Magnetic resonance camera device and method of controlling the same

Abstract

This invention provides a magnetic resonance imaging device and its control method. The objective of this invention is to suppress the extension of imaging time and perform high-precision N / 2 ghosting correction in imaging using EPI signal acquisition. A pre-scan is performed in the EPI sequence, with each subsequent scan involving a single excitation of a blip-shaped gradient magnetic field, generating N / 2 ghost-free data with different positive and negative values. Pseudo-N / 2 ghosting data is then created based on this N / 2 ghost-free data. Using the pseudo-N / 2 ghosting data, a correction kernel is calculated based on the signal of the measurement point of the object to be corrected and the signals of the measurement points on the surrounding even and odd rows. The final scan data is corrected using the signals of the measurement point of the object to be corrected and the signals of the measurement points on the surrounding even and odd rows, and through the correction kernel.

Description

Technical Field

[0001] This invention relates to a magnetic resonance imaging device (hereinafter referred to as an MRI device), and more particularly to a technique for reducing artifacts specific to pulse sequences when using echo-plane imaging (EPI) for imaging. Background Technology

[0002] When collecting MRI signals from a subject, the MRI device controls the intensity and timing of the gradient magnetic field applied to the subject, as well as the timing of MRI signal collection, according to a prescribed pulse sequence, and performs imaging. Various pulse sequences are known, but EPI (Excitation Per Injection) is a method that rapidly reverses the polarity of the readout gradient magnetic field after a single excitation, generating a large amount of MRI signal as gradient echoes for collection at each reversal. It is used as a high-speed imaging pulse sequence for various imaging applications. For example, in DWI (Diffusion-weighted imaging) to image tissues with diffusive properties such as blood and cerebrospinal fluid by applying multiple high-intensity gradient magnetic field (MPG) pulses, multiple imaging sessions are required, with different MPG pulse sizes or application axes. Because this is easily affected by body movement, the EPI sequence, an ultra-high-speed imaging sequence capable of scanning the entire k-space, is used as the imaging sequence.

[0003] However, due to the B0 inhomogeneity generated in the readout direction (x-direction), the output delay of the gradient magnetic field power supply driving the gradient magnetic field coil, and the resulting sluggish rise and fall of the gradient magnetic field pulses or the influence of eddy currents, EPI produces a shift in the applied area of ​​the readout gradient magnetic field when the polarity of the readout gradient magnetic field is positive and negative. This results in a change in the scan trajectory (k-trajectory) configured in k-space between the echoes configured on odd-numbered rows and those configured on even-numbered rows in k-space. This deviation between odd and even rows in k-space manifests as N / 2 ghosting (also known as artifacts) in real space (image space).

[0004] The aforementioned DWI provides diagnostic images for minor cerebral hemorrhages or cerebral infarctions by repeatedly performing EPI and processing the resulting multiple images (k-space). It is also used to calculate diffusion coefficients, etc. Therefore, the artifacts generated by EPI have a significant impact, and eliminating this impact is indispensable for high-precision imaging.

[0005] Various methods have been proposed to eliminate artifacts associated with EPI (Electronic Image Processing). For example, Non-Patent Document 1 discloses a method that corrects the scan (hereinafter referred to as the formal scan) used to acquire an image of the photographic object using reference data obtained from a single excitation without the application of a blip gradient magnetic field. In this method, the phase difference between the odd and even rows in the x-ky space is calculated to be linearly approximated, and the phase difference between the odd and even rows of the formal scan data is corrected in the x-ky space using the phase difference of the linearly approximated reference data. However, this method suffers from the problem of being unable to cope with higher-order phase changes generated in the readout direction, and the problem of being unable to correct the phase changes of the odd and even rows in the phase encoding direction generated when a blip gradient magnetic field is applied.

[0006] In contrast, Patent Documents 1 and 2 propose methods (2D correction) capable of handling higher-order phase changes and phase differences generated in the two-dimensional direction of k-space. For example, in the method described in Patent Document 1, a third type of data (pre-scan data) that does not produce N / 2 ghosting is obtained, different from the even-echo and odd-echo data of the formal scan. This third type of data is used to estimate the relationship information between each even-echo and each odd-echo. Based on this relationship information, the k-space data of the even-echo and the k-space data of the odd-echo are corrected respectively, and then synthesized. In the method described in Patent Document 2, instead of using the pre-scan data as described in Patent Document 1, a convolution kernel is calculated based on the measurement data of the formal scan. This convolution kernel is applied to the measurement data as synthetic data, and combined data is generated based on this synthetic data and the measurement data.

[0007] These methods can correct higher-order phases, and the method in Patent Document 2, in particular, does not have the problem of prolonged imaging time because it does not require pre-scanning. However, all of them suffer from the problem of reduced correction accuracy as the magnification rate increases in parallel imaging methods.

[0008] In contrast, Non-Patent Document 2 discloses the following technology: introducing a blip gradient magnetic field, performing two pre-scans to reverse the polarity of the readout gradient magnetic field (reversing the reversal of positive and negative polarity), generating data without N / 2 ghosting, data with only positive polarity, and data with only negative polarity based on the obtained reference data of the two excitations, calculating a correction kernel for correcting N / 2 ghosting and a kernel for interpolating undersampled points based on these three types of data, and interpolating undersampled points simultaneously with the correction of odd and even rows.

[0009] This technique eliminates N / 2 ghosting caused by higher-order phases and enables parallel imaging processing, avoiding the problems found in the two patent documents mentioned above. However, to obtain reference data, two pre-scanning excitations are required, inevitably extending the imaging time and resulting in residual N / 2 ghosting if there is body movement between the two pre-scannings.

