Magnetic resonance imaging apparatus and image processing method

Abstract

To perform suitable body motion correction while avoiding an SN ratio deterioration due to body motion correction to the utmost.SOLUTION: An arithmetic unit for performing body motion correction estimates an SN ratio of an image before the body motion correction and an SN ratio of the image after the body motion correction, determines deterioration of the SN ratio due to the body motion correction, and selects suitable correction processing according to it. In order to properly perform body motion correction processing like this, a user interface which displays a parameter concerning the body motion correction and receives user designation is provided.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a magnetic resonance imaging apparatus (hereinafter referred to as an MRI apparatus) and a method for processing measurement data collected by the MRI apparatus, and more particularly to an image processing method that involves body motion correction for reducing the influence of body motion. [Background technology]

[0002] During imaging with an MRI device, the body movement of the subject (examinee) during imaging causes artifacts and degrades image quality. Body movement includes periodic movements of the subject, such as breathing and heartbeat, as well as sudden movements of the subject, and various techniques have been proposed to reduce body movement artifacts depending on the type of body movement.

[0003] A conventional body motion correction technique has been proposed in which body motion information obtained from a body motion detection device such as a camera installed separately from the MRI device, or body motion information detected from a nuclear magnetic resonance signal or measurement data for body motion detection called a navigator or navigator echo in the MRI device, is used to identify data collected when body motion is detected (hereinafter referred to as body motion affected data) from the measurement data used to generate an image of the subject, and the body motion affected data is removed or corrected to reconstruct the image (Patent Document 1).

[0004] Various methods are known for processing body motion effect data, such as a method of removing the body motion effect data from measurement data and then using the measurement data remaining after removal to generate an image by applying image reconstruction methods using compressed sensing or parallel imaging, and a method of correcting the body motion effect data (removed data) by using surrounding data that is not affected by body motion, or by using separately obtained reference data (Patent Document 2).

[0005] Patent document 3 also describes a method for detecting moving parts from measurement data for non-rigid body movements such as eyeball and skin movements, removing the data for the moving parts, and reconstructing a body movement corrected image using the remaining measurement data that was not removed and an image reconstructed from the measurement data before removal. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Publication No. 2023-22669 [Patent Document 2] Japanese Patent Publication No. 2020-130867 [Patent Document 3] Special Publication No. 2015-536746 Summary of the Invention [Problem to be solved by the invention]

[0007] When reconstructing an image by removing data affected by body motion, there is a problem of a decrease in the signal-to-noise ratio (hereinafter abbreviated as SN ratio or SNR). The deterioration of the SN ratio varies depending on the configuration of the receiving coil and the amount of data to be removed, but in the past, the deterioration of the SN ratio after body motion correction has not been fully studied. For example, Patent Document 1 discloses a method of changing the reconstruction method and determining whether or not remeasurement is necessary depending on the type of body motion (such as periodic body motion or sudden body motion), but does not describe a body motion correction method that takes into account the deterioration of the SN ratio after body motion correction.

[0008] Patent Document 3 proposes reducing motion artifacts while maintaining a high S / N ratio by reconstructing an image using an image that contains motion artifacts but has a high S / N ratio (i.e., an image reconstructed from measurement data before removal of data containing motion) and data after removal of the data containing motion, but this technology is premised on specific local motion such as non-rigid motion, and is not suitable for correcting rigid motion such as sudden movements of the subject. Furthermore, it is desirable for body motion correction to be flexible in response to the imaging conditions and the type of body motion occurring, but the prior art does not provide a means for flexibly suppressing degradation of the S / N ratio in accordance with the imaging conditions, etc.

[0009] An object of the present invention is to provide a technology that enables appropriate motion correction while minimizing the degradation of the S / N ratio due to motion correction, and also to provide a technology that can flexibly deal with the degradation of the S / N ratio according to the situation. [Means for solving the problem]

[0010] In the present invention, an MRI apparatus includes a calculation unit that performs motion correction, which estimates the S / N ratio of an image before motion correction and the S / N ratio of an image after motion correction, determines the deterioration of the S / N ratio due to motion correction, and selects an appropriate correction process accordingly. To appropriately perform such motion correction, the apparatus is also provided with a user interface that displays parameters related to motion correction and accepts user specifications.

