Magnetic resonance imaging apparatus, image processing apparatus, and image processing method

Abstract

To solve such problems that many repetitions are required and the reconstruction time becomes long when performing image reconstruction of measurement data acquired by an MRI apparatus without full sampling with the sequential approximate operation.SOLUTION: A magnetic resonance imaging apparatus estimates the density in the middle result of repetitions and performs the density correction by using the estimated density. The apparatus can perform convergence in a less number of times by performing not only the density correction using an initial value but also the density correction in each repetition processing.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present invention relates to a magnetic resonance imaging apparatus (hereinafter referred to as an MRI apparatus), and more particularly to a technique for reconstructing an image by iterative approximation using insufficient data. [Background technology]

[0002] There is a technology that reconstructs images by iterative approximation when data is insufficient, and this technology is used in MRI for compressed sensing (CS) (Non-Patent Document 1, Patent Document 1). By reducing the amount of data acquired during imaging, the imaging time can be shortened, but there is an issue that image reconstruction requires many repeated processes involving Wavelet transform, Curvelet transform, and their inverse transform, which lengthens the reconstruction time.

[0003] There are several methods for speeding up iterative processing. For example, there is an excellent method called the Alternating Split Bregman method, which achieves high-speed processing even when multiple transformations such as Wavelet transformation and Curvelet transformation are used in iterative processing (Non-Patent Document 2). In addition, density correction is also performed on the initial value of the iteration to speed up convergence. Density correction is a process of weighting data according to the density distribution of the measurement data. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] U.S. Patent Publication No. 2015 / 0126850 [Non-patent literature]

[0005] [Non-Patent Document 1] Michael Lustig, David Donoho, and John M. Pauly. Sparse MRI: The Application of Compressed Sensing for Rapid MR Imaging. Magnetic Resonance in Medicine 58:1182-1195 (2007). [Non-Patent Document 2] Gerlind Plonka, Jianwei Ma. Curvelet-Wavelet Regularized Split Bregman Iteration for Compressed Sensing. International Journal of Wavelets, Multiresolution and Information Processing|Vol. 09, No. 01, pp. 79-110 (2011) Summary of the Invention [Problem to be solved by the invention]

[0006] Although the efficiency of the iterative processing has been improved considerably by the Alternating Split Bregman method described in Non-Patent Document 2 and prior density correction, many iterations are still required, and the problem of long reconstruction time remains, so that further reduction in reconstruction time is desired. [Means for solving the problem]

[0007] In order to solve the above problem, the present invention performs density correction on data that is an intermediate result of repetition. Density correction is premised on the density of the data to be corrected, but the density of the intermediate result of the repetitive calculation changes when data that was not acquired is created. In the present invention, density correction of the intermediate result is performed according to the changing density.

[0008] That is, the MRI apparatus of the present invention comprises an imaging unit that acquires measurement data consisting of nuclear magnetic resonance signals, and an image creation unit that creates an image by successive approximation calculation using the measurement data acquired by the imaging unit through undersampling, and the image creation unit comprises an iterative approximation calculation unit that updates the measurement data by successive approximation calculation for the measurement data, and a density correction unit that performs density correction on the measurement data. And repeat The density correction is performed while changing the amount of density correction for each iteration.

[0009] The image processing device of the present invention has the functions of the image creation section.

[0010] The image processing method of the present invention is an image processing method for creating an image by successive approximation calculation of measurement data, and includes a process of converting the measurement data into processing space data of the successive approximation calculation, a process of converting the processing space data obtained by the successive approximation calculation into measurement data and updating the measurement data, and a process of repeating the conversion process and the updating process using the updated measurement data, and at that time, performing density correction on the updated measurement data, wherein the density correction is characterized in that the amount of density correction is changed for each repetition. Effect of the Invention

[0011] According to the present invention, by performing density correction according to the density of the target data in each iterative process, rather than just using the initial density value of the measurement data, it is possible to make the iterative approximation calculation converge in fewer iterations.

[0012] Furthermore, in the present invention, by performing density correction by estimating density at the end, it is possible to obtain a reconstructed image closer to the correct answer. Accordingly, the number of repetitions may be reduced, and the processing speed may be increased. [Brief description of the drawings]

