on the k-space modulation amount information output in step SC2 by implementing the correction function 535 (step SC3). The corrected MRSk data is called corrected MRSk data.

[0132] After step SC3 is performed, the processing circuitry 51C generates an MRS spectrum based on the corrected MRSk data generated in step SC3 by implementing the signal processing function 533 (step SC4).

[0133] FIG. 24 is a diagram schematically showing the correction process and spectrum generation process in steps SC3 and SC4. As shown in FIG. 24, MRSk data OK1 to be corrected is acquired. As the MRSk data OK1, the first MRSk data and / or the second MRSk data acquired in step SC1 may be selected, or other MRSk data may be selected. As the other MRSk data, for example, MRSk data other than the first MRSk data and the second MRSk data acquired in step SC1 may be selected from the MRSk data for the number of accumulations. Furthermore, average MRSk data generated by performing an addition process on the MRSk data for the number of accumulations may be selected as the MRSk data OK1. Hereinafter, the MRSk data OK1 is assumed to be average MRSk data.

[0134] 24, k-space modulation amount information regarding the f0 shift and the phase shift is output in step SC2. The processing circuit 51C corrects the average MRSk data OK1 based on the k-space modulation amount information regarding the f0 shift and the phase shift to generate corrected MRSk data CK1 (step SC3). The corrected MRSk data CK1 has reduced f0 shift and phase shift.

[0135] Next, as shown in FIG. 24, the processing circuitry 51C performs a Fourier transform on the corrected MRSk data CK1 to convert the signal intensity values into digital data expressed as a frequency function, and then performs post-processing such as phase correction and baseline correction on the converted digital data to generate an MRS spectrum CS1 (step SC4). The MRS spectrum CS1 has reduced noise due to k-space modulation compared to an MRS spectrum generated from the average MRSk data OK1 without the correction process (step SC3). That is, according to the third embodiment, even when average MRSk data obtained with a smaller number of integrations than conventional ones is used, the accuracy of the MRS spectrum can be improved by correction based on k-space modulation amount information. Therefore, according to the third embodiment, the number of integrations can be reduced compared to when the correction process (step SC3) is not performed.

[0136] After step SC4 is performed, the processing circuitry 51C displays the MRS spectrum generated in step SC4 (step SC5) by implementing the display control function 549. This allows the user to observe the denoised MRS spectrum.

[0137] With the above, the signal processing by the MR signal processing device 50 according to the third embodiment is completed.

[0138] As in the above-described embodiment, the MR signal processing device 50 according to the third embodiment includes a processing circuit 51C. The processing circuit 51C applies a trained model to multiple MRSk data collected by MR spectroscopy for the same subject, and outputs k-space modulation amount information for MRS reconstruction. The processing circuit 51C corrects the MRSk data based on the k-space modulation amount information. Furthermore, the processing circuit 51C generates an MRS spectrum based on the corrected MRSk data.

[0139] According to the above configuration, since k-space modulation amount information is obtained based on a plurality of MRSk data, the accuracy of the k-space modulation amount information can be improved. Also, since MRSk data and an MRS spectrum can be generated based on such k-space modulation amount information, the accuracy of the MRSk data and an MRS spectrum can be improved.

[0140] (Fourth embodiment) Next, an MR signal processing device according to a fourth embodiment will be described. In the following description, components having substantially the same functions as those in the first embodiment will be given the same reference numerals and will be described only when necessary.

[0141] Fig. 25 is a diagram showing an example of the configuration of a magnetic resonance imaging apparatus 1 according to the fourth embodiment. As shown in Fig. 25, the magnetic resonance imaging apparatus 1 has a gantry 11, a bed 13, a gradient magnetic field power supply 21, a transmission circuit 23, a reception circuit 25, a bed driving device 27, a sequence control circuit 29, and an MR signal processing device (host computer) 50. The MR signal processing device 50 is a computer having a processing circuit 51D, a memory 53, a display 55, an input interface 57, and a communication interface 59.

[0142] The processing circuitry 51D has a processor such as a CPU as a hardware resource. The processing circuitry 51D functions as the core of the magnetic resonance imaging apparatus 1. For example, the processing circuitry 51D executes various programs to realize an acquisition function 541, a condition setting function 542, a signal processing function 543, a substance quantity output function 544, a synthesis function 545, a modulation amount output function 546, a correction function 547, a learning function 548, and a display control function 549. The acquisition function 541 is an example of an acquisition unit, the condition setting function 542 is an example of a setting unit, the signal processing function 543 is an example of a spectrum generation unit, the substance quantity output function 544 is an example of a parameter output unit, the synthesis function 545 is an example of a synthesis unit, the modulation amount output function 546 is an example of a parameter output unit, the correction function 547 is an example of a correction unit, the learning function 548 is an example of a learning unit, and the display control function 549 is an example of a display unit.

[0143] The acquisition function 541, the condition setting function 542, the signal processing function 543, the substance amount output function 544, and the synthesis function 545 are functions that approximately correspond to the acquisition function 511, the condition setting function 512, the signal processing function 513, the substance amount output function 514, and the synthesis function 515 according to the first embodiment, respectively. The modulation amount output function 546 and the correction function 547 are functions that approximately correspond to the modulation amount output function 534 and the correction function 535 according to the third embodiment, respectively. The learning function 548 is a function that approximately corresponds to the learning function 516 according to the first embodiment and the learning function 536 according to the third embodiment. The display control function 549 is a function that approximately corresponds to the display control function 517 according to the first embodiment and the display control function 537 according to the third embodiment.

