Method for obtaining an MR image dataset of at least two slices

Abstract

The present invention relates to a method for acquiring magnetic resonance (MR) image datasets of at least two slices (S1, S2) by simultaneous multi-slice excitation. The method comprises: a) performing an MR imaging sequence using a multi-band radio frequency (RF) excitation pulse to simultaneously excite at least two slices (S1, S2) in at least two repetitions (Rep1, Rep2), wherein the repetitions are performed according to a phase modulation scheme, wherein each of the simultaneously excited slices is assigned a phase, and the phase of at least one of the simultaneously excited slices varies from one repetition to the next, thereby acquiring MR datasets of folded images (C1, C2) in each repetition; b) performing spatial registration (16) between at least two folded images (C1, C2); c) performing motion correction (18) on at least one MR dataset of the folded images (C1, C2); and d) reconstructing (20) MR images of at least two slices from the corrected MR datasets of the folded images.

Description

Field of the Invention

[0001] The present invention relates to a method for acquiring magnetic resonance (MR) image datasets of at least two slices by simultaneous multi-slice excitation, a magnetic resonance device, and a non-transitory computer-readable data storage medium. Background Art

[0002] Magnetic resonance imaging (MRI) is an important imaging modality in modern medicine and biology. However, its main drawbacks are the relatively long scan time required for spatially encoding signals and the relatively low signal-to-noise ratio, especially for low-field systems operating at <1T, such as 0.5T. Hadamard-encoded simultaneous multi-slice (SMS) imaging was proposed by Souza et al. (Souza SP, Szumowski J.: "SIMA: Simultaneous multislice acquisition of MR images by Hadamard-encoded excitation" J. Comput. Assist. Tomogr. 12(6):1026-1030 (1988)), and is currently experiencing a renaissance in low-field MRI systems. SMS imaging typically requires slice-selective excitation of two or more slices simultaneously, which can be achieved by multi-band radiofrequency (RF) excitation pulses. The superimposed signals from two or more slices can be disentangled by phase manipulation of the signals. When using a phase modulation scheme such as Hadamard encoding, the imaging sequence must be repeated N times in order to disentangle the signals from N slices, but each excited slice has a different phase pattern. As described below, the use of such addition / subtraction makes the imaging method vulnerable to motion because the signal for each slice is obtained from several repetitions. However, compared to the recently introduced slice-GRAPPA-based simultaneous multi-slice imaging method (Setsompop K., Gagoski B.A. et al.: "Blipped-Controlled Aliasing in Parallel Imaging for Simultaneous Multislice Echo Planar Imaging with Reduced g-factor Penalty" Magnetic Resonance in Medicine 67:1210-1224 (2012)), Hadamard-encoded SMS imaging can be performed using a single-channel RF coil, while the slice-GRAPPA method exploits the different sensitivity profiles of the individual coils in a multi-channel RF coil.

[0003] Accordingly, an object of the present invention is to provide a multi-slice imaging method that is robust and provides improved image quality even in a low-field MRI system and under difficult imaging conditions. Summary of the Invention

[0004] This object and other objects are achieved by a method for acquiring MR image datasets of at least two slices by simultaneous multi-slice excitation according to claim 1. The method comprises the following steps:

[0005] a) Performing an MR imaging sequence using a multi-band radiofrequency (RF) excitation pulse to simultaneously excite at least two slices, wherein the MR imaging sequence comprises at least two repetitions, wherein the repetitions are performed according to a phase modulation scheme, wherein each of the slices simultaneously excited in each repetition is assigned a phase, and the phase of at least one of the slices simultaneously excited changes from one repetition to the next, whereby an MR dataset of a folded image is acquired in each repetition;

[0006] b) Performing spatial registration between at least two folded images so as to obtain translation and / or rotation correction parameters;

[0007] c) Performing motion correction on at least one MR dataset of the folded images based on the correction parameters so as to obtain a corrected MR dataset; and

[0008] d) Reconstructing MR images of at least two slices from the corrected MR datasets of the folded images.

[0009] The present invention also relates to an MR device adapted to perform the method, and a computer-readable data storage medium encoded with programming instructions. Any feature described with respect to the method also applies to the MR device and the data storage medium, and vice versa.

[0010] The present invention proposes to incorporate motion correction into a simultaneous multi-slice (SMS) imaging method. Such SMS imaging relies on the repeated execution of an MR imaging sequence, using a multi-band RF excitation pulse to simultaneously excite at least two slices. The repetitions are performed according to a phase modulation scheme, wherein each of the slices simultaneously excited in each repetition is assigned a phase, and the phase of at least one slice changes from one repetition to the next. In each repetition, an MR dataset comprising superimposed signals from two or more slices is acquired, wherein the corresponding MR dataset (in image space or k-space) is referred to as a "folded image" because the image data from two or more slices overlap each other. During the SMS reconstruction of the MR images of the individual slices, signals from the individual slices are separated from each other with the motion-corrected MR datasets of the folded images using, for example, parallel acquisition reconstruction techniques, also known as slice multiplexing.

