MR imaging using partial echo acquisition

Abstract

A method is provided that allows for improved MR images using partial echo acquisition, comprising the steps of subjecting an object 10 to an imaging sequence including a series of RF excitations, where echo signals are generated in the presence of a readout magnetic field gradient at each repetition time interval between successive RF excitations, where a phase-encoding magnetic field gradient is switched at each repetition time interval to sample a predetermined region of k-space, acquiring echo signals from the object 10 as partial echoes having different readout polarities, where each echo signal is associated with either a first direction of the readout magnetic field gradient or a second direction of the readout magnetic field gradient opposite the first direction, where the echo signals associated with the first and / or second directions are sparsely sampled in at least one phase-encoding direction, and reconstructing an MR image from the acquired echo signals.

Description

[Technical field]

[0001] The present invention relates to the field of magnetic resonance (MR) imaging. The present invention relates to a method for MR imaging of an object placed in an examination volume of an MR system. The present invention also relates to an MR system and to a computer program executed on the MR system. [Background technology]

[0002] MR imaging, which utilizes the interaction between a magnetic field and atomic nuclear spins to form two- or three-dimensional images, is widely used today, especially in the field of medical diagnostics, for imaging of soft tissues, since it is superior in many ways to other imaging methods, does not require ionizing radiation, and is usually non-invasive.

[0003] In MR imaging, fractional sampling of k-space along the readout (frequency encoding) direction, herein referred to as partial echo (PE) acquisition, is a well-known strategy to reduce echo time (TE) (see M. Bydder et al., in NMR in Biomedicine, 2021; 34:e4458). This helps to acquire MR signals from short T2 species, but also reduces the repetition time (TR) and, if applicable, the echo spacing of the imaging sequence used. PE sampling is of particular interest in magnetic resonance angiography (MRA) to mitigate flow-induced signal loss.

[0004] Theoretically, partial Fourier (PF) methods such as PE acquisition allow sampling only half of the k-space. The other half of the k-space is synthesized in the reconstruction under the assumption of conjugate symmetry, which only holds for real (as opposed to complex) MR signals in image space. Reconstruction by simple Fourier transformation is discouraged from PE sampling. However, various algorithms have been developed to produce relatively artifact-free images, based on the assumption that there is a smooth phase variation across the image. In practice, PE sampling is slightly extended to completely cover the central k-space region, thus capturing low-frequency phase variations.

[0005] PE methods are fundamentally limited by missing k-space data, which prevents recovery of high-frequency phase variations and causes signal loss. Acceleration methods such as parallel imaging (PI) and compressed sensing (CS) generally cannot compensate for this shortcoming, mainly due to the large size of the continuous k-space region given by the incomplete sampling. Summary of the Invention [Problem to be solved by the invention]

[0006] It is an object of the present invention to provide a method that allows improved MR imaging using PE acquisitions. [Means for solving the problem]

[0007] According to the invention, a method for MR imaging of an object placed in an examination volume of an MR system is disclosed, the method comprising the steps of: exposing the object to an imaging sequence having a series of RF excitations, where echo signals are generated in the presence of a readout magnetic field gradient at each repetition time interval between successive RF excitations, and where a phase encoding magnetic field gradient is switched at each repetition time interval to sample a predefined region of k-space; acquiring the echo signals from the object as partial echoes having different readout polarities, each echo signal being associated with either a first direction of the readout magnetic field gradient or a second direction of the readout magnetic field gradient opposite to the first direction, the echo signals associated with the first and / or second directions sparsely sampling k-space in at least one phase encoding direction; reconstructing an MR image from the acquired echo signals; has.

[0008] The present invention proposes to acquire PE readouts with two gradient polarities in opposite directions, i.e. bipolar PE acquisitions. Previously, two PE readouts with gradient polarities in opposite directions and identical phase encoding were combined to obtain one (virtual) full echo (FE) readout. In this way, a complete sampling of the k-space along the readout direction is obtained, which is obtained with a short TE of the PE acquisition, but only at approximately twice the acquisition time (see MB Scheidegger et al. In Magnetic Resonance Imaging 1991; 9:517-524). The present invention proposes that the PE readouts with positive and negative gradient polarities sample each k-space sparsely in at least one of the phase encoding directions. Sparse sampling (also called subsampling) means that the sampling density in the phase encoding direction is lower than required by the Nyquist criterion according to the field of view (FOV). The present invention also proposes that the PE readouts with positive and negative polarities do not necessarily share the same phase encoding. The same phase encoding does not imply that PE readouts with positive and negative polarity cover k-space differently not only along the readout direction but also in at least one of the phase encoding directions.

