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

Abstract

This invention provides a magnetic resonance imaging apparatus, an image processing apparatus, and an image processing method. A simple technique is used to perform appropriate phase correction on a complex image to be subjected to complex addition operations, suppressing artifacts in the signal region and reducing noise in the background region. Two or more smoothed phase images with different smoothing degrees are obtained by applying two or more smoothing processes to the phase image of the complex image. Weights for these smoothed phase images are calculated based on the signal-to-noise ratio (SNR) of the intensity image, and addition operations are performed with each signal value weight to create a smoothed phase image for correction. After correcting the phase of the complex image using this smoothed phase image, complex addition operations are performed on the phase-corrected complex image. Thus, phase correction equivalent to a scheme that reduces smoothing in the signal region and enhances smoothing in the background region is achieved.

Description

Technical Field

[0001] This invention relates to a magnetic resonance imaging apparatus and an image processing apparatus and method for processing images obtained from the magnetic resonance imaging apparatus, and particularly to a technique for correcting the phase of each complex image during the addition of complex images. Background Technology

[0002] In magnetic resonance imaging (MRI), to improve the signal-to-noise ratio (SNR), multiple images are often taken and the resulting images are added together. This is particularly true in diffusion-intensity imaging (DWI), which emphasizes the diffusion of water. There is a problem where the signal is reduced due to changes in the phase of the active magnetization caused by the application of a strong, tilted magnetic field pulse called a motion probing gradient (MPG) pulse. Therefore, techniques to improve SNR through addition are important.

[0003] Images obtained in MRI are typically complex images containing phase information. Among the techniques for adding these complex images to obtain a single image, there are complex addition, which preserves the complexity of the images, and absolute addition, which adds the images based on their absolute values. Complex addition is suitable when the images are in phase, while absolute addition is suitable when they are out of phase. In DWI, due to physiological movements such as blood flow and cerebrospinal fluid flow, the phases are not consistent; therefore, absolute addition is often used.

[0004] Generally, noise in an image exhibits a Gaussian distribution. However, when set as an absolute value image, in regions with low SNR, such as the background area, it approximates a Rayleigh distribution, becoming non-Gaussian. Therefore, performing addition operations on multiple absolute value images cannot adequately reduce noise.

[0005] On the other hand, while complex addition does not present the problem caused by the non-Gaussian nature of noise, it still requires phase correction to achieve a certain degree of consistency across multiple images. Phase correction techniques for this purpose have been proposed. For example, Non-Patent Document 1 describes performing complex addition on multiple images after correcting global phase variations using low-pass filters, etc. Non-Patent Document 1 also reports on how the results of complex addition vary depending on the filter size (smoothness), noting that there is a trade-off between artifact generation and noise reduction effects depending on the filter size.

[0006] Existing technical documents

[0007] Non-patent literature

[0008] Non-patent document 1: Diffusion Tensor Imaging (DTI) With Retrospective MotionCorrection for Large-Scale Pediatric Imaging, Samantha J. Holdsworth, JOURNAL OFMAGNETIC RESONANCE IMAGING 36: 961-971 (2012)

[0009] As described above, if a smoothing filter with high noise reduction effect (i.e., high smoothness) is used, artifacts are easily generated in the region where there is a signal from the subject (called the signal region). Conversely, if a filter that suppresses the generation of such artifacts is used, the noise reduction effect will be reduced. There is such a trade-off problem, so it is impossible to solve both issues. Summary of the Invention

[0010] The objective of this invention is to provide a method that allows for appropriate phase correction using a simple technique, suppressing artifact generation in the signal region and reducing noise in the background region. Furthermore, in this specification, the signal region refers to a region with high SNR (e.g., SNR of 5 or higher), and the background region refers to a region with low SNR (e.g., SNR of 1 or lower).

[0011] The present invention, which addresses the aforementioned issues, performs two or more smoothing processes with different degrees of smoothing on a complex image (or an image containing its phase information). The resulting smoothed phase images are then added using weights based on signal values ​​from the intensity image, etc., to create a phase image for correction. Thus, for example, phase correction equivalent to a scheme that reduces smoothness in the signal region and increases smoothness in the background region is achieved, resolving the issue of one filter gaining at the expense of the other.