[0011] The idea of the present invention is to perform spatial registration of the folded slices between each required repetition before SMS reconstruction. The method is shown for two repetitions and two simultaneously acquired slices, but is clearly also applicable to three or more repetitions / slices. Although the method is also applicable to overlapping slices, two or more simultaneously excited slices preferably do not overlap. The imaging method may include multiple iterations of method steps a to d, preferably with additional simultaneously excited slices in order to cover the entire volume to be imaged.

[0012] A multi-band RF pulse is any pulse used to simultaneously excite or otherwise manipulate (e.g., to refocus or saturate) two or more slices. Such a multi-band RF pulse may be a multiplexing (superposition) of individual RF pulses, which would be used to individually manipulate a single slice. In order to be able to separate the signals obtained from the individual slices, a phase is assigned to each simultaneously excited slice. The phase can be assigned, for example, by manipulating the phase of the multi-band RF excitation pulse used, in particular the phases of the individual RF pulses that make them up, or by an additional gradient to be switched. In a preferred embodiment, the multi-band RF pulse used in each repetition is multiplexed from an additive pulse form of such individual RF pulses according to a phase modulation scheme. Preferably, the phase modulation scheme is predetermined for each SMS imaging session and is determined, for example, by the total number of slices to be imaged and the number of slices to be imaged per repetition.

[0013] Spatial encoding of the acquired MR signals can be achieved by standard gradient switching in two directions, e.g., in the read direction and the phase encoding direction (two-dimensional gradient encoding). A single-channel or multi-channel RF coil can be used to acquire the resulting MR signals so that the signals from all excited slices are folded into a single k-space dataset.

[0014] Spatial registration between the folded images is performed to obtain at least transformation and / or rotation correction parameters in order to correct for translation or rotation of the object to be imaged that occurs between the repetitions required for SMS imaging. Spatial registration and motion correction can be performed by the method described in the article by M.V. Wyawahare: "Image Registration Technique: An overview", International Journal of Signal Processing, Image Processing and Pattern Recognition, Vol. 2, No. 3, September 2009.

[0015] Generally, spatial registration and motion correction may include the following steps:

[0016] - Feature detection, where unique image features, such as closed boundary regions, edges, contours, line intersections, corners, etc., are detected in the images to register them with each other.

[0017] - Feature matching: Establishing correspondences between features in two or more images.

[0018] - Transformation model estimation: Estimating the type and parameters of a so-called mapping function that aligns one image with another (e.g., a sensed image with a reference image). These mapping functions are also referred to as calibration parameters in this document, especially translation and / or rotation calibration parameters.

[0019] - Image resampling and transformation: One of the images (e.g., the sensed image) is transformed by means of the mapping function, thereby performing the actual motion correction and obtaining a set of corrected images.

[0020] Typically, this requires image resampling, where pixel values are resampled, e.g., by interpolation to new pixel positions.

[0021] According to an embodiment of the present invention, the spatial registration is performed by means of a registration method that is insensitive to contrast changes, especially a mutual information-based method, as described in the article by Wyawahare. The mutual information-based registration starts with the estimation of the joint probability of the corresponding voxel intensities in two images. The mutual information can be used to parameterize and solve the correspondence problem in feature-based registration. The advantage of this method is that it is insensitive to image contrast, which may vary strongly between folded MR data sets due to different pulse superpositions. Appropriate optimization methods using gradients or other methods can be used to maximize the mutual information. A review of mutual information-based image registration methods can be found in J. Pluim et al., "Mutual information-based registration of medical images: A survey" IEEE Transactions on Medical Imaging, Vol. 22(8), 986–1004 (2003).

[0022] According to a preferred embodiment, the registration of the folded images with each other is performed in the image space (instead of in the k-space). Therefore, it is preferred to first reconstruct the MR data sets of the folded images, where the reconstruction includes a Fourier transform to obtain the images from the signals acquired in the k-space. Further steps, namely performing the motion correction and reconstructing the individual images, can be performed in the k-space or the image space, as described below.

[0023] According to one embodiment, motion correction includes translation and / or rotation in the image space. Thus, the translation and / or rotation correction parameters obtained in spatial registration may include a translation vector and / or a rotation transformation. Alternatively, motion correction may also be performed in k-space, where the translation transformation corresponds to a multiplication with a phase ramp, and the rotation in the image space corresponds to a rotation in k-space.