[0009] In one embodiment of the present invention, the echo signals associated with a first readout direction sample a sub-region of k-space that is not sampled by the echo signals associated with a second readout direction, and vice versa. In other words, the k-space region provided by the echo signals associated with a first readout direction is covered by the echo signals associated with a second readout direction. Thus, the present invention achieves synthesis of missing k-space data without the assumption of a smooth phase variation across the image. To this end, instead of the conventional PE strategy, the present invention proposes to apply strategies as known in the art for reconstructing artifact-free images from sparsely sampled k-space data, such as Sensitivity Encoding (SENSE), Compressive Sensing (CS), and combinations thereof, for synthesis of missing k-space data.

[0010] In a further embodiment, the k-space regions sampled by the echo signals associated with the first and second directions, respectively, overlap in the central portion of the k-space. In another embodiment, the echo signals associated with the first and second directions combine to completely sample the k-space in the central portion. In yet another embodiment, they sample the k-space in an interleaved pattern in at least one phase encoding direction. This results in a higher sampling density, or even complete sampling, in the center of the k-space, but without increasing the TE or overall acquisition time compared to conventional PE methods.

[0011] Reversing the readout direction during acquisition causes distortions induced by main field inhomogeneities to change direction. These distortions can be addressed by adding a step of mapping the main field inhomogeneities, which are then taken into account in the step of reconstructing the MR image. Alternatively, the distortions can be estimated from two sparsely sampled individual data sets, one associated with the first readout direction and the other with the second readout direction, requiring separate intermediate reconstruction of both and the application of registration to the resulting intermediate MR images.

[0012] Similarly, reversing the readout direction during acquisition potentially changes the linear and higher order gradient fields. These changes can be addressed by mapping the spatial or spatially varying phase difference between the echo signals associated with the first and second directions, respectively, which is taken into account in the step of reconstructing the MR image. This phase difference can be again estimated from the two sparsely sampled individual data sets, requiring separate intermediate reconstructions of both, or it can be estimated from separate measurements.

[0013] Furthermore, reversing the readout gradient during acquisition potentially changes the integral of the readout gradient from TR to TR. In a further embodiment, a spoiler magnetic field gradient in the readout direction, applied after acquiring the echo signal, is adapted to the imaging sequence such that this integral remains the same from TR to TR. In this way, for example, a steady-state spoiling condition can be maintained.

[0014] In non-contrast brain MRA, a TE where the MR signals from water and fat have opposite phases is often preferred. However, this leads to low scan efficiency of the gradient echo sequence used, especially at 1.5T, while only providing insufficient fat suppression. In a further embodiment, a TE where the MR signals from water and fat are nearly orthogonal, i.e. have a phase difference of ±90 degrees, is selected, and the one-point Dixon method known in the art for separating MR signals from water and fat from echo signals acquired with a single TE is applied.

[0015] The use of a two-point Dixon method, i.e. dual acquisition Dixon imaging, primarily promises better fat suppression and allows to further shorten the TE, but usually involves introducing dead times and shifting the readout magnetic field gradient for chemical shift encoding, which also negatively affects the scan efficiency. Therefore, according to a further aspect, the invention proposes that echo signals are acquired with different TEs, each of which is associated with either a first TE or a second TE such that chemical shift encoding is applied to the acquired echo signal, and the contributions from water and fat are then separated in a step of reconstructing the MR image according to the (as such known) two-point Dixon method. Therefore, the invention proposes to combine dual acquisition Dixon imaging with bipolar PE acquisition. In this way, the invention makes it possible to improve the image quality by shortening the TE, removing fat signals, avoiding partial echo artifacts and increasing the signal-to-noise ratio (SNR).

