Image processing device, image processing method, and image processing program

Abstract

To reduce a distortion in an image generated along the phase encode direction while keeping an anatomical structure.SOLUTION: An image processing device according to an embodiment comprises: an acquisition unit; a shift map generation unit; and an image generation unit. The acquisition unit acquires two magnetic resonance images corresponding to the two mutually-opposite phase encode directions. The shift map generation unit optimizes a cost function using a first difference in the two magnetic resonance images and a second difference in the two edge images generated on the basis of the two magnetic resonance images to generate a shift map about the shift of a plurality of pixels in the two magnetic resonance images. The image generation unit generates a correction image obtained by correcting a distortion in the two magnetic resonance images on the basis of the two magnetic resonance images and the shift map.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The embodiments disclosed in this specification and the drawings relate to an image processing device, an image processing method, and an image processing program. [Background technology]

[0002] To reduce image distortion in the phase encoding direction in magnetic resonance imaging, a method has been proposed that utilizes the difference in the distortion direction between two images acquired in the forward and reverse phase encoding directions. Typically, the difference in distortion between the two images is expressed as a cost function, and a shift map used to correct the distortion in the two images is obtained by solving an optimization problem to minimize this cost function. Examples of cost functions that have been proposed include the least-squares difference between the two images and a cost function with a constraint term added to suppress the change in the shift map.

[0003] However, using only the cost function described above poses a problem in that there is a high risk of destroying anatomical structures when there is strong image distortion or when unexpected differences occur between two images due to imperfections in the collected images (e.g., pulsation or body movement) (e.g., large differences due to pixel compression caused by distortion, or changes in magnetic resonance signals of cerebrospinal fluid (CSF) or blood due to pulsation). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2019-50937 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 reduce image distortion that occurs along the phase encoding direction while preserving anatomical structures. 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 each configuration shown in the embodiments described below can also be positioned as other problems. [Means for solving the problem]

[0006] An image processing device according to an embodiment includes an acquisition unit, a shift map generation unit, and an image generation unit. The acquisition unit acquires two magnetic resonance images corresponding to two opposite phase encoding directions. The shift map generation unit generates a shift map relating to shifts of multiple pixels in the two magnetic resonance images by optimizing a cost function using a first difference between the two magnetic resonance images and a second difference between two edge images generated based on the two magnetic resonance images. The image generation unit generates a corrected image in which distortion in the two magnetic resonance images is corrected, based on the two magnetic resonance images and the shift map. [Brief explanation of the drawings]

[0007] [Figure 1] FIG. 1 is a block diagram illustrating an example of an image processing apparatus according to an embodiment. [Figure 2] FIG. 2 is a diagram showing an example of an MRI apparatus according to an embodiment. [Figure 3] FIG. 3 is a diagram showing an example of the relationship between the coefficient distribution of the fitting model and a shift map according to the embodiment. [Figure 4] FIG. 4 is a flowchart showing an example of a procedure of a correction process according to the embodiment. [Figure 5] FIG. 5 is a diagram showing an example of a brightness-corrected forward image generated from a position-corrected forward image according to the embodiment. [Figure 6] FIG. 6 is a diagram showing an example of a distribution of relative errors and threshold values for phase encoding lines according to a second application example of the embodiment. [Figure 7]FIG. 7 is a diagram showing an example of a distribution of relative errors and threshold values with respect to slice numbers according to a second application example of the embodiment. [Figure 8] FIG. 8 is a flowchart illustrating an example of a procedure of a correction process according to a third application example of the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] Hereinafter, embodiments of an image processing device, an image processing method, and an image processing program will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing an example of an image processing device 1. The image processing device 1 is installed, for example, in various modalities equipped with various functions of the image processing device 1, or in a server in a hospital. Note that the various functions of the image processing device 1 may also be installed in various server devices, such as a server for a medical image management system (hereinafter referred to as a PACS (Picture Archiving and Communication Systems)) or a server for a hospital information system (hereinafter referred to as an HIS (Hospital Information System)).

[0009] Furthermore, modalities equipped with the various functions of the image processing device 1 include, for example, a magnetic resonance imaging (hereinafter referred to as an MRI) device, a PET (Positron Emission Tomography)-MRI device, a SPECT (Single Photon Emission Computed Tomography)-MRI device, etc. For the sake of concrete explanation, it is assumed below that the image processing device 1 is installed in an MRI device. In this case, the MRI device has the various functions of the processing circuitry 15.

[0010] (Embodiment) 2 is a diagram showing an example of an MRI apparatus 100 according to this embodiment. As shown in FIG. 2, the image processing apparatus 1 in the MRI apparatus 100 further includes an input / output interface 17. Note that the image processing apparatus 1 may not include the input / output interface 17, as shown in FIG. 1. As shown in FIG. 2, the MRI apparatus 100 includes a static magnetic field magnet 101, a gradient magnetic field coil 103, a gradient magnetic field power supply 105, a bed 107, a bed control circuit (bed control unit) 109, a transmission circuit 113, a transmission coil 115, a reception coil 117, a reception circuit 119, an imaging control circuit (acquisition unit) 121, a system control circuit (system control unit) 123, a storage device 125, and the image processing apparatus 1.

[0011] The static magnetic field magnet 101 is a magnet formed in a hollow, approximately cylindrical shape. The static magnetic field magnet 101 generates a substantially uniform static magnetic field in the internal space. For example, a superconducting magnet or the like is used as the static magnetic field magnet 101.

[0012] The gradient magnetic field coil 103 is a hollow, approximately cylindrical coil and is disposed on the inner surface of the cylindrical cooling vessel. The gradient magnetic field coil 103 receives current individually from a gradient magnetic field power supply 105 to generate gradient magnetic fields whose magnetic field strength varies along the mutually orthogonal X, Y, and Z axes. The gradient magnetic fields of the X, Y, and Z axes generated by the gradient magnetic field coil 103 form, for example, a slice selection gradient magnetic field, a phase encoding gradient magnetic field, and a frequency encoding gradient magnetic field (also referred to as a readout gradient magnetic field). The slice selection gradient magnetic field is used to arbitrarily determine an imaging cross section. The phase encoding gradient magnetic field is used to change the phase of a magnetic resonance signal (hereinafter referred to as an MR (Magnetic Resonance) signal) according to a spatial position. The frequency encoding gradient magnetic field is used to change the frequency of the MR signal according to a spatial position.

[0013] The gradient magnetic field power supply 105 is a power supply device that supplies current to the gradient magnetic field coil 103 under the control of the imaging control circuit 121 .

[0014] The bed 107 is a device equipped with a top plate 1071 on which the subject P is placed. The bed 107 inserts the top plate 1071 on which the subject P is placed into the bore 111 under the control of a bed control circuit 109.

[0015] The bed control circuit 109 is a circuit that controls the bed 107. The bed control circuit 109 drives the bed 107 in response to instructions from the operator via the input / output interface 17, thereby moving the tabletop 1071 in the longitudinal direction, the up-down direction, and in some cases the left-right direction.

[0016] The transmission circuit 113 supplies radio frequency pulses modulated at the Larmor frequency to the transmission coil 115 under the control of the imaging control circuit 121. For example, the transmission circuit 113 includes an oscillator, a phase selection unit, a frequency conversion unit, an amplitude modulation unit, an RF amplifier, and the like. The oscillator generates an RF pulse at a resonance frequency specific to the target atomic nucleus in a static magnetic field. The phase selection unit selects the phase of the RF pulse generated by the oscillator. The frequency conversion unit converts the frequency of the RF pulse output from the phase selection unit. The amplitude modulation unit modulates the amplitude of the RF pulse output from the frequency conversion unit according to, for example, a sinc function. The RF amplifier amplifies the RF pulse output from the amplitude modulation unit and supplies it to the transmission coil 115.