[0012] That is, the MRI apparatus of the present invention includes: an imaging unit for measuring magnetic resonance signals generated from a subject; and an image processing unit for generating an image of the subject using the magnetic resonance signals acquired by the imaging unit. The image processing unit includes: a phase correction unit for performing phase correction on each of a plurality of complex images obtained through multiple imaging operations; and a complex addition unit for performing addition operations on the phase-corrected plurality of complex images. The image processing unit further includes: a smoothing unit for performing two or more smoothing processes with different smoothing degrees on the complex images or images having their phase information; and a weighted addition unit for performing weighted addition operations on the results of smoothing with different smoothing degrees in the smoothing unit. The phase correction unit uses the phase of the complex image obtained after the weighted addition operation by the weighted addition unit to perform phase correction on the complex image.

[0013] Furthermore, the image processing method of the present invention is an image processing method for processing multiple complex images obtained by a magnetic resonance imaging device through multiple imaging operations, wherein the processing of each of the multiple complex images includes the following steps.

[0014] The smoothing step involves performing two or more smoothing processes with different degrees of smoothing; calculating weights based on the signal values ​​of the intensity images of the complex images; performing a weighted addition operation on the images smoothed according to each of the smoothing processes using the weights; using the images obtained by performing the weighted addition operation to correct the phase of the complex images; and performing a complex addition operation on the multiple complex images with corrected phases.

[0015] The multiple complex images are, for example, images obtained through diffusion-intensity imaging.

[0016] Invention Effects

[0017] According to the present invention, by applying filters with different smoothness to complex images and performing weighted addition operations on the results with weights having a given distribution, the same result can be obtained as the scheme of processing with filters with different smoothness according to the region of the image.

[0018] For example, by performing weighted addition using weights calculated from the signal values ​​corresponding to the intensity image, phase images that have been appropriately smoothed for regions with different signal-to-noise ratios (e.g., signal regions and background regions) can be obtained. By performing complex addition on a complex phase image obtained by correcting the smoothed phase image, artifact generation can be suppressed in the image after addition, while noise reduction can be achieved. Attached Figure Description

[0019] Figure 1 This is a diagram showing an overall outline of the MRI device using the present invention.

[0020] Figure 2 This is a functional block diagram of the computer of the MRI device according to the implementation method.

[0021] Figure 3 This is a diagram illustrating an example of a DWI sequence performed by the camera unit.

[0022] Figure 4 This is a flowchart illustrating the processing of the image processing unit in Embodiment 1.

[0023] Figure 5 This is an explanatory diagram showing the processing of the phase correction unit in the implementation method.

[0024] Figure 6 This is a diagram representing an example of weights.

[0025] Figure 7 This is an explanatory diagram showing the processing of the smoothing section and the weighted addition operation section in the implementation method.

[0026] Figure 8 This is a diagram illustrating the effect of the image processing unit in the embodiment.

[0027] Figure 9 This is a diagram showing an example of a GUI displayed on a display device.

[0028] Figure 10 This is a structural diagram illustrating an embodiment of the image processing apparatus of the present invention.

[0029] Explanation of reference numerals in the attached figures

[0030] 1: MRI device; 2: Image processing device; 5: Subject; 10: Imaging unit; 20: Computer; 22: Image processing unit; 30: Storage device; 40: UI unit; 211: Measurement control unit; 212: Display control unit; 221: Noise level calculation unit; 222: Weight calculation unit; 223-1: Smoothing unit; 223-2: Smoothing unit; 224: Weighted addition unit; 225: Phase correction unit; 226: Complex addition unit. Detailed Implementation

[0031] The embodiments of the MRI device of the present invention will be described below with reference to the accompanying drawings.

[0032] First, the structure of the MRI device using the present invention will be described. The structure of MRI device 1 is as follows: Figure 1As shown, similar to the structure of a typical MRI device, it includes: a magnet 11 that generates a uniform static magnetic field in the examination space where the subject is placed; a tilting magnetic field coil 12 that applies a magnetic field gradient to the static magnetic field generated by the magnet 11; a probe 13 equipped with a transmitting coil that applies a pulsed high-frequency magnetic field to the subject to induce nuclear magnetic resonance in the atomic nuclei of the atoms constituting the subject's tissue, and a receiving coil that receives the nuclear magnetic resonance signal generated from the subject; a receiver 14 connected to the receiving coil; a high-frequency magnetic field generator 15 connected to the transmitting coil; a tilting magnetic field power supply 16 connected to the tilting magnetic field coil 12; a sequence generator 17 that controls the receiver 14, the high-frequency magnetic field generator 15, and the tilting magnetic field power supply 16 according to a given pulse sequence; and a computer 20. All elements except the computer 20 are collectively referred to here as the imaging unit 10.

[0033] The nuclear magnetic resonance signal received by the receiver 14 of the camera unit 10 is digitized and sent to the computer 20 as measurement data.