[0013] [Figure 1] 1 is a diagram showing the overall configuration of an example of an MRI apparatus to which the present invention is applied. [Diagram 2] FIG. 2 is a functional block diagram of a signal processing unit of the MRI apparatus according to the first embodiment. [Diagram 3] FIG. 2 is a flowchart showing an outline of processing performed by the MRI apparatus according to the embodiment. [Figure 4] 11 is a flow diagram showing an example of a procedure of iterative reconstruction according to the first embodiment. [Diagram 5] FIG. 4 is a diagram showing an example of a density estimation method according to the first embodiment. [Figure 6] 5A to 5C are diagrams illustrating the effects of the first embodiment. [Figure 7] 13 is a flow diagram showing an example of a procedure of iterative reconstruction according to the fifth modification of the first embodiment. [Figure 8] FIG. 13 is a diagram showing an example of a UI screen according to Modification Example 5. [Figure 9] A flow chart showing the procedure for iterative reconstruction using the Alternating Split Bregman method. [Figure 10] 13 is a flow diagram showing an example of a procedure of iterative reconstruction according to the second embodiment. [Figure 11] 13 is a flow chart showing another example of the procedure of iterative reconstruction according to the second embodiment. [Figure 12] 13 is a flow diagram showing an example of a procedure of iterative reconstruction according to a modified example of the second embodiment. [Figure 13] 13 is a flow chart showing another example of the procedure of iterative reconstruction according to the modified example of the second embodiment. [Figure 14] FIG. 1 is a diagram showing the configuration of an image processing device. [Figure 15] FIG. 4 is a diagram showing an example of a UI screen according to the first embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] First, an overall outline of an MRI apparatus to which the present invention is applied will be described with reference to the drawings. Fig. 1 is a block diagram showing the overall configuration of an embodiment of an MRI apparatus according to the present invention. This MRI apparatus obtains a tomographic image of a subject by utilizing the NMR phenomenon, and includes a static magnetic field generating unit 2, a gradient magnetic field generating unit 3, a transmitting unit 5, a receiving unit 6, a signal processing unit 7, a sequencer 4, and a central processing unit (CPU) 8. In the following description, the static magnetic field generating unit 2, the gradient magnetic field generating unit 3, the sequencer 4, the transmitting unit 5, and the receiving unit 6 are collectively referred to as an imaging unit.

[0015] The static magnetic field generating unit 2 generates a uniform static magnetic field in the space around the subject 1, and is equipped with a static magnetic field generating device of a permanent magnet type, a resistive conductive type, or a superconductive type. Depending on the direction of the generated magnetic field, there are vertical magnetic field types, horizontal magnetic field types, etc., and the present invention can be applied to any of these types.

[0016] The gradient magnetic field generating unit 3 is composed of a gradient magnetic field coil 9 that applies gradient magnetic fields in three axial directions of X, Y, and Z, which are the coordinate system (stationary coordinate system) of the MRI apparatus, and a gradient magnetic field power supply 10 that drives each gradient magnetic field coil. The gradient magnetic field power supply 10 of each coil is driven according to an instruction from the sequencer 4, thereby applying gradient magnetic fields Gx, Gy, and Gz in three axial directions of X, Y, and Z. A gradient magnetic field in any direction can be generated by combining gradient magnetic fields in these three axial directions. For example, during imaging, a slice direction gradient magnetic field pulse (Gs) is applied in a direction perpendicular to the slice plane (imaging cross section) to set a slice plane for the subject 1, and a phase encoding direction gradient magnetic field pulse (Gp) and a frequency encoding direction gradient magnetic field pulse (Gf) are applied in the remaining two directions perpendicular to the slice plane and perpendicular to each other, to encode position information in each direction in the echo signal.

[0017] The sequencer 4 is a control means that repeatedly applies radio frequency magnetic field pulses (hereinafter referred to as "RF pulses") and gradient magnetic field pulses in a certain predetermined pulse sequence, and operates under the control of the CPU 8, and sends various commands required for collecting data on tomographic images of the subject 1 to the transmitting unit 5, the gradient magnetic field generating unit 3, and the receiving unit 6.

[0018] The transmitter 5 irradiates the subject 1 with an RF pulse to cause nuclear magnetic resonance in the nuclear spins of atoms constituting the biological tissue of the subject 1, and includes a high-frequency oscillator 11, a modulator 12, a high-frequency amplifier 13, and a transmitting-side high-frequency coil (transmitting coil) 14a. The RF pulse output from the high-frequency oscillator 11 is amplitude-modulated by the modulator 12 at a timing according to a command from the sequencer 4, and the amplitude-modulated RF pulse is amplified by the high-frequency amplifier 13 and then supplied to the transmitting coil 14a arranged close to the subject 1, whereby the RF pulse is irradiated to the subject 1.

[0019] The receiving unit 6 detects an echo signal (NMR signal) emitted by nuclear magnetic resonance of atomic spins constituting the biological tissue of the subject 1, and includes a receiving-side high-frequency coil (receiving coil) 14b, a signal amplifier 15, a quadrature phase detector 16, and an A / D converter 17. An NMR signal of the response of the subject 1 induced by the electromagnetic wave irradiated from the transmitting coil 14a is detected by the receiving coil 14b arranged close to the subject 1, amplified by the signal amplifier 15, and then split into two orthogonal systems of signals by the quadrature phase detector 16 at a timing according to a command from the sequencer 4. Each of the signals is converted into a digital quantity by the A / D converter 17 and sent to the signal processing unit 7.

[0020] The signal processing unit 7 performs various data processing, and displays and stores the processing results, and some of the functions are realized by the CPU 8. The processing performed by the signal processing unit 7 (CPU 8) includes image reconstruction of the subject 1, calculation of values ​​indicating the characteristics of the subject, and correction of signals and processing results using device characteristics such as receiving coil sensitivity.

[0021] The signal processing unit 7 includes an external storage device 18 such as an optical disk or a magnetic disk, a display 19, and an operation unit 20 consisting of a trackball, a mouse, a keyboard, etc. When data from the receiving unit 6 is input to the CPU 8, the CPU 8 executes processing such as signal processing and image reconstruction, and displays the resulting tomographic image of the subject 1 on the display 19 and records it on the magnetic disk or the like of the external storage device 18. The CPU 8 functions as a calculation unit that performs the above-mentioned processing, and also functions as a control unit that controls the sequencer 4 and the entire device.