[0017] The transmission coil 115 is an RF (Radio Frequency) coil arranged inside the gradient magnetic field coil 103. In response to the output from the transmission circuit 113, the transmission coil 115 generates an RF pulse corresponding to a high frequency magnetic field.

[0018] The receiving coil 117 is an RF coil arranged inside the gradient magnetic field coil 103. The receiving coil 117 receives MR signals emitted from the subject P by a high frequency magnetic field. The receiving coil 117 outputs the received MR signals to a receiving circuit 119. The receiving coil 117 is, for example, a coil array having one or more, typically a plurality of coil elements. For the sake of concreteness, the receiving coil 117 will be described below as a coil array having a plurality of coil elements.

[0019] The receive coil 117 may be configured by a single coil element. Although the transmit coil 115 and receive coil 117 are shown as separate RF coils in Fig. 2, the transmit coil 115 and receive coil 117 may be implemented as an integrated transmit / receive coil. The transmit / receive coil corresponds to the imaging region of the subject P and is, for example, a local transmit / receive RF coil such as a head coil.

[0020] The receiving circuit 119 generates digital MR signals (hereinafter referred to as MR data) based on the MR signals output from the receiving coil 117 under the control of the imaging control circuit 121. Specifically, the receiving circuit 119 performs various signal processing on the MR signals output from the receiving coil 117, and then performs analog-to-digital (A / D) conversion on the data that has been subjected to various signal processing to generate MR data. The receiving circuit 119 outputs the generated MR data to the imaging control circuit 121. For example, MR data is generated for each coil element and output to the imaging control circuit 121 together with a tag that identifies the coil element.

[0021] The imaging control circuit 121 controls the gradient magnetic field power supply 105, the transmission circuit 113, the reception circuit 119, etc. in accordance with the imaging protocol output from the processing circuit 15 to perform imaging of the subject P. The imaging protocol has a pulse sequence according to the type of examination. The imaging protocol defines the magnitude of the current supplied to the gradient magnetic field coil 103 by the gradient magnetic field power supply 105, the timing at which the gradient magnetic field power supply 105 supplies the current to the gradient magnetic field coil 103, the magnitude and duration of the radio frequency pulse supplied to the transmission coil 115 by the transmission circuit 113, the timing at which the radio frequency pulse is supplied to the transmission coil 115 by the transmission circuit 113, the timing at which the MR signal is received by the reception coil 117, etc. The imaging control circuit 121 drives the gradient magnetic field power supply 105, the transmission circuit 113, the reception circuit 119, etc. to image the subject P, and then receives MR data from the reception circuit 119 and transfers the received MR data to the image processing device 1, etc.

[0022] The imaging control circuit 121 is realized by, for example, a processor. The imaging control circuit 121 executes a pulse sequence for performing imaging along two opposite phase encoding directions, thereby acquiring two sets of MR data corresponding to the two opposite phase encoding directions. The two opposite phase encoding directions are, for example, a direction in which the phase encoding line increases (hereinafter referred to as the antegrade direction) and a direction in which the phase encoding line decreases (hereinafter referred to as the retrograde direction) in k-space. For the sake of concrete explanation, it is assumed that the sequence for acquiring MR data along the antegrade direction and the sequence for acquiring MR data along the retrograde direction are each performed by echo planar imaging (EPI).

[0023] Note that the sequence for acquiring MR data along the antegrade direction and the sequence for acquiring MR data along the retrograde direction are not limited to EPI, and may be realized by, for example, a field echo method (also called a gradient echo method). Hereinafter, MR data acquired by EPI along the antegrade direction will be referred to as antegrade MR data. MR data acquired by EPI along the retrograde direction will be referred to as retrograde MR data. In this case, two MR data corresponding to two opposite phase encoding directions correspond to antegrade MR data and retrograde MR data.

[0024] In the above description, an example has been described in which the "processor" reads out and executes a program corresponding to each function from memory 13, but the embodiment is not limited to this. The term "processor" refers to a circuit such as a CPU, a GPU (Graphics Processing Unit), 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)).

[0025] If the processor is a CPU, for example, the processor realizes its functions by reading and executing a program stored in memory 13. On the other hand, if the processor is an ASIC, instead of storing a program in memory 13, the function is directly incorporated into the processor circuit as a logic circuit. Note that each processor in this embodiment is not limited to being configured as a single circuit, but may be configured as a single processor by combining multiple independent circuits to realize its functions. Also, although the description has been given assuming that a single storage circuit stores a program corresponding to each processing function, multiple storage circuits may be distributed and arranged, and the processing circuits may read corresponding programs from individual storage circuits.

[0026] The system control circuit 123 includes hardware resources such as a processor, a read-only memory (ROM), a random access memory (RAM), and the like (not shown), and controls the MRI apparatus 100 using a system control function. Specifically, the system control circuit 123 reads a system control program stored in the storage device 125, loads it into the memory, and controls each circuit of the MRI apparatus 100 according to the loaded system control program. For example, the system control circuit 123 loads an imaging protocol from the storage device 125 based on imaging conditions input by an operator via the input / output interface 17. The system control circuit 123 transmits the imaging protocol to the imaging control circuit 121 and controls imaging of the subject P. The system control circuit 123 is realized by, for example, a processor. The system control circuit 123 may be incorporated into the processing circuitry 15. In this case, the system control function is executed by the processing circuitry 15, and the processing circuitry 15 functions as a substitute for the system control circuit 123.

[0027] The storage device 125 stores various programs executed by the system control circuit 123, various imaging protocols, imaging conditions including a plurality of imaging parameters that define the imaging protocols, etc. The storage device 125 is, for example, a semiconductor memory element such as a RAM or a flash memory, a hard disk drive (HDD), a solid state drive (SSD), an optical disk, etc. The storage device 125 may also be a drive that reads and writes various information from and to a portable storage medium such as a compact disc (ROM) drive, a digital versatile disc (DVD) drive, or a flash memory. Note that the data stored in the storage device 125 may also be stored in the memory 13. In this case, the memory 13 functions as a substitute for the storage device 125.

[0028] The image processing device 1 includes a communication interface 11, a memory 13, and a processing circuit 15. As shown in FIGS. 1 and 2, the communication interface 11, the memory 13, and the processing circuit 15 are electrically connected via a bus in the image processing device 1. As shown in FIGS. 1 and 2, the image processing device 1 is connected to a network via the communication interface 11. The network is communicatively connected to, for example, various modalities and information processing systems within a medical institution, such as an HIS and a Radiology Information System (RIS). Note that the image processing device 1 shown in FIG. 1 may include, as an input / output interface 17, an input interface for inputting various user data and a display (output interface) for displaying medical images generated by an image generation function 157, as shown in FIG. 2.

[0029] The communication interface 11 performs data communication with, for example, various modalities that image the subject P during an examination of the subject P, an HIS, a PACS, etc. Any standard may be used for communication between the communication interface 11 and various modalities and hospital information systems, such as HL7 (Hearth Level 7), DICOM, or both. When the image processing device 1 functions separately from the modalities, the communication interface 11 receives, from the modalities, a magnetic resonance image generated based on forward MR data (hereinafter referred to as a forward MR image) and a magnetic resonance image generated based on retrograde MR data (hereinafter referred to as a retrograde MR image). The forward MR image and the retrograde MR image correspond to two magnetic resonance images (MR images) corresponding to two opposite phase encoding directions.