[0034] The structure and function of each part constituting the camera unit 10 are the same as those of known MRI devices. Furthermore, since the present invention can be applied to various types of known MRI devices and elements, a detailed description of the camera unit 10 is omitted here.

[0035] Computer 20 can be composed of a computer or workstation equipped with a CPU, GPU, and memory. It has control functions for the operation of camera unit 10 and image processing functions for performing various operations on measurement data acquired by camera unit 10 and images reconstructed from the measurement data. The various functions of computer 20 are implemented, for example, by programs uploaded and executed by a CPU or similar device. Some of the functions of computer 20 can also be implemented using hardware such as programmable ICs (ASICs, FPGAs). Furthermore, sometimes the functions of computer 20 can be implemented in a remote computer or a cloud computer connected wirelessly or wiredly to MRI device 1.

[0036] The computer 20 includes a storage device (storage medium 30) for storing data and results (including intermediate results) required for control and calculation; and a UI (user interface) 40 for displaying a GUI, calculation results, or receiving user-specified information. The UI 40 includes a display device and an input device (figures omitted).

[0037] The MRI apparatus of this embodiment is characterized by processing in the computer 20, particularly the processing of the magnetic resonance signals acquired by the imaging unit 10 or the images (multiplex images) reconstructed therefrom. Specifically, the computer 20 includes: a multiplex addition unit for performing addition operations on multiple multiplex images obtained by multiple imaging of the same subject; and a phase correction unit for correcting the phase of each multiplex image supplied for the addition operation. In the phase correction unit, smoothing processing of different intensities is performed, and the processing results are weighted and added to perform phase correction of the multiplex images.

[0038] exist Figure 2 An example of the structure of a computer 20 that implements the above functions is shown. As shown, the computer 20 includes a measurement control unit 211, a display control unit 212, and an image processing unit 22. The image processing unit 22 includes a noise level calculation unit 221, a weight calculation unit 222, smoothing units 223-1 and 223-2, a weighted addition operation unit 224, a phase correction unit 225, and a complex addition operation unit 226. Furthermore, Figure 2 Although not illustrated, the image processing unit 22 includes image reconstruction functions typical of computers in MRI devices, such as Fourier transform or inverse transform functions, or successive approximation operations. Furthermore, additional functions may be added depending on the imaging method, and sometimes these functions may be omitted. Figure 2 It is part of the functional components contained therein.

[0039] The measurement control unit 211 calculates the pulse sequence used in the camera based on the pulse sequence set by the user or pre-set in the inspection protocol and the camera conditions (camera parameters), sets the sequence generator 17, and controls the operation of the camera unit 10 via the sequence generator 17.

[0040] The display control unit 212 performs controls for displaying images obtained through image processing on the display device provided with the UI unit 40, and for accepting GUIs specified by the user.

[0041] The image processing unit 22 performs image reconstruction using measurement data, phase correction of the reconstructed image (complex image), and complex addition operations on the phase-corrected complex image.

[0042] The following describes in detail the operation of the MRI device with the above structure and the processing of the image processing unit 22.

[0043] When the MRI apparatus 1 starts imaging under the control of the measurement control unit 211, it performs imaging according to the pulse sequence set for the sequence generator 17. Here, as an example, it is assumed to perform imaging (DWI) using a pulse sequence that utilizes the aforementioned phase correction effect. Figure 3 This shows an example of a typical pulse sequence for DWI. Figure 3 In the diagram, RF and Sig. represent the timing of the application of the high-frequency magnetic field pulse (RF pulse) and the timing of the acquisition of the echo signal, respectively, while Gs, Gp, and Gr represent the timing of the application of the tilted magnetic field pulse in the slice direction, phase encoding direction, and readout direction, respectively.

[0044] like Figure 3 The pulse sequence shown is a spin-echo EPI sequence. After applying an excitation RF pulse 301 and an inverted RF pulse 303 along with slice selection tilt magnetic fields 302 and 304, a spike-shaped phase-encoding tilt magnetic field pulse 305 is applied, simultaneously reversing the polarity of the readout tilt magnetic field pulse 306, generating an echo signal 307. In DWI, by further applying a pair of high-intensity MPG pulses 308 and 309 before and after the inverted RF pulse 303, the stationary spin recovers the phase that has changed significantly due to MPG pulse 308 via MPG pulse 309. Conversely, the transitioned spin experiences a large phase change corresponding to the direction of the transition. Therefore, in regions where water molecules diffuse in random directions, the signal is reduced due to the phase deviation, resulting in a high-contrast image with a large difference between diffused and diffused tissues.