[0011] That is, the MRI apparatus of the present invention comprises an imaging unit that measures nuclear magnetic resonance signals generated from a subject and collects measurement data for generating an image of the subject, a calculation unit that performs calculations using the measurement data, and a control unit that controls the imaging unit and the calculation unit. The calculation unit comprises a body-motion-affected data identifying unit that identifies nuclear magnetic resonance signals collected when body movement of the subject is detected in the measurement data as body-motion-affected data, an SNR comparing unit that compares the signal-to-noise ratio of a first image generated from the first measurement data before removing the body-motion-affected data with the signal-to-noise ratio of a second image generated from the second measurement data after removing the body-motion-affected data, and an image generating unit that selects a correction process to be applied to at least one of the second measurement data and the second image according to the comparison result of the SNR comparing unit, and generates an image of the subject by performing the selected correction process.

[0012] An image processing method of the present invention is an image processing method for processing measurement data consisting of nuclear magnetic resonance signals collected by an MRI apparatus to generate an image of a subject, and includes the following steps.

[0013] A step of identifying a nuclear magnetic resonance signal collected when the subject's body movement is detected as body movement affected data; a step of comparing the signal-to-noise ratio of a first image generated from the measurement data before the body movement affected data is removed from the measurement data with the signal-to-noise ratio of a second image generated from the measurement data after the body movement affected data has been removed; a step of selecting a correction process from a plurality of body movement correction processes to be applied to at least one of the measurement data after the body movement affected data has been removed and the second image according to the comparison result; and a step of performing the selected correction process to generate an image of the subject. [Effects of the Invention]

[0014] According to the present invention, by selecting and applying motion correction processing to be performed on measurement data based on a comparison of the S / N ratios of images reconstructed from measurement data before and after removal of motion-affected data, it is possible to suppress S / N ratio loss and obtain images that have been appropriately motion-corrected. [Brief explanation of the drawings]

[0015] [Figure 1] 1 is a diagram showing an outline of an MRI apparatus to which the present invention is applied; [Figure 2] Functional block diagram of the calculation unit of the first embodiment [Figure 3] Flow showing the processing of the calculation unit [Figure 4] A diagram showing an example of a navigator sequence [Figure 5] FIG. 10 shows an example of body movement information [Figure 6] Diagram showing an example of excluded data [Figure 7] Diagram illustrating the application of filters with different strengths [Figure 8] Flow showing the processing of the second embodiment [Figure 9] FIG. 10 is a diagram for explaining the processing of the second embodiment. [Figure 10] FIG. 10 is a diagram showing an example of a GUI according to a third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0016] An outline of an MRI apparatus to which the present invention is applied will be described below. MRI devices are classified into horizontal magnetic field types and vertical magnetic field types depending on the direction of the static magnetic field they generate, and various types of MRI devices are known, including permanent magnet types, paramagnetic electromagnet types, and superconducting magnet types for the magnets that generate the static magnetic field. The present invention can be applied to known MRI devices. The shape of the static magnetic field space in which the subject is placed can also be cylindrical, flat, sandwiched between upper and lower magnets, and the present invention can be applied to any of these.

[0017] First, an outline of an MRI apparatus to which the present invention is applied will be described with reference to FIG. As shown in the figure, the MRI apparatus 1 includes a static magnetic field generator (static magnetic field generating magnet) 101 that generates a static magnetic field, an RF transmitter 106 that applies radio frequency magnetic field pulses to a subject 50 placed in a space in which the static magnetic field is generated by the static magnetic field magnet, an RF receiver 107 that receives nuclear magnetic resonance signals emitted by the subject 50, a gradient magnetic field coil 102 and a gradient magnetic field power supply 105 (together referred to as a gradient magnetic field generator) that generate gradient magnetic field pulses that impart a magnetic field gradient to the static magnetic field, a sequencer 108 that controls the gradient magnetic field power supply 105, the RF transmitter 106, and the RF receiver 107 according to a predetermined pulse sequence, and a processor 20 that controls the entire apparatus including the sequencer 108. An RF transmitter coil 103 that applies RF pulses generated by the RF transmitter 106 to the subject and an RF receiver coil 104 that detects nuclear magnetic resonance signals generated by the subject 50 are arranged close to the subject 50. Hereinafter, the RF transmitting unit 106, the RF receiving unit 107, the gradient magnetic field generating unit, etc. will also be collectively referred to as the imaging unit 10.