[0016] To improve the scan efficiency of the resulting gradient echo imaging sequence, the strength of the readout magnetic field gradient applied during the acquisition of the echo signal associated with the shorter of the two TEs can be smaller than the strength of the readout magnetic field gradient applied during the acquisition of the echo signal associated with the longer of the two TEs. Furthermore, the partial echo factor (PEF) can be adapted to maximize the acquisition time under the constraint imposed by the phase encoding magnetic field gradient. Alternatively or additionally, a spoiler readout magnetic field gradient with a larger area is associated with a shorter TE. That is, the region of the spoiler readout magnetic field gradient associated with the shorter echo time is larger than the region of the spoiler readout magnetic field gradient associated with the longer echo time.

[0017] The reconstruction of MR images from conventional PE methods, or conventional PF methods in general, usually involves weighting of the acquired echo signals in k-space, Fourier transformation, and optionally phase correction and subsequent removal of the imaginary parts in image space. The weighting takes care of the truncation of the echo signals in k-space and suppresses (Gibbs) ringing artifacts that would otherwise occur. For example, a simple Hanning window filter can be used as a weighting, which smoothly varies from zero to one over an appropriately selected distance in k-space. However, it has been observed that the use of such weighting also causes signal losses in MRA, which can be misinterpreted as stenosis. Thus, according to yet another aspect, the invention proposes a method for MR imaging of an object placed in an examination volume of an MR system, the method comprising: exposing the object to an imaging sequence including a series of RF excitations, where echo signals are generated in the presence of a readout magnetic field gradient at each repetition time interval between successive RF excitations, and where a phase encoding magnetic field gradient is switched at each repetition time interval to sample a predefined region of k-space; acquiring an echo signal from an object as a partial echo; applying a shift in k-space in the readout direction to the acquired echo signal data, weighting the echo signal data, and reconstructing intermediate MR images for each of a plurality of different shift values; calculating a final MR image from the intermediate MR images by selecting, for each voxel of the final MR image, the voxel value having the maximum intensity from the set of intermediate MR images at each voxel location; has.

[0018] This aspect of the invention, which can be used in combination with the other aspects mentioned above or separately, is based on the insight that signal loss in PF imaging can result from rapid variations in phase in image space. To approximately compensate for such rapid variations in phase in image space, the acquired echo signals are first shifted by different ranges in k-space according to the number of shift values, and then weighting is applied to them. For each shift value, a separate intermediate reconstruction is then performed, and finally, signal recovery in image space is achieved by selecting, voxel by voxel, the maximum signal intensity over all resulting intermediate MR images. The weighting suppresses ringing artifacts, but also affects the acquired echo signals. Since the phase in image space is a priori unknown, the effect of the weighting on the acquired echo signals is also unknown. Therefore, the invention aims to minimize this effect by considering the range of relative shifts between the weighting signal and the acquired echo signal. This is motivated by the fact that applying a shift to the acquired echo signals in k-space barely changes their magnitude in image space if the weighting is ignored. Furthermore, this is motivated by the observation that applying weighting to the acquired echo signals in k-space almost reduces their magnitude in image space.

[0019] The inventive method described so far can be performed by an MR system comprising at least one main magnet coil for generating an essentially homogeneous static magnetic field B0 in an examination volume, a plurality of gradient coils for generating switched magnetic field gradients in different spatial directions in the examination volume, at least one body RF coil for generating RF pulses in the examination volume and / or for receiving MR signals from a patient's body arranged in the examination volume, a control unit for controlling the time sequence of the RF pulses and the switched magnetic field gradients, and a reconstruction unit for reconstructing an MR image from the received MR signals. The inventive method can be implemented by a corresponding program of the reconstruction unit and / or the control unit of the MR system.

[0020] The method of the present invention can be advantageously implemented in most MR systems currently in clinical use. For this purpose, it is simply necessary to utilize a computer program that controls the MR system to carry out the above-mentioned method steps of the present invention. The computer program may be present on a data carrier or in a data network so as to be downloaded for installation in the control unit of the MR system.