[0030] The memory 13 is realized by a storage circuit that stores various types of information. For example, the memory 13 is a storage device such as an HDD, an SSD, or an integrated circuit storage device. The memory 13 corresponds to a storage unit. Note that, in addition to an HDD or an SSD, the memory 13 may also be a drive that reads and writes various types of information from a semiconductor memory element such as a random access memory (RAM) or a flash memory, an optical disc such as a compact disc (CD) or a digital versatile disc (DVD), a portable storage medium, or a semiconductor memory element such as a RAM.

[0031] The memory 13 stores an acquisition function 151, a preprocessing function 153, a shift map generation function 155, and an image generation function 157, which are realized by the processing circuitry 15, in the form of a program executable by a computer. The memory 13 stores various data received by the acquisition function 151 via the communication interface 11. Specifically, the memory 13 stores, for example, two MR images (a forward MR image and a retrograde MR image) corresponding to two opposite phase encoding directions acquired by the acquisition function 151.

[0032] The memory 13 also stores a provisional shift map (hereinafter referred to as a provisional shift map) provisionally generated by the shift map generation function 155 in the optimization process of the cost function. The optimization process of the cost function is, for example, to determine a shift map so as to minimize a cost value calculated by the cost function. The shift map is a map used to reduce image distortion along the phase encoding direction for forward MR images and retrograde MR images. Specifically, the shift map is a map indicating the shift amount of each of a plurality of pixels in the forward MR images and retrograde MR images along the phase encoding direction.

[0033] Note that a fitting model of the shift map may be used to represent the shift map. The fitting model corresponds to a model of the shift map using a fitting function. The fitting function is, for example, Cubic Spline Fitting, but is not limited to this and any known function may be used. In this case, the shift map is reproduced by, for example, convolving a predetermined kernel with the coefficient distribution of the fitting model. For these reasons, the amount of data of the fitting model is smaller than the amount of data of the shift map.

[0034] FIG. 3 is a diagram showing an example of the relationship between the coefficient distribution of a fitting model and a shift map. As shown in FIG. 3, the shift map and the coefficient distribution of the fitting model have a one-to-one correspondence via convolution using a kernel. When a shift map and a fitting model are used, the distribution of multiple coefficients in the fitting model corresponds to variables in an optimization problem related to a cost function, which will be described later. Also, as shown in FIG. 3, the coefficient distribution of the fitting model has a reduced amount of data compared to the shift map. When the fitting model is used to represent the shift map, the calculation cost and calculation time required to generate the fitting model are reduced compared to the shift map.

[0035] The memory 13 also stores conditions (hereinafter referred to as optimization judgment conditions) used to judge the end of the optimization process of the cost function performed by the shift map generation function 155. The optimization judgment conditions are, for example, the number of iterative calculations related to the optimization process (hereinafter referred to as the number of optimizations), a value (hereinafter referred to as the cost threshold) to be compared with the value calculated by the cost function, etc. Note that the optimization judgment conditions are not limited to those described above, and judgment conditions related to known optimization problems may also be used.

[0036] The memory 13 also stores a determination condition for repeating the optimization process (hereinafter referred to as a repetition determination condition). The repetition determination condition is, for example, a smoothing level or resolution for the forward MR image and the reverse MR image, the number of times the optimization process is repeated, a rate of change of the shift map accompanying the repetition of the optimization process, etc. The repetition determination condition is not limited to the above, and any known determination condition may be used.

[0037] The memory 13 also stores MR images generated by the image generation function 157. The MR images generated by the image generation function 157 include, for example, corrected images obtained by correcting distortions in the forward MR images and the retrograde MR images based on the forward MR images, the retrograde MR images, and the shift map, forward MR images, and retrograde MR images.

[0038] The processing circuit 15 controls the entire image processing device 1. The processing circuit 15 is realized by the above-mentioned processor or the like. The processing circuit 15 includes an acquisition function 151, a pre-processing function 153, a shift map generation function 155, an image generation function 157, and the like. The processing circuit 15, which realizes the pre-processing function 153, the shift map generation function 155, and the image generation function 157, respectively, corresponds to an acquisition unit, a pre-processing unit, a shift map generation unit, and an image generation unit. Each function, such as the pre-processing function 153, the shift map generation function 155, and the image generation function 157, is stored in the memory 13 in the form of a computer-executable program. The processing circuit 15 is a processor. For example, the processing circuit 15 realizes a function corresponding to each program by reading and executing the program from the memory 13. In other words, the processing circuit 15, after reading each program, has each function, such as the pre-processing function 153, the shift map generation function 155, and the image generation function 157.

[0039] The processing circuitry 15 acquires two MR images corresponding to two opposite phase encoding directions using the acquisition function 151. Specifically, the acquisition function 151 generates a forward MR image based on the forward MR data and generates a retrograde MR image based on the retrograde MR data, thereby acquiring a forward MR image and a retrograde MR image corresponding to the two MR images. Note that the generation of the forward MR image and the retrograde MR image may be realized by the image generation function 157. In this case, the image generation function 157 generates the forward MR image by performing a Fourier transform on the forward MR data and generates the retrograde MR image by performing a Fourier transform on the retrograde MR data.

[0040] 1, when the processing circuitry 15 is installed in a standalone image processing device 1, the acquisition function 151 acquires antegrade MR images and retrograde MR images from a PACS, modality, or the like via the network and communication interface 11. The acquisition function 151 stores the acquired antegrade MR images and retrograde MR images in the memory 13.

[0041] The processing circuitry 15 generates edge images of the two MR images by the preprocessing function 153 through repeated calculations of the cost function optimization process.

[0042] The preprocessing function 153 may also generate an edge image by performing intensity correction on the two MR images using a temporary shift map. Specifically, the preprocessing function 153 calculates the Jacobian of the temporary shift map. The map of the Jacobian of the temporary shift map (hereinafter referred to as the shift Jacobian map) corresponds to a map indicating the degree to which the intensity values of a region that has become high due to compression of intensity values associated with image distortion along the phase encoding direction are reduced in accordance with the shift of pixels due to the shift amount. The preprocessing function 153 multiplies the forward MR image and the retrograde MR image by the shift Jacobian map, respectively. As a result, the preprocessing function 153 generates an image obtained by performing intensity correction on the forward MR image before position correction (hereinafter referred to as the intensity-corrected forward image) and an image obtained by performing intensity correction on the retrograde MR image before position correction (hereinafter referred to as the intensity-corrected retrograde image).

[0043] The pre-processing function 153 generates two edge images by performing edge extraction processing on the brightness-corrected forward image before position correction and the brightness-corrected backward image before position correction. The two edge images include a forward edge image corresponding to the forward MR image and a backward edge image corresponding to the backward MR image. The two edge images generated by the above processing correspond to images from which anatomical structures in the forward MR image and the backward MR image can be suitably extracted as edges because high brightness caused by image distortion along the phase encoding direction has been reduced.