[0024] According to one embodiment, motion correction includes rigid-in plane motion correction within the plane of one or more of at least two slices. The most straightforward variant is to correct only for rigid-in plane motion, and the simplest variant includes only translation. As described above, these effects can be corrected by translation / rotation in the image space or multiplication / rotation with a phase ramp in k-space. Since important applications of SMS (especially Hadamard) imaging are brain or joint imaging, this assumption is usually reasonable.

[0025] In another embodiment, motion correction may also include elastic motion correction. In this case, the correction parameters include a motion vector field. According to a more advanced embodiment, motion / rotation through the plane may also be considered. This can be performed by registering folded images obtained from different slice pairs and performing signal interpolation between these folded images by, for example, cubic spline interpolation in the image space or applying a Fourier transform along the slice direction, adding a linear phase ramp, and applying an inverse Fourier transform to obtain intermediate slices. Methods based on convolutional neural networks, as described by Wu et al. (https: / / arxiv.org / pdf / 2001.11698.pdf), may also be used.

[0026] Motion correction through the plane is particularly feasible when the motion between different slices is not independent, but includes, for example, dependent motion through the slices, such as dilation and / or contraction of the object (e.g., possibly caused by breathing). Thus, in certain cases where the motion vector fields in two simultaneously acquired slices are similar, motion correction, even through the plane and / or elastic motion correction, is possible.

[0027] According to one embodiment, the multi-band radio frequency (RF) excitation pulses each include first and second single-band pulse shapes, where at least one of the single-band pulse shapes is phase-shifted between one repetition and the next. The phase shift is determined by a phase modulation scheme.

[0028] Alternatively, the phase assigned to simultaneously excited slices varies from one repetition to the next by switching an additional gradient in the slice direction, as described, for example, in the paper by Setsompop et al. above. Thus, the phase assigned to simultaneously excited slices can be imparted either by RF excitation or by gradients or gradient blips switched in the slice direction to add the desired phase from each simultaneously excited slice to the spins. The manner in which the phase is implemented in two or more simultaneously excited slices may be independent of the phase modulation scheme used.

[0029] According to one embodiment, as described by S.B. Souza et al., the phase modulation scheme uses Hadamard coding. In this case, the phase of each slice excitation frequency is modulated in a binary pattern, for example, given by a Hadamard matrix whose size is equal to the number of slices. All repetitions are used to reconstruct each slice. This technique is particularly well-suited for a moderate number of slices, such as 2 - 16. The Hadamard matrix is its own inverse matrix, so a Hadamard transform can be defined for any positive integer of order N. For example, in the case of N = 2, the Hadamard matrix is

[0030]

[0031] Thus, the addition of the folded images obtained in two repetitions will yield an image of the first slice. The subtraction of the two folded images will yield an image of the second slice. This principle can be extended to a larger number of slices. For example, when N = 4, the corresponding excitation matrix is

[0032]

[0033] Although this process is closely related to the Fourier transform, a distinction must be made between this technique and existing 3D - FT (three - dimensional Fourier transform) methods. In Hadamard coding, spatial encoding of the third dimension (slice direction) is achieved by phase - modulating the excitation envelope in discrete steps of π, rather than by adding phase - encoding magnetic field gradient pulses. Thus, the slices do not have to be equidistant. Instead, arbitrary slice placement is achieved by appropriate selection of the phase - modulation pattern.

[0034] However, the present invention can also be applied to other SMS averaging methods, such as those described in US 10,557,903B2. This slice multiplexing method follows a similar approach that utilizes multiple repetitions with different CAIPIRINHA blip patterns to obtain a fully sampled three-dimensional (3D) k-space for direct Fourier reconstruction or for calibration of slice or in-plane GRAPPA kernels without a separate reference scan. Thus, the method of the present invention can also be applied to slice multiplexing methods where the k-space in the slice direction is undersampled. This requires at least two RF receive coils, preferably with different sensitivity profiles in the slice direction. In fact, the method of the present invention is applicable to any slice multiplexing method where multiple N slices are acquired simultaneously using M repetitions, where N≥2 and M≥2, but where N does not have to be equal to M. For Hadamard encoding, N = M, but when techniques such as slice-GRAPPA or other parallel imaging methods are used, the number of slices N may also be higher than the number of repetitions M.