[0018] The processor 20 may be configured as a general-purpose computer including one or more processors, a memory, and a CPU, and functions as a calculation unit 20A and a control unit 20B. For example, as shown in the functional block diagram of FIG. 1 , the calculation unit 20A may include an image generation unit 220 that generates an image of the subject 50 using the nuclear magnetic resonance signals received by the RF receiving unit 107, and a body movement processing unit 240 that processes signals from a detection unit that detects body movement of the subject during the examination, acquires the magnitude and occurrence time of the body movement (body movement information), and performs calculations and control to eliminate the effects of the body movement using the body movement information. The control unit 20B may also include an imaging control unit 210 that controls the operation of the imaging unit 10 as a whole, and a display control unit 250 that displays the images generated by the image generation unit 220. These functions of the processor 20 are realized by the CPU uploading a predetermined program. Some of the functions may also be realized using a programmable IC.

[0019] Furthermore, the functions of the processor 20 relating to the calculation unit can also be realized by an image processing device independent of the MRI apparatus 1.

[0020] The processor 20 is connected to an input device 30 through which the operator inputs commands and data necessary for imaging, a display device 40 that displays images generated by the image generation unit 220, a storage device 60, and the like. The storage device 60 includes an internal storage device and an external storage device, and the external storage device may be a storage device such as a cloud connected via the Internet, etc. The MRI apparatus 1 can also exchange data with an external database (not shown) such as a PACS. The input device 30 and the display device 40 are installed close to each other and function as a user interface (UI unit 70).

[0021] The MRI apparatus 1 may also be equipped with an optical detection means such as a monitoring camera 80 for monitoring the state of the subject 50 placed in the static magnetic field space, and it is also possible to detect the body movement of the subject 50 using the image from the monitoring camera 80. In this case, the body movement processing unit 240 receives information from the monitoring camera 80 and performs predetermined processing such as body movement correction on the measurement data acquired by the imaging unit 10.

[0022] The configuration of the imaging unit 10 and the imaging flow are the same as those of conventional imaging units, so a description thereof will be omitted. The following describes the embodiment of the present invention, focusing on the body movement processing unit 240.

[0023] <Embodiment 1> In the body motion processing, the MRI apparatus of this embodiment identifies body motion-affected data, and then estimates the degree of degradation in the S / N ratio when image reconstruction is performed after removing the body motion-affected data compared to when image reconstruction is performed without removing the body motion-affected data, i.e., the S / N loss, and selects subsequent processing depending on the degree of S / N loss. Subsequent processing includes correction processing using a filter, half estimation processing, etc., and the filter processing includes filter processing with different strengths.

[0024] A functional block diagram of the calculation unit 20A that executes the above functions is shown in Fig. 2. As shown in the figure, in addition to the image generation unit 220, the calculation unit 20A includes a body motion influence data identification unit 241, an SNR comparison unit 243, and a correction processing selection unit 245. As shown in the figure, these may be functions included in the body motion processing unit 240, or each may be configured as an independent processor.

[0025] Hereinafter, the operation including the body motion processing of the MRI apparatus of this embodiment will be described with reference to the flow of FIG.

[0026] <Image capture: S1> First, under the control of the imaging control unit 210, imaging is performed by the imaging unit 10 (S1). The imaging is performed under the control of the sequencer 108 based on an imaging pulse sequence that is registered in advance or set by the user via the UI unit 70, and imaging parameters such as TE, TR, FOV, and speed ratio (R factor). By performing the imaging, measurement data for generating an image of the subject 50, i.e., k-space data, is collected.

[0027] The pulse sequence may include a navigator sequence for detecting the body motion of the subject 50. In this case, navigator echo data (navigator data) is also acquired. The navigator echo is a nuclear magnetic resonance signal acquired from a predetermined region without phase encoding, separate from the nuclear magnetic resonance signal (referred to as the actual imaging echo) used to generate an image of the subject 50. The navigator sequence for acquiring the navigator echo and the method for acquiring body motion information from the navigator data are well-known sequences and methods, and therefore detailed description is omitted in this specification. However, an example is shown in FIG. 4. FIG. 4 shows an example in which the actual imaging echo and the navigator echo are acquired within the same TR (repetition time) after application of an RF pulse. The upper side shows an example in which the navigator echo is acquired before the actual imaging echo, and the lower side shows an example in which the navigator echo is acquired after the actual imaging echo. In addition, a navigator sequence for acquiring only the navigator echo may be added separate from the pulse sequence for acquiring the actual imaging echo.

[0028] Furthermore, if the MRI apparatus 1 is equipped with a body motion detection device such as a monitoring camera 80 for detecting body motion, signals and images from the body motion detection device are sent to the body motion processing unit 240 in parallel with imaging. Note that various types of body motion detection devices are known, such as the monitoring camera 80, an infrared sensor, and a balloon attached to the subject 50, and any of these can be used alone or in combination. Depending on the imaging purpose and imaging sequence, it may be difficult to execute the navigator sequence, so a configuration in which the user can select an appropriate sequence is also possible.