[0044] The processing circuitry 15 sets parameters related to the correction process described below using the shift map generation function 155. The parameters related to the correction process include parameters of a fitting model for the shift map and the above-mentioned judgment conditions. The shift map generation function 155 generates a shift map related to the shifts of multiple pixels in the two magnetic resonance images by optimizing a cost function using a first difference between the two MR images and a second difference between two edge images generated based on the two MR images. The cost function has a regularization parameter by which the second difference is multiplied. When performing an iterative calculation related to the optimization of the cost function, the shift map generation function 155 increases the value of the regularization parameter according to the number of iterations of the iterative calculation. As a result, the shift map generation function 155 can strengthen the constraint due to the squared error between the two edge images as the cost function is reduced, thereby generating a shift map that maintains the anatomical structure in the MR image.

[0045] The processing circuitry 15 generates, through the image generation function 157, corrected images in which distortions in the two magnetic resonance images have been corrected based on the two magnetic resonance images and the shift map generated by the shift map generation function 155. Specifically, the image generation function 157 performs position correction on the two magnetic resonance images using the shift map to generate a position-corrected forward MR image and a position-corrected backward MR image. Next, the image generation function 157 generates a Jacobian map based on the shift map. Subsequently, the image generation function 157 multiplies the position-corrected forward MR image and the position-corrected backward MR image by the Jacobian map to generate a position- and intensity-corrected forward MR image and a backward MR image. Finally, the image generation function 157 generates a corrected image by calculating the average of the position- and intensity-corrected forward MR image and the backward MR image. The corrected image is an MR image in which image distortion between the forward MR image and the backward MR image has been corrected. In other words, the correction of image distortion corresponds to a combination of pixel position correction using a shift map and luminance correction using a Jacobian map based on the shift map.

[0046] The correction process executed by the MRI apparatus 100 of this embodiment configured as described above will be described with reference to FIGS. 4 to 7. The correction process is a process for reducing image distortion occurring along the phase encoding direction while maintaining the anatomical structure in the image to be corrected. The correction process corrects two magnetic resonance images to generate a corrected image. For the sake of concreteness, the following description will be given assuming that the two MR images used in the correction process are a forward MR image and a backward MR image obtained by EPI.

[0047] Fig. 4 is a flowchart showing an example of the procedure of the correction process. Prior to the execution of the correction process shown in Fig. 4, the imaging control circuitry 121 executes two EPIs on the subject P corresponding to two opposite phase encoding directions. By executing the two EPIs, the imaging control circuitry 121 acquires forward MR data and retrograde MR data. Next, the processing circuitry 15 generates forward MR images and retrograde MR images using the image generation function 157. The memory 13 stores the generated forward MR images and retrograde MR images.

[0048] (correction processing) (Step S401) The processing circuitry 15 acquires antegrade MR images and retrograde MR images from the memory 13 using the acquisition function 151. When the processing circuitry 15 is installed in a standalone image processing device 1, the acquisition function 151 acquires the antegrade MR images and retrograde MR images from a PACS, modality, or the like via the network and the communication interface 11.

[0049] (Step S402) The processing circuit 131 sets parameters related to the correction process, such as parameters of the fitting model of the iteration count shift map, parameters (coefficients) of the fitting model of the shift map, and the iteration count in the optimization problem, using the shift map generation function 155. The parameters may be set, for example, by a user's instruction via the input / output interface 17 before the correction process is executed. Parameters related to the correction, such as optimization judgment conditions and iteration judgment conditions, are stored in the memory 13. The optimization judgment conditions are used in step S410, which will be described later, and the iteration judgment conditions are used in step S411, which will be described later.

[0050] (Step S403) If the number of repetitions for generating the shift map is less than 1 (NO in step S403), the process of step S404 is executed. If the number of repetitions for generating the shift map is 1 or more (YES in step S403), the process of step S405 is executed. The determination in this step is executed by, for example, the shift map generating function 155.

[0051] (Step S404) The processing circuitry 15 generates a forward edge image and a backward edge image based on the forward MR image and the backward MR image using the preprocessing function 153. As the edge extraction process for generating the edge image, known methods such as the Canny method, thinning, and binarization (cutting off pixel values at a low threshold and clipping pixel values at a high threshold) can be used as appropriate, and therefore a description thereof will be omitted.

[0052] (Step S405) The processing circuit 15 generates a Jacobian map based on the temporary shift map by the preprocessing function 153. Specifically, the preprocessing function 153 calculates a Jacobian for each pixel using the shift amount for each of the multiple pixels in the temporary shift map (partial differentiation of a variable (position coordinate) with respect to the shift amount), and generates the Jacobian map.

[0053] (Step S406) The processing circuitry 15 generates an intensity-corrected forward image and an intensity-corrected reverse image based on the forward MR image, the reverse MR image, and the Jacobian map using the preprocessing function 153. Specifically, the preprocessing function 153 multiplies the forward MR image by the Jacobian map to generate an intensity-corrected forward image. In addition, the preprocessing function 153 multiplies the reverse MR image by the Jacobian map to generate an intensity-corrected reverse image.

[0054] FIG. 5 illustrates an example of an intensity-corrected forward image generated from a forward MR image. As shown in FIG. 5, each of the pixels in the Jacobian map has a value (Dv(y+d(x,y)v)) that reduces the intensity value of the high intensity generated by distortion (pixel compression) along the phase encoding direction y in accordance with the pixel shift due to the shift amount (d(x,y)). Therefore, as shown in FIG. 5, the intensity-corrected forward image is generated by reducing the intensity value in the forward MR image using the Jacobian map in accordance with the shift amount (d(x,y)) corresponding to the pixel position. As shown in FIG. 5, the high intensity areas (white areas) in the intensity-corrected forward image due to distortion along the phase encoding direction are improved compared to the forward MR image.

[0055] (Step S407) The processing circuit 15 generates a forward edge image and a backward edge image based on the brightness-corrected forward image and the brightness-corrected backward image using the preprocessing function 153. Specifically, the preprocessing function 153 performs edge extraction processing on the brightness-corrected forward image to generate the forward edge image. The preprocessing function 153 also performs edge extraction processing on the brightness-corrected backward image to generate the backward edge image.

[0056] Because forward and reverse MR images exhibit brightness compression due to distortion, extracting edges from the forward and reverse MR images results in extracted edges that do not actually exist as structures. On the other hand, as shown in step S407, the effects of brightness compression are reduced in the brightness-corrected forward and reverse images before position correction, resulting in edge images that are more in line with the anatomical structure. In other words, in step S407, edges are extracted from the brightness-corrected forward and reverse images before position correction, thereby obtaining edge images that reduce the effects of brightness value changes in the grid before position correction. Preparing edge images with a grid before position correction in advance simplifies the Jacobian of the cost function in the optimization calculation performed in the optimization problem OPP.

[0057] (Step S408) The processing circuitry 131 calculates a cost value using a cost function by the shift map generation function 155. The cost function Cost is expressed by, for example, the following equation (1).

[0058]

number

[0059] As shown in Equation (1), the cost function Cost includes, for example, a first sum of squares error (SSE) based on a first difference between a forward MR image and a retrograde MR image, a second sum of squares error based on a second difference between a forward edge image and a retrograde edge image, a constraint term B indicating the degree of curvature of the calculated shift map, a first regularization parameter λ1 by which the constraint term B is multiplied, and a second regularization parameter λ2 by which the second sum of squares error is multiplied. The first term on the right side of Equation (1) is the first sum of squares error. The second term on the right side of Equation (1) is the first regularization parameter λ1 multiplied by the constraint term B. The third term on the right side of Equation (1) is the second regularization parameter λ2 multiplied by the second sum of squares error.