[0144] An example of the operation of the MR signal processing device 50 according to the fourth embodiment will be described in detail below.

[0145] The fourth embodiment is a combination of the first and third embodiments. That is, k-space modulation amount information is output by a modulation amount output function 546 based on a composite MRS spectrum generated by a synthesis function 545 based on substance amount information output by a substance amount output function 544. As one means for achieving this, the substance amount output function 544, the synthesis function 545, and the modulation amount output function 546 may be implemented by a single machine learning model (hereinafter referred to as a second integrated model).

[0146] FIG. 26 is a diagram illustrating an example of a network configuration of a second integrated model NN30 according to the fourth embodiment. As illustrated in FIG. 26, the second integrated model NN30 is a deep neural network including a substance amount information output layer NN31, a synthesis layer NN32, an inverse transformation layer NN33, a substance amount information output layer NN34, a synthesis layer NN35, an inverse transformation layer NN36, and a k-space modulation amount information output layer NN37. The substance amount information output layer NN31 and the substance amount information output layer NN34 are a type of neural network layer that outputs substance amount information from MRS signals using a substance amount output function 544. The synthesis layer NN32 and the synthesis layer NN35 are neural network layers that perform processing using a synthesis function 545 to generate a synthetic MRS spectrum based on the substance amount information. The inverse transformation layer NN33 and the inverse transformation layer NN36 are neural network layers that perform processing using a signal processing function 543 to convert the synthetic MRS spectrum into MRSk data. The k-space modulation amount information output layer NN37 is a neural network layer that performs processing to output k-space modulation amount information from two MRSk data using a modulation amount output function 546.

[0147] 26, the substance amount information output layer NN31 receives a first MRS spectrum as an input and outputs first substance amount information for propylene glycol, ethanol, acetate, and acetone corresponding to the first MRS spectrum. The synthesis layer NN32 receives the first substance amount information for propylene glycol, ethanol, acetate, and acetone output from the substance amount information output layer NN31 and outputs a first synthesized MRS spectrum. The inverse transformation layer NN33 receives the first synthesized MRS spectrum output from the synthesis layer NN32 and generates first MRSk data, which is k-space data corresponding to the first synthesized MRS spectrum.

[0148] Similarly, the substance amount information output layer NN34 receives the second MRS spectrum as input and outputs second substance amount information for propylene glycol, ethanol, acetate, and acetone corresponding to the second MRS spectrum. The synthesis layer NN35 receives the second substance amount information for propylene glycol, ethanol, acetate, and acetone output from the substance amount information output layer NN34 and outputs a second synthesized MRS spectrum. The inverse transformation layer NN36 receives the second synthesized MRS spectrum output from the synthesis layer NN35 and generates second MRSk data, which is k-space data corresponding to the second synthesized MRS spectrum.

[0149] The k-space modulation amount information output layer NN37 receives the first MRSk data output from the inverse transformation layer NN33 and the second MRSk data output from the inverse transformation layer NN36 as input, and outputs k-space modulation amount information relating to the f0 shift and phase shift between data acquisitions of the first and second MRS spectra. As described above, the combination of the substance quantity information output layer NN31 and the synthesis layer NN32 and the combination of the substance quantity information output layer NN34 and the synthesis layer NN35 can be considered equivalent to denoising of MRS spectra. The k-space modulation amount information output layer NN37 outputs k-space modulation amount information from MRSk data based on the synthesis MRS spectrum, which is the denoised MRS spectrum, and is therefore expected to improve the accuracy of the k-space modulation amount information.

[0150] The second integrated model NN30 is trained by the processing circuit 51D through the implementation of the learning function 548. The second integrated model NN30 can be generated, for example, by pre-training or end-to-end learning. The pre-training can be performed, for example, according to the following procedure. First, the processing circuit 51D trains the substance amount information output layer NN31, the synthesis layer NN32, the inverse transformation layer NN33, the substance amount information output layer NN34, the synthesis layer NN35, the inverse transformation layer NN36, and the k-space modulation amount information output layer 37 individually. Next, the processing circuit 51D connects the substance amount information output layer NN31, the synthesis layer NN32, the inverse transformation layer NN33, the substance amount information output layer NN34, the synthesis layer NN35, the inverse transformation layer NN36, and the k-space modulation amount information output layer 37 as shown in FIG. 26 and trains them as a single deep neural network. This allows the second integrated model NN30 to be generated. In the end-to-end type, the processing circuitry 51D trains one deep neural network so that it receives two MRS spectra as input and outputs k-space modulation amount information of the f0 shift and phase shift.

[0151] The second integrated model NN30 shown in Fig. 26 can be modified in various ways. For example, two or more unit layers of the substance amount information output layer NN31 and the synthesis layer NN32 may be connected in series. Similarly, two or more unit layers of the substance amount information output layer NN34 and the synthesis layer NN35 may be connected in series.