[0010] Non-patent literature 1: Bruder H, Fischer H, Reinfelder HE, Schmitt F. Image reconstruction for echo planar imaging with nonequidistant k-space sampling. Magn Reson Med 1992; 23:311-323.

[0011] Non-patent literature 2: W. Scott Hoge1* and Jonathan R. Polimeni. Dual-PolarityGRAPPA for Simultaneous Reconstruction and Ghost Correction of Echo PlanarImaging Data. Magn Reson Med 2016; 76:32-44

[0012] Patent Document 1: US Patent No. 11002815

[0013] Patent Document 2: US Patent No. 10,557,907

[0014] As mentioned above, 1D correction has the problem of not being able to eliminate higher-order phase differences between odd and even rows. While 2D correction can solve higher-order phase differences, it has problems such as reduced correction accuracy at high magnification, longer imaging time, and the effects caused by body motion. Summary of the Invention

[0015] The objective of this invention is to provide an MRI technique that can resolve higher-order phase differences, maintain correction accuracy even at high magnification rates, and minimize the duration of imaging while minimizing the effects of body motion.

[0016] To address the aforementioned issues, this invention performs a pre-scan by applying a blip gradient magnetic field with a single excitation every other controlled interval, and uses this pre-scan data to generate simulated data without N / 2 ghosting. A correction kernel is calculated using this simulated data and odd-numbered and even-numbered rows of reference data, and then applied to the final scan data.

[0017] That is, the MRI apparatus of the present invention includes: an imaging unit that measures magnetic resonance signals generated from an examination subject according to a predetermined pulse sequence; and a processor that performs calculations including image reconstruction using the magnetic resonance signals and motion control of the imaging unit. The processor performs the following processing: as a formal scan, it executes a pulse sequence based on echo-plane imaging, which, after applying an excitation RF pulse, applies a readout gradient magnetic field that vibrates polarity between positive and negative and a blip-shaped phase-encoded gradient magnetic field, and measures the echo signal to collect k-space data each time the polarity of the applied readout gradient magnetic field reverses; as a pre-scan, it executes a pre-scan sequence based on echo-plane imaging, which changes the application of the blip-shaped phase-encoded gradient magnetic field in the formal scan to one time every two polarity reversals of the readout gradient magnetic field. Furthermore, the processor performs the following processing: dividing the reference k-space data collected in the pre-scan into a first reference data consisting of a positive signal obtained when the applied polarity of the readout gradient magnetic field is positive, and a second reference data consisting of a negative signal obtained when the applied polarity of the readout gradient magnetic field is negative; generating at least two pseudo-N / 2 ghost data with alternating positive and negative signals using the first and second reference data; calculating a correction kernel for correcting N / 2 ghost using the pseudo-N / 2 ghost data; correcting the formal scan k-space data collected by performing the formal scan using the correction kernel; and reconstructing the image using the corrected k-space data.

[0018] Furthermore, in the control method of the MRI device of the present invention, the MRI device includes: an imaging unit that measures magnetic resonance signals generated from the subject of examination according to a predetermined pulse sequence; and a computing unit that includes image reconstruction using the magnetic resonance signals. The control method of the MRI device controls the imaging unit to execute a formal scan sequence based on echo-plane imaging and a pre-scan sequence based on echo-plane imaging. In the formal scan sequence based on echo-plane imaging, after applying an excitation RF pulse, a readout gradient magnetic field with polarity oscillation between positive and negative and a blip-shaped phase-encoded gradient magnetic field are applied. Each time the polarity of the applied readout gradient magnetic field reverses, the echo signal is measured to collect k-space data. In the pre-scan sequence based on echo-plane imaging, the application of the blip-shaped phase-encoded gradient magnetic field in the formal scan sequence is changed so that the polarity of the readout gradient magnetic field reverses twice, which constitutes one time. Furthermore, the control and calculation unit divides the reference k-space data collected in the pre-scan into first reference data consisting of a first echo signal (positive signal) obtained when the applied polarity of the readout gradient magnetic field is positive, and second reference data consisting of a second echo signal (negative signal) obtained when the applied polarity of the readout gradient magnetic field is negative. Using the first reference data and the second reference data, at least two pseudo-N / 2 ghost data with alternating positive and negative signals are generated. Using the pseudo-N / 2 ghost data, a correction kernel is calculated. Using the correction kernel, the formal scan k-space data collected by performing the formal scan is corrected. Finally, the corrected k-space data is used for image reconstruction.

[0019] -Invention Effects-

[0020] According to the present invention, only single-shot EPI data is used as reference data to calculate the correction kernel for correcting the formal scan data obtained through EPI, thereby reducing the extension of imaging time. Furthermore, by creating the correction kernel using pseudo-N / 2 ghost data created based on the reference data, and applying the correction kernel to the signal at the target point and the signals (including multiple signals, including positive and negative signals) on the surrounding odd and even rows, the correction processing for each odd and even row and subsequent synthesis processing can be eliminated, simplifying the processing. Moreover, since the signal at the target point itself is included in the correction, high-precision correction is possible. Attached Figure Description

[0021] Figure 1 This is a diagram showing an overall overview of an MRI device.

[0022] Figure 2 This is a functional block diagram of the processor.

[0023] Figure 3 This is a diagram illustrating the processing flow of Implementation Method 1.

[0024] Figure 4 This is a diagram illustrating the process of N / 2 ghosting correction.

[0025] Figure 5 This is a diagram representing a DWI sequence as an example of a formal scan.