[0021] The accompanying drawings disclose preferred embodiments of the invention, but it is to be understood that the drawings are designed for illustrative purposes only and are not intended as a definition of the limits of the invention. [Brief description of the drawings]

[0022] [Figure 1] 1 shows an MR system for carrying out the method of the present invention. [Figure 2a] 1 shows a k-space diagram of the partial Fourier acquisition concept. [Figure 2b] 1A-1D show respective views of k-space illustrating an example of a bipolar readout strategy of the present invention. [Figure 2c] 1A-1D show respective views of k-space illustrating an example of a bipolar readout strategy of the present invention. [Diagram 3] FIG. 1 shows a schematic (simplified) pulse sequence diagram according to a first embodiment of the present invention. [Figure 4] FIG. 1 shows a schematic (simplified) pulse sequence diagram according to a second embodiment of the present invention. [Diagram 5] FIG. 13 shows a schematic (simplified) pulse sequence diagram according to a third embodiment of the present invention. [Figure 6] FIG. 13 shows a schematic (simplified) pulse sequence diagram according to a fourth embodiment of the present invention. [Figure 7] 1 shows MR images illustrating another embodiment of the PF method according to the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] Referring to Figure 1, an MR system 1 is shown as a block diagram. The apparatus comprises superconducting or resistive main magnet coils 2 such that a substantially uniform and time-constant main magnetic field B0 is generated along the z-axis through the examination volume. The apparatus further comprises a set of (primary, secondary and, if applicable, tertiary) shimming coils 2', the current through each individual shimming coil of the set 2' being controllable to minimize B0 deviations within the examination volume.

[0024] The magnetic resonance generation and manipulation system applies a series of RF pulses and switched magnetic field gradients to invert, excite, refocus, and / or spatially encode magnetic resonance and otherwise perform MR imaging.

[0025] More specifically, gradient pulse amplifiers 3 apply current pulses to selected ones of whole-body gradient coils 4, 5, and 6 along the x, y, and z axes of the examination volume. A digital RF frequency transmitter 7 transmits RF pulses or pulse packets via a transmit / receive switch 8 to an RF body coil 9, which transmits RF pulses into the examination volume. A typical MR imaging sequence consists of packets of short duration RF pulse segments which, together with any applied magnetic field gradients, achieve selected manipulation of nuclear magnetic resonance. The RF pulses are used to manipulate the magnetic resonance and select a portion of a body 10 located within the examination volume. The MR signals are also picked up by the RF body coil 9.

[0026] To generate MR images of limited regions of the body 10, a set of local RF array coils 11, 12, 13 are positioned adjacent to the region selected for imaging. The array coils 11, 12, 13 may be used to receive MR signals induced by the transmission of the RF body coil.

[0027] The resulting MR signals are picked up by an RF body coil 9 and / or RF array coils 11, 12, 13 and demodulated by a receiver 14, which preferably includes a preamplifier (not shown). The receiver 14 is connected to the RF coils 9, 11, 12, 13 via a transmit / receive switch 8.

[0028] A host computer 15 controls the shimming coils 2' as well as the gradient pulse amplifiers 3 and the transmitter 7 to generate the imaging sequences of the present invention. For a selected sequence, the receiver 14 receives MR signal data in rapid succession from single or multiple k-space lines following each RF excitation pulse. The data acquisition system 16 performs analog-to-digital conversion of the received signals and converts the MR data into a digital format suitable for further processing. In most modern MR systems, the data acquisition system 16 is a separate computer dedicated to acquiring raw image data.

[0029] Finally, the digital MR signal data is reconstructed into an image representation by a reconstruction processor 17, which applies a Fourier transform or other appropriate reconstruction algorithm, such as detection. The MR image may represent a planar slice through the patient, an array of parallel planar slices, a three-dimensional volume, etc. The image is then stored in an image memory, where it can be accessed to convert slices, projections, or other portions of the image representation into an appropriate format for visualization, for example, via a video monitor 18, which provides a human-readable display of the resulting MR image.

[0030] The host computer 15 and the reconstruction processor 17 are configured by corresponding programs to carry out the methods of the invention described above and below.