[0152] FIG. 27 is a diagram illustrating an example of a network configuration of another second integrated model NN40 according to the fourth embodiment. As illustrated in FIG. 27, the second integrated model NN40 is a deep neural network including a substance amount information output layer NN41, a synthesis layer NN42, an inverse transformation layer NN43, a substance amount information output layer NN44, a synthesis layer NN45, an inverse transformation layer NN46, and a k-space modulation amount information output layer NN47. The substance amount information output layer NN41 is a type of neural network layer that outputs substance amount information from MRS signals using a substance amount output function 544. The substance amount information output layer NN44 is a type of neural network layer that outputs substance amount information from two MRS signals using the substance amount output function 544. The synthesis layers NN42 and NN45 are neural network layers that perform processing to generate a synthetic MRS spectrum based on substance amount information using a synthesis function 545. The inverse transformation layer NN43 and the inverse transformation layer NN46 are neural network layers that perform processing to convert the composite MRS spectrum into MRSk data using a signal processing function 543. The k-space modulation amount information output layer NN47 is a neural network layer that performs processing to output k-space modulation amount information from two MRSk data using a modulation amount output function 546.

[0153] 27, the substance amount information output layer NN41 receives a first MRS spectrum as an input and outputs first substance amount information for propylene glycol, ethanol, acetate, and acetone corresponding to the first MRS spectrum. The synthesis layer NN42 receives the first substance amount information for propylene glycol, ethanol, acetate, and acetone output from the substance amount information output layer NN41 and outputs a first synthesized MRS spectrum. The inverse transformation layer NN43 receives the first synthesized MRS spectrum output from the synthesis layer NN42 and generates first MRSk data, which is k-space data corresponding to the first synthesized MRS spectrum.

[0154] Similarly, the substance amount information output layer NN44 receives the first MRS spectrum and the second MRS spectrum as input, and outputs second substance amount information for propylene glycol, ethanol, acetate, and acetone corresponding to the first MRS spectrum and the second MRS spectrum. The synthesis layer NN45 receives the second substance amount information for propylene glycol, ethanol, acetate, and acetone output from the substance amount information output layer NN44 as input, and outputs a second synthesized MRS spectrum. The inverse transformation layer NN46 receives the second synthesized MRS spectrum output from the synthesis layer NN45 as input, and generates second MRSk data, which is k-space data corresponding to the second synthesized MRS spectrum.

[0155] The k-space modulation amount information output layer NN47 receives the first MRSk data output from the inverse transformation layer NN43 and the second MRSk data output from the inverse transformation layer NN46 as inputs and outputs k-space modulation amount information relating to the f0 shift and phase shift between data acquisitions of the first and second MRS spectra. As described above, the combination of the substance quantity information output layer NN41 and the synthesis layer NN42 and the combination of the substance quantity information output layer NN44 and the synthesis layer NN45 can be considered equivalent to denoising of MRS spectra. The k-space modulation amount information output layer NN47 outputs k-space modulation amount information from MRSk data based on the synthesis MRS spectrum, which is the denoised MRS spectrum, and is therefore expected to improve the accuracy of the k-space modulation amount information. Furthermore, the synthesis MRS spectrum, which is output from the combination of the substance quantity information output layer NN44 and the synthesis layer NN45, functions as an ideal MRS spectrum for the measurement target region. It can be said that the k-space modulation amount information output layer NN47 outputs k-space modulation amount information from MRSk data based on the denoised first MRS spectrum and MRSk data based on the ideal MRS spectrum. This k-space modulation amount information can be said to represent the modulation amount of k-space modulation resulting from modulation of the actual magnetic field at the time of data acquisition for the first MRS spectrum relative to the ideal magnetic field. By correcting the MRSk data based on such k-space modulation amount information, it is possible to generate more accurate MRSk data and an MRS spectrum based on the MRSk data.

[0156] The second integrated model NN40 is trained by the processing circuit 51D through the implementation of the learning function 548. The second integrated model NN40 can be generated, for example, by pre-training or end-to-end learning. In the case of pre-training, the processing circuit 51D first trains the substance amount information output layer NN41, the synthesis layer NN42, the inverse transformation layer NN43, the substance amount information output layer NN44, the synthesis layer NN45, the inverse transformation layer NN46, and the k-space modulation amount information output layer NN47 individually. Next, the processing circuit 51D connects the substance amount information output layer NN41, the synthesis layer NN42, the inverse transformation layer NN43, the substance amount information output layer NN44, the synthesis layer NN45, the inverse transformation layer NN46, and the k-space modulation amount information output layer NN47 as shown in FIG. 27 and trains them as a single deep neural network. This allows the second integrated model NN40 to be generated. In the end-to-end type, the processing circuitry 51D trains one deep neural network so that it receives three MRS spectra as input and outputs k-space modulation amount information of the f0 shift and phase shift.

[0157] The second integrated model NN40 shown in FIG. 27 can be modified in various ways. For example, the substance amount information output layer NN44 is assumed to receive, together with the second MRS spectrum, a first MRS spectrum identical to the input to the substance amount information output layer NN41. However, the substance amount information output layer NN44 may also receive, together with the second MRS spectrum, a third MRS spectrum different from the input to the substance amount information output layer NN41. Furthermore, instead of the substance amount information output layer NN44 with one channel input, a substance amount information output layer with two channels input may be provided. Two or more unit layers of the substance amount information output layer NN41 and the synthesis layer NN42 may be connected in series. Similarly, two or more unit layers of the substance amount information output layer NN44 and the synthesis layer NN45 may be connected in series.

[0158] (Fifth embodiment) The fifth embodiment is an application example of the third embodiment. The magnetic resonance imaging apparatus 1 and the MR signal processing apparatus 50 according to the fifth embodiment have the same configuration as the magnetic resonance imaging apparatus and the MR signal processing apparatus 50 according to the third embodiment.