[0060] The vector f in Eq. (1) + is a vertical vector whose elements are pixel values in a forward MR image. - is a vertical vector having pixel values in a retrograde MR image as vector elements. An index of the scalar value B in equation (1) is the bending energy of a shift map.

[0061] Vector E in Equation (1) + Before the optimization problem OPP is executed, is a vertical vector having pixel values in the forward edge image as elements. After the optimization problem OPP is executed, the vector E +is a vertical vector having, as elements, pixel values in the forward edge image for which position correction has been performed using the temporary shift map generated by the shift map generation function 155. The position correction using the temporary shift map is performed by the shift map generation function 155 prior to the calculation of the cost function in this step.

[0062] Vector E in Equation (1) - Before the optimization problem OPP is executed, is a vertical vector having pixel values in the inverse edge image as vector elements. After the optimization problem OPP is executed, the vector E in Equation (1) - is a vertical vector having, as elements, pixel values in the backward edge image for which position correction has been performed using the temporary shift map generated by the shift map generation function 155. The position correction using the temporary shift map is performed by the shift map generation function 155 before the calculation of the cost function in this step.

[0063] In optimizing the cost function, the first square sum error acts on the cost function to reduce the first difference. Also, in optimizing the cost function, the second square sum error acts on the cost function to reduce the second difference. In optimizing the cost function, the constraint term B acts on the cost function to reduce the curvature of the shift map and smooth the change in the shift amount in the shift map. The first regularization parameter is a parameter indicating the degree of contribution of the constraint term to the cost function. The second regularization parameter is a parameter indicating the degree of contribution of the second square sum error to the cost function.

[0064] (Step S409) If the optimization problem is not completed, that is, if the optimization determination condition is not satisfied (No in step S409), the process of step S408 is repeated. The determination of completion of the optimization problem is made, for example, when the cost value converges or the number of calculations of the cost value reaches a preset number of repetitions, and is executed by the shift map generation function 155. At this time, the shift map generation function 155 changes the variables in the shift map, that is, the shift amount, based on, for example, the Jacobian of the shift map in the cost function. If the optimization problem is completed, that is, if the optimization condition is satisfied (Yes in step S409), the process of step S410 is executed. At this time, the shift map generation function 155 determines the latest shift map as the tentative shift map.

[0065] (Step S410) The processing circuit 15 determines, through the shift map generation function 155, whether or not iteration of the optimization problem OPP needs to be terminated according to the iteration determination condition. If it is determined that iteration of the optimization problem OPP should be terminated, that is, if the iteration determination condition is satisfied (YES in step S410), the processing of step S411 is executed. The shift map generation function 155 determines the tentative shift map as the shift map to be applied to the forward MR image and the retrograde MR image. If it is not determined that iteration of the optimization problem OPP should be terminated, that is, if the iteration determination condition is not satisfied (NO in step S410), the processing of step S402 and subsequent steps is repeated. At this time, the shift map generation function 155 may increase the second regularization parameter λ2 in step S402. At this time, in the next processing of step S408, the contribution of the second sum of squares error to the cost value becomes larger.

[0066] (Step S411) The processing circuitry 15 generates a corrected image based on the forward MR image, the retrograde MR image, and the shift map using the image generation function 157. Specifically, the image generation function 157 applies the shift map to the forward MR image to shift multiple pixel values in the forward MR image. As a result, the image generation function 157 generates a position-corrected forward MR image. Furthermore, the image generation function 157 applies the shift map to the retrograde MR image to shift multiple pixel values in the retrograde MR image.

[0067] Furthermore, the image generation function 157 generates a Jacobian map based on the shift map. The image generation function 157 multiplies the position-corrected forward MR image and the position-corrected reverse MR image by the Jacobian map to generate a forward MR image and a reverse MR image whose positions and intensities have been corrected. The forward MR image and the reverse MR image whose positions and intensities have been corrected correspond to the forward MR image and the reverse MR image whose distortion has been corrected. The image generation function 157 generates a corrected image by calculating the average of the forward MR image and the reverse MR image whose positions and intensities have been corrected. This completes the correction process.

[0068] The image processing device 1 and MRI device 100 according to the above-described embodiment acquire two magnetic resonance images corresponding to two opposite phase encoding directions, generate a shift map relating to the shifts of multiple pixels in the two magnetic resonance images by optimizing a cost function using a first difference between the two magnetic resonance images and a second difference between two edge images generated based on the two magnetic resonance images, and generate a corrected image in which distortions in the two magnetic resonance images are corrected based on the two magnetic resonance images and the shift map. Thus, the image processing device 1 and MRI device 100 according to the embodiment can add the edge difference between the two images to the cost function as a new constraint term that preserves anatomical structures, and solve the optimization problem of the cost function. Therefore, the image processing device 1 and MRI device 100 according to the embodiment can obtain a corrected image (MR image) in which distortions are reduced while preserving anatomical structures.

[0069] Furthermore, in the image processing device 1 and the MRI device 100 according to the embodiment, the cost function further includes a second regularization parameter λ2 multiplied by the second difference, and when the cost function is optimized multiple times, the value of the second regularization parameter λ2 is increased according to the number of times the optimization is repeated. As a result, according to the image processing device 1 and the MRI device 100 according to the embodiment, the contribution of the difference in anatomical structure between two images (the second sum-of-squares error shown as the difference between two edge images) to the cost function can be increased according to the repetition, in line with the improvement in the accuracy of distortion estimation according to the repeated generation of the shift map. Therefore, according to the image processing device 1 and the MRI device 100 according to the embodiment, it is possible to obtain a corrected image in which distortion is further reduced while further preserving the anatomical structure.

[0070] Furthermore, the image processing device 1 and the MRI device 100 according to the embodiment generate an edge image using a provisional shift map that is provisionally generated in the repeated optimization of the cost function. As a result, the image processing device 1 and the MRI device 100 according to the embodiment use, in the cost function, an edge image that has undergone provisional distortion correction (brightness correction and position correction) using the provisional shift map, and therefore, a corrected image can be obtained without reducing the correction accuracy.

[0071] Furthermore, the image processing device 1 and the MRI device 100 according to the embodiment perform brightness correction on two magnetic resonance images using a provisional shift map to generate an edge image. As a result, the image processing device 1 and the MRI device 100 according to the embodiment can generate an edge image by reducing high-brightness areas due to pixel value compression caused by image distortion to appropriate brightness values after distortion correction. That is, in evaluating edge differences, it is possible to perform edge extraction worthy of evaluation for areas where the cost function evaluation based on edge differences is inappropriate, such as areas where brightness values have changed due to distortion. Therefore, the image processing device 1 and the MRI device 100 according to the embodiment can generate an edge image by temporarily reducing high brightness caused by distortion (pixel compression) along the phase encoding direction y after distortion correction using a provisional shift map, and use the edge image in the cost function. This allows a corrected image with further reduced distortion while further preserving anatomical structures to be obtained.

[0072] From the above, according to the image processing device 1 and the MRI device 100 of the embodiment, even when two MR images to be corrected having different phase encoding directions have strong image distortion or when unexpected differences occur between the two images due to imperfections in the collected images (for example, pulsation, body movement, etc.) (for example, large differences due to pixel compression caused by distortion, or changes in magnetic resonance signals of cerebrospinal fluid (CSF), blood, etc. due to pulsation), it is possible to generate a shift map that can maintain the anatomical structure.