[0022] The operation unit 20 is used to input various control information for the MRI apparatus and control information for processing performed by the signal processing unit 7. It is placed close to the display 19, and the operator interactively controls various processes of the MRI apparatus through the operation unit 20 while looking at the display 19.

[0023] 1, the high frequency coil 14a on the transmitting side and the gradient magnetic field coil 9 are placed in the static magnetic field space of the static magnetic field generating unit 2 into which the subject 1 is inserted, facing the subject 1 in the case of a vertical magnetic field type, or surrounding the subject 1 in the case of a horizontal magnetic field type. The high frequency coil 14b on the receiving side is placed facing or surrounding the subject 1.

[0024] The transmitting coil 14a and the receiving coil 14b may be separate high-frequency coils, or a single high-frequency coil may serve as both the transmitting coil 14a and the receiving coil 14b. Also, a whole-body coil that serves both transmission and reception may be used in combination with another high-frequency coil such as a local coil.

[0025] The MRI apparatus of this embodiment includes the above-mentioned functions of the signal processing unit 7, particularly the image generation function, which includes a function of performing successive approximation calculation such as CS. An overview of the functions of the signal processing unit 7 related to image generation is shown in FIG.

[0026] 2, the signal processing unit 7 mainly includes an imaging setting unit 71 that sets imaging conditions and creates a pulse sequence, and an image creating unit 73. The image creating unit 73 includes an iterative approximation calculation unit (sometimes simply referred to as a calculation unit) 731, a density estimation unit 732, and a density correction unit 733.

[0027] The imaging setting unit 71 calculates a pulse sequence according to the imaging method and imaging conditions set by the user via the operation unit 20 or the like, and sets the pulse sequence in the sequencer 4. The pulse sequence includes imaging methods that scan along the axes of a Cartesian coordinate system as well as non-Cartesian coordinate system scans such as radial scans, and the method of thinning out the measurement space, the density of measured data points, and the like are determined by the scanning method and the speed ratio set as an imaging condition.

[0028] The iterative approximation calculation unit 731 repeatedly performs inverse Fourier transform on the measurement data, transformation into a space for regularization such as wavelet transform, minimization in the space for regularization, and inverse transformation of the Fourier transform and transformation into a space for regularization while updating the measurement data, to create an image.

[0029] The density estimation unit 732 estimates the density of data updated during the repeated calculation. The density can be estimated from the difference between data or based on a predetermined density change curve. Details of the estimation method will be described in the embodiment described later. The density correction unit 733 uses the density estimated by the density estimation unit 732 to perform density correction on the data during the repeated processing.

[0030] The above-mentioned functions of the signal processing unit 7 can be realized by executing software such as a program installed in a memory (storage unit) included in the CPU 8 or in the CPU 8. In addition, some of the functions may be realized by hardware such as an ASIC or an FPGA.

[0031] Next, an overview of imaging and signal processing by the above-mentioned MRI apparatus will be described with reference to the flow of FIG.

[0032] First, the subject 1 is placed in the static magnetic field space of the MRI apparatus (S101), and imaging is performed (S102). Imaging is performed by the imaging unit applying RF pulses and gradient magnetic field pulses to the subject and collecting NMR signals generated from the subject according to a pulse sequence calculated by the imaging setting unit 71. The NMR signals become measurement data (also called k-space data) arranged in a matrix of k-space. In this embodiment, data that fills the entire k-space is not collected, but undersampled measurement data is obtained. The density of the obtained measurement data is determined by imaging parameters such as the encoding step by the gradient magnetic field pulse and imaging conditions such as the thinning rate (or double speed rate).

[0033] Next, the image creation unit 73 uses the measurement data (NMR signals) obtained by imaging to update unmeasured data through repeated calculations by the iterative approximation calculation unit 731, and finally creates a target image (S103). Target images include weighted images such as proton density images and diffusion weighted images, and calculated value images derived from these images. In this image creation step S103, the density correction unit 733 performs density correction (hereinafter referred to as density correction processing) on ​​the data being processed in accordance with the density of the data that changes during the repeated calculations. The density of the changing data is the density estimated by the density estimation unit 732. Methods of density estimation include a method of estimation from the difference in data and a method based on a predetermined density change curve.

[0034] Specific embodiments of the image creation method according to the present invention will be described below. Note that the configuration of the device shown in Figures 1 and 2 and the outline of the operation shown in Figure 3 are common to each embodiment, so these figures will be referred to as appropriate.

[0035] First Embodiment In this embodiment, when performing reconstruction by updating unacquired data through repeated calculations for measurement data that includes thinned-out, i.e., unacquired, measurement points, the density estimation unit 732 estimates the density from the progress of updating the unacquired data in each repetition. The density correction unit 733 performs density correction based on the estimated density.