[0159] 28 is a diagram schematically illustrating an input / output relationship of a trained model (modulation amount estimation NN) according to the fifth embodiment. As shown in FIG. 28, the modulation amount estimation NN is a machine learning model in which parameters are trained so as to receive a first MR acquisition signal and a second MR acquisition signal as input and output k-space modulation amount information related to k-space modulation resulting from magnetic field modulation occurring between data acquisitions for the first MR acquisition signal and the second MR acquisition signal. The processing circuitry 51C applies the modulation amount estimation NN to the first MR acquisition signal and the second MR acquisition signal of the same imaging subject to be processed, and outputs k-space modulation amount information related to k-space modulation resulting from magnetic field modulation occurring between data acquisitions for the first MR acquisition signal and the second MR acquisition signal of the same imaging subject.

[0160] The MR acquisition signal is a general term for signals obtained by MR imaging or chemical shift measurement by the sequence control circuit 29. Specifically, MR imaging according to the fifth embodiment can be T1-weighted imaging, T2-weighted imaging, T2*-weighted imaging, diffusion-weighted imaging, MR angiography, or any other MR imaging. The MR imaging pulse sequence is not particularly limited, and spin echo sequences, gradient echo sequences, inversion recovery, echo planar imaging, parallel imaging, compressed sensing, and any other pulse sequences are applicable. The k-space filling method is also not particularly limited, and any method, regardless of the number of dimensions, such as Cartesian scan, radial scan, spiral, stack-of-stars, or Cushball, is applicable. Chemical shift measurement can be applied to MR spectroscopy and chemical shift imaging applied in the above embodiments, as well as CEST (Chemical Exchange Spectroscopy) or ZAPPED (Z-Spectrum Analysis Provides Proton Environment Data).

[0161] The MR acquisition signal is a concept including, for example, k-space data and an MR image based on the k-space data. The first MR acquisition signal and the second MR acquisition signal are MR acquisition signals targeted at the same subject P. The data acquisition conditions for the first MR acquisition signal and the second MR acquisition signal may be the same or different. The MR acquisition signal input to the modulation amount estimation NN can be k-space data or an MR image. Although the number of MR acquisition signals input to the modulation amount estimation NN shown in FIG. 28 is two, this is an example, and any number greater than or equal to two may be used. The MR acquisition signal includes the MRS signals according to the first to fourth embodiments.

[0162] 29 is a diagram schematically illustrating a specific example of an input / output relationship of the modulation amount estimation NN according to the fifth embodiment. As illustrated in FIG. 29, the modulation amount estimation NN is a machine learning model in which parameters are trained so as to receive first non-MRS k-space data and second non-MRS k-space data as input and output k-space modulation amount information between data acquisitions of the first non-MRS k-space data and the second non-MRS k-space data. Specific examples of the k-space modulation amount information include an f0 shift and a phase shift. The non-MRS k-space data is k-space data acquired by chemical shift measurement or MR imaging other than MR spectroscopy.

[0163] As described above, the non-MRS k-space data may be acquired by any MR imaging method, pulse sequence, or k-space filling method. As the first non-MRS k-space data and the second non-MRS k-space data, k-space data for each k-space trajectory may be used. A k-space trajectory corresponds to, for example, a k-space line in a Cartesian scan, a spoke in a radial scan, or a spiral in a spiral scan. Hereinafter, a case where a spoke in a radial scan is used as the non-MRS k-space data will be described.

[0164] FIG. 30 is a diagram schematically illustrating a further specific example of the input / output relationship of the modulation amount estimation NN shown in FIG. 29. As shown in FIG. 30, the modulation amount estimation NN is a machine learning model in which parameters are trained so as to input multiple spokes SK1-SK4 of a radial scan and output k-space modulation amount information related to f0 shift and phase shift. The multiple spokes SK1-SK4 are arbitrary spokes selected from a set SK0 of multiple spokes collected by radial scan. FIG. 30 illustrates an example in which all spokes included in the set SK0 are input to the modulation amount estimation NN as multiple spokes SK1-SK4. However, it is not necessary for all spokes included in the set SK0 to be input to the modulation amount estimation NN; it is sufficient if any two spokes from the set SK0 are input to the modulation amount estimation NN.

[0165] The processing circuit 51C selects multiple spokes SK1-SK4 to be processed from the set SK0, applies a modulation amount estimation NN to the selected multiple spokes SK1-SK4, and outputs k-space modulation amount information related to the f0 shift and phase shift. The processing circuit 51C corrects the set SK0 based on the output k-space modulation amount information to generate a corrected set, and reconstructs an MR image based on the generated corrected set. The MR image has reduced f0 shift and phase shift.

[0166] According to the fifth embodiment, k-space modulation amount information can be obtained using MR acquisition signals other than MRS signals, thereby enabling highly accurate correction of k-space modulation for various MR acquisition signals.

[0167] (Sixth embodiment) The sixth embodiment is a modified example that can be applied to any of the above embodiments. An MR signal processing device according to the sixth embodiment will be described below. In the following description, components having substantially the same functions as those in the above embodiments will be given the same reference numerals and will be described only when necessary.