[0031] According to the invention, echo signals are acquired from the body 10 as partial echoes. Each partial echo represents a k-space line. This is shown in Fig. 2a,b,c. In the depicted 2D k-space diagram, the readout direction is indicated by M and the phase encoding direction by P. In a conventional PE acquisition (Fig. 2a) the peripheral parts of k-space are covered in one direction but not in the opposite direction. Thus, a large continuous k-space region is saved. To avoid insufficient alignment resulting from the MR image reconstruction, the direction in which the partial echoes are acquired is reversed during the scan according to the invention. This makes it possible to generate the alternative sampling patterns shown (among others) in Fig. 2b and 2c. The PE acquisition is performed with different readout polarities, each echo signal being associated either with a first direction (right side of Fig. 2) or with a second direction opposite to the first direction (left side of Fig. 2), and the echo signals associated with the first and second directions sparsely sample the k-space in the phase encoding direction P. In principle, the complete k-space can be covered in this way by performing the acquisition twice with opposite readout directions without extending the TE. In the example of FIG. 2b, the readout direction is flipped between adjacent k-space lines, resulting in two uniformly sparsely sampled (i.e. subsampled) data sets in the phase encoding direction P. In the example of FIG. 2c, two non-uniformly subsampled data sets are obtained in the phase encoding direction. In the examples of FIG. 2b and FIG. 2c, the k-space lines associated with the first and second readout directions overlap in the central k-space portion in the phase encoding direction such that the central k-space portion in the phase encoding direction is fully sampled, i.e. not subsampled. This further improves the conditioning of the reconstruction. In FIG. 2c, the central k-space portion in the frequency encoding direction is also fully sampled.

[0032] A strategy for efficiently implementing the sampling patterns shown in Figures 2b and 2c is provided in Figure 3. The frequency encoding direction is reversed after each iteration, for example, to cover k-space sequentially from bottom to top.

[0033] FIG. 3 shows a pulse sequence diagram of a 3D spoiled partial gradient echo sequence constituting an imaging sequence according to the invention. The diagram shows switched magnetic field gradients in the frequency encoding direction M, the phase encoding direction P, and the slice selection direction S. The diagram also shows RF excitation and refocusing pulses, indicated by ACQ1 and ACQ2, as well as the time intervals during which echo signals are acquired. One gradient echo is generated at each repetition time between successive RF excitations. The diagram shows the acquisition of the first two echo signals. At the subsequent repetition time TR, the phase encoding magnetic field gradients in the directions P and S are scaled to sample a given region of k-space. Two echo signals are generated at echo time TE. From odd (left) to even (right) repetitions, the polarity of the first two gradient lobes in the readout direction M is reversed, while the area of ​​the third gradient lobe is adjusted at the even repetition to maintain a steady-state spoiled regime. This does not necessarily result in a longer TR, since a longer TR is often chosen anyway to optimize the signal-to-noise ratio (SNR).

[0034] In an alternative approach, the readout direction is reversed only once. In this case, two subsampled echo signal data sets are acquired consecutively. This breaks the steady-state spoiling regime and allows to preserve, except for the sign, the region of the third gradient lobe in the readout direction M, but the two subsampled echo signal data sets are potentially less consistent, for example in the presence of motion.

[0035] Reversing the readout direction of the partial echo signals during a scan changes the direction of distortions induced by the main magnetic field inhomogeneities. These distortions can be addressed by mapping the main magnetic field inhomogeneities in a separate measurement and taking them into account in the reconstruction step. Alternatively, they can be estimated from two subsampled sets of echo signals, requiring separate intermediate reconstructions of both and the application of registration to the resulting images.

[0036] Furthermore, reversing the readout direction of the partial echo signals during a scan potentially requires considering linear and higher order gradient variations, including mapping the spatial or spatial-temporal varying phase difference between echo signals acquired with opposite readout directions, which may also be estimated from two subsampled sets of echo signals.

[0037] The final reconstruction of the MR image from the acquired echo signals is performed using the model Solve the inverse problem corresponding to TIFF2025506478000002.tif23145. where y denotes the two subsampled sets of echo signals, E denotes the corresponding encoding matrix, D denotes the corresponding distortion matrix, P denotes the corresponding phase error matrix, and x denotes the image. Now, the missing data is always surrounded by available data in k-space, so the adjustment of the inverse problem is greatly improved.