[0035] The MR imaging sequence for obtaining the folded image can be any MR imaging sequence, such as a spin echo or turbo spin echo sequence, or a gradient echo sequence, such as steady state free precession (SSFP), balanced steady state free precession (bSSFP) or spoiled gradient echo sequence, such as FLASH (Fast Low Angle SHot). It can also be an echo planar imaging (EPI), inversion recovery or diffusion weighted imaging sequence. Generally, acquiring the MR data set of the folded image in each repetition will be performed by sampling the k-space, for example, by Cartesian sampling where the k-space is sampled in multiple lines. While collecting the signal, a magnetic field gradient is applied along the frequency encoding direction. To sample multiple lines, an additional phase encoding gradient is briefly applied along a direction perpendicular to the frequency encoding direction, thereby imparting a position-dependent phase. By repeating the process of RF excitation, phase encoding, and frequency encoding multiple times and gradually performing different phase encoding gradient values, a two-dimensional (2D) image (in the k-space) can be formed.

[0036] The required spatial resolution and field of view (FOV) determine how much k-space data should be acquired. The spacing between adjacent k-space lines is inversely proportional to the FOV:

[0037] FOV = 1 / Δk (3)

[0038] To increase the FOV in one direction, the spacing between the sampled k-space points must be decreased. If the FOV in the phase encoding direction is smaller than the object to be imaged, the object will fold in, and this effect is called aliasing. This is related to the Nyquist sampling theorem, according to which if the sampling frequency is too low, high-frequency signals will be incorrectly displayed as low-frequency signals.

[0039] The spatial resolution is inversely proportional to the distance between the origin and the maximum extent of the k-space (k max ).

[0040] Δx = 1 / 2k max (4)

[0041] Therefore, in order to improve the spatial resolution, it is necessary to sample the k-space points farther from the origin.

[0042] Therefore, the scan time in magnetic resonance imaging can be reduced by sampling a smaller number of phase-encoding lines in the k-space; however, if no further processing is performed, the quality of the resulting image will be degraded due to aliasing artifacts. Most MR scanners use multi-channel RF coils, i.e., the RF coil consists of an array of multiple independent receiver coils. Since these multiple coils have different sensitivity curves, it is possible to utilize this property of the coil array to separate the aliased pixels in the image domain, or use the knowledge of the k-space points acquired nearby to estimate the missing k-space data. These methods are generally referred to as "parallel imaging", as described, for example, in J. Hamilton et al., "Recent Advances in Parallel Imaging for MRI", Prog. Nucl. Magn. Reson. Spectrosc; 101:71-95 (2017). If the phase-encoding lines are skipped at regular intervals, the undersampling in the phase-encoding direction will reduce the effective FOV, resulting in coherent aliasing artifacts, where replicas of the object will appear at regular intervals in the reduced FOV image. The amount of undersampling is described by the acceleration factor R, which is defined as the ratio between the number of k-space points in the fully sampled data and the undersampled data. The acceleration factor of R results in R image replicas along the phase-encoding direction spaced apart from each other by a distance of FOV / R.

[0043] In this article, the parallel imaging technique is also referred to as an "in-plane acceleration method". According to an embodiment of the present invention, the MR imaging sequence uses an in-plane parallel imaging technique, particularly in-plane GRAPPA or in-plane SENSE. In SENSE (Sensitivity Encoding), the aliased pixels are separated in the image domain, while in GRAPPA (Generalized Autocalibrating Partially Parallel Acquisition), the missing phase-encoding lines are reconstructed in the k-space.

[0044] According to one embodiment, when using parallel imaging techniques, spatial registration is performed on folded images that include image replicas caused by aliasing. This embodiment is particularly useful if the motion to be corrected mainly consists of in-plane translations, since translations are also visible on the aliased images. Thus, method steps b and c are performed on the MR data set of the folded images reconstructed from the uncorrected MR data set (i.e., the folded images reconstructed from the incomplete k-space). However, when reconstructing individual slice images from the motion-corrected MR data set of the folded images, the missing k-space data points (referred to as target points) are synthesized as a linear combination of the acquired adjacent k-space points, referred to as source points. The spatial arrangement of the source points and the target points is referred to as the GRAPPA kernel. Each acquired source point is multiplied by a coefficient or GRAPPA weight, and the results are then added together to estimate the target point. The source points from all the other coils are used to reconstruct a single target point for one coil. For Cartesian acquisitions, the weights are invariant for first-order approximation, and thus the same GRAPPA weights can be applied throughout the k-space. Thus, in many GRAPPA techniques, GRAPPA requires additional data to estimate the set of GRAPPA weights. In most embodiments, GRAPPA is considered to be auto-calibrated because a number of additional phase-encoding lines, referred to as auto-calibration signals, are collected near the origin of the k-space for calculating the weights. Then, the set of GRAPPA weights can be determined and applied throughout the k-space. The synthesis of the missing k-space data points can be performed before or after adding / subtracting the MR data set of the folded images to unravel the signals related to the individual slices.