[0168] FIG. 31 is a diagram schematically illustrating the input / output relationship of a trained model NN according to the sixth embodiment. The trained model NN according to the sixth embodiment is a machine learning model trained to input a plurality of MRS signals and output MRS reconstruction parameters. The trained model NN according to the sixth embodiment may be any of the machine learning models according to the first to fifth embodiments. Specifically, the trained model NN may be a substance amount estimation NN that outputs substance amount information as shown in FIG. 2 of the first embodiment, a basis component amount estimation NN that outputs basis component amount information as shown in FIG. 14 of the second embodiment, or a modulation amount estimation NN that outputs k-space modulation amount information as shown in FIG. 21 of the third embodiment. The substance amount information, basis component amount information, and k-space modulation amount information are types of MRS reconstruction parameters. The MRS reconstruction parameters are parameters obtained by MRS reconstruction. MRS reconstruction generally refers to processing for generating an MRS spectrum. For example, MRS reconstruction includes a process of generating an MRS spectrum from a spectral model by fitting, as in the first and second embodiments, and a process of generating an MRS spectrum from MRSk data by Fourier transform, as in the third embodiment.

[0169] As shown in Fig. 31, the trained model NN receives input of multiple MRS signals with different numbers of additions. The number of MRS signals may be any number greater than or equal to two, but Fig. 31 illustrates an example of 96 MRS signals. The number of additions refers to the number of MRS signals to be added. It can also be expressed as resolution.

[0170] In the sixth embodiment, the processing circuitry 51 acquires a number of MRS signals (e.g., 128) corresponding to NEX. Next, the processing circuitry 51 generates 96 MRS signals from the number of MRS signals corresponding to NEX. Specifically, the processing circuitry 51 sets 96 patterns with different combinations of the number of MRS signals to be added (number of additions) and the numbers of the MRS signals to be added. The 96 patterns may be set by an arbitrary algorithm or may be set artificially. Then, the processing circuitry 51 generates 96 MRS signals by extracting and adding MRS signals for each of the 96 patterns.

[0171] As described above, by inferring MRS reconstruction parameters using a trained model NN based on multiple MRS signals with different summation times, it is expected that the estimation accuracy of MRS reconstruction parameters will be improved.

[0172] In the above embodiment, multiple MRS signals with different summation counts are input to the trained model NN. However, multiple MRS signals with different qualities may also be input. The quality depends on the degree of body movement of the subject P when the MRS signals are acquired. A small degree of body movement indicates high quality, while a large degree of body movement indicates low quality. As an example, the degree of body movement can be determined based on measurement data measured by an external measuring device in parallel with the acquisition of the MRS signals. Examples of external measuring devices that can be used include an electrocardiograph that measures the electrocardiogram waveform of the subject P and a respirometer that measures the respiratory waveform. During periods when the peak values of the electrocardiogram waveform or respiratory waveform or the temporal fluctuations of the peak values are large, the subject P is moving rapidly, and therefore the quality of the MRS signals acquired during those periods can be said to be relatively low. Therefore, the processing circuitry 51 monitors the peak values of the electrocardiogram waveform and the respiratory waveform or the temporal fluctuation of the peak values, and sets the quality of the MRS signals collected during that period to a higher value as the peak value or the temporal fluctuation increases, and sets the quality of the MRS signals collected during that period to a lower value as the peak value or the temporal fluctuation decreases. Note that the degree of body movement may be determined from images captured by an optical camera in parallel with the collection of MRS signals, or may be determined from the output of any other measuring device capable of measuring the body movement of the subject P.

[0173] The quality of the MRS signal may be measured not only by the output from the external measuring instrument but also calculated based on the MRS signal. In this case, the quality of the MRS signal means the error between the MRS signal and an arbitrary reference. As the reference, any MRS signal such as a previously collected MRS signal or an artificially generated MRS signal may be used. When the MRS signal is an MRS spectrum, the error is defined by the degree of frequency shift of another MRS spectrum with respect to the reference MRS spectrum. The processing circuit 51 calculates the difference in the frequency values of the signal peaks derived from the same substance between the target MRS spectrum and the reference MRS spectrum as the error. Since the chemical shift value of water changes when the subject P moves, the error can be used as an index for measuring quality. The processing circuit 51 sets the quality of the MRS signal collected during the period to a higher value as the error is smaller, and sets the quality of the MRS signal collected during the period to a lower value as the error is larger.

[0174] (Seventh Embodiment) The seventh embodiment is a modified example applicable to any of the above embodiments. Hereinafter, the MR signal processing apparatus according to the seventh embodiment will be described. In the following description, components having substantially the same functions as those in the above embodiments are denoted by the same reference numerals, and redundant description will be given only when necessary.

[0175] FIG. 32 is a diagram schematically showing a processing example according to the seventh embodiment. As shown in FIG. 32, the processing circuit 51 collects a plurality of MRS signals with different qualities, rearranges the plurality of collected MRS signals according to the quality (step SD1), and applies the plurality of rearranged MRS signals to the learned model NN to estimate the MRS reconstruction parameters. The learned model NN and the MRS reconstruction parameters according to the seventh embodiment are the same as those in the sixth embodiment. The number of MRS signals input to the learned model NN may be any number as long as it is two or more, but 96 MRS signals are illustrated in FIG. 32. The quality according to the seventh embodiment is the same as that in the sixth embodiment. #n (1 < n ≦ N = 96) in FIG. 32 represents the collection order of the MRS signals. It is assumed that the collection order has no correlation with the quality order No.n (1 < n ≦ N = 96).