[0045] exist Figure 3 As an example, the case where the MPG pulse is applied along the readout direction Gr is shown, but the application axis of the MPG pulse can be any of the three axes (Gs, Gp, Gr) or a combination thereof. Furthermore, in Figure 3 The image shows a two-dimensional sequence with phase encoding on the 1 axis, but it could also be a three-dimensional sequence with phase encoding added to the Gs axis.

[0046] The echo signal 307 collected by executing the above-mentioned DWI sequence is received by receiver 14 and sent to computer 20 as digitized measurement data.

[0047] The sequence generator 17 repeats the aforementioned pulse sequence to collect multiple measurement data (k-space data) consisting of the number of echo signals required for one image from the same subject. That is, multiple images are captured to collect multiple image data. The computer 20 performs addition operations on the multiple image data to reconstruct one image.

[0048] Next, the implementation method of image reconstruction (processing of image processing unit 22) in computer 20 will be described.

[0049] <Implementation Method 1>

[0050] refer to Figure 4 The processing of this embodiment will be explained below.

[0051] First, the image processing unit 22 transforms the measurement data (k-space data) obtained from each capture into image data by performing an inverse Fourier transform (S401). Thus, one image is obtained for each capture. This image is a complex image containing both intensity and phase information. That is, as shown... Figure 7 As shown, the complex image 500 is composed of an intensity image 500M and a phase image 501.

[0052] The image processing unit 22 performs complex addition on multiple complex images to create a single image (S407). Before this, phase correction is performed on each complex image (phase image). Phase correction is performed... Figure 5 As shown, the uncorrected complex image (or phase image) 501 is smoothed to correct global phase variations. The phase difference between the smoothed complex image (or phase image) 503 and the uncorrected complex image 501 is taken to obtain the corrected complex image (the phase-corrected image), which corrects global phase variations, and serves as the difference image 505. For example, the corrected complex image 505 can be obtained by performing a complex division operation as follows.

[0053]

Mathematical Formula 1

[0054]

[0055] Here, C corr (x, y) represents the corrected complex image, and C(x, y) represents the original complex image. The phase image was characterized by smoothing.

[0056] In this smoothing process, different degrees of smoothing are performed, and a weighted addition operation is performed on the smoothed phase image to produce a smoothed phase image 502.

[0057] Furthermore, smoothing can be performed using several methods, including smoothing a complex image containing both phase and intensity information unchanged, extracting the phase component from the complex image and smoothing a complex image containing only the phase component, and creating a phase image from the complex image and then smoothing it. The result obtained in the weighted addition operation is the phase smoothing result of the original complex image; however, depending on the method used, the processing up to creating the resulting image or the phase image 502 used in the phase correction described above differs. Details of the methods will be described later; in this embodiment, the case of smoothing a complex image composed only of a complex image or phase components will be used as an example.

[0058] The weights assigned to the results obtained through smoothing with varying intensities are weights distributed according to the intensity image. The weights multiplied by multiple smoothing results vary depending on the location in the image. The basis for calculating the weights is not limited; for example, they can be calculated based on the signal value (signal-to-noise ratio) of the intensity image. For instance, the weights for smoothed images with high smoothing intensity result in a larger background area compared to the region containing the signal from the subject (signal region), while the weights for smoothed images with low smoothing intensity, conversely, result in a smaller background area compared to the region containing the signal from the subject (signal region).

[0059] Therefore, firstly, the noise level calculation unit 221 calculates the SNR (distribution) of the image. If the intensity distribution (pixel value) of the intensity image (absolute value image) is set as M, and the noise level is set as η, then the SNR can be characterized by equation (2) (S402).

[0060]

Mathematical Formula 2

[0061] SNR(x,y)=M(x,y) / η (2)

[0062] In equation (2), x and y represent the positions of pixels.

[0063] Regarding methods for obtaining the noise level η, various approaches can be used, such as: calculating it based on the standard deviation of pixel values; calculating it based on the average and variance of pixel values ​​after removing image edges; and calculating it based on the standard deviation of the background region (e.g., the four corners) of the image. Any of these methods can be used. Furthermore, in this embodiment, the spatial noise amount is set to be the same, and the spatial distribution of noise can be assumed based on information such as the G factor calculated in parallel imaging reconstruction. Additionally, regarding the intensity image M(x, y), in order to reduce the influence of noise in the weighted image, an intensity image smoothed by a median filter or the like can be used. Alternatively, to shorten the computation time, the intensity component of a complex image that has undergone weak smoothing by the smoothing unit can be used unchanged.