[0045] According to another embodiment, when using in-plane parallel imaging, especially methods that include undersampling in the phase-encoding direction, the in-plane field of view of the MR image data set in the phase-encoding direction is larger than the object to be imaged. This is usually the case, and it can be easily arranged if the object being imaged is, for example, the head or limbs, which are imaged in substantially axial slices. In this case, the method preferably comprises the steps of: identifying segments of the field of view in the phase-encoding direction that do not include overlapping image replicas, performing spatial registration between the identified segments of the field of view between at least two folded images, thereby obtaining translation and / or rotation correction parameters. This embodiment is particularly advantageous if valid rotation correction parameters must be obtained, i.e., when rotational motion may occur between multiple repetitions. In this case, FOV segments or subsets are defined that are limited in their spatial extent along the phase-encoding direction. They are selected so that they do not include overlapping image replicas, i.e., those segments are not affected by aliasing. Such segments will exist if the object is at least somewhat smaller than the field of view extension in the phase-encoding direction.

[0046] In an alternative, segments are automatically determined based on the folded images. For example, a segmentation algorithm may identify the outer perimeter of the imaged object for each image copy and thereby determine the amount of overlap caused by aliasing. According to another embodiment, segments of the FOV may be predetermined. For example, at an acceleration factor of 2, the segments may be predetermined as (1) a segment / strip at the center of the FOV of a predetermined wit w, and (2) a segment of wit w / 2 at one end of the FOV, and (3) a segment of wit w / 2 at the other end of the FOV.

[0047] According to the invention, motion correction can be performed by spatial registration of the folded slices prior to Hadamard / SMS imaging reconstruction. This results in improved image quality, reduced artifact corruption of the final MR image, and thus reduced rescan, which may otherwise be required when the patient moves during image acquisition.

[0048] The invention further relates to a magnetic resonance (MR) device, comprising

[0049] a) an MR scanner adapted to acquire an MR data set from an object placed inside the MR scanner, and

[0050] b) a computer configured to send control signals to the MR scanner to cause the MR scanner to perform the method according to any one of the preceding claims. The MR scanner can be any commercially available MR scanner, in particular a low-field scanner. The MR scanner includes all common devices, in particular a main magnet, gradient coils, and an RF coil for radiating RF excitation pulses and receiving MR signals. The RF coil may include a coil array. The MR scanner is connected to a computer configured to send control signals. The computer can be part of a console through which the MR scanner can be controlled. The computer can be any computing device, such as a laptop computer, PC, workstation, cloud computer, or mobile device.

[0051] The invention further relates to a computer program product comprising program instructions adapted to be loaded into a computer of an MS device including an MR scanner, wherein the program instructions cause the computer to perform the method of the invention.

[0052] According to another aspect of the invention, a non-transitory computer-readable data storage medium encoded with programming instructions adapted to be loaded into a computer of a magnetic resonance (MR) device including an MR scanner, the programming instructions causing the computer to perform the method of the invention together with the MR scanner by sending control signals to the MR scanner and receiving data from the MR scanner. The storage medium can be in the cloud or can be any digital data storage medium, such as a CD-ROM, hard disk, SD card, SSD card, USB card, etc. Description of the Drawings

[0053] Embodiments of the present invention will now be described with reference to the drawings, wherein:

[0054] Figure 1 is a schematic diagram of a magnetic resonance device according to an embodiment of the present invention;

[0055] Figure 2 is a schematic diagram of simultaneously acquiring two slices using the Hadamard method;

[0056] Figure 3 is a schematic diagram of reconstructing a single slice using the Hadamard method;

[0057] Figure 4 is a flowchart of an embodiment of the method of the present invention;

[0058] Figure 5 is a schematic diagram of an embodiment of the method of the present invention using Hadamard encoding in combination with in-plane GRAPPA with a GRAPPA factor of 2;

[0059] Figure 6 is a schematic diagram of multiple slices imaged using an embodiment of the method of the present invention. Detailed Description of the Invention

[0060] Figure 1 The magnetic resonance (MR) device 1 of the present invention is schematically shown. The MR device 1 has an MR data acquisition scanner 2, which has a basic field magnet 3 for generating a constant magnetic field, a gradient coil device 5 for generating a gradient field, a radio frequency antenna 7 for radiating and receiving radio frequency signals, and a control computer 9 configured to execute the method of the present invention. In Figure 1 only such subunits of the magnetic resonance device 1 are schematically outlined. The radio frequency antenna 7 may consist of multiple subunits, in particular at least two coils, such as the coils 7.1 and 7.2 schematically shown, which may be configured to only transmit radio frequency signals or only receive trigger radio frequency signals (MR signals), or both.