[0029] <Analysis of body movement information: S2, Identification of body movement influence data: S3> When imaging begins, the calculation unit (body movement processing unit 240) analyzes the video from the navigator echo or monitoring camera 80 (S2) and identifies data (body movement influence data) collected when body movement is detected from the measurement data (S3). The body movement information obtained by the body movement detection means varies, but is obtained as displacement or fluctuation (change) of the subject along the time axis. An example of body movement information obtained from the navigator echo or monitoring camera 80 is shown in Figure 5. The example shown is information (graph) indicating the displacement of a specific part of the subject 50 along the time axis, and by analyzing this graph, the magnitude of the displacement (absolute value or relative value), the time when a displacement of a specific magnitude occurred, the period of the displacement, etc. can be obtained.

[0030] In order to identify the body movement influence data, thresholds can be set in the body movement processing unit 240 (body movement influence data identification unit 241) for the magnitude, duration, frequency, etc. of the body movement, and these thresholds are used to identify the body movement that affects the data and to identify the measurement data collected during the time when that body movement occurred.

[0031] Figure 6 shows an example of identified motion-affected data. The left side of Figure 6 shows k-space data (kx-ky space) for 2D imaging, and the right side shows k-space data (ky-kz space) for 3D imaging. Although this differs depending on the k-space data acquisition method (sampling order), in 2D imaging, the ky direction is the phase encoding direction, which usually corresponds to the time axis, and kx-ky line data is identified as motion-affected data. In 3D imaging, when echoes are acquired while changing the phase encoding within one slice encoding, some of the ky-kz line data may be identified as motion-affected data.

[0032] <Image SNR estimation before and after motion correction: S4> When the SNR comparison unit 243 removes the identified body movement influence data from the measurement data, it calculates the loss in the SNR of the image reconstructed from the measurement data after the removal.

[0033] In motion correction, for example, after removing motion-affected data, if the motion-affected data is not low-frequency data and sufficient k-space data remains to be reconstructed even after removing the motion-affected data, the removed data is zero-filled to perform image reconstruction. However, the image reconstructed in this manner has a degraded SNR due to the reduced number of data. The SNR comparator 243 calculates the SNR loss by comparing the SNR of an image reconstructed after removing the motion-affected data with the SNR of an image reconstructed using measurement data (k-space data) without the motion-affected data, i.e., without removing the motion-affected data.

[0034] Generally, the signal-to-noise ratio (SNR) of an image can be expressed by the following equation, using the g factor of the receiving coil used for imaging and the R factor (speed ratio) during imaging, assuming that the signal-to-noise ratio of an image when fully sampled is SNR-0. SN ratio=SNR-0 / (g-factor*√(R-factor))

[0035] The SNR-0 and g-factor are the same for the image before and after removing the motion-affected data, and the difference in the SNR between the two is caused by the change in the R-factor due to the removal of the data. That is, by removing the motion-affected data, the k-space thinning rate (the ratio of the number of unmeasured data to the number of fully sampled data: the reciprocal of the R-factor) becomes a value obtained by adding the number of motion-affected data to be removed as the denominator.

[0036] Therefore, by using the thinning rate before and after excluding the body motion-affected data, the loss of the S / N ratio in the image after excluding (the rate of S / N ratio degradation when the S / N ratio before excluding is 100) can be estimated using the following formula: S / N ratio loss={1 / g(R1)*√(1 / R1)-1 / g(R2)*√(1 / R2)}g(R1) / √(1 / R1) In the formula, R1 is the R-factor before excluding data affected by body movement, R2 is the inverse of the thinning rate after excluding data affected by body movement (equivalent to the R-factor), and g() is a function of the g-factor that depends on the R-factor.

[0037] Here, the S / N ratio loss was calculated simply using the thinning rate, but it is also possible to use known S / N ratio estimation techniques to estimate the S / N ratio for both an image without the influence of body motion and an image after removing the data affected by body motion, and then calculate the S / N ratio loss. However, because this method does not perform estimation processing using the image itself, there is no computational load and estimation and selection of the next filter processing can be easily performed.