[0026] Figure 6 It is a diagram representing the encoded portion of the formal scan and the k-space data obtained in the formal scan.

[0027] Figure 7 It is a diagram representing the encoded portion of the pre-scan and the k-space data obtained during the pre-scan.

[0028] Figure 8 This is a diagram illustrating the generation of pseudo-N / 2 ghosting data.

[0029] Figure 9 This diagram illustrates the calculations for the correction kernel that uses pseudo-N / 2 ghosting data.

[0030] Figure 10 This diagram illustrates the correction of formal scan k-space data using a correction kernel.

[0031] Figure 11 It is a diagram representing the various forms of the calibration kernel.

[0032] Figure 12 This is a diagram illustrating the processing flow of Implementation Method 2.

[0033] Figure 13 This is a diagram illustrating an example of the first pre-scan used in Implementation 2.

[0034] Figure 14 This is a diagram illustrating the 1D correction process.

[0035] Figure 15 This is a graph showing the correction effect of Implementation Method 1 performed at a rate of 1 times.

[0036] Figure 16 This is a graph showing the correction effect of Implementation Method 1 performed at a rate of 2 times.

[0037] Figure 17 This is a diagram showing the correction effect of Implementation Method 2.

[0038] Symbol Explanation

[0039] 10 - MRI device, 100 - camera unit, 200 - processor, 210 - measurement and control unit, 220 - N / 2 ghosting correction unit, 221 - pseudo-N / 2 ghosting data generation unit, 222 - correction kernel calculation unit, 223 - formal scan data correction unit, 224 - 1D linear phase correction unit, 230 - image reconstruction unit. Detailed Implementation

[0040] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0041] First, refer to Figure 1 An embodiment of the MRI apparatus to which the present invention is applied will be described. The MRI apparatus 10 is generally divided into an imaging unit 100 and a processor 200. The structure of the imaging unit 100 is the same as that of known MRI apparatuses, and detailed description is omitted, but it includes a static magnetic field generating device 102 that generates a uniform magnetic field in the imaging space where the subject 300 is placed, a gradient magnetic field coil 103 that imparts a gradient magnetic field to the static magnetic field, a shimming coil 104 that corrects the uniformity of the static magnetic field, a transmitting coil 105 that applies a high-frequency magnetic field to the atoms (atomic nuclei) of the tissue constituting the subject 300 to generate nuclear magnetic resonance, and a receiving coil 106 that detects the nuclear magnetic resonance signal generated from the subject 300.

[0042] The transmitting coil 105 is connected to a transmitter 107 equipped with a high-frequency transmitter or high-frequency amplifier, and the receiving coil 106 is connected to a receiver 108 equipped with a quadrature detector circuit or an A / D converter. In the case where the receiving coil 106 is a multi-channel receiving coil, each channel is connected to a separate receiver 108. The gradient magnetic field coil 103 and the shimming coil 104 are respectively connected to the gradient magnetic field power supply 112 and the shimming power supply 113.

[0043] The camera unit 100 also includes a sequence control circuit 114, which is connected to the transmitter 107, receiver 108, gradient magnetic field power supply 112, and shimming power supply 113, and controls their operation. The sequence control circuit 114 causes each part of the camera unit 100 to operate according to a predetermined pulse sequence. Thus, the camera unit 100 performs imaging. Furthermore, various pulse sequences are stored in storage devices described later. For example, the user selects a predetermined pulse sequence based on the purpose or object of the imaging, and sets imaging parameters such as echo time (TE), repetition time (TR), number of slices, and field of view (FOV), thereby determining the pulse sequence used in actual imaging. The selection of the pulse sequence or the setting of the imaging parameters are not performed by the user each time, but are sometimes determined in advance as a check protocol, and may also be set or changed through the control of the processor described later.

[0044] The processor 200 has the functions of performing image reconstruction and correction processing using the nuclear magnetic resonance signals collected by the camera unit 100, and controlling the operation of the entire device including the camera unit 100. It is connected to devices such as the display 201, the input device 202, and the external storage device 203. Furthermore, although not shown, it can also be connected to external databases, external processors, etc., via networks such as the Internet or an intra-facility network.

[0045] Processor 200 in Figure 1 A processor can be represented by a single box, but it can also consist of one or more hardware components, or a combination of hardware and programs. The type of hardware is not limited; for example, a processor can be composed of programmable logic devices such as CPUs (Central Processing Units), MPUs (Micro Processing Units), FPGAs (Field Programmable Gate Arrays), application-specific integrated circuits (ASICs) for performing specific processes, GPUs (Graphics Processing Units), or NPUs (Neural Processing Units). Furthermore, the type of hardware can be a combination of different types of hardware. When multiple hardware components are configured to perform one or more processes of a particular processor, these multiple hardware components can reside in physically separate devices or in the same device.

[0046] When the processor 200 is implemented by a combination of hardware and a program, the program can be software such as firmware or microcode. Furthermore, the program can be, for example, a group of program modules, each of which can be implemented by a processor configured to perform its respective function. The program can be program code or multiple code segments stored in one or more non-transitory computer-readable media (e.g., storage media or other storage devices).

[0047] In the MRI apparatus of this embodiment, the processor 200 has the following functions as a function: when the imaging unit 100 executes an EPI pulse sequence (hereinafter referred to as an EPI sequence), in addition to the formal scan which collects data used in image reconstruction of the subject 300, it also performs a pre-scan that uses a single excitation different from the same EPI sequence but with a different application method of the blip gradient magnetic field (control of the imaging unit); and uses the data obtained in the pre-scan (reference data) to create data including pseudo N / 2 ghosting (pseudo N / 2 ghosting data, hereinafter also referred to as simulated data), calculates a correction kernel to correct the formal scan data based on the simulated data, and uses the correction kernel to correct the formal scan data (formal scan data correction function).