[0064] Next, the weight calculation unit 222 uses the SNR calculated by the noise level calculation unit 221 to calculate the weight W when performing a weighted addition operation on the phase image (smoothed phase) that has been smoothed by the two smoothing units 223-1 and 223-2 (described later). S W L (S403). There are no special restrictions on the weights as long as they monotonically increase (or decrease) relative to the change in SNR, such as... Figure 6 As shown, a function can be used to set the weight of pixels with an SNR below the lower threshold t1 to zero and the weight of pixels with an SNR above the upper threshold t2 to 1.

[0065] For example, W S It can be calculated using the following formula (3).

[0066]

Mathematical Expression 3

[0067]

[0068] Conversely, using a function (W) that sets the weights to 1 below t1 and to zero above t2... L =1-W S To calculate W L The thresholds t1 and t2 for the start and end of the rising edge are not limited; for example, they can be set to t1 = 1 and t2 = 5. By setting these values, the non-Gaussianity of the noise distribution is increased, and a weighted average can be used to strictly distinguish between background regions that roughly follow a Rayleigh distribution and signal regions where the noise distribution roughly follows a Gaussian distribution. Furthermore, the thresholds can be set by default or can be set or adjusted by the user via the GUI described later.

[0069] On the other hand, smoothing units 223-1 and 223-2 use filters with different smoothing intensities to smooth the complex image or phase image 501 (S404). The phase image 501 is smoothed by dividing the complex image C, which is the object of processing, by its absolute value image |C|. Thus, the phase component is extracted. The complex image containing only the phase component can be used as the phase image 501. Smoothing can be performed on the complex image C. When only the phase component is extracted, there is a characteristic that information of absolute values ​​is not mixed in; conversely, when the complex image is used unchanged, there is a characteristic that information of absolute values ​​is preserved. Appropriate selection can be made according to the location of the object, the disease to be diagnosed, etc.

[0070] As a filter, known smoothing filters such as Gaussian filters and low-pass filters can be used. Furthermore, for example, when applying a Gaussian filter to a complex image, a smoothed complex image can be calculated by performing convolution operations with Gaussian kernels of appropriate kernel sizes on the real components (real image) and imaginary components (imaginary image) of the complex image respectively. The strength of the smoothing can be enhanced or weakened by varying the kernel size (also called the filter size) of the filter function used; increasing the kernel size increases the strength of the smoothing, and decreasing the kernel size decreases the strength of the smoothing.

[0071] Furthermore, in this specification, the parameters that result in different smoothing intensities are collectively referred to as kernel size. For example, in the case of image smoothing through regularization, the regularization parameter is equivalent to the kernel size in this specification. Moreover, in this specification, it is defined that a larger kernel size results in greater smoothing intensity. However, the relationship between kernel size and smoothing intensity generally varies depending on the smoothing method and the kernel size, and is therefore not limited to this definition.

[0072] Like the threshold for weights, kernel size can be set by default or can be set by the user via the GUI. When the smoothing intensity is preset, an appropriate value can be set based on factors such as its relationship with noise reduction in the background region and the amount of artifacts generated in the signal region. Specifically, for filters with high intensity, since increasing the filter size increases the noise reduction effect in the background region, and this effect becomes fixed above a certain size, the minimum value within the range that becomes fixed can be set. On the other hand, for filters with low intensity, the maximum filter size that no longer generates artifacts in the signal region can be set as the size of the low-intensity filter. However, this is not a limitation; the filter size can also be set by default based on empirically predictable noise levels and then adjusted by the user.

[0073] Next, the weighted addition operation unit 224 uses the weight W calculated by the weight calculation unit 222 to process the smoothed phase processed by the two smoothing units 223-1 and 223-2. S W L To perform weighted addition (S405). Specifically, as follows... Figure 7 As shown, for filters that utilize low strength (filter size σ) S The smoothing part 223-1 smoothed the phase 502-1 by multiplying it by the weight W. S The filter utilizes a high-intensity filter (filter size σ). L The smoothing part 223-2 smoothed the phase 502-2 by multiplying it by the weight W. L They are then added together to obtain the smoothed phase 503. In addition, the "phases" 502-1, 502-2, and 503 mentioned above refer to the results of smoothing. Here, they are phase images composed only of phase components, but when the complex image itself has been smoothed, they are referred to as smoothed complex images.

[0074] By performing a weighted addition operation on the two smoothing results (smoothed phase images), a weak smoothing is performed in the high SNR region, i.e. the signal region, without compromising the signal value, while a smoothing with a high noise reduction effect is performed in the low SNR region, i.e. the background region.