[0038] <Body motion correction process selection: S5> In this embodiment, motion correction is achieved by removing motion-affected data and filtering, and the filter strength is varied based on the comparison result of the signal-to-noise ratio. The filtering process may be performed in either image space or k-space. For image space filters, smoothing filters such as a Guanzian filter, a bilateral filter, or a median filter can be used. As k-space filters (frequency domain filters), low-pass filters (LPFs), Fermi filters, etc. can be used. Depending on the circumstances, these filters can also be combined with high-pass filters (HPFs) or band-pass filters (BPFs). The k-space filter multiplies k-space data by a predetermined window function, and the window functions can be one or a combination of Hanning, Hamming, Gaussian, Kaiser-Blackman, Fermi, etc.

[0039] Regarding the type and combination of filters, a filter configuration for suppressing truncation artifacts (also called ringing artifacts) that occur due to data truncation may be adopted. As a filter for suppressing truncation artifacts, for example, a smoothing filter used in image space, as well as a technique for combining an HPF or the like based on an LPF and BPF as a k-space filter, are known, and these can be applied.

[0040] The filter strength is selected by setting, for example, two or more thresholds for the SNR loss calculated in S4 above, and setting the strength to "low" if the SNR loss is at the lower threshold (for example, 20% or more), "high" if the SNR loss is at the higher threshold (40% or more), and "medium" if the SNR loss is between 20% and 40%. In either case, the filter strength is changed for high-frequency data.

[0041] Figure 7 shows a conceptual diagram of a k-space filter. In the figure, the left side shows 2D k-space data, with data identified as motion-affected data in the high-frequency range. In this example, three filters with different strengths are provided, with stronger filtering applied in the high-frequency range as the strength increases from "low" to "high." Processing the measurement data after removing motion-affected data with such a filter recovers the SNR loss and smooths the data with a strength according to the magnitude of the SNR loss. Reconstructing this measurement data makes it possible to obtain an image with no degradation in the SNR and with the effects of motion eliminated.

[0042] In the above description, the correction process selection unit 245 automatically selects the filter strength using a preset threshold value, but it is also possible to accept a selection by the user via the UI unit 70. Furthermore, the type of filter and the application method may also be selected according to criteria established for the number of body motion influence data, the arrangement in k-space, etc.

[0043] <Image reconstruction: S6> For example, when a k-space filter is used, the image generation unit 220 zero-fills the measurement data after removing the body motion-affected data using the filter strength selected by the correction processing selection unit 245, and then performs filter processing on the measurement data to reconstruct an image. The reconstruction method is not particularly limited, and known methods can be used, such as image reconstruction based on the PI method using the receiver sensitivity distribution of a receiver coil consisting of multiple small coils, or image reconstruction using iterative calculations such as compressed sensing. It is also possible to perform filter correction processing and image reconstruction within a single reconstruction method.

[0044] As described above, the MRI apparatus of this embodiment has a calculation unit with a motion correction function, and performs motion correction using a method that suppresses degradation of the S / N ratio. To this end, the S / N ratios of images before and after motion correction are estimated, the degradation of the S / N ratio (S / N ratio loss) due to motion correction is determined, and a correction processing method is selected. This allows for the acquisition of reconstructed images that do not include motion-affected data and have no degradation of the S / N ratio.

[0045] In the first embodiment, a case has been described in which a relatively small amount of body motion-affected data is excluded from the measurement data, and image reconstruction is performed with correction processing of the measurement data after exclusion. However, in cases in which a large amount of body motion-affected data is excluded or the data is concentrated in the low range of k-space, the data or the entire k-space data may be remeasured without correction processing.

[0046] <Embodiment 2> In the first embodiment, the body motion processor 240 selected the strength of the filtering process to be performed as body motion correction by comparing the SNR of the images before and after body motion correction. However, in this embodiment, the correction process and image reconstruction process are selected in accordance with the position of the body motion-affected data in k-space.

[0047] The configuration of body movement processing unit 240 in this embodiment is the same as the configuration of body movement processing unit 240 in embodiment 1 shown in Fig. 2, and the processing flow of body movement processing unit 240 in this embodiment will be described below with reference to the flow shown in Fig. 8. In Fig. 8, processes with the same content as the processes shown in Fig. 3 are indicated by the same reference numerals, and detailed descriptions of those processes will be omitted.

[0048] First, while imaging is being performed (S1), the presence or absence of body movement of the subject is detected using body movement of at least one of the monitoring camera and navigator data (S2). When body movement is detected, for example, when there is a change in body movement position equal to or greater than a predetermined threshold, the k-space data collected during that time is identified as body movement-affected data (S3).