[0073] In addition, according to the image processing device 1 and the MRI device 100 of the embodiment, the only variable component in the optimization problem (shift map) is the one due to the predetermined shift of the edge image, so the Jacobian of this cost function can be expressed by a simple formula. As a result, the convergence of the optimization problem is faster than when cost calculation is performed on edge images extracted from images corrected for distortion in the optimization problem. Furthermore, according to the image processing device 1 and the MRI device 100 of the embodiment, preprocessing related to the generation of edge images can improve the accuracy of edge extraction in distorted regions, as shown in FIG. 7.

[0074] For these reasons, the image processing device 1 and the MRI device 100 according to the embodiment can generate a corrected image in which image distortion (artifacts) occurring along the phase encoding direction is reduced while maintaining the anatomical structure. Therefore, the image processing device 1 and the MRI device 100 according to the present embodiment can improve, for example, the efficiency of image interpretation by radiologists, and can improve the throughput of diagnoses for the subject P.

[0075] (First application example) In this application example, the second difference is multiplied by a weight that is reduced according to the magnitude of the shift amount in the temporary shift map. That is, the cost function in this application example has a weight that is multiplied according to the shift amount for each of multiple pixels in the second difference. The processing circuitry 15 determines a weight for each pixel in the second difference according to the magnitude of the shift amount in the temporary shift map using the shift map generation function 155. That is, the shift map generation function 155 multiplies the second difference in equation (1) by a weight that corresponds to the magnitude of the shift amount in the latest temporary shift map. Areas with large shift amounts in the temporary shift map correspond to areas estimated to have strong distortions by the optimization problem in at least one pre-correction grid of the forward MR image and the retrograde MR image.

[0076] The weights are set so that pixel values with large shift amounts have small weights, and pixel values with small shift amounts have large weights. Specifically, when performing repeated calculations related to the optimization of the cost function, the shift map generation function 155 sets the weights for areas in which the shift amount exceeds a predetermined threshold in the provisional shift map provisionally generated in the repeated calculations to be smaller than the weights for areas in which the shift amount is equal to or smaller than the predetermined threshold. Note that the weight values for the shift amounts may be set in advance using a correspondence table.

[0077] In the correction process of this application example, from step S405 to step S407, the shift map generation function 155 determines a weight to be multiplied for each of a plurality of pixels in the second difference based on the temporary shift map and a predetermined threshold. Also, in this application example, in step S408, the third term in equation (1) is calculated using the determined weight. Other processes in the correction process are the same as those in the embodiment, so descriptions are omitted.

[0078] The cost function in the image processing device 1 and the MRI device 100 according to the first application example of the embodiment described above has a weight that is multiplied for each pixel in the second difference according to the amount of shift. For example, when optimizing the cost function multiple times, the image processing device 1 and the MRI device 100 according to the first application example set a weight for a region in which the shift amount exceeds a predetermined threshold in a provisional shift map provisionally generated for each optimization of the cost function to be smaller than the weight for a region in which the shift amount is equal to or smaller than the predetermined threshold. As a result, the image processing device 1 and the MRI device 100 according to the first application example can set a smaller weight for a region in which the shift amount is large, thereby further improving the effect of Example 1. Other effects are similar to those of the embodiment, and therefore will not be described.

[0079] (Second application example) In this application example, for pixel values in the first difference that are greater than a threshold value among the pixel values, the cost function places more importance on the second squared error than on the first squared error. For example, in this application example, when the difference between the sums of the pixel values along the phase encoding direction in two magnetic resonance images exceeds a predetermined threshold value, the weight multiplied by the second difference for a region where the difference between the sums exceeds the threshold value is set to be greater than the weight for a region along the phase encoding direction where the difference between the sums is equal to or less than the threshold value.

[0080] Specifically, the processing circuitry 15 calculates, for each line in the phase encoding direction (hereinafter referred to as a phase encoding line), a sum of multiple pixel values in a forward MR image (hereinafter referred to as a forward sum) and a sum of multiple pixel values in a retrograde MR image (hereinafter referred to as a retrograde sum) using the shift map generation function 155. Next, the shift map generation function 155 calculates a relative error between the forward sum and the retrograde sum for each phase encoding line. The relative error is, for example, a value obtained by dividing the absolute value of the difference between the forward sum and the retrograde sum by the forward sum.

[0081] The shift map generation function 155 compares the relative error with a threshold and identifies a phase encoding line for which the relative error exceeds the threshold. The shift map generation function 155 sets a weight for the identified phase encoding line (hereinafter referred to as the "specific line") that is heavier than weights for other encoding lines different from the specific line. That is, when the difference between the sums of multiple pixel values along the phase encoding direction in two magnetic resonance images exceeds a predetermined threshold, the shift map generation function 155 sets a weight for a region that exceeds the threshold that is greater than the weight for a region along the phase encoding direction for which the difference between the sums is equal to or less than the threshold. The shift map generation function 155 multiplies the second difference by the set weight.

[0082] 6 is a diagram showing an example of the distribution of relative errors for phase encoding lines and thresholds. As shown in FIG. 6, the shift map generation function 155 assigns a weight to multiple phase encoding lines HW that exceed the relative error, which is greater than the weight for other phase encoding lines. As a result, in the cost function, the pixel value of the second difference in pixels related to a specific line contributes significantly to the cost value. In other words, in the cost function, the cost related to a specific line is emphasized among the costs due to the difference in the edge image.

[0083] In addition, when EPIs are imaged for multiple slices, the shift map generation function 155 may calculate the relative error for each of the multiple slices and perform weight setting in this application example for slices related to relative errors that exceed a threshold.

[0084] 7 is a diagram showing an example of the distribution of relative errors with respect to slice numbers and threshold values. As shown in FIG. 7, the relative errors for slice numbers 1, 11, and 29 exceed the threshold values. Therefore, the shift map generation function 155 executes the process of setting the weights for slice numbers 1, 11, and 29.

[0085] Note that the setting of the weight to be multiplied for each pixel in the second difference is not limited to the above description. For example, for a region (x-α≦x≦x+α, y-α≦y≦y+α, hereinafter referred to as a set region) where α is a predetermined value between two images, a forward MR image and a backward MR image, positionally corrected by the temporary shift map, the shift map generating function 155 sets a weight corresponding to a region where the sum of squares of differences is larger than a threshold A and the difference between the sums is larger than a threshold C to be larger than the weights of other regions.

[0086] That is, in the set region between the forward MR image and the reverse MR image after the position correction, when the distortion defined by the threshold A is large and the difference in pixel value defined by the threshold C is large, the shift map generating function 155 sets a weight larger than that of other regions. As a result, in the cost function, the contribution of the second difference becomes larger in the region where the distortion is large and the difference in MR signal is large.

[0087] In the above description, the difference between the sum of squares of the differences and the sum in the set region is described as a condition for setting the weight, but this is not limiting. That is, the shift map generation function 155 may set a weight corresponding to a region in the set region where the sum of squares of the differences is larger than the threshold A or the difference in the sum is larger than the threshold C to be larger than the weight of other regions.

[0088] In the correction process in this application example, from step S405 to step S409, the shift map generation function 155 determines a weight to be multiplied for each of a plurality of regions in the second difference based on the forward MR image and the backward MR image whose positions have been corrected by the temporary shift map and the threshold A and / or the threshold C. Also, in this application example, in step S409, the third term in equation (1) is calculated using the determined weight. Other processes in the correction process are the same as those in the embodiment, and therefore description thereof will be omitted.