[0075] Phase correction part 225 Figure 5 As shown, the difference between the original complex image's phase (phase image) 501 and the smoothed phase 503 is taken to obtain the corrected phase 505 (S406). Then, while the complex image is smoothed unchanged, the complex image C after the weighted addition operation is... F (C F =C s W s +C L W L The phase image is transformed using the following equation (4). Used to smooth phase 503.

[0076]

[0077] The difference between the original phase image 501 and the smoothed phase image 503 is obtained by using the complex division operation of Equation (1), as described above.

[0078] After performing the above processing S402 to S406 on all complex images corresponding to the addition operation, the complex addition operation unit 226 replaces the phase image 501 of the unprocessed complex image 500 with the corrected phase 505, and performs complex addition operation together with the intensity image 500A (equation (5) below) to obtain one addition operation image X. cmp (S407).

[0079]

Mathematical Expression 5

[0080]

[0081] In equation (5), C corr i Represent each (i-th) complex image that constitutes n corrected complex images.

[0082] After the obtained addition operation images are transformed into intensity images and phase images, they are displayed on the display device of UI unit 40 by display control unit 212, and stored in storage device 30 as needed.

[0083] Because the noise is effectively reduced by not excessively correcting the phase in the signal region, the addition image obtained through the above processing suppresses artifact generation and becomes an image with reduced noise.

[0084] According to the MRI apparatus of this embodiment, when performing complex addition operations on multiple complex images, as a process to correct the phase of each complex image, multiple smoothing results obtained by smoothing with different smoothing intensities are weighted according to the noise characteristics that vary depending on the region and then added. This can achieve the same result as a scheme that has undergone smoothing processing corresponding to noise characteristics, and can simultaneously achieve the noise reduction and artifact suppression that are mutually exclusive in conventional smoothing processing.

[0085] Especially in the complex addition operation of DWI, which is important for SNR improvement, it can perform phase correction with good accuracy on each complex image of the addition operation, and can provide DWI images that are helpful for diagnosis.

[0086] Furthermore, the MRI apparatus according to this embodiment can solve the aforementioned problems through a simple structure that includes two smoothing units and performs a weighted addition operation on the smoothed phase. This method is particularly effective for areas with large phase variations caused by physiological body movements, such as the head region affected by cerebrospinal fluid flow and the abdominal region affected by respiratory activity.

[0087] Figure 8 The images shown are (A) and (B) images obtained by performing phase correction and complex addition according to this embodiment, and (B) images obtained by performing complex addition using a conventional method. In the conventional method, the image is obtained by smoothing, performing phase correction, and performing complex addition at a filter size. In this example, a kernel-sized filter that can sufficiently reduce the noise of the background signal is used.

[0088] In images based on such existing methods, phase correction becomes insufficient in areas of significant physiological movement, i.e., areas of large spatial phase change, resulting in phase inconsistency between images. Therefore, in the image after addition, as indicated by the arrow in (B), artifacts such as signal loss will occur.

[0089] In contrast, in the image (A) after phase correction using the method of the present invention, the signal region is corrected with a weakly smoothed phase and the background region is corrected with a strongly smoothed phase because an adaptive kernel size corresponding to SNR is set. Therefore, the phase of the signal region can be made consistent between cameras, and the noise in the background region can be sufficiently reduced (right figure).

[0090] <Modification of Implementation Method 1>

[0091] The structure and operation of the MRI device of this embodiment have been described above. However, the present invention is not limited to the above embodiment and can be modified in various ways, such as changing the imaging method, replacing the above elements with other elements, or adding new elements.

[0092] For example, in Implementation 1, the case of using the complex image itself or a complex image composed only of phase components taken from the complex image as the object of smoothing is described. However, it is also possible to generate a phase image from the complex image and perform smoothing processing and weighted addition on the phase image.

[0093] In this case, the arg of the complex image C is first calculated in the same way as in equation (4), and the phase image is then constructed. The generated phase image is smoothed after phase unrolling. The same filters used for smoothing complex images can be used for phase image smoothing. Alternatively, a Fourier transform can be performed on the phase image, and only the low-frequency region of the Fourier transform data can be extracted for inverse Fourier transform. The degree of smoothing can be varied by differentiating the extent of the extracted low-frequency region.

[0094] Next, a weighted addition operation is performed on the smoothed phase image, just like in Implementation 1. This variation requires the expansion processing that is not needed in Implementation 1, but since it can be processed as a real number, it has the advantage of being easy to use various existing smoothing techniques (regularization processing, etc.).