[0048] Figure 2 The figure shows a functional block diagram of the processor 200 that implements these functions. As shown, the processor 200 includes a measurement control unit 210, an N / 2 ghosting correction unit 220, and an image reconstruction unit 230. The N / 2 ghosting correction unit 220 includes a pseudo N / 2 ghosting data generation unit 221, a correction kernel calculation unit 222, and a formal scan correction unit 223. Furthermore, the N / 2 ghosting correction unit 220 can also have the function of correcting N / 2 ghosting, etc., using methods other than correction methods using a correction kernel. Figure 2 As an example, a correction unit (1D linear phase correction unit 224) performing 1D correction is shown. However, this is not necessary.

[0049] Although Figure 2 Although not shown in the diagram, the processor 200 may also have display control functions for controlling the display 201, user interface functions for exchanging information with the input device 202 or external devices, and other functions.

[0050] Next, based on the above structure, the operation overview of the MRI device of this embodiment will be explained. Figure 3 and Figure 4 The action flow is shown in the diagram. For example... Figure 3 As shown, under the control of the processor 200, the camera unit 100 first performs a single-excitation pre-scan and collects reference data (S1001). Next, it performs a formal scan including an EPI sequence and collects formal scan data (S1002). The pre-scan includes the same single-excitation EPI sequence as the formal scan, but unlike the formal scan, a blip-shaped gradient magnetic field is applied every other EPI sequence.

[0051] Processor 200 uses the collected reference data to correct for N / 2 ghosting (S1003) and uses the corrected formal scan data to reconstruct the image (S1004). In the correction of N / 2 ghosting, such as... Figure 4As shown, the processor 200 performs the following steps: using reference data to create pseudo N / 2 ghost data (S1101), calculating the correction kernel (S1102), and correcting the formal scan data (S1103).

[0052] Regarding the creation of pseudo-N / 2 ghost data (S1101), the reference data is divided into two N / 2 ghost-free data sets (first reference data and second reference data) with different positive and negative values, and pseudo-N / 2 ghost data is created based on these N / 2 ghost-free data sets. Furthermore, a correction kernel is calculated (S1102) using the two pseudo-N / 2 ghost data sets, based on the signals of the measurement points of the object to be corrected and the signals of the measurement points on the even and odd rows surrounding it. Regarding the correction of the formal scan data (S1103), correction is performed by applying the correction kernel to multiple signals, including positive and negative signals (i.e., the signals of the measurement points of the object to be corrected and the signals of the measurement points on the even and odd rows surrounding it), and performing a convolution operation on the formal scan data.

[0053] According to the MRI apparatus of this embodiment, a correction kernel is calculated by using reference data obtained in a pre-scan of a single excitation to create N / 2 ghost-free positive electrode data, negative electrode data, and pseudo-N / 2 ghost-free data required for correction. This suppresses the increase in imaging time required to obtain the reference data. Furthermore, since a correction kernel based on multiple positive and negative electrode signals, including the signal to be corrected, is created based on these three data sets, unlike the case where positive and negative electrodes are corrected separately, subsequent synthesis or other processing is unnecessary, and high-precision correction is possible. Moreover, in the case of undersampling, processing can be performed in the same way as in the case of full sampling, thus maintaining the same correction effect when applicable to parallel imaging.

[0054] Based on the above overview, the specific implementation methods are described below.

[0055] <Implementation Method 1>

[0056] The N / 2 ghosting processing in this embodiment is characterized by performing a pre-scan by executing a single excitation of a blip-shaped gradient magnetic field every other interval in the EPI sequence to generate N / 2 ghost-free data with different positive and negative values, creating pseudo-N / 2 ghosting data based on the N / 2 ghost-free data, using the pseudo-N / 2 ghosting data to calculate a correction kernel based on the signal of the measurement point of the calibration object and the signals of the measurement points on the even and odd rows around it, and using the signal of the measurement point of the calibration object and the signals of the measurement points on the even and odd rows around it to correct the formal scan data through the correction kernel.

[0057] The following is a detailed description of each process. Please refer to the relevant sources as appropriate in the following explanations. Figures 1-4 .

[0058] <Formal Scan: S1002>

[0059] Before explaining the pre-scan, the formal scan, which is the object of calibration, and the formal scan data will be explained. Here, as an example, the case of DWI will be explained, but the present invention is not limited to DWI, and can be applied to any imaging that includes an EPI sequence.

[0060] like Figure 5 As shown, DWI applies an excitation RF pulse RF1 and a 180-degree RF pulse RF2, and applies high-intensity MPG pulses (MPGf1, MPGf2, MPGp1, MPGp2, MPGs1, MPGs2) before and after the 180-degree RF pulse RF2. Next, a vibrational readout gradient magnetic field and a blip-shaped phase-encoded gradient magnetic field are applied. During the application of the gradient magnetic field, almost the entire echo generated each time the readout gradient magnetic field reverses is sampled. Furthermore, in parallel imaging, this becomes k-space data that has been undersampled at a predetermined ratio according to the rate of change. Additionally, in... Figure 5 In the diagram, Gf, Gp, and Gs represent the axes of the frequency encoding direction (readout direction) gradient magnetic field, the phase encoding direction gradient magnetic field, and the slice direction gradient magnetic field, respectively, and AD represents the sampling time of the echo data.