[0061] In order to acquire MR data from an examination object U, such as a patient or a phantom, the examination object U is introduced into the measurement space of the scanner on a table L. Slices S1 and S2 are examples of two different slices of the examination object from which MR data can be acquired simultaneously. The control computer 9 centrally controls the magnetic resonance device and can control the gradient coil device 5 via the gradient controller 5' and the radio frequency transmit / receive controller 7' to control the radio frequency antenna 7. The radio frequency antenna 7 has a plurality of channels in which signals can be transmitted or received. The radio frequency antenna 7 and its radio frequency transmit / receive controller 7' are responsible for generating and radiating (transmitting) a radio frequency alternating field to manipulate the nuclear spins in the region to be examined of the examination object U (especially in different slices S1 and S2). The center frequency of the radio frequency alternating field, also known as the B1 field, should be close to the resonance frequency of the nuclear spins to be manipulated here. To generate the B1 field, a current controlled by the radio frequency transmit / receive controller 7' is applied to the RF coil in the radio frequency antenna 7. The control computer 9 also has a phase determination processor 15 which determines the phase φ1 to be additionally assigned according to the present invention. The calculation processor 13 of the control computer 9 is configured to perform the required measurements and all the calculation operations required. Intermediate results and final results obtained for this purpose or determined during the process can be stored in the memory 17 of the control computer 9. The units shown here do not necessarily have to be regarded as physically separate units, but only represent a subdivision into functional units which can also be implemented with fewer physical units or only one physical unit. A user can input control commands into the magnetic resonance device 1 via the input / output interface E / A and / or view the displayed results, such as image data, from the control computer 9. A non-transitory data storage medium 26 can be loaded into the control computer 9 and encoded with programming instructions (program code) for the control computer 9 and its various functional units described above to implement the method according to any or all embodiments of the present invention, as described above.

[0062] Figure 2Shows the Hadamard encoding of two slices S1 and S2. This encoding requires repetitions, Rep1 and Rrep2. In the first repetition, the single-band pulse shapes of slices 1 and 2 (S1, S2) are added. In the second repetition rep2, they are subtracted from each other, as shown in 14. Subtracting from each other is achieved by a 180° phase shift of the excitation of the second slice, for example by assigning an appropriate phase shift to the single-band pulse shape of S2 in the second repetition Rep2. Imaging in each repetition can be performed by any MR imaging sequence, such as a Turbo Spin-Echo sequence. This sequence can include in-plane acceleration techniques, such as in-plane GRAPPA (also known as parallel imaging technique). Thus, the result of each repetition is a folded slice, i.e., an MR data set related to a 2D image, which includes image data from each slice S1, S2. In the first repetition, a folded image C1 is acquired, and in the second repetition, a folded image C2 is acquired. It can be seen that C1 is the superposition of the two slices S1 and S2, while C2 is an image in which the signal intensity of S2 has been subtracted from the signal intensity of S1 for each pixel. The MR data sets related to C1 and C2 are available in k-space, but of course can be transferred to image space (as Figure 2 shown). Figure 3 Shows how to reconstruct the images of the individual slices S1, S2 from the folded images. In particular, S1 can be restored by adding the two folded images C1 and C2, while the second slice S2 can be restored by subtracting C2 from C1. Note that this results in doubling of the signal intensity and thus an increase in the signal-to-noise ratio (SNR) by approximately √2, which is particularly advantageous in low-field MR systems. For more than two slices, the individual slices can also be restored by adding and subtracting the simultaneously acquired folded images according to the Hadamard encoding scheme and by a Hadamard transform along the repetition dimension.

[0063] As Figure 3 shown, since the slices are reconstructed only from two repetitions, this method is prone to patient motion, especially in cases where long-term averaging has to be performed. A mismatch between the folded slices C1 and C2 will result in image artifacts. The same is true for the SMS method described in US 10,557,903 B2, even when a common reference scan for obtaining two averages is performed, performing separate reconstructions and averaging these results will at least erase the case of motion artifacts.

[0064] Accordingly, the present invention proposes to perform motion correction between repetitions, thereby incorporating motion correction into the Hadamard / SMS averaging reconstruction chain. The idea is to perform spatial registration of the folded slices C1, C2 between each acquired repetition Rep1, Rep2 before Hadamard / SMS reconstruction. In the illustrated embodiment, a method for Hadamard imaging is described; however, SMS imaging is also covered by the present invention.