[0095] As other variations, for example, in the above embodiment, the smoothing section is shown to have two types of smoothing sections: one with high smoothing degree and one with low smoothing degree. However, it is also possible to further increase the smoothing section with an intermediate smoothing degree, and set their weights according to the signal-to-noise ratio (SNR) of the intensity image. Furthermore, in the above embodiment, the imaging method is described as DWI, but the present invention can be applied when multiple images are added together regardless of the imaging method.

[0096] <Implementation Method 2>

[0097] This embodiment is an embodiment for accepting user-specified GUI for processing in the image processing unit 22, particularly for the weight calculation unit 222 and the smoothing units 223-1 and 223-2. Other structures are the same as in Embodiment 1, and repeated descriptions are omitted.

[0098] exist Figure 9 (A) and (B) show examples of the GUI displayed by the display control unit 212 on the display device of the UI unit 40. Figure 9The examples shown are GUIs for setting phase smoothing conditions in the camera condition setting screen 900 when an image addition operation method (complex addition or absolute value addition) is performed and when complex addition is performed. (A) is an example where boxes 901 and 902 are set to receive the filter size used in the smoothing units 223-1 and 223-2 when complex addition is selected. (B) is an example where boxes 903 and 904 are further set to receive the threshold of the weight function calculated by the weight calculation unit 222. Values ​​set as initial values ​​can be displayed in these boxes. Additionally, in... Figure 9 In the example shown, the input field is set to a box, but it could also be set to a GUI that displays a dropdown menu when the complex addition operation is selected. The display method is not limited to this. Figure 9 As shown in the diagram.

[0099] For example, users can further confirm from the image displayed on the display device (the image after complex addition) whether it is necessary to reduce background noise or whether artifacts are sufficiently suppressed. If noise is reduced, the set filter size (large) value can be changed to a larger value, or if artifact suppression is insufficient, the filter size (small) can be made smaller, or the value of the GUI threshold (e.g., threshold 2) can be adjusted.

[0100] The image processing unit 22 changes to receive new filter size and weight thresholds via the GUI, and then performs... Figure 4 The processes shown are S402 to S407. The processing content is the same as in Implementation Method 1.

[0101] Furthermore, if the image processing unit 22 has more types of smoothing units 223 than two types with different smoothing degrees, it can also be configured so that the user can select any one of them. As a GUI, a box for selecting the smoothing unit 223 can be displayed.

[0102] According to this embodiment, users can obtain more useful images for diagnosis while confirming the images.

[0103] <Implementation Method 3>

[0104] In the above embodiments, the image processing unit 22 of the MRI apparatus 1 performs complex addition operations and phase correction, etc. However, all or part of the processing performed by the image processing unit 22 can also be performed in an image processing unit 2 independent of the MRI apparatus 1. This reduces the computational burden on the image processing unit 22 of the MRI apparatus 1. Furthermore, post-processing can be performed in a location different from the MRI apparatus 1.

[0105] exist Figure 10An example of the structure of the image processing apparatus 2 is shown. This image processing apparatus 2 is a device that can transmit and receive data with the MRI apparatus 1 via a known communication unit, storage medium, etc., and has the function of performing given phase correction processing on multiple complex images received from the MRI apparatus 1.

[0106] To perform this function, the image processing device 2 is equipped with a... Figure 2 The image processing unit 22 of the computer 20 of the MRI apparatus 1 shown has the same functional units as the MRI apparatus 1, namely, a noise level calculation unit 221, a weight calculation unit 222, smoothing units 223-1 and 223-2, a weighted addition unit 224, a phase correction unit 225, and a complex addition unit 226. Their functions are the same as those of the image processing unit 22 of the MRI apparatus 1 described above, and repeated descriptions are omitted. A user interface (UI) unit may also be provided in this image processing apparatus 2 to receive user instructions regarding the filter size and weight thresholds as described in Embodiment 2.

[0107] In addition, Figure 10 In this context, the image processing device 2 is shown to have the same function as the image processing unit 22 of the MRI device. However, the image processing device 2 may also perform only a portion of the processing of the image processing unit 22, such as the smoothing unit and the weighted addition unit, and send the results back to the MRI device to perform phase correction and complex addition operations on the MRI device side.