[0176] In the sorting process (step SD1), the processing circuit 51 sorts the multiple MRS signals in a predetermined order. For example, the predetermined order is, as shown in FIG. 32, in descending order of quality. The processing circuit 51 inputs the multiple MRS signals to the trained model NN in the sorted order. Specifically, each channel of the input layer of the trained model NN is associated with a quality ranking, and the processing circuit 51 inputs the MRS signal to the channel corresponding to the quality ranking of the input MRS signal. Since the MRS signals are sorted in order of quality before being input to the trained model NN, they are always input in a constant order of quality. This makes the quality of MRS signals input to the same channel relatively constant, which is expected to improve the estimation accuracy of the MRS reconstruction parameters. Note that the above-mentioned predetermined order is not limited to this, and any constant order, such as ascending order of quality, may be used. Furthermore, although the MRS signals are sorted according to quality, they may also be sorted according to the number of additions.

[0177] The trained model NN according to the seventh embodiment is trained as follows. First, the number of additions or quality is set for input data in the training samples using the above-described method. Meanwhile, the number of additions or quality ranking is associated with each channel of the input layer of the machine learning model. Then, as described in the first to fifth embodiments, the processing circuit 51 trains the machine learning model based on multiple training samples. At this time, the input data in each training sample is input to the channel corresponding to the number of additions or quality. In this way, the trained model NN according to the seventh embodiment is trained.

[0178] (Eighth embodiment) The eighth embodiment is a modified example that can be applied to any of the above-mentioned embodiments. An MR signal processing device according to the eighth embodiment will be described below. In the following description, components having substantially the same functions as those in the above-mentioned embodiments will be given the same reference numerals and will be described only when necessary.

[0179] FIG. 33 is a diagram schematically showing a processing example according to the eighth embodiment. As shown in FIG. 33, the processing circuit 51 collects a plurality of MRS signals with different qualities, and distributes the collected plurality of MRS signals into a use group and a rejection group based on the quality (step SE1). The MRS signals distributed to the use group are applied to the learned model NN to estimate MRS reconstruction parameters, and the MRS signals distributed to the rejection group are rejected (step SE2). In other words, the processing circuit 51 rejects MRS signals according to the quality, and applies MRS signals other than the rejected MRS signals to the learned model NN. The learned model NN and the MRS reconstruction parameters according to the eighth embodiment are the same as those in the fifth embodiment. The quality according to the eighth embodiment is the same as that in the sixth embodiment. Let #n (1 < n ≦ N) represent the collection order of the MRS signals. Note that the number N of MRS signals used for distribution may be any number as long as it is two or more, but in FIG. 33, it is assumed that N = 96.

[0180] In the sorting process (step SE1), the processing circuit 51 classifies each MRS signal into a use group and a rejection group according to the quality. The use group is a group to which MRS signals having a quality exceeding the threshold belong. The rejection group is a group to which MRS signals having a quality below the threshold belong. The processing circuit 51 compares the quality of each MRS signal with the threshold for each MRS signal, and determines whether the quality exceeds or is below the threshold. When the quality exceeds the threshold, the processing circuit 51 distributes the MRS signal to the use group, and when the quality is below the threshold, the processing circuit 51 distributes the MRS signal to the rejection group. Then, the processing circuit 51 applies the MRS signals distributed to the use group to the learned model NN to estimate MRS reconstruction parameters. By limiting the MRS signals used for the estimation of the MRS reconstruction parameters to MRS signals having a quality exceeding the threshold in this way, an improvement in the estimation accuracy of the MRS reconstruction parameters is expected. Also, although it has been described that MRS signals are used or rejected according to the quality, they may be used or rejected according to the number of addition times.

[0181] The learned model NN according to the eighth embodiment is learned as follows. First, the number of additions or quality is set for the input data among the learning samples by the above method. The processing circuit 51 trains a machine learning model based on a plurality of learning samples as described in the first to fifth embodiments. At this time, each learning sample will be limited to those having the number of additions or quality above the threshold. Thereby, the learned model NN according to the eighth embodiment is learned.

[0182] (The ninth embodiment) The ninth embodiment is a modification applicable to any of the above embodiments. Hereinafter, the MR signal processing apparatus according to the ninth embodiment will be described. In the following description, components having substantially the same functions as those in the above embodiments are denoted by the same reference numerals, and duplicate description will be made only when necessary.

[0183] FIG. 34 is a diagram schematically showing a processing example according to the ninth embodiment. As shown in FIG. 34, the processing circuit 51 collects a plurality of MRS signals with different qualities, bins the collected plurality of MRS signals according to the quality (step SF1), and applies the MRS signals in units of quality levels (bins) to the learned model NN to estimate MRS reconstruction parameters. The learned model NN and the MRS reconstruction parameters according to the ninth embodiment are the same as those in the sixth embodiment. The quality according to the ninth embodiment is the same as that in the sixth embodiment. #n (1 < n ≦ N) represents the collection order of the MRS signals. Note that the number N of MRS signals subjected to binning may be any number as long as it is 2 or more, but in FIG. 34, it is assumed that N = 96. The number of quality levels may be any number as long as it is 2 or more, but in FIG. 34, it is exemplarily 3 levels. It is assumed that the quality decreases in the order of quality levels "1", "2", and "3".