[0061] Through this EPI sequence, in a single excitation (one-time stimulus), such as Figure 6 As shown on the right, the data for the entire k-space is collected in rows arranged in the ky direction, corresponding to the number of echoes. The row spacing Δky corresponds to the amount of a blip-shaped gradient magnetic field pulse applied. Figure 6 The left side indicates that it has been taken out. Figure 5 A diagram of the encoded portion of the EPI sequence. (See diagram below.) Figure 6 As shown, if a thick solid line represents the positive polarity of the readout gradient magnetic field and a thin solid line represents the negative polarity, then in the k-space data, the data for positive times are in odd-numbered rows, and the data for negative times are in even-numbered rows. In this case, the deviation at the echo peak position in the odd and even rows is the cause of the N / 2 ghosting.

[0062] In this embodiment, a pre-scan is performed before the formal scan, and the N / 2 ghosting is corrected using a correction kernel calculated based on reference data collected in the pre-scan.

[0063] <Pre-scan: S1001>

[0064] The pre-scan uses the same EPI sequence as the formal scan, including a vibratory readout gradient magnetic field and a blip-shaped phase-encoded gradient magnetic field. However, in the formal scan, a blip gradient magnetic field is applied each time the polarity of the readout gradient magnetic field reverses; in contrast, a blip gradient magnetic field is applied every other readout gradient magnetic field in the pre-scan. This assigns the same phase encoding to a pair of positive and negative data (odd echo and even echo). In the case of DWI with MPG applied in the formal scan, the phase deviation between the positive and negative poles does not change even if MPG is omitted in the pre-scan; therefore, it is preferable to omit MPG for a high SNR.

[0065] Figure 7 The image shows a portion of the pre-scanned pulse sequence (the coded portion) and the k-space configuration of the echo signals collected in that sequence. (See image from...) Figure 6 As shown in the comparison with the formal scan, in the pre-scan, a blip-shaped gradient magnetic field (phase-encoded gradient magnetic field) is applied only when the readout gradient magnetic field is negative (or only positive). The amount of the blip-shaped gradient magnetic field applied (equivalent to Δky) is the same as in the formal scan. Figure 7 As shown on the right, the data 600 collected through this pre-scan is encoded in the same phase, i.e., in the same row, to obtain data when the readout gradient magnetic field is positive (positive data 610) and data when it is negative (negative data 620). The row spacing Δky becomes the same as the row spacing of the formal scan.

[0066] Furthermore, although there are no restrictions, the FOV of the pre-scan is set to be the same as the number of echoes in the pre-scan and the formal scan. However, by making the FOVs of the pre-scan and the formal scan the same, a high-precision correction kernel can be calculated.

[0067] in addition, Figure 6 and Figure 7 This indicates the case where the formal scan and pre-scan are performed at a dual rate of 1, but it is also possible to perform them at a specified dual rate (>1). In this case, although there is no limitation, by performing the formal scan and pre-scan at the same rate and with the same sampling mode, it is possible to calculate the high-precision correction kernel described later.

[0068] <N / 2 Ghost Correction: S1003>

[0069] <Creation of pseudo-N / 2 ghosting data:> Figure 4 S1101>

[0070] The processor 200 (pseudo-N / 2 ghost data generation unit 221) generates pseudo-N / 2 ghost data (hereinafter, referred to as analog data) based on data obtained in the pre-scan. Therefore, as... Figure 8As shown, the reference data 600 is first divided into data consisting of odd-numbered rows (positive data 610) and data consisting of even-numbered rows (negative data 620). The odd-numbered and even-numbered rows of the positive data 610 and negative data 620 are then interchanged to generate simulated data with alternating positive and negative data configurations. Two sets of simulated data are generated: simulated data 631, which combines the odd-numbered rows of the positive data and the even-numbered rows of the negative data, and simulated data 632, which combines the odd-numbered rows of the negative data and the even-numbered rows of the positive data.

[0071] Thus, in this embodiment, data information for calculating the correction kernel can be created based on reference data from a single excitation. Therefore, compared with the method of acquiring reference data twice as described in Non-Patent Document 2, the extension of the imaging time can be suppressed, and the reduction in correction accuracy caused by the influence of body movement can be prevented.

[0072] <Calculation of the calibration kernel: S1102>

[0073] The correction kernel is a matrix of weights multiplied by each of the signals used to convert the target signal (hereinafter, the target signal) into the corrected signal. The corrected signal is obtained by convolving the source signal with the weights corresponding to the correction kernel.

[0074] In this embodiment, the calibration kernel calculation unit 222 calculates the calibration kernel using two analog data sets 631 and 632. Here, as an example, the case where the calibration kernel size is 3×3 will be described. That is, here, for the calibration target signal, the signals on both sides of its kx direction, and these three signals, a total of 9 data sets are used, three data sets each on the up and down directions (or the up and down positive directions when the calibration target signal is the negative line).

[0075] The following is for reference. Figure 9 The calculation of the correction kernel is explained. Figure 9 This shows the calculation of the correction kernel for each channel CH of the receiving coil.

[0076] like Figure 9 As shown, simulation data 631 and 632 are k-space data arranged alternately, with solid lines representing positive data (data with a right-facing trajectory) and dashed lines representing negative data (data with a left-facing trajectory). The rows corresponding to simulation data 631 and 632 in the ky direction are opposite in sign. The N / 2 correction in this embodiment uses... Figure 9 The negative pole signal of the calibration object, represented by "0", and the positive pole signal around it (hereinafter referred to as the source signal (S)) are represented by "0". l src(In this example, there are 9 target signals) are converted into the positive pole signal of the calibration object indicated by "●" (hereinafter referred to as the target signal (S)). j trg The transformation is performed using a matrix of weights for the target signal, which forms the correction kernel. Therefore, the correction kernel is calculated by weighting the signals as shown in the following equation, using the least squares method.