Claims

1. A magnetic resonance imaging apparatus, characterized by, have: The imaging unit measures the nuclear magnetic resonance signals generated from the subject; and The image processing unit uses the nuclear magnetic resonance signals acquired by the imaging unit to generate an image of the subject. The image processing unit has: The phase correction unit performs phase correction on each of the multiple complex images obtained through multiple imaging operations. The complex addition unit performs addition operations on multiple complex images after phase correction. The smoothing unit performs two or more smoothing processes with different degrees on the complex image or the image having its phase information; and The weighted addition operation unit performs weighted addition operations on the results of smoothing with different degrees of smoothing in the smoothing unit. The phase correction unit uses the phase of the complex image obtained after weighted addition by the weighted addition unit to perform phase correction on the complex image. The magnetic resonance imaging device also features: The weight calculation unit calculates the weights used by the weighted addition operation unit based on the intensity information of the complex image. The image processing unit also includes: The noise level calculation unit calculates the signal-to-noise ratio based on the intensity information of the complex image. The weight calculation unit calculates the weight based on the signal-to-noise ratio.

2. The magnetic resonance imaging device according to claim 1, characterized in that, The weight calculation unit calculates weights that increase in regions with relatively low signal-to-noise ratios and decrease in regions with relatively high signal-to-noise ratios, and uses these weights as weights in the smoothing unit for processing results with high smoothing degrees.

3. The magnetic resonance imaging device according to claim 1, characterized in that, The weight calculation unit calculates the weights that make the weight of the background region 1 and the weight of the signal region with the signal from the subject zero, and uses these weights as the weights for the smoothing result with a high degree of smoothing in the smoothing unit.

4. The magnetic resonance imaging device according to claim 1, characterized in that, The smoothing unit smooths the complex image. After the phase correction unit transforms the complex image obtained by the weighted addition operation performed by the weighted addition operation unit into a phase image, it uses the phase image to perform phase correction on the complex image before correction.

5. The magnetic resonance imaging device according to claim 4, characterized in that, The smoothing unit smooths the complex image composed of phase components extracted from the complex image.

6. The magnetic resonance imaging device according to claim 1, characterized in that, The smoothing unit smooths the phase image created from the complex image. The phase correction unit uses the phase image obtained after weighted addition by the weighted addition unit to perform phase correction on the complex image before correction.

7. The magnetic resonance imaging device according to claim 1, characterized in that, The smoothing unit uses a Gaussian filter, a low-pass filter, and one or more filters selected from the regularization process to perform smoothing.

8. The magnetic resonance imaging apparatus according to claim 7, characterized in that, The smoothing section uses two or more filters with different core sizes.

9. The magnetic resonance imaging device according to claim 1, characterized in that, The multiple complex images are images obtained through diffusion-intensity imaging.

10. The magnetic resonance imaging apparatus according to claim 1, characterized in that, The magnetic resonance imaging device also features: The display control unit causes the display device to display a user-specified GUI that accepts the smoothness of the filter used by the smoothing unit.

11. The magnetic resonance imaging apparatus according to claim 1, characterized in that, The magnetic resonance imaging device also features: The display control unit causes the display device to display a user-specified GUI that accepts weights related to those calculated by the weight calculation unit.

12. An image processing apparatus for processing images captured by a magnetic resonance imaging device. The image processing device is characterized by having: The smoothing section performs two or more smoothing processes with different degrees of smoothing on complex images or images containing their phase information. The weighted addition operation unit performs a weighted addition operation on the results of smoothing with different degrees of smoothing in the smoothing unit. The weight calculation unit calculates the weights used by the weighted addition operation unit based on the intensity information of the complex image; and The phase correction unit performs phase correction on the complex image using the phase of the complex image obtained after weighted addition by the weighted addition unit. The image processing device also has: The noise level calculation unit calculates the signal-to-noise ratio based on the intensity information of the complex image. The weight calculation unit calculates the weight based on the signal-to-noise ratio.

13. The image processing apparatus according to claim 12, characterized in that, The image processing apparatus further includes: The complex addition unit performs complex addition operations on multiple complex images. The complex addition unit performs addition operations on multiple complex images that have undergone phase correction by the phase correction unit.

14. An image processing method for processing multiple complex images obtained by a magnetic resonance imaging device through multiple imaging operations. The image processing method is characterized in that... The following steps are included for each of the multiple complex images: The smoothing step involves performing two or more smoothing processes with different degrees of smoothing. The weights are calculated based on the intensity information of the complex image; The weights are used to perform a weighted addition operation on the images obtained in each of the two or more types of smoothing processes. The phase of the complex image is corrected using the image obtained by performing a weighted addition operation; and Complex addition is performed on the multiple complex images that have been phase-corrected. The image processing method further includes, for each of the plurality of complex images, a step of calculating the signal-to-noise ratio based on the intensity information of the complex images. The weights are calculated based on the signal-to-noise ratio.

15. The image processing method according to claim 14, characterized in that, The multiple complex images are obtained through diffusion-intensity imaging.

Images