[0065] Figure 4 is a simplified flow chart of an embodiment of the method of the present invention. Thus, the method uses two folded images C1, C2. In step 16, they are registered to obtain translation and / or rotation motion correction parameters, e.g., a motion vector field in the most general case. The most straightforward embodiment is to correct only rigid in-plane motion, and the simplest variant includes only translation. These effects can be easily corrected by translation / rotation in the image space or multiplication with phaseramps / rotation in the k-space. Since the main applications of Hadamard imaging are brain or joint imaging, this assumption is reasonable. In step 18, motion correction is performed in the image space / k-space on one of the folded datasets, e.g., on the second folded image C2. The result is the corrected folded dataset C2'. Thus, in step 20, Hadamard reconstruction is performed on the corrected dataset according to Figure 3 but using dataset C1 and the motion-corrected dataset C2' in this example.

[0066] As Figure 5 shown, this method can also be employed if the Hadamard / SMS imaging technique is combined with an in-plane acceleration method (e.g., planar GRAPPA) that includes undersampling in the phase encoding direction P. The GRAPPA factor of R results in R image replicas along the phase encoding direction P within the FOV / R distance. In Figure 5 the example, R = 2. In Figure 5 the images of slices S1 and S2 show two repetitions Rep1, Rep2, as Figure 2As shown. The phase encoding direction P in the plane is the vertical direction, and the readout direction is the horizontal direction. The object 30 to be imaged can be a head imaged in a substantially axial direction, and the field of view FOV is slightly larger than the extent of the head 30. The head 30 has moved slightly between the first and second repetitions, resulting in the slices S1' and S2' in the second repetition Rep2 not matching exactly with the images S1, S2 of the first repetition, that is, the head 30 has moved between them. Therefore, the folded images C1, C2 from each repetition also do not exactly correspond to each other, and the reconstruction according to Hadamard encoding will result in significant artifacts. If only translational motion has occurred between Rep1 and Rep2, an effective correction factor can be obtained through rigid motion correction between C1 and C2, as described herein. However, in order to obtain effective rotation correction parameters, it is advantageous to perform motion registration only on Figure 5 those segments of the FOV shown shaded. These segments T1, T2, T3 are limited in their spatial extent along the phase encoding direction P. They are located around the center and the edges of the field of view, and have a width w in the center and a width w / 2 at the edges. Ideally, the width w is related to the ratio of the field of view in the phase encoding direction to the maximum extent of the imaged object 30 and the acceleration factor R. At R = 2, for example, assuming the object 30 is at the center of the field of view, if the object 30 covers 80% of the FOV in the phase encoding direction, then there will be 20% of the segments in the middle and 10% of the segments at the edges without aliasing artifacts. Therefore, if the coverage rate of the field of view in the P direction is O = 80%, then for R = 2, the maximum expansion of the entire field of view available for motion correction is 2w = (1 - O) x 2. Therefore, spatial registration between the folded images C1, C2 is performed only based on the shaded regions, that is, T1 is registered with T1', T2 is registered with T2', and T3 is registered with T3' (shown by arrow 22).

[0067] If the assumptions of rigid motion and major in-plane motion are valid, it can be expected to obtain the best results. Elastic motion correction can still be performed in the case where the motion vector fields in two simultaneously acquired slices are similar.

[0068] Figure 6 A method is shown in which more than two images are acquired to cover a larger field of view in the slice direction. In this case, 4 pairs of images are acquired one after another, where the slices S1.1 and S2.1 are acquired simultaneously (represented by 28). Similarly, the slices S1.2 and S2.2, S1.3, S2.3, and S1.4, S2.4 are acquired simultaneously. As Figure 6As shown, pairs of slices are interleaved such that slices acquired simultaneously are not directly adjacent to each other. This allows for more advanced implementations, including motion correction by planar motion. Thus, motion registration is performed simultaneously (e.g., after acquisition is complete) on all folded slice pairs SX.1, SX.2, SX.3, and SX.4. If motion can be determined to have occurred, e.g., between the acquisitions of SX.1 and SX.2, this may still be correctable because the slices are adjacent to each other. Thus, motion correction can be performed by signal interpolation between these folded slices.