[0077] First, as described above, the odd-numbered rows of the analog data 631 are set as the target signal (S). j trg Points around the same coordinates of the simulated data 632 are set as the source signal (S). l src ), set the even-numbered rows of analog data 632 as the target signal (S) j trg Points around the same coordinates of the simulated data 631 are set as the source signal (S). l src If the kernel weights of each source signal are set to W... l Then the target signal (S) j trg The weight W (matrix representation) can be represented by the following equation (1). As shown in equation (4), the weight W is calculated by the least squares method. Here, the subscript H in equation (4) denotes the complex conjugate transpose of the matrix. Here, the receiving coil is multi-channel, and the correction kernel is calculated for each channel ch.

[0078] [Formula 1]

[0079]

[0080] In addition, the symbols in the formula are as follows.

[0081] [Formula 2]

[0082]

[0083] Therefore, for a set of simulation data 631 and 632, a correction kernel is calculated. For multiple channels, the correction kernel for each channel is calculated.

[0084] in addition, Figure 9 The diagram illustrates a case where the negative signal of the calibration object is converted (corrected) to a positive signal using the negative signal of the calibration object and its surrounding negative and positive signals. However, conversely, a positive signal can also be converted to a negative signal. In this case, the positive signal of the calibration object is converted to a negative signal using the positive signal of the calibration object and its surrounding positive and negative signals.

[0085] Furthermore, if the pre-scan and the formal scan are at the same rate and the same sampling mode (row extraction method), then only the row interval is different, and the correction kernel can be calculated in the same way as the data with a rate of 1, and can be used for the correction of the formal scan in the same way.

[0086] In existing methods, such as those in Patent Documents 1 and 2, the kernel of the estimated positive signal is calculated using the positive signal surrounding the negative signal of the calibration object. This is essentially the same method used to estimate the positive signal as in parallel imaging (GRAPPA), which easily leads to a decrease in SNR and a deterioration in the G factor. However, according to this embodiment, not only is the positive signal surrounding the negative signal of the calibration object used, but the negative signal of the calibration object itself is used to estimate the positive signal. Therefore, the calibration accuracy can be improved without a decrease in SNR, and the decrease in accuracy at high speeds in parallel imaging can be suppressed.

[0087] <Correction of formal scan data: S1103>

[0088] The formal scan includes the EPI sequence, similar to the pre-scan described above. It is not limited to this; for example, it could be... Figure 5 The DWI sequence shown can obtain results such as Figure 6 The official scan data is shown on the right.

[0089] like Figure 10 As shown, in the correction of the formal scan data, each signal constituting the formal scan data 500 is used as the target signal. As shown in the aforementioned equation (1), the correction kernel 640 calculated based on the reference data is used to perform convolution integration to obtain the corrected formal scan data 510. The corrected formal scan data 510 is converted into formal scan data where all signals are positive (or negative), and therefore does not contain N / 2 ghosting. Furthermore, by using the correction kernel to perform convolution operation, the higher-order components of the phase are also corrected, so as with 1D correction, there is no residual N / 2 ghosting, and the correction accuracy is high.

[0090] <Image Reconstruction: S1004>

[0091] Image reconstruction of the corrected formal scan data can be performed using known methods such as high-speed FT-based reconstruction, parallel imaging-based reconstruction, and reconstruction using successive computation.

[0092] According to this embodiment, since the data for calculating the correction kernel is created based on reference data from a single excitation, the extension of imaging time can be minimized. Furthermore, since the correction kernel is created using the positive and negative signals surrounding the negative signal of the target object, and the negative signal of the target object itself, the number of points used in the correction can be increased compared to using surrounding signals, thus improving correction accuracy. Moreover, according to this embodiment, correction with guaranteed accuracy can be performed regardless of the magnification rate of the actual scan data of the target object; even at high magnification rates, the correction effect will not decrease.

[0093] <Modifications of Implementation Method 1>

[0094] In Implementation 1, the case where the correction kernel size is 3×3 was described, but the kernel size is not limited to 3×3. The correction kernel can be calculated using the correction target signal and one or more adjacent signals. As long as the kernel size is 1×2 or 2×1 or larger, the dimensions in the ky direction and the kx direction can be the same or different. Furthermore, the shape of the matrix is ​​not limited to a square or rectangle, for example, as shown in the figure. Figure 11 As shown, a correction kernel of a diamond shape can also be set with the correction target signal as the center.

[0095] <Implementation Method 2>

[0096] In Implementation 1, a correction kernel is calculated based on reference data obtained in the pre-scan, and the N / 2 ghosting of the formal scan is corrected. However, it is also possible to perform 1D correction on the reference data and the formal scan data before using the correction kernel, and then calculate the correction kernel and perform correction on the formal scan data using the correction kernel.

[0097] The processing flow of this embodiment is shown below. Figure 12 In. Figure 12 In, with Figure 1 Processes that involve the same content are represented by the same symbols, and repeated descriptions are omitted. As shown in the figure, in this embodiment, the characteristic is that a pre-scan for 1D correction is performed (S2001), and 1D correction is performed using the data obtained in the pre-scan (S2002). For distinction, the pre-scan performed for 1D correction is referred to as the first pre-scan, and the pre-scan performed for calculating the correction kernel is referred to as the second pre-scan. 1D correction can be performed using the same method as described in Non-Patent Document 1. Hereinafter, a brief description will be given.