[0049] Next, the motion-affected data is removed, and the S / N ratio of the image reconstructed using the k-space data obtained by zero-filling the data in that portion is calculated to determine how much S / N ratio loss there is compared to the S / N ratio of the image reconstructed assuming no motion (S4). The method for calculating the S / N ratio loss in this case is the same as in the first embodiment, and the S / N ratio loss is calculated based on the thinning rate of the k-space data excluding the motion-affected data.

[0050] Next, when the number of body motion affected data is relatively large and the SNR loss is significant, the position of the body motion affected data in k-space is determined, and correction processing or reconstruction processing after removing the body motion affected data is determined (S51, S52).

[0051] Specifically, for example, if the signal-to-noise ratio loss is 30% or more, or if the body motion effect data is a predetermined ratio or more of the k-space data (S51), it is determined whether the body motion effect data is high-frequency data in k-space and whether the high-frequency data on the opposite side does not contain body motion effect data (S52).

[0052] When the motion-affected data is high-frequency data in k-space and there is symmetrical high-frequency data, the image is reconstructed by performing half estimation using the properties of k-space. That is, the high-frequency data, which has a different phase encoding polarity from the motion-affected data to be removed, is subjected to processing to invert only the phase, and is then replaced with the motion-affected data.

[0053] On the other hand, if there is no complementary high-frequency data, i.e., if body motion-affected data exists in a high-frequency region complementary to the high-frequency region containing the body motion-affected data, correction is performed using a filter rather than half estimation. In this case, it is possible to simply increase the filter strength, as in the case of the first embodiment where the SNR loss is equal to or greater than a higher threshold (e.g., 40%), but in this embodiment, a truncation artifact filter is used as the filter. For example, the default filter is switched to a truncation artifact filter. The state in which high-frequency data is thinned here is similar to the situation in which truncation artifacts occur due to data truncation, and by using a truncation artifact filter as the filter, it is possible to effectively suppress the occurrence of artifacts caused by removing high-frequency data.

[0054] If the loss in the SNR is equal to or less than a predetermined threshold in S51, the strength of the filter may be selected according to the degree of the loss in the SNR (S5, S6), as in embodiment 1. Furthermore, if the position in the k-space of the body motion effect data is determined in step S52, and the body motion effect data exists in the low frequency range rather than the high frequency range, remeasurement is performed as in embodiment 1.

[0055] Note that the flow in FIG. 8 is premised on estimating the loss in the SNR (FIG. 8: S4, S51), but if the proportion of data affected by body motion in the high frequency range of k-space is equal to or greater than a predetermined proportion, it is also possible to directly perform process S52 (process for determining whether complementary high frequency data exists in k-space) without calculating the SNR loss, and then perform half estimation reconstruction or reconstruction after correction processing using a truncation artifact filter.

[0056] According to this embodiment, by determining the position in k-space of the body motion effect data and the presence or absence of complementary data, and selecting the filter type and image reconstruction method, it is possible to obtain images without degradation in image quality while suppressing loss of the S / N ratio.

[0057] <Modification 1 of Embodiment 2> In the second embodiment, when a relatively large amount of high-frequency data (for example, 30% or more) must be removed as data affected by body motion, a filter for truncation artifacts is used or half estimation is performed. However, it is also possible to employ filter selection or image reconstruction using a machine learning model such as DCNN (Deep Convolutional Neural Network).

[0058] As a learning model (CNN) that can be used for measurement data after the removal of the influence of body movement, for example, a model that uses a set of a large number of images reconstructed from k-space data in which high-frequency data has been randomly removed at a predetermined rate and an image reconstructed from full-sampled k-space data as training data, and is trained to output a full-sampled image in response to input of an image of measurement data with missing high-frequency data, or a model that is trained to output measurement data in which the missing data has been interpolated in response to input of measurement data with missing high-frequency data.

[0059] Regarding filter selection, a method can be used in which the system is trained using a set of images reconstructed after processing k-space data from which high-frequency data has been randomly removed at a predetermined rate using various filters or combinations of filters, and images reconstructed from full-sampled k-space data, and in response to input measurement data with defects in the high frequency range, the system outputs the optimal filter or measurement data processed with that filter.

[0060] In applying CNN, CNNs with noise reduction functions are also known, and these known CNNs may be combined.

[0061] According to this modification, although it is necessary to construct a separate CNN, even when the high-frequency data contains a large amount of data affected by body movement, it is possible to select an optimal filter to suppress SNR loss or perform image reconstruction, and it is possible to easily reconstruct an image with reduced SNR loss.