[0038] An example of dual acquisition Dixon imaging with bipolar PE acquisition is shown in Figure 4. Two sub-acquisitions with opposite polarity of the readout magnetic field gradient M are performed in series with different echo times TE1 and TE2 for chemical shift encoding. One repetition from each sub-acquisition is shown. The first (shorter) echo time TE1 is associated with the first readout gradient direction and the second (longer) echo time TE2 is associated with the second readout gradient direction. In the first repetition on the left, a dead time is introduced at the end. In the second repetition on the right, a dead time is introduced after slab selection. For simplicity, flow compensation is only applied in the readout direction in this example.

[0039] Compared to conventional non-contrast brain MRA with anti-phase TE, the dual acquisition Dixon imaging sequence shown in FIG. 4 provides superior fat suppression and also reaches a shorter TE. However, the introduced dead time leads to a longer repetition time TR and a lower scan efficiency, defined as the ratio of the duration of the acquisition windows ACQ1 and ACQ2 and TR. To improve this, i.e. to improve the scan efficiency, two different strategies are applied to the two sub-acquisitions, as shown in FIG. 5, which shows a serial dual acquisition Dixon imaging sequence with individually optimized bipolar PE acquisition. In the first repetition on the left, the readout magnetic field gradient and the acquisition window ACQ1 are extended at the end and the spoiler and rewinder magnetic field gradients are shifted. In addition, the strength of the readout magnetic field gradient is reduced. This leads to a smaller PEF (PEF indicates the proportion of the k-space sampled by PE to the complete k-space) close to 0.5, which is feasible in combination with the bipolar PE acquisition according to the invention, since the fully sampled central k-space is no longer required to perform phase correction. In the second iteration on the right, the readout magnetic field gradient and the acquisition window are extended forward and the phase encoding gradient is shifted. This leads to a larger PEF. Preferably, the TE1 of the first partial acquisition is chosen to be more out of phase (with respect to the relative phase of the water and fat signals) than the TE2 of the second partial acquisition. In this way, flow-induced signal loss between the first and second partial acquisitions due to the longer TE is less likely to be misinterpreted as water-fat dephasing.

[0040] The two sub-acquisitions in the above embodiment may be performed in series. Since spoiling is interrupted at the transition from the first to the second sub-acquisition, several additional repetition time intervals may be inserted at this transition to re-establish spoiling. This causes only a small increase in scan time.

[0041] Alternatively, it may be advantageous to interleave the two sub-acquisitions, for example by switching between them after each repetition. In this case, the integral of the magnetic field gradient in the readout direction M must remain constant, which requires increasing the area of ​​the spoiler magnetic field gradient in the readout direction M in one of the sub-acquisitions, as shown in Fig. 6, which shows an interleaved dual-acquisition Dixon imaging sequence with bipolar partial echo acquisitions and an optimized allocation of the two echo times TE1 and TE2 to the two polarities of the readout magnetic field gradient. Preferably, the area of ​​the spoiler readout magnetic field gradient is increased in the sub-acquisitions with a shorter echo time, in this case TE2, to optimize the scan efficiency.

[0042] Furthermore, the present invention proposes to (almost) compensate for the undesired signal loss in PF imaging resulting from rapid variations in phase in the image space, which may lead to misdiagnosis. The acquired echo signals are first shifted by different extents in k-space according to several shift values, and then a weighting is applied to them. For each shift value, a separate intermediate reconstruction is performed and the maximum signal intensity is selected voxel-by-voxel across all resulting intermediate MR images. Mathematically, the proposed PF reconstruction method is: Described by TIFF2025506478000003.tif20143. Here, S represents the acquired echo signal, W represents the weighting, F represents the Fourier transform, p represents the set of shift values, and s represents the final reconstructed MR image. This aspect of the invention is illustrated in a time-of-flight (TOF) MRA in FIG. 7, where selected slices from a 3D axial partial echo acquisition, reconstructed from echo signals from one channel of the employed head coil, are shown once with the conventional PE method (left image), once with the PF method proposed by the invention (middle image), as well as their difference (right image). Using the conventional PF method with simple weighting results in a signal loss in the left internal carotid artery marked in the left image, which can easily be misinterpreted as a stenosis. However, the PF method proposed by the invention, using a set of 20 shift values ​​in this example, successfully recovers the signal intensity in this vessel, and it can be seen that there is indeed no sign of stenosis.