[0036] FIG. 4 shows an example of the repetitive process in the image creation process (S103) of this embodiment. The data s0(k) acquired by imaging (S102) is density corrected (S1031) and an image is created by inverse Fourier transform (S1032). The image is wavelet transformed (S1033), and only large coefficients are extracted by threshold processing (S1034), then the image is restored by inverse wavelet transform (S1035), and the k-space (space of acquired data) is restored by Fourier transform (S1036). Density estimation (S1037) is performed from this updated data. Of the updated data, the data acquired by imaging is replaced by s0(k) (S1038) to obtain the final updated data s(k). If the updated data has not converged, the process is repeated from the process of performing density correction (S1031) with the estimated density for s(k). If the updated data has converged, the repetition is terminated, and an image is created by inverse Fourier transform (S1040), and the process is terminated.

[0037] Here, density correction (S1031) performed in the first iteration is based on the density (initial value) of data s0(k). From the second iteration onwards, it is performed based on the density of the updated data s(k), i.e., the density estimated in S1037. Density correction is a process of weighting data according to the density distribution of the data; the higher the density, the smaller the weight, and the lower the density, the larger the weight. The reciprocal of the density is called the density correction amount.

[0038] The processing of each step other than step S1037 is similar to the processing of known successive approximation calculations such as compressed sensing, and detailed description thereof will be omitted here. In addition, in Fig. 4, the wavelet transform is shown as an example of transformation into a space for regularization, but a basis function other than the wavelet transform, such as a curvelet transform, may be used, or multiple basis functions may be combined.

[0039] Hereinafter, density estimation in the repetitive processing (S1037) will be specifically described with reference to Fig. 5. Fig. 5(a) is a diagram showing the presence of measurement data in k-space 500, and the position of acquired data (acquired data) is shown in black. Initially, it is determined by the sampling pattern at the time of imaging. The density ρ0(k) of this sampling pattern may be directly obtained from the density of data around position k, or if the probability of acquiring data at each position k is given as p0(k) as shown by the contour line 502 in Fig. 5(b), the probability p may be taken as the density ρ.

[0040] For data that is being updated, the density is estimated according to the progress of the update. For this purpose, we first define the insufficiency of data updates, α(k), as follows:

[0041] [Number 1] α(k) = (b(k) - c(k)) ÷ b(k) (1) In equation (1), b(k) is the average obtained by dividing the sum of the absolute values ​​of the signal value s0(k) of the data acquired by imaging (the black area in Fig. 5(a), called the data acquisition area) by the number of points of the acquired data, and c(k) is the average obtained by dividing the sum of the absolute values ​​of the signal value s(k) of the updated data in the area where no data has been acquired (the white area, called the data non-acquired area) by the number of points in the data non-acquired area. α(k) is 1 when data has not been updated, and becomes 0 when the average value of the data non-acquired area is updated to the magnitude of the average value of the acquired data.

[0042] Next, using the insufficiency of updating α(k) calculated by equation (1) and the initial value ρ0(k) of the density distribution at the time of data acquisition, the density ρ(k) that has changed due to data updating is estimated by the following equation (2).

[0043] [Number 2] ρ(k)=α(k)×(ρ0(k)-1)+1 (2) As can be seen from equation (2), when α(k) = 1 (data has not been updated), the density is estimated to be ρ0(k), and when α(k) = 0 (data is estimated to have been completely updated), the density is estimated to be 1.

[0044] In this example, b(k) and c(k) are averaged over the entire data, and α(k) is a constant value regardless of k. However, b(k) and c(k) may be defined for each small region divided into k-space. That is, as shown by a small square in FIG. 5(a), the sum of the absolute values ​​of the signal values ​​s0(k) of the acquired data included in the region 501 around the position k is divided by the number of data points acquired in the region 501 to obtain the average, and c(k) is defined as the sum of the absolute values ​​of the signal values ​​s(k) of the data updated in the region 501 divided by the number of data points updated in the region 501. In this case, α(k) is a value that changes depending on k.

[0045] In this case, the region 501 for taking the average when calculating b(k) and c(k) need not be the region around k, but may be the region near the edge of a square passing through k (region shown in black) 503 as shown in FIG. 5(c), or when a data acquisition probability p0(k) as shown in FIG. 5(b) is given, it may be the region 504 with a close value of probability p as shown in FIG. 5(d), and may be freely determined.

[0046] The density correction unit 733 performs density correction on the data during the repeated calculation using the density estimated in step S1037 described above (S1031). That is, the density correction is performed on the measurement data with a density correction amount according to the density, and the process proceeds to each repeated process. According to this embodiment, by estimating the density based on the progress of data updating, even during repetition, density correction can be performed with high accuracy based on the probable density at that time.

[0047] A comparison between an image with density correction according to this embodiment and an image without density correction is shown in Fig. 6. In Fig. 6, (A) is an image with no data shortage (full sampling), (B) is an image reconstructed by performing iterative approximation calculation without density correction in a state where data is insufficient, and repeating it 10 times, and (C) is an image reconstructed by iterative approximation calculation while performing density correction midway through the repetition in a state where data is insufficient like (B), and the number of repetitions was 5.

[0048] (D) and (E) are the differences between (A) and (B) and between (A) and (C). Comparing (B) and (C) or (D) and (E) shows that in this embodiment, the number of repetitions is half that of the case where density correction is not performed, and the image approaches the image obtained by full sampling. In other words, this embodiment improves the accuracy of image reconstruction, and it is possible to approach the original image in a shorter time than with conventional image reconstruction. The application of this embodiment may be set in advance, but the user may select whether or not to perform density correction during repetition via a user interface (FIG. 1: operation unit 20). An example of a user interface screen is shown in FIG. 15. In this example, the user can select whether or not to perform density correction during repetition using a pull-down selection method.