[0062] <Modification 2 of Embodiment 2> In the second embodiment, a case has been shown in which part of the high frequency data of k-space is identified as body motion affected data and removed, but for example, when detecting body motion, if the presence or absence of body motion is determined based on the amount of variation rather than the absolute value of the displacement of the subject, and it is determined that body motion has occurred from time t1 to t2, as shown in Fig. 9, the k-space data collected from t1 to t2 is identified as body motion affected data. However, if the subject position subsequently changes from the original position, it is necessary to remove the data as body motion affected data even if the body motion has disappeared.

[0063] Although it is possible to deal with such data by phase correction, in this modification, data collected after the subject position has changed following collection of motion-affected data is also identified as motion-affected data, and subsequent correction processing is performed according to the flow of Fig. 8. That is, in this case as well, the image SNR loss before and after motion correction is estimated (S4, S51), and the data remaining in the high frequency range of the k-space data is taken into consideration (S52), and the subsequent reconstruction processing is determined. According to this modification, even when the body movement influence data specified differs due to differences in body movement detection means, etc., it is possible to reliably remove data that influences the image.

[0064] <Embodiment 3> This embodiment relates to the UI unit 70 of the MRI apparatus 1, and can be commonly applied to the above-described embodiments.

[0065] In the first and second embodiments, the selection of the filter strength or the selection of whether to use half estimation or a filter for truncation artifacts is automatically performed by the body movement processing unit 240 using a threshold value. In this embodiment, however, a GUI for prompting the user to make these selections is displayed on the display device 40 of the UI unit 70, and processing is performed according to the user selection.

[0066] The GUI allows the user to select parameters or conditions related to body motion correction (collectively referred to as body motion parameters), and allows the user to set at least one of the following: body motion detection means, filter strength used for body motion correction, threshold for filter selection, filter type, and selection of half estimation processing.

[0067] Fig. 10 shows an example of a GUI displayed on the display device 40. Fig. 10 shows a screen 1000 for inputting body movement parameters. Such screen 1000 is displayed when body movement correction "required" is selected on a screen for setting imaging conditions such as imaging parameters (TE, TR, FOV, magnification rate, etc.).

[0068] In the example shown in FIG. 10, there are set up a block 1010 for selecting the strength of body motion correction, i.e., the strength of the filter used for body motion correction, a block 1020 for selecting the type of filter, a block 1030 for setting a threshold for determining the filter strength, a block 1040 for selecting a means for detecting body motion, and a block 1050 for displaying body motion information, etc.

[0069] As explained in the first embodiment, the body movement detection means can be the surveillance camera 80 ("VISUAL"), navigator data ("NAVI"), or a combination of these ("COMBINATION"), and the user can select any of these via block 1040. Depending on the imaging purpose and imaging sequence, it may be difficult to execute the navigator sequence. There may also be cases where the surveillance camera is malfunctioning or no surveillance camera is provided. The user can set the most appropriate body movement detection means according to the imaging purpose and system conditions.

[0070] In the first embodiment, the filter strength is set according to the signal-to-noise ratio loss, but the user can set any strength via block 1010. Here, the GUI allows the user to select one of "LIGHT" (weak), "MEDIUM" (medium), or "HEAVY" (strong) as the strength. However, the strength may also be set or selected using a numerical value such as "0 to 3." As shown in the figure, the user may also select whether or not to perform filter correction in addition to the strength selection. When selecting the strength, for example, the body motion processor 240 may display the selection result, and the user may determine whether or not a change is necessary. The user may then change the selection as necessary. In this case, one or more thresholds (TH1, TH2) used by the body motion processor 240 may be displayed in block 1030, allowing the user to change them. Regarding the filter, a default filter may also be displayed in block 1020, allowing the user to change it.

[0071] Furthermore, a diagram showing body motion influence data in k-space data or body motion information detected by a body motion detection means, such as the diagram shown in Fig. 9, may be displayed in block 1050 or in a separate window. This allows the user to make a decision, including whether or not to perform remeasurement. Along with this block 1050, a button (GUI) 1060 for selecting half estimation processing may be displayed.

[0072] By providing a GUI specialized for body motion processing, the user can flexibly handle body motion correction, and as post-processing, can acquire images after body motion correction processing with different filter conditions, etc. Such multiple images with different conditions can be used as training data for CNN, and can be used to improve the accuracy of CNN as described in Modification 1 of Embodiment 2.