[0089] In addition, in the above explanation, an example was shown in which the difference between the sum of squares of the differences and the sum was calculated using a forward MR image and a backward MR image whose positions have been corrected using a temporary shift map. However, when the number of iterations is 0, i.e., when the first-stage optimization problem is solved, there is no temporary shift map, and therefore the difference between the sum of squares of the differences and the sum may be calculated using a forward MR image and a backward MR image acquired by the acquisition function 151.

[0090] At this time, before or after step S404, the shift map generation function 155 determines a weight to be multiplied for each of a plurality of regions in the second difference based on the forward MR image and the reverse MR image and the threshold value A and / or the threshold value C. From these facts, when performing an iterative calculation related to optimization of the cost function, the shift map generation function 155 sets a weight based on at least one of the difference between the above sums based on two magnetic resonance images in which position correction has been performed on the two magnetic resonance images using a provisional shift map provisionally generated in the iterative calculation, and the difference between the above sums in predetermined regions in the two magnetic resonance images.

[0091] In the above description, the weight to be multiplied by the difference (second difference) between the forward edge image and the backward edge image is adjusted (for example, a region where the difference between the forward edge image and the backward edge image is large is weighted larger than other regions), but the present invention is not limited to this. For example, the shift map generating function 155 may set a smaller weight for pixel values corresponding to a region where the weight is to be increased by the above determination in the difference (first difference) between the forward MR image and the backward MR image than for other regions. That is, the shift map generating function 155 sets a weight in the cost function so that the second difference is given more importance than the first difference for a region where the difference between the forward edge image and the backward edge image is large.

[0092] The image processing device 1 and the MRI device 100 according to the second application example of the above-described embodiment set a weight for a region where the difference between the sums of a plurality of pixel values along the phase encoding direction in two magnetic resonance images exceeds a predetermined threshold value, so as to be greater than a weight for a region along the phase encoding direction where the difference between the sums is equal to or less than the threshold value. For example, when performing an iterative calculation for optimizing a cost function, the image processing device 1 and the MRI device 100 according to the second application example set a weight based on at least one of the difference between the sums based on two magnetic resonance images obtained by performing position correction on the two magnetic resonance images using a provisional shift map provisionally generated in the iterative calculation, and the difference between the sums in predetermined regions in the two magnetic resonance images.

[0093] The image processing device 1 and the MRI device 100 according to the second application example can achieve the same effect as the brightness correction in the embodiment. For example, if there is a region in the forward MR image and the retrograde MR image where the brightness value differs depending on the imaging timing due to the flow of CSF and / or blood, adding a weight to the region makes it possible to execute the optimization problem by placing more importance on the difference 2 between the forward edge image and the retrograde edge image than on the difference 1 between the forward MR image and the retrograde MR image.

[0094] From the above, according to the image processing device 1 and the MRI device 100 of the second application example, even if there is a difference in brightness value not caused by distortion between a forward MR image and a reverse MR image, it is possible to generate a shift map with high correction accuracy, and to generate a corrected image in which image distortion (artifacts) occurring along the phase encoding direction is reduced while maintaining the anatomical structure. Other effects are the same as those of the embodiment and the first application example, and therefore description thereof will be omitted.

[0095] (Third application example) This application example is to execute various processes in steps S404 to S407 using a temporary shift map in the correction process described in the embodiment while solving the optimization problem OPP. For example, the processing circuitry 15 performs distortion correction on two magnetic resonance images (a forward MR image and a backward MR image) using a shift map generated during the optimization process OPP by the preprocessing function 153 to generate an edge image (a forward edge image E + and the retrograde edge image E - )

[0096] Fig. 8 is a flowchart showing an example of the procedure of the correction process according to this application example. Prior to the execution of the correction process shown in Fig. 8, the imaging control circuitry 121 executes two EPIs on the subject P, which correspond to two opposite phase encoding directions. By executing the two EPIs, the imaging control circuitry 121 acquires forward MR data and retrograde MR data. Next, the processing circuitry 15 generates forward MR images and retrograde MR images using the image generation function 157. The memory 13 stores the generated forward MR images and retrograde MR images.

[0097] Steps S101, S102, S106, and S108 in Fig. 8 correspond to steps S401, S402, S410, and S411 in Fig. 4, respectively, and therefore a description thereof will be omitted. Also, in step S103 in Fig. 8, in addition to the processing in step S408 in Fig. 4, the processing in steps S404 to S407 (i.e., processing of distortion correction for two magnetic resonance images (forward MR image and retrograde MR image)) is executed. Therefore, the processing content in step S103 is similar to steps S404 to S408, and therefore a description thereof will be omitted.

[0098] (correction processing) (Step S104) The processing circuit 15 determines a temporary shift map so as to reduce the cost value in the cost function using the shift map generation function 155. A known method can be applied to the calculation of the optimization problem in step S104, and therefore a description thereof will be omitted.

[0099] (Step S107) If the determination in step S106 is No, the processing circuit 15 sets the second regularization parameter λ2 in the cost function by the shift map generation function 155 to, for example, a value larger than the value of the second regularization parameter used in the previous optimization process. If the second regularization parameter λ2 is set to a fixed value in step S102, this step is unnecessary. Alternatively, the shift map generation function 155 may set the second regularization parameter λ2 according to the reduction rate of the second difference.

[0100] For example, if the second difference corresponding to the difference between the edge images does not decrease below a predetermined value from the nth to the (n+1)th optimization processing iteration by the shift map generation function 155, the processing circuit 15 increases the second regularization parameter λ2 in the next optimization processing because it is necessary to reduce the cost in the cost function due to the second difference. At this time, the second squared sum error (error of the edge image) in the cost function is given more importance than the first squared sum error (error of the two images). Also, if the second squared sum error (error of the edge image) is sufficiently reduced but the first squared sum error (error of the two images) is not sufficiently reduced from the nth to the (n+1)th optimization processing iteration, the processing circuit 15 decreases the second regularization parameter λ2 or increases the regularization parameter multiplied by the first squared sum error (error of the two images) in the next optimization processing.

[0101] The image processing device 1 and the MRI device 100 according to the third application example of the above-described embodiment perform distortion correction on two magnetic resonance images using a shift map generated during the process of optimizing the cost function to generate an edge image. At this time, the image processing device 1 and the MRI device 100 according to the third application example can adjust the regularization parameter λ2 by which the first sum of squares error is multiplied or the second sum of squares error is multiplied, depending on the degree of reduction in the cost of the first sum of squares error based on the first difference between the forward MR image and the reverse MR image and the degree of reduction in the second sum of squares error based on the second difference between the forward edge image and the reverse edge image.

[0102] As a result, according to the image processing device 1 and the MRI device 100 of the third application example, even if there is a difference in brightness between a forward MR image and a reverse MR image that is not caused by distortion (in other words, caused by CSF or blood), it is possible to generate a shift map with high correction accuracy, and to generate a corrected image in which image distortion (artifacts) occurring along the phase encoding direction is reduced while maintaining the anatomical structure. Other effects are similar to those of the embodiment and other application examples, and therefore description thereof will be omitted.

[0103] The embodiment and the first to third application examples may be realized singly as appropriate. Also, the embodiment and the first to third application examples may be realized in combination as appropriate. The embodiment and the first to third application examples can be combined in any desired manner.