[0049] <Modification 1 of the first embodiment> In the first embodiment, as a density estimation method, the progress of the data update is obtained directly from the update data in the data-not-acquired area. However, instead of using the update data in the data-not-acquired area, the progress of the data update may be obtained indirectly from the update data in the data-acquired area.

[0050] In this modification, the insufficiency of data updating α(k) is defined as the following equation (3), and the density ρ(k) is estimated using equation (2) in the first embodiment.

[0051] [Number 3] α(k)=Δb(k)÷Δb0(k) (3) Here, Δb(k) and Δb0(k) are expressed by the following equations (4) and (5), respectively. Δb(k) represents the summation, in the data acquisition area, of the absolute values ​​of the difference between the signal value s(k) of the updated data at position k in the data acquisition area and the signal value s0(k) of the acquired data at the same position k. Δb0(k) represents the summation, in the data acquisition area, of the absolute values ​​of the difference between the signal value s0(k) of the acquired data at position k in the data acquisition area, which is density-corrected by dividing the signal value s0(k) by the density ρ0(k) at that position, and the signal value s0(k).

[0052] [Number 4] Δb(k)=Σ|s(k)-s0(k)| (4) [Number 5] Δb0(k)=Σ|s0(k)÷ρ0(k)-s0(k)| (5) The data acquisition region may be the entire k-space, or may be a part of the data acquisition region including the position k.

[0053] In equation (3), when updates are insufficient, s(k) deviates from s0(k), so α(k) becomes larger and approaches 1. When the data is completely updated, s(k) does not deviate from s0(k), so α(k) becomes smaller and approaches 0. Then, the density can be estimated using equation (2).

[0054] <Modification 2 of the First Embodiment> In the first embodiment, the density ρ(k) estimated by the formula (2) is not distinguished between the data acquisition area and the data non-acquired area, but the density may be different between the data acquisition area and the data non-acquired area. For example, the density in the data acquisition area is set to 1 without changing, and only the data non-acquired area is calculated by using the insufficiency of update α(k) defined by the formula (1),

[0055] [Number 6] ρ(k)=1-α(k) (6) It is also possible to use the following.

[0056] Alternatively, the density may be defined as shown in the following formula (7) and may be changed depending on the region. [Number 7] ρ(k)=w(k)×α(k)×((1-w(k))×ρ0(k)-1)+1 (7) In equation (7), w(k) is a weight according to the region. For example, w(k)=0 in the data acquisition region. Areas where data is not available So if w(k)=1, ρ(k)=1(data acquisition area) ρ(k)=1-α(k)(data not acquired area) This results in the case of equation (6).

[0057] <Modification 3 of the First Embodiment> The density estimated in the first embodiment or its modified examples 1 and 2 may be further adjusted. Examples of adjustment methods include the following. α(k) is adjusted to be 0 to 1, and ρ(k) is adjusted to be always in the range of ρ0(k) to 1. Alternatively, ρ(k) may be adjusted to increase monotonically, that is, if the density estimated in the nth iteration is ρn, ρ(k) may be adjusted so that ρn≧ρn-1 for the density ρn-1 estimated in the previous iteration.

[0058] By making such adjustments, even if the signal values ​​of the acquired data or updated data contain abnormal values ​​due to artifacts, noise, etc., it is possible to perform density estimation and density correction based on the estimation that is not influenced by the abnormal values.

[0059] <Fourth Modification of the First Embodiment> In the first embodiment, the update progress α is calculated using the data being updated, and the density of the data being updated is estimated based on this, but the update progress α may be defined as a function with the number of iterations n of the calculation as a variable, and the density of the data being updated may be estimated using this function. For example, the following formula (8) may be used as the function.

[0060] [Number 8] α(n)=0.5^n (8) In this case, the density ρ(k) can be estimated using the above-mentioned formula (2). Note that the function is not limited to formula (8) as long as it has an initial value of 1 and gradually decreases.

[0061] The above-mentioned modified example 2 can also be applied to this modified example. For example, the density of the data acquisition area may be fixed to 1, and the density may be changed only in the data non-acquired area.

[0062] According to this modification, the density can be estimated based on a predetermined function without using updated data, so that the density can be easily estimated.

[0063] <Fifth Modification of the First Embodiment> In the first embodiment, the case where s(k) is subjected to an inverse Fourier transform (S1040) after the repetition is completed to obtain the final image has been described, but it is also possible to perform density estimation on s(k), perform density correction using the estimated density, and then perform an inverse Fourier transform to obtain the final image.

[0064] The process flow of this modified example is shown in Fig. 7. The process up to S1038 is the same as in Fig. 4, but in this modified example, density estimation is performed on the measurement data obtained after the repeated calculation is completed, and density correction is performed based on the estimated density. The density estimation method may be any of the methods in the first embodiment and modified examples 1 to 4 thereof.