[0184] As shown in FIG. 34, a trained model NN is prepared for each quality level. Each trained model NN is a machine learning model trained to input an MRS signal of the corresponding quality level and output an MRS reconstruction parameter. The processing circuitry 51 applies the MRS signal of each quality level to the trained model NN corresponding to the corresponding quality level to estimate the MRS reconstruction parameter. The processing circuitry 51 calculates the final MRS reconstruction parameter based on multiple MRS reconstruction parameters corresponding to multiple quality levels. For example, the processing circuitry 51 may calculate the average value of the multiple MRS reconstruction parameters as the final MRS reconstruction parameter, or may calculate the arithmetic average value of MRS reconstruction parameters weighted according to the quality level as the final MRS reconstruction parameter. Alternatively, the processing circuitry 51 may use the MRS reconstruction parameter corresponding to a user-specified quality level as the final MRS reconstruction parameter and discard the MRS reconstruction parameters corresponding to other quality levels. Furthermore, although the MRS signal is binned according to quality, it may also be binned according to the number of additions.

[0185] The trained model NN according to the ninth embodiment is trained as follows. First, the number of additions or quality is set for input data among the training samples using the above-described method. Then, multiple machine learning models corresponding to multiple levels of the number of additions or quality levels are prepared. As described in the first to fifth embodiments, the processing circuit 51 trains each machine learning model based on multiple training samples. At this time, the machine learning model corresponding to each level of the number of additions or quality level is trained based on the training samples belonging to that level. In this way, the trained model NN according to the ninth embodiment is generated.

[0186] (Tenth embodiment) In the above-described embodiments, the MR signal processing device 50 is incorporated into the magnetic resonance imaging apparatus 1. However, the MR signal processing device 50 does not have to be incorporated into the magnetic resonance imaging apparatus 1. The MR signal processing device 50 according to the tenth embodiment will be described below. In the following description, components having substantially the same functions as those in the first to ninth embodiments are denoted by the same reference numerals, and will be described repeatedly only when necessary.

[0187] Fig. 35 is a diagram showing an example of the configuration of a magnetic resonance imaging system 100 including an MR signal processing device according to the tenth embodiment. As shown in Fig. 35, the magnetic resonance imaging system 100 includes a magnetic resonance imaging device 1, an MR signal processing device 50, and a PACS server 90, which are communicably connected to each other via a network. The MR signal processing device 50 is applicable to any of the MR signal processing devices according to the first to ninth embodiments.

[0188] The magnetic resonance imaging apparatus 1 acquires MR acquired signals such as MRS signals and transmits the MR acquired signals to a PACS server 90. The PACS server 90 stores the MR acquired signals. The PACS server 90 stores the MR acquired signals and medical signals acquired by other medical image diagnostic apparatuses such as an X-ray computed tomography apparatus in a searchable manner. The MR signal processing apparatus 50 acquires the MR acquired signal to be processed from the MR acquired signals stored in the PACS server 90 and executes the various processes described in the first to ninth embodiments.

[0189] According to the tenth embodiment, even if an imaging mechanism such as MR spectroscopy or MR imaging is not provided, the accuracy of MR acquisition signals such as MRS signals can be improved by performing the various processes described in the first to ninth embodiments.

[0190] (Additional remarks) According to at least one of the above-described embodiments, the MR signal processing device 50 includes a processing circuitry 51. The processing circuitry 51 applies a trained model to multiple MRS signals acquired by MR spectroscopy for the same subject and outputs parameters for MRS reconstruction. MRS reconstruction generally refers to processing for generating an MRS spectrum. For example, MRS reconstruction includes processing for generating an MRS spectrum by fitting from a spectral model, as in the first and second embodiments, and processing for generating an MRS spectrum by Fourier transform from MRSk data, as in the third embodiment. The parameters for MRS reconstruction are spectral parameters in the first and second embodiments, and are k-space modulation amounts such as f0 shift and phase shift in the third embodiment.

[0191] According to the above configuration, it is possible to obtain highly accurate parameters, and as a result, it is possible to obtain highly accurate MRS signals.

[0192] According to at least one of the embodiments described above, the accuracy of the MRS signal can be improved.

[0193] The term "processor" used in the above description refers to a circuit such as a CPU, a GPU, an application specific integrated circuit (ASIC), a programmable logic device (e.g., a simple programmable logic device (SPLD), a complex programmable logic device (CPLD), and a field programmable gate array (FPGA)). A processor realizes its function by reading and executing a program stored in a memory circuit. Note that instead of storing a program in a memory circuit, the processor may be configured so that the program is directly embedded in the circuit. In this case, the processor realizes its function by reading and executing the program embedded in the circuit. Furthermore, instead of executing a program, a function corresponding to the program may be realized by a combination of logic circuits. Note that each processor in this embodiment is not limited to being configured as a single circuit for each processor, but may be configured as a single processor by combining multiple independent circuits to realize its function. Furthermore, a plurality of components in FIGS. 1, 13, 20, 25, and 35 may be integrated into one processor to realize the functions thereof.

[0194] Although several embodiments have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, modifications, and combinations of embodiments can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, as well as within the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0195] 1. Magnetic resonance imaging device 11 Mounting stand 13 berths 21 Gradient magnetic field power supply 23 Transmitting circuit 25 Receiving circuit 27 Bed drive unit 29 Sequence control circuit 41 Static magnetic field magnet 43 Gradient magnetic field coil 45 Transmitting coil 47 receiving coil 50 MR signal processing device 51A, 51B, 51C, 51D Processing circuit 53 Memory 55 Display 57 Input Interface 58 Communication Interface 131 Top plate 133 Foundation

Claims

1. a parameter output unit that applies a trained model that has been trained to input a plurality of MRS signals and output MRS reconstruction parameters to a plurality of MRS signals related to the same subject and collected by MR spectroscopy, and outputs the MRS reconstruction parameters related to the subject; An MR signal processing apparatus comprising:

2. Further comprising a synthesis unit, The MRS reconstruction parameters include a plurality of parameters, the plurality of parameters are physical quantity information of a plurality of substances included in the object, the synthesis unit generates an MRS spectrum for the object based on the plurality of pieces of physical quantity information and a plurality of spectral models corresponding to the plurality of substances, respectively.