[0104] When the technical idea of the embodiment is realized as an image processing method, the image processing method acquires two magnetic resonance images corresponding to two opposite phase encoding directions, optimizes a cost function using a first difference between the two magnetic resonance images and a second difference between two edge images generated based on the two magnetic resonance images, thereby generating a shift map related to shifts of multiple pixels in the two magnetic resonance images, and generates a corrected image in which distortion in the two magnetic resonance images is corrected based on the two magnetic resonance images and the shift map. Furthermore, when the image processing method optimizes the cost function multiple times, the image processing method may perform brightness correction on the two magnetic resonance images using a provisional shift map provisionally generated in each iteration of optimization to generate an edge image. Furthermore, the cost function in the image processing method may further include a weight that is multiplied according to the amount of shift for each pixel in the second difference. When the image processing method optimizes the cost function multiple times, the weight for a region in the provisional shift map provisionally generated for each optimization where the shift amount exceeds a predetermined threshold may be set smaller than the weight for a region where the shift amount is equal to or less than the predetermined threshold. The procedure and effects of the correction process according to this image processing method are the same as those described in the embodiment and the first to third application examples, and therefore a description thereof will be omitted.

[0105] When the technical idea of the embodiment is realized by an image processing program, the image processing program causes a computer to acquire two magnetic resonance images corresponding to two opposite phase encoding directions, optimize a cost function using a first difference between the two magnetic resonance images and a second difference between two edge images generated based on the two magnetic resonance images, thereby generating a shift map related to shifts of multiple pixels in the two magnetic resonance images, and generate a corrected image in which distortion in the two magnetic resonance images is corrected based on the two magnetic resonance images and the shift map. Furthermore, the image processing program may cause the computer to, when optimizing the cost function multiple times, perform brightness correction on the two magnetic resonance images using a provisional shift map provisionally generated in each iteration of optimization to generate an edge image. Furthermore, the cost function in the image processing program may further include a weight that is multiplied according to the amount of shift for each pixel in the second difference. The image processing program may, when optimizing the cost function multiple times, set a weight for a region in the provisional shift map where the shift amount exceeds a predetermined threshold to be smaller than a weight for a region where the shift amount is equal to or smaller than the predetermined threshold.

[0106] For example, the correction process can be realized by installing an image processing program in a computer in a modality such as the MRI apparatus 100, a PACS server, or various image processing servers, and expanding the program in memory. In this case, the program that can cause the computer to execute the method can also be stored and distributed on a storage medium such as a magnetic disk (such as a hard disk), an optical disk (such as a CD-ROM or DVD), or a semiconductor memory. The procedure and effects of the correction process using the image processing program are the same as those described in the embodiment and the first to third application examples, so a description thereof will be omitted.

[0107] According to at least one of the embodiments described above, it is possible to reduce artifacts that occur due to abrupt phase changes.

[0108] 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]

[0109] 1. Image processing device 11 Communication Interface 13. Memory 15 Processing circuit 17 Input / Output Interface 100 Magnetic resonance imaging device 101 Static Magnetic Field Magnet 103 Gradient magnetic field coil 105 Gradient magnetic field power supply 107 Sleeper 109 Bed control circuit 111 Bore 113 Transmitting Circuit 115 Transmitting Coil 117 Receiving Coil 119 Receiving circuit 121 Imaging control circuit 123 System Control Circuit 125 Storage device 151 Acquisition Function 153 Pre-processing function 155 Shift map generation function 157 Image generation function

Claims

1. an acquisition unit for acquiring two magnetic resonance images corresponding to two mutually opposite phase encoding directions; a shift map generating unit that generates a shift map relating to shifts of a plurality of pixels in the two magnetic resonance images by optimizing a cost function using a first difference between the two magnetic resonance images and a second difference between two edge images generated based on the two magnetic resonance images; an image generating unit that generates a corrected image in which distortion in the two magnetic resonance images is corrected based on the two magnetic resonance images and the shift map; An image processing device comprising:

2. the cost function further comprises a regularization parameter multiplied by the second difference; When the optimization of the cost function is performed a plurality of times, the shift map generation unit increases the value of the regularization parameter according to the number of times the optimization is repeated. The image processing device according to claim 1 .

3. the cost function has a weight that is multiplied for each pixel in the second difference according to the amount of the shift; 3. The image processing device according to claim 1.

4. a pre-processing unit that generates the edge image using a provisional shift map that is provisionally generated during the iteration of the optimization, or that performs distortion correction on the two magnetic resonance images using a shift map that is generated during the optimization process, to generate the edge image; The image processing device according to claim 2 .

5. the preprocessing unit performs brightness correction on the two magnetic resonance images using the temporary shift map to generate the edge image. The image processing device according to claim 4 .

6. When optimizing the cost function a plurality of times, the shift map generating unit sets the weight for a region in which a shift amount in a provisional shift map provisionally generated for each optimization exceeds a predetermined threshold to be smaller than the weight for a region in which the shift amount is equal to or smaller than the predetermined threshold. The image processing device according to claim 3 .

7. When a difference between sums of a plurality of pixel values along a phase encoding direction in the two magnetic resonance images exceeds a predetermined threshold, the shift map generator sets the weight for a region where the difference exceeds the threshold to be larger than a weight for a region along the phase encoding direction where the difference between the sums is equal to or smaller than the threshold. The image processing device according to claim 3 .

8. The shift map generation unit When performing an iterative calculation for optimizing the cost function, the difference between the sums based on two magnetic resonance images in which position correction for the two magnetic resonance images has been performed using a tentative shift map temporarily generated in the iterative calculation; and a difference between the sums in a predetermined region in the two magnetic resonance images; and setting the weight based on at least one of the following: The image processing device according to claim 7 .

9. acquiring two magnetic resonance images corresponding to two opposite phase encoding directions; generating a shift map relating to shifts of a plurality of pixels in the two magnetic resonance images by optimizing a cost function using a first difference between the two magnetic resonance images and a second difference between two edge images generated based on the two magnetic resonance images; generating a corrected image in which distortion in the two magnetic resonance images is corrected based on the two magnetic resonance images and the shift map; An image processing method comprising:

10. When the optimization of the cost function is performed a plurality of times, the edge image is generated by performing brightness correction on the two magnetic resonance images using a temporary shift map temporarily generated in the repetition of the optimization. The image processing method according to claim 9.

11. the cost function further comprises a weight for each pixel in the second difference that is multiplied according to the amount of the shift; When the optimization of the cost function is performed a plurality of times, the weight for a region in which a shift amount in a provisional shift map provisionally generated for each optimization exceeds a predetermined threshold is set smaller than the weight for a region in which the shift amount is equal to or smaller than the predetermined threshold.

11. The image processing method according to claim 9 or 10.

12. On the computer, acquiring two magnetic resonance images corresponding to two opposite phase encoding directions; generating a shift map relating to shifts of a plurality of pixels in the two magnetic resonance images by optimizing a cost function using a first difference between the two magnetic resonance images and a second difference between two edge images generated based on the two magnetic resonance images; generating a corrected image in which distortion in the two magnetic resonance images is corrected based on the two magnetic resonance images and the shift map; An image processing program that achieves this.

13. The computer, When the optimization of the cost function is performed a plurality of times, the edge image is generated by performing brightness correction on the two magnetic resonance images using a temporary shift map temporarily generated in the repetition of the optimization. The image processing program according to claim 12, further comprising:

14. the cost function further comprises a weight for each pixel in the second difference that is multiplied according to the amount of the shift; The computer, When the optimization of the cost function is performed a plurality of times, the weight for a region in which a shift amount in a provisional shift map provisionally generated for each optimization exceeds a predetermined threshold is set smaller than the weight for a region in which the shift amount is equal to or smaller than the predetermined threshold.

14. The image processing program according to claim 12, further comprising:

Images