[0065] The application of this modified example may be set in advance, but the user may select whether or not to add density correction after the repetition ends via a user interface (FIG. 1: operation unit 20). An example of a user interface screen is shown in FIG. 8. In this example, as shown in (A) and (B), the user can select whether or not to perform density correction during the repetition and density correction after the repetition ends, using a pull-down selection method.

[0066] According to this modification, even if the iterative processing does not sufficiently converge, image quality can be improved by performing density correction, so that a good approximate image can be obtained with even fewer iterations. Also, by allowing the user to select, image reconstruction can be performed according to the user's preference, such as prioritizing image quality or time until image creation.

[0067] Second Embodiment In the first embodiment, a reconstruction was performed in which a process for density correcting data s(k) was added for each repetition to a general iterative approximation iterative calculation, but in this embodiment, the iterative approximation calculation unit 731 uses a high-speed Alternating Split Bregman algorithm for the iterative processing. In this embodiment, density correction is added in the iterative processing as in the first embodiment, but the difference is that the entire processing is changed rather than adding a process (S1031) for density correcting data s(k) to the flow as in FIG. 4.

[0068] First, the iterative approximation calculation using the Alternating Split Bregman method employed in this embodiment will be described.

[0069] This image reconstruction by iterative approximation is a process in which measurement data converted into image space data by inverse Fourier transform is subjected to wavelet transform and curvelet transform to minimize the L1 norm of the measurement data to obtain a reconstructed image, and is a process of solving the problem expressed by the following equation (9).

[0070]

number

[0071] In equation (9), f is the imaging data, u is the reconstructed image, Φ is a Fourier transform to extract only the data acquisition area, Ψ is a curvelet transform, Ψ is a wavelet transform, θ is a curvelet transform coefficient, θ is a wavelet transform coefficient, and Λ is a weight set to the curvelet coefficient. Λw is The weights to be set for the wavelet coefficients.

[0072] This problem can be reduced to repeatedly solving the following equation (10) using the Bregman method. In the following equation, the superscript "n" indicates the number of iterations (hereinafter the same).

[0073]

number

[0074] Equation (10) can be further replaced by the Split Bregman method as follows:

number

[0075] Equation (11) can be further replaced by the Alternating Split Bregman method as follows:

number

[0076] The image creation process (Fig. 3: S103) using the Alternating Split Bregman method is as shown in Fig. 9. First, set the initial values ​​of u, θw, θc, bc, and bw (S2031). The initial value can be 0, for example. Next, update u using equation (12-1) (S2032). Next, update θc using equation (12-2) and update θw using equation (12-3) (S2033). Next, update bc using equation (12-4) and update bw using equation (12-5) (S2034). Next, update f using equation (12-6) (S2035). If the solution u has not converged, return to S2032 and repeat the process. If the solution u has converged, the process ends.

[0077] In this embodiment, density correction is introduced into this process. To achieve this, first, equation (10) for determining u, θc, and θw is modified to the following equation (13).

[0078]

number

[0079] In equation (13), F represents the Fourier transform, and Φ is a transformation that does not include the Fourier transform and extracts only the data acquisition area, (Ω n ) -1 represents the density correction process. The method of density correction is the same as that described in the first embodiment, and the data obtained in the nth repetition (F n ) or the density estimated by the number of iterations n. n Since can be determined arbitrarily, it is assumed that it does not depend on u. μ is the weight of the term and takes a value greater than 0.

[0080] Here, (Ω n ) -1 In density estimation to determine the density correction amount, if the density becomes 1 after a certain n, and no density correction is performed, then it will converge to the same solution as equation (9), and equation (9) can be solved by equation (13). Even if the density does not become 1 no matter how large n is, after a certain n, Ω n becomes Ω∞, the solution will converge to the solution of equation (14), which is a modification of equation (9). n As long as we make sure that converges, the solution to equation (13) is guaranteed to converge.

[0081]

number

[0082] In this embodiment, by further replacing the formula (13) with the following formula (15), the same solution combination (sequence) {(u 1 ,θ 1 c,θ 1 w)(u 2 ,θ 2 c,θ 2 w)(u 3 ,θ 3 c,θ 3Note that u, θc, θw, f, bc, and bw can be real numbers, vectors, or complex numbers, or a matrix of n rows and 2 columns using a matrix representation of complex numbers.

[0083]

number

[0084] Furthermore, the fixed point obtained by this iteration is indeed the solution of equation (13).

[0085] Furthermore, equation (15-1) is changed to equations (16-1) to (16-3), and optimized in order by u, θc, and θw. Equations (16-4) to (16-6) are the same as equations (15-2) to (15-4).

number

[0086] Specifically, formula (16-1) is

number

[0087] Specifically, equations (16-2) and (16-3) are as follows:

number

[0088] FIG. 10 shows the flow of the image creation process (S103) of this embodiment in which the above-mentioned density correction process (processing based on equation (16)) is introduced.

[0089] First, set the initial values ​​of u, θw, θc, bc, and bw (S3031). This step is the same as step S2031 in FIG. 9, and the initial value may be, for example, 0. Next, update u with equation (16-1) including density correction (S3032). Next, update θc with equation (16-2) including density correction, and update θw with equation (16-3) (S3033). Next, update bc with equation (16-4) including density correction, and update bw with equation (16-5) (S3034). Next, update f with equation (16-6) including density correction (S3035). If the solution u has not converged, return to S3032 and repeat the process. If the solution u has converged, end the process.