[0069] List of reference numerals

[0070] 1 Magnetic resonance device

[0071] 2 MR scanner

[0072] 3 Fundamental field magnet 3

[0073] 5 Gradient coil device

[0074] 5' Gradient controller

[0075] 7 RF antenna

[0076] 7' RF transmit / receive controller

[0077] 7.1 First coil of coil array

[0078] 7.2 Second coil of coil array

[0079] S1 First slice

[0080] S2 Second slice

[0081] U Examination object / patient

[0082] L Hospital bed

[0083] 13 Computational processor

[0084] 15 Determination processor

[0085] 17 Memory

[0086] 26 Storage medium

[0087] E / A Input / output device

[0088] Rep1 First repetition

[0089] Rep2 Second repetition

[0090] C1 First folded image

[0091] C2 Second folded image

[0092] 12 Addition

[0093] 14 Division

[0094] 16 Registration step

[0095] 18 Motion correction step

[0096] 20 Hadamard / SMS reconstruction step

[0097] S1', S2' The first and second slices after motion correction

[0098] 30 Object

[0099] FOV Field of view

[0100] T1, T2, T3 Segments of the field of view

[0101] 22 Registration

[0102] P Phase encoding direction

[0103] R Read direction

[0104] S1.1, S2.1 Simultaneously acquired slices

Claims

1. A method for obtaining a magnetic resonance (MR) image dataset of at least two slices (S1, S2), the MR image dataset being obtained by simultaneous multi-slice excitation, the method comprising the following steps: a) Performing an MR imaging sequence using multi-band radiofrequency (RF) excitation pulses to simultaneously excite at least two slices (S1, S2), wherein the MR imaging sequence comprises at least two repetitions (Rep1, Rep2), wherein the repetitions are performed according to a phase modulation scheme, wherein in each repetition, each of the simultaneously excited slices is assigned a phase, and the phase of at least one of the simultaneously excited slices changes from one repetition to the next, whereby an MR dataset of folded images (C1, C2) is obtained in each repetition; b) Performing spatial registration (16) between at least two folded images (C1, C2) to obtain translation and / or rotation correction parameters; c) Performing motion correction (18) on at least one MR dataset of the folded images (C1, C2) based on the correction parameters to obtain a corrected MR dataset; and d) Reconstructing (20) MR images of at least two slices from the corrected MR dataset of the folded images.

2. The method according to claim 1, wherein, The registration (16) is performed by means of a registration method insensitive to contrast variations, in particular a method based on mutual information.

3. The method according to claim 1 or 2, wherein, The registration of the folded images with respect to each other is performed in image space.

4. The method according to any one of the preceding claims, wherein, Performing motion correction (18) comprises translation and / or rotation in image space or multiplication and / or rotation with a phase ramp in k-space.

5. The method according to any one of the preceding claims, wherein, Performing motion correction (18) comprises performing rigid in-plane motion correction in the plane of one or more of the at least two slices.

6. The method according to any one of the preceding claims, wherein, Motion correction (18) through the plane is performed by registering folded images (C1, C2) obtained from different pairs of slices and performing signal interpolation between these folded images.

7. The method according to any one of the preceding claims, wherein, The multi-band radiofrequency (RF) excitation pulses each comprise first and second single-band pulse shapes, wherein at least one of the single-band pulse shapes is a phase shift between one repetition and the next.

8. The method according to any one of the preceding claims, wherein, By switching an additional gradient in the slice direction, the phase assigned to the simultaneously excited slices changes from one repetition to the next.

9. The method according to any one of the preceding claims, wherein, The phase modulation scheme for simultaneous multi-slice excitation uses Hadamard coding.

10. The method according to any one of the preceding claims, wherein, The MR imaging sequence uses an in-plane acceleration method, in particular in-plane GRAPPA or in-plane SENSE technology, wherein the undersampling factor in the phase direction is given by an acceleration factor R, and the folded images (C1, C2) comprise R image replicas spaced apart from each other by the field of view distance divided by R along the phase encoding direction.

11. The method according to claim 10, wherein, Spatial registration is performed on the folded images (C1, C2) comprising image replicas caused by aliasing.

12. The method according to claim 10, wherein, The in-plane field of view of the MR image dataset in the phase encoding direction is larger than the object to be imaged, and wherein the method comprises: Identifying segments (T1, T2, T3, T1', T2', T3') of the field of view in the phase encoding direction that do not comprise overlapping image replicas, Spatial registration is performed between segments of the fields of view of at least two folded images to obtain translation and / or rotation correction parameters.

13. A magnetic resonance (MR) device (1), comprising a) an MR scanner (2) adapted to acquire MR data sets (C1, C2) from an object (U) located inside the MR scanner, and b) a computer (9) configured to send control signals to the MR scanner (2) to cause the MR scanner to perform the method according to any one of the preceding claims.

14. A non-transitory computer-readable data storage medium (26) encoded with programming instructions, adapted to be loaded into a computer of a magnetic resonance (MR) device (1) comprising an MR scanner (2), by sending control signals to the MR scanner and receiving data from the MR scanner, the programming instructions cause the computer to perform, in conjunction with the MR scanner (2), the method according to any one of claims 1-11.

Images