[0098] First, the measurement control unit 210 of the processor 200 executes each scan in the order of first pre-scan, second pre-scan, and final scan, and collects first reference data, second reference data, and final scan data. Similar to Embodiment 1, the second pre-scan and final scan are performed, for example, in the following manner... Figure 6 , Figure 7 The EPI sequence shown is used to scan the entire k-space. The first pre-scan is performed to perform linear phase correction of the data obtained in the second pre-scan and the formal scan, as shown... Figure 13 As shown, the scan is the same as the second pre-scan, except that no blip-shaped gradient magnetic field is applied. It is not limited to this, but the imaging conditions such as FOV are set to be substantially the same as those of the second pre-scan and the formal scan.

[0099] Next, 1D correction is performed using three data points measured by the 1D linear phase correction unit 224 (S2002). In the 1D correction, as... Figure 14 As shown, the data obtained through the first pre-scan is first converted into x-ky space data. The phase difference between the odd and even rows of the x-ky space data is calculated and linearly approximated (S2101). Next, using the linearly approximated phase difference, one-dimensional phase correction is performed on the second reference data and the formal scan data respectively (S2102, S2103). Regarding one-dimensional phase correction, the second reference data and the formal scan data are converted into x-ky space data, and the odd or even rows of data are corrected using a linearly approximated phase difference, aligning one with the other. Here, in the subsequent 2D correction (correction using a correction kernel), when the positive data is odd-row data and the negative data is converted to positive data, the process is set to correspond to the odd-row data. When the data used as the reference data in the 2D correction is even-row, the 1D correction also corresponds to it.

[0100] After performing the aforementioned 1D correction, the N / 2 ghosting correction unit 220 uses the second reference data to perform N / 2 ghosting correction on the formal scan data (S1003). Similar to Embodiment 1, the N / 2 ghosting correction uses the 1D-corrected second reference data (kx-ky data) to create simulated data, uses the simulated data to calculate the correction kernel, and uses the correction kernel to correct the 1D-corrected formal scan data (…). Figure 4 (S1101~S1103). Similar to Embodiment 1, the correction kernel can be any shape, such as a 3×3 square matrix, a 1×3, 2×3, 3×4 rectangular matrix, or a rhombus matrix. However, in this embodiment, since 1D phase correction is performed, high correction accuracy can be obtained even when using a correction kernel with a smaller matrix size.

[0101] exist Figure 12The illustrated embodiment shows a pre-scan without applying the blip-shaped gradient magnetic field used for 1D correction (first pre-scan), but it is also possible to use the linear phase obtained from the data of the second pre-scan without performing the first pre-scan for 1D correction. In this case, the positive and negative poles of the zeros in the phase encoding direction are converted to x-ky space using a one-dimensional Fourier transform, and the phase difference between them is linearly approximated. 1D correction is then performed on the formal scan data using this linear approximation, as described above.

[0102] Furthermore, similar to Embodiment 1, this embodiment is also applicable to cases where the rate of change is greater than 1.

[0103] As explained above, according to this embodiment, by performing linear phase correction on the reference data and actual scan data used in the correction kernel before N / 2 ghosting correction, correction accuracy can be maintained even with a small kernel size. By reducing the kernel size, computation time can be shortened.

[0104] <Effects of the Implementation Method>

[0105] DWI imaging was performed using a phantom (b-value=0, number of channels=6, number of slices=8), and corrections were performed based on Implementation Method 1 and Implementation Method 2, and the results were confirmed.

[0106] <Effects of Implementation Method 1>

[0107] Figure 15 This is a graph representing the absolute value image captured at a rate of 1x (number of channels = 6, slice = 1, phase encoding direction = AP direction, b value = 0). From left to right, it shows (a) the image of the first echo acquired when the polarity of the vibration readout gradient magnetic field is positive (forward image), (b) the image of the first echo acquired when the polarity of the vibration readout gradient magnetic field is reversed (a) and the polarity of the vibration readout gradient magnetic field is negative (reverse image), (c) the complex average of the forward and reverse images, and (d) the complex difference between the forward and reverse images. Furthermore, from top to bottom, it shows the images before correction, after 1D correction, and after correction using the method of Embodiment 1. From this graph, it can be confirmed that by eliminating the positive and negative phase difference using the method of Embodiment 1, even data from a single excitation can sufficiently achieve an N / 2 ghosting reduction effect.

[0108] Figure 16 It means in relation to Figure 15 The images were taken under the same photographic conditions, with the magnification set to 2. The comparisons are between the images before correction, after 1D correction, and after correction according to Implementation Method 1. Figure 15Similarly, from left to right, are shown (a) the forward-rotated image, (b) the reverse-rotated image, (c) the complex average of the forward-rotated and reverse-rotated images, and (d) the complex difference of the forward-rotated and reverse-rotated images. This figure confirms that sufficient correction is achieved even when the magnification rate is set to 2x.

[0109] <Effects of Implementation Method 2>

[0110] The image obtained by squaring and taking the square root of the image for each channel is shown below. Figure 17 middle. Figure 17 The left side shows the forward-rotated image, and the right side shows the reversed image. Furthermore, Figure 17 The images shown from top to bottom are the results before correction, after 1D correction, and after correction based on the method of Embodiment 2. As shown, the N / 2 ghosting that appeared in the reconstructed image without correction almost disappeared after 1D correction, but a small amount remained at the top and bottom. In contrast, it was confirmed that when the correction of this embodiment was performed, the residual ghosting disappeared, and a high N / 2 ghosting correction effect was obtained.