2. The MR signal processing apparatus according to claim 1.

3. a first MRS signal of the plurality of MRS signals is acquired by performing a data acquisition including application of a region-selective pulse; a second MRS signal of the plurality of MRS signals is acquired by performing data acquisition that does not include application of a region-selective pulse; 3. The MR signal processing apparatus according to claim 2.

4. The MR signal processing apparatus according to claim 3 , wherein the parameter output unit applies the plurality of MRS signals and frequency information of a frequency band selected by applying the region-selective pulse to the trained model.

5. a first MRS signal of the plurality of MRS signals is acquired by performing data acquisition under a first combination of TR and TE; a second MRS signal among the plurality of MRS signals is acquired by performing data acquisition under a second combination of TR and TE different from the first combination of TR and TE; 3. The MR signal processing apparatus according to claim 2.

6. the plurality of spectral models are a plurality of basis spectra respectively corresponding to a plurality of bases obtained by performing data compression on a specific MRS spectrum; The plurality of parameters includes component amount information of the plurality of bases related to the object.

3. The MR signal processing apparatus according to claim 2.

7. The MR signal processing apparatus according to claim 6 , wherein the plurality of bases includes a first basis based on MRS spectra of a healthy subject and a second basis based on MRS spectra of an unhealthy subject.

8. Further comprising a determination unit, the parameter output unit applies the trained model to the plurality of MRS signals to output first substance amount or component amount information regarding the first basis and second substance amount or component amount information regarding the second basis; the determination unit determines whether the object is normal or abnormal based on the first substance amount or component amount information and the second substance amount or component amount information.

8. The MR signal processing apparatus according to claim 7.

9. The MR signal processing apparatus according to claim 8 , further comprising a display unit that displays the determination result by said determining unit.

10. The MR signal processing apparatus according to claim 2 , wherein one or more of the plurality of MRS signals are MRS signals generated by the synthesis unit.

11. The MR signal processing apparatus according to claim 1 , wherein the MRS reconstruction parameters further include reliability information.

12. 2. The MR signal processing apparatus according to claim 1, wherein the MRS reconstruction parameter is modulation amount information of k-space modulation caused by magnetic field modulation occurring during data acquisition for the plurality of MRS signals.

13. the plurality of MRS signals are a plurality of k-space data acquired by MR spectroscopy; a correction unit that corrects average k-space data based on at least one k-space data of the plurality of k-space data based on the modulation amount information, 13. The MR signal processing apparatus according to claim 12.

14. The MR signal processing apparatus according to claim 13 , further comprising a generating unit that generates an MRS spectrum based on the corrected average k-space data.

15. Further comprising a plurality of synthesis units, the parameter output unit has a plurality of first parameter output units and a second parameter output unit; the plurality of first parameter output units apply a first trained model to the plurality of MRS signals, respectively, to output a plurality of quantity information items related to materials or bases; the plurality of synthesis units respectively generate a plurality of synthesis MRS signals related to the object based on the quantity information; the second parameter output unit applies a second learned model to the plurality of composite MRS signals and outputs modulation amount information of k-space modulation caused by magnetic field modulation occurring during data acquisition for the plurality of MRS signals.

2. The MR signal processing apparatus according to claim 1.

16. further comprising a first synthesis unit and a second synthesis unit; the parameter output unit has a first parameter output unit, a second parameter output unit, and a third parameter output unit; the first parameter output unit applies a first trained model to one MRS signal among the plurality of MRS signals to output first quantity information related to a material or a basis; the second parameter output unit applies a first trained model to the plurality of MRS signals to output second quantity information related to a material or a basis; the first synthesis unit generates a first synthesis MRS signal related to the object based on the first quantity information; the second synthesis unit generates a second synthesis MRS signal related to the object based on the second quantity information; the third parameter output unit applies a second learned model to the first composite MRS signal and the second composite MRS signal, and outputs modulation amount information of k-space modulation caused by magnetic field modulation occurring during data acquisition for the plurality of MRS signals.

2. The MR signal processing apparatus according to claim 1.

17. The parameter output unit Applying the plurality of MRS signals sorted according to the number of summations or quality to the trained model; or Applying MRS signals other than the MRS signals rejected according to the number of additions or the quality among the plurality of MRS signals to the trained model; 2. The MR signal processing apparatus according to claim 1.

18. The trained model has a plurality of models corresponding to a plurality of quality levels, The parameter output unit bins each of the plurality of MRS signals according to quality and applies the MRS signals to a model in units of quality levels.

2. The MR signal processing apparatus according to claim 1.

19. a parameter output unit that applies the learned model to a plurality of MR acquisition signals related to the same subject and outputs a modulation amount related to k-space modulation caused by magnetic field modulation occurring between data acquisitions of the plurality of MR acquisition signals; An MR signal processing apparatus comprising:

Images