[0090] The variables (θc, θw, bc, bw) to be updated at each iteration may be changed as follows. The left side of the third and fourth equations is b(~) n c , b(~) n w It is written as follows.

number

[0091] In that case, equations (16-1) to (16-6) become as follows.

number

[0092] And equation (17) becomes as follows:

number

number

[0093] In this case, the flow of the image creation process (S103) is shown in FIG. n c , b(~) n wThe initial value of is set (S4031). The initial value may be, for example, 0. Next, u is updated by equation (20-1) including density correction (S4032). Next, uc is updated by equation (20-2) including density correction, and uw is updated by equation (20-3) including density correction (S4033). Next, b(~) is updated by equation (20-4) including density correction. n c b(~) in equation (20-5) including density correction n w is updated (S4034). Next, f is updated using equation (20-6) including density correction (S4035). If the solution u has not converged, the process returns to S4032 and is repeated. If the solution u has converged, the process ends.

[0094] <Modification 1 of the second embodiment> In the second embodiment, as shown in equation (14), a density correction (Ω) is applied before performing a wavelet transform or a curvelet transform (i.e., a transform into a space that minimizes the L1 norm). -1 However, density correction may be performed on the measurement data under the condition of f = Φu to match the measurement data. In this case, instead of adding the density correction Ω in the formula (13) of the second embodiment to the terms of the curvelet transform and the wavelet transform, it is added to the term of ||ΦFu-f|| as shown in the following formula.

[0095]

number

[0096] In this case as well, the process flow is the same as that shown in FIGS.

[0097] <Modification 2 of the second embodiment> In the second embodiment, as in variant example 5 of the first embodiment, before obtaining a final image from the measurement data after the repetition is completed, density correction may be performed on the measurement data obtained at the end of the repetition, and then an inverse Fourier transform may be performed to obtain the final image.

[0098] In this case, as shown in Figures 12 and 13, a process of Fourier transforming u (S5038), density correction (S1039), and inverse Fourier transform (S1040) are added after the repetition of Figures 10 and 11 is completed. The density estimation method may be any of the methods in the first embodiment and its modified examples 1 to 4.

[0099] According to this modification, even if the iterative processing does not sufficiently converge, the image quality can be improved by performing density correction, so that a good approximate image can be obtained with even fewer iterations.

[0100] Although an embodiment in which the image processing of the present invention is performed by the signal processing unit 7 (image creation unit) of the MRI apparatus has been described above, it is also possible to perform the functions of the image creation unit 73 in an image processing device 200 independent of the MRI apparatus 100, as shown in Fig. 14. The image processing device 200 is connected to the MRI apparatus 100 via a wired, wireless, or internet line, or receives measured data acquired by the imaging unit of the MRI apparatus 100 via a portable medium, and performs image reconstruction calculations. The created image may be displayed on a display attached to the image processing device 200, stored in a storage device, or sent to the MRI apparatus 100. [Explanation of symbols]

[0101] 2: static magnetic field generator, 3: gradient magnetic field generator, 4: sequencer, 5: transmitter, 6: receiver, 7: signal processing unit, 8: central processing unit (CPU), 14a: transmitter coil, 14b: receiver coil, 18: external storage device, 19: display, 20: operation unit, 71: imaging setting unit, 73: image creation unit, 731: iterative approximation calculation unit, 732: density estimation unit, 733: density correction unit, 100: MRI device, 200: image processing device.

Claims

1. An imaging unit that acquires measurement data consisting of nuclear magnetic resonance signals, and an image creation unit that creates an image by successive approximation calculation using the measurement data acquired by the imaging unit through undersampling, the image creation unit includes an iterative approximation calculation unit that updates the measurement data by iterative approximation calculation on the measurement data, and a density correction unit that performs density correction on the measurement data, a density correction unit that divides the measurement data into a plurality of regions, determines a density correction amount from the density obtained for each region, and performs density correction while changing the density correction amount for each iteration of the iterative approximation calculation.

2. A system comprising: an imaging unit that acquires measurement data consisting of nuclear magnetic resonance signals; and an image creation unit that creates an image by successive approximation calculation using the measurement data acquired by the imaging unit through undersampling; the image creation unit includes an iterative approximation calculation unit that updates the measurement data by iterative approximation calculation on the measurement data, and a density correction unit that performs density correction on the measurement data, The density correction unit performs density correction while changing the amount of density correction for each iteration of the iterative approximation calculation, and at that time, among the measurement data to be corrected, performs density correction with different weights on the data of the measurement points acquired by the imaging unit and the data of the measurement points estimated by the iterative approximation calculation.

3. A system comprising: an imaging unit that acquires measurement data consisting of nuclear magnetic resonance signals; and an image creation unit that creates an image by successive approximation calculation using the measurement data acquired by the imaging unit through undersampling; the image creation unit includes an iterative approximation calculation unit that updates the measurement data by iterative approximation calculation on the measurement data, and a density correction unit that performs density correction on the measurement data, The magnetic resonance imaging apparatus according to claim 1, wherein the density correction unit performs density correction while changing the amount of density correction for each iteration of the iterative approximation calculation, and performs density correction based on a preset change in density.