Claims

1. 1. A method of MR imaging of an object placed in an examination volume of an MR system, comprising: exposing the object to an imaging sequence having a series of RF excitations, where echo signals are generated in the presence of a readout magnetic field gradient at each repetition time interval between successive RF excitations, and where a phase encoding magnetic field gradient is switched at each repetition time interval to sample a predefined region of k-space; acquiring the echo signals from the object as partial echoes having different readout polarities, each echo signal being associated with either a first direction of the readout magnetic field gradient or a second direction of the readout magnetic field gradient opposite to the first direction, the echo signals associated with the first and / or second directions sparsely sampling k-space in at least one phase encoding direction; reconstructing an MR image from the acquired echo signals; The method comprising:

2. 2. The method of claim 1, wherein echo signals associated with the first direction sample a subregion of k-space that is not sampled by echo signals associated with the second direction, or vice versa.

3. 3. The method of claim 1, wherein the k-space regions sampled by echo signals associated with the first and second directions, respectively, overlap in a central portion of k-space.

4. 4. The method of claim 1, wherein echo signals associated with the first and second directions respectively sample k-space in an interleaved pattern in the at least one phase-encoding direction.

5. 5. The method of claim 1, wherein echo signals associated with the first and second directions are combined to completely sample k-space in a central portion of k-space.

6. 6. The method according to claim 1, wherein the reconstruction of MR images from the acquired MR signals relies on sensitivity encoding, compressive detection, or a combination thereof, instead of the assumption of smooth phase variation between the MR images.

7. 7. The method according to claim 1, further comprising the step of mapping the main magnetic field inhomogeneity, which is taken into account in reconstructing the MR image.

8. 8. The method according to claim 1 , further comprising the step of mapping in image space a spatial or spatially and time-varying phase difference between echo signals associated with the first and second directions, respectively, wherein the phase difference is taken into account in the step of reconstructing the MR image.

9. 9. The method according to claim 1, wherein a spoiler magnetic field gradient in the readout direction, applied after acquiring the echo signals, is adapted such that an integrated area of ​​the readout magnetic field gradient is the same for each repetition time interval for both the first and second directions.

10. 10. The method according to claim 1, wherein the echo signals are acquired at echo times at which the signal contributions from water and fat are approximately in quadrature, and the contributions from water and fat are separated in the step of reconstructing the MR image.

11. 10. The method of claim 1, wherein the echo signals are acquired at different echo times, each echo signal being associated with either a first echo time or a second echo time such that chemical shift encoding is applied to the acquired echo signals, and contributions from water and fat are separated in the step of reconstructing the MR image.

12. 12. The method of claim 11, wherein a strength of a readout magnetic field gradient applied during acquisition of the echo signals associated with the shorter echo times is less than a strength of a readout magnetic field gradient applied during acquisition of the echo signals associated with the longer echo times.

13. 13. The method of claim 9, wherein an area of ​​a spoiler readout magnetic field gradient associated with the shorter echo time is larger than an area of ​​a spoiler readout magnetic field gradient associated with the longer echo time.

14. 14. An MR system comprising at least one main magnet coil for generating a homogeneous static magnetic field in an examination volume, a plurality of gradient coils for generating switched magnetic field gradients in different spatial directions in the examination volume, at least one RF coil for generating RF pulses in the examination volume and / or for receiving MR signals from an object placed in the examination volume, a control unit for controlling a time sequence of the RF pulses and the switched magnetic field gradients, and a reconstruction unit for reconstructing MR images from the received MR signals, wherein the MR system is configured to perform the steps of the method according to any one of claims 1 to 13.

15. A computer program comprising instructions which, when said program is executed by a computer, preferably a control unit and / or a reconstruction unit of an MR system, cause said computer to carry out the method according to any one of claims 1 to 13.

Images