[0111] The various embodiments of the present invention have been described above. However, the present invention is not limited to the above embodiments or their variations. As long as it does not contradict the technology, it is also possible to add known structures or correction processes, and it is also possible to recombine the structure of the processor or add other functions. These added and modified devices or methods are also included in the present invention.

Claims

1. A magnetic resonance imaging device, characterized in that, It includes: an imaging unit that measures nuclear magnetic resonance signals generated from an object under examination according to a prescribed pulse sequence; and a processor that performs calculations including image reconstruction using the nuclear magnetic resonance signals and control the operation of the imaging unit. The processor performs the following processing: As a formal scan, a pulse sequence based on echo-plane imaging is performed. After applying an excitation RF pulse, a readout gradient magnetic field with polarity oscillation between positive and negative and a blip-shaped phase-encoded gradient magnetic field are applied. The echo signal is measured each time the polarity of the applied readout gradient magnetic field is reversed in order to collect k-space data. As a pre-scan, a pre-scan sequence based on echo-plane imaging is performed, wherein the application of the blip-shaped phase-encoded gradient magnetic field in the formal scan is changed to occur once every two reversals of the polarity of the readout gradient magnetic field. The reference k-space data collected in the pre-scan is divided into a first reference data consisting of a positive signal obtained when the applied polarity of the readout gradient magnetic field is positive, and a second reference data consisting of a negative signal obtained when the applied polarity of the readout gradient magnetic field is negative. At least two pseudo-N / 2 ghosting data are generated using the first reference data and the second reference data, with the positive and negative signals alternately configured. The correction kernel for correcting N / 2 ghosting is calculated using the pseudo-N / 2 ghosting data; and The calibration kernel is used to calibrate the formal scan k-space data collected by performing the formal scan, and the calibrated k-space data is used for image reconstruction.

2. The magnetic resonance imaging device according to claim 1, characterized in that, The processor performs the following processing: The calibration kernel is a calibration kernel that calculates and corrects the signal of the calibration object using at least one of the positive and negative signals around the calibration object. The formal scan k-space data is corrected using the calibration kernel and at least one of the positive and negative signals around the calibration object.

3. The magnetic resonance imaging device according to claim 2, characterized in that, The processor performs the following processing: The calibration kernel is calculated as follows: using the signal of the calibration object and at least one of the positive and negative signals around it, if the signal of the calibration object is a negative signal, it is converted to a positive signal, and if the signal of the calibration object is a positive signal, it is converted to a negative signal.

4. The magnetic resonance imaging device according to claim 1, characterized in that, The processor performs the following processing: The FOV of the pre-scanned k-space data is controlled to be the same as the FOV of the formally scanned k-space data.

5. The magnetic resonance imaging device according to claim 1, characterized in that, The size of the correction kernel is N×M. N and M are any integers greater than or equal to 1, except where N = M = 1.

6. The magnetic resonance imaging device according to claim 1, characterized in that, The processor performs the following processing: The phase difference between the positive and negative echo signals is made linearly approximate. After the phase difference between the positive and negative signals of the formal scan k-space data is corrected using the linearly approximate phase difference, the formal scan k-space data is corrected using the correction kernel.

7. The magnetic resonance imaging device according to claim 1, characterized in that, The camera unit has one or more receiving coils for receiving nuclear magnetic resonance signals, and collects k-space data for each receiving coil. The processor calculates the correction kernel for each receiving coil and corrects the formal scan k-space data for each receiving coil.

8. The magnetic resonance imaging device according to claim 1, characterized in that, The processor performs the following processing: Perform a formal scan to collect the k-space data at a rate greater than 1, and control the pre-scan to perform the scan at the same rate as the formal scan. After correcting the formal scan k-space data using the correction kernel, image reconstruction is performed with interpolation of unmeasured data.

9. The magnetic resonance imaging device according to claim 1, wherein, The processor performs the following control: As the specified pulse sequence, a diffusion-weighted sequence including the application of MPG pulses is executed before the pulse sequence based on the echo-plane imaging method.

10. A control method for a magnetic resonance imaging (MRI) device, the MRI device comprising: The imaging unit measures the nuclear magnetic resonance signals generated from the object being examined according to a prescribed pulse sequence; The magnetic resonance imaging device control method, which includes a computing unit and an image reconstruction unit using the nuclear magnetic resonance signal, is characterized in that... The camera unit is controlled to execute a formal scan sequence and a pre-scan sequence based on echo-plane imaging. In the formal scan sequence based on echo-plane imaging, after applying an excitation RF pulse, a readout gradient magnetic field with polarity oscillation between positive and negative and a blip-shaped phase-encoding gradient magnetic field are applied. Each time the polarity of the applied readout gradient magnetic field reverses, the echo signal is measured to collect k-space data. The pre-scan sequence based on echo-plane imaging changes the application of the blip-shaped phase-encoding gradient magnetic field in the formal scan sequence to a single application where the polarity of the readout gradient magnetic field reverses twice. The arithmetic unit is controlled to divide the reference k-space data collected in the pre-scan into first reference data consisting of a first echo signal obtained when the applied polarity of the readout gradient magnetic field is positive, and second reference data consisting of a second echo signal obtained when the applied polarity of the readout gradient magnetic field is negative. At least two pseudo-N / 2 ghost data are generated using the first reference data and the second reference data, which alternately configure the first echo signal and the second echo signal; The correction kernel for correcting the N / 2 artifacts contained in the pseudo-N / 2 ghost data is calculated using the pseudo-N / 2 ghost data. and The calibration kernel is used to calibrate the formal scan k-space data collected by performing the formal scan, and the calibrated k-space data is used for image reconstruction.

Images