[0073] According to this embodiment, by providing a GUI for setting body motion parameters, it is possible to perform appropriate body motion correction according to the actual imaging situation, in addition to automatic processing selection by the system. Furthermore, since it is possible to generate images under various body motion correction conditions using measurement data after imaging, it is effective for evaluating body motion correction processing and building CNN. [Explanation of symbols]

[0074] 1: MRI device, 10: imaging unit, 20: processor, 20A: calculation unit, 20B: control unit, 30: input device, 40: display device, 60: storage device, 70: UI unit, 80: surveillance camera, 240: body motion processing unit, 241: body motion influence data identification unit, 243: SNR comparison unit, 245: correction processing selection unit

Claims

1. The imaging device includes an imaging unit that measures nuclear magnetic resonance signals generated from a subject and collects measurement data for generating an image of the subject, a calculation unit that performs calculations using the measurement data, and a control unit that controls the imaging unit and the calculation unit. The calculation unit a body movement-affected data identifying unit that identifies, in the measurement data, a nuclear magnetic resonance signal collected when a body movement of the subject is detected, as body movement-affected data; an SNR comparison unit that compares the signal-to-noise ratio of a first image generated from the first measurement data before removing the body motion effect data with the signal-to-noise ratio of a second image generated from the second measurement data after removing the body motion effect data; an image generating unit that selects a correction process to be applied to at least one of the second measurement data and the second image according to a comparison result of the SNR comparing unit, and generates an image of the subject by performing the selected correction process.

2. 2. The magnetic resonance imaging apparatus according to claim 1, the SNR comparator calculates a loss in signal-to-noise ratio of the second image; The magnetic resonance imaging apparatus according to claim 1, wherein the image generating unit selects one of a plurality of preset correction processes in accordance with the loss of the signal-to-noise ratio.

3. 2. The magnetic resonance imaging apparatus according to claim 1, a signal-to-noise ratio (SNR) comparison unit that estimates a signal-to-noise ratio (SNR) of the first image and a signal-to-noise ratio (SNR) of the second image using a multiplication rate of the first measurement data and a thinning rate of the second measurement data.

4. 3. The magnetic resonance imaging apparatus according to claim 2, The magnetic resonance imaging apparatus according to claim 1, wherein the plurality of preset correction processes include a process using a filter and a process for estimating missing data.

5. 3. The magnetic resonance imaging apparatus according to claim 2, The magnetic resonance imaging apparatus according to claim 1, wherein the plurality of preset correction processes are processes using a plurality of filters with different strengths or types.

6. 6. The magnetic resonance imaging apparatus according to claim 5, The magnetic resonance imaging apparatus is characterized in that the filter is a filter having any one of a Hanning, Hamming, Gaussian, Kaiser-Blackman, and Fermi window functions, or a filter having a combination of two or more of these window functions.

7. 2. The magnetic resonance imaging apparatus according to claim 1, The magnetic resonance imaging apparatus according to claim 1, wherein the image generating unit selects a correction process depending on the comparison result of the SNR comparing unit and the arrangement of the body motion effect data in k-space.

8. 8. The magnetic resonance imaging apparatus according to claim 7, a truncation filter for correcting the motion of the body motion affected data; a correction filter for correcting the motion of the body motion affected data;

9. 8. The magnetic resonance imaging apparatus according to claim 7, a magnetic resonance imaging apparatus characterized in that the image generation unit selects half estimation processing of the measurement data as the correction processing when a proportion of the body motion effect data present in the high frequency range of k-space is equal to or greater than a predetermined value.

10. 2. The magnetic resonance imaging apparatus according to claim 1, 1. A magnetic resonance imaging apparatus, comprising: a UI unit that displays parameters relating to body motion correction performed on measurement data and accepts settings of the parameters by a user.

11. 11. The magnetic resonance imaging apparatus according to claim 10, The parameters include one or more of the type of body movement detection method, the strength of body movement correction, the type of correction process, and a threshold for selecting the correction process, and the UI unit presents at least one of the parameters to the user.

12. 1. An image processing method for generating an image of a subject by processing measurement data consisting of nuclear magnetic resonance signals collected by a magnetic resonance imaging device, comprising: identifying the nuclear magnetic resonance signals collected when the subject's body movement is detected as body movement influence data; comparing a signal-to-noise ratio of a first image generated from the measurement data before the body motion influence data is removed with a signal-to-noise ratio of a second image generated from the measurement data after the body motion influence data is removed; An image processing method characterized by selecting a correction process to be applied to at least one of the measurement data after excluding the body movement influence data and the second image from among a plurality of correction processes depending on the comparison result, and performing the selected correction process to generate an image of the subject.

Images