4. An imaging unit that acquires measurement data consisting of nuclear magnetic resonance signals, and an image creation unit that creates an image by successive approximation calculation using the measurement data acquired by the imaging unit through undersampling, the image creation unit includes an iterative approximation calculation unit that updates the measurement data by iterative approximation calculation on the measurement data, and a density correction unit that performs density correction on the measurement data, The density correction unit performs density correction by changing the density correction amount for each repetition of the iterative approximation calculation, and adjusts the density correction amount to perform density correction under the condition that the insufficiency of updating the measurement data updated by the iterative approximation calculation is greater than or equal to 0 and less than 1.

5. 4. A magnetic resonance imaging apparatus according to claim 1, 4. A magnetic resonance imaging apparatus, comprising: a density correction unit that performs density correction by changing the amount of density correction under the condition that the change in density increases monotonically.

6. 5. A magnetic resonance imaging apparatus according to claim 1, 4. A magnetic resonance imaging apparatus according to claim 3, wherein the density correction unit determines an initial value of a density correction amount in accordance with a data acquisition method in the imaging unit.

7. A method for detecting a subject, comprising: an imaging unit that acquires measurement data consisting of nuclear magnetic resonance signals; and an image creation unit that creates an image by successive approximation calculation using the measurement data acquired by the imaging unit through undersampling; the image creation unit includes an iterative approximation calculation unit that updates the measurement data by iterative approximation calculation on the measurement data, and a density correction unit that performs density correction on the measurement data, the density correction unit performs density correction while changing a density correction amount for each iteration of the iterative approximation calculation, The magnetic resonance imaging apparatus according to claim 1, wherein the density correction unit performs density correction after the iterative approximation calculation unit completes the repetition of the iterative approximation calculation.

8. A method for detecting a subject, comprising: an imaging unit that acquires measurement data consisting of nuclear magnetic resonance signals; and an image creation unit that creates an image by successive approximation calculation using the measurement data acquired by the imaging unit through undersampling; the image creation unit includes an iterative approximation calculation unit that updates the measurement data by iterative approximation calculation on the measurement data, and a density correction unit that performs density correction on the measurement data, the iterative approximation calculation unit includes a Fourier transform unit that performs a conversion from the measurement data to real space data and an inverse conversion thereof, and a space transformation unit that performs a conversion from the real space data to space data for regularization and an inverse conversion thereof, The magnetic resonance imaging apparatus according to claim 1, wherein the density correction unit performs density correction while changing an amount of density correction for each iteration of the iterative approximation calculation.

9. 5. A magnetic resonance imaging apparatus according to claim 1, a Fourier transform unit that performs a conversion from the measurement data to real space data and an inverse conversion thereof, and a spatial transformation unit that performs a conversion from the real space data to spatial data for regularization and an inverse conversion thereof.

10. A method of imaging a subject, comprising: an imaging unit that acquires measurement data consisting of nuclear magnetic resonance signals; and an image creation unit that creates an image by successive approximation calculation using the measurement data acquired by the imaging unit through undersampling; the image creation unit includes an iterative approximation calculation unit that updates the measurement data by iterative approximation calculation on the measurement data, and a density correction unit that performs density correction on the measurement data, the iterative approximation calculation unit performs a calculation based on an Alternating Split Bregman algorithm as the iterative approximation calculation, The magnetic resonance imaging apparatus according to claim 1, wherein the density correction unit performs density correction while changing an amount of density correction for each iteration of the iterative approximation calculation.

11. 5. A magnetic resonance imaging apparatus according to claim 1, The magnetic resonance imaging apparatus according to claim 1, wherein the iterative approximation calculation unit performs, as the iterative approximation calculation, a calculation based on the Alternating Split Bregman algorithm.

12. An image processing device including a calculation unit that receives measurement data and creates an image by successive approximation calculation, 5. An image processing apparatus comprising: a density correction unit that performs density correction of the measurement data while changing a density correction amount for each repetition of the iterative approximation calculation, the density correction unit being the density correction unit according to claim 1.

13. The image processing device according to claim 12, 2. An image processing apparatus according to claim 1, wherein the measurement data is measurement data consisting of nuclear magnetic resonance signals measured by a magnetic resonance imaging apparatus.

14. 1. An image processing method for creating an image by successive approximation of measurement data, comprising: The iterative approximation calculation includes a conversion process including a conversion from the measurement data to real space data and an inverse conversion thereof, and a conversion from the real space data to space data to be regularized and an inverse conversion thereof; A process of converting the data in the processing space that has been subjected to the iterative approximation calculation into measurement data and updating the measurement data; repeating the conversion process and the updating process using the updated measurement data; At that time, a process of performing density correction on the updated measurement data is included. The image processing method according to the present invention, characterized in that the density correction is performed by varying an amount of density correction for each repetition.

15. An image processing method for creating an image by successive approximation of measurement data, comprising the steps of: The iterative approximation calculation is performed based on the Alternating Split Bregman algorithm, and the measurement data is repeatedly updated; At that time, a process of performing density correction on the updated measurement data is included. The image processing method according to the present invention, characterized in that the density correction is performed by varying an amount of density correction for each repetition.

Images