RU2773486C2 - Magnetic resonance imaging with water/fat separation by dixon method - Google Patents

Abstract

FIELD: magnetic resonance imaging.

SUBSTANCE: invention is used for magnetic resonance imaging (MRI) by the Dixon method. Following stages are performed: an object is subjected to an imaging sequence containing at least one RF excitation pulse and switchable magnetic field gradients, wherein two echo signals, the first echo signal and the second echo signal, generate echo with different times (TE1, TE2), echo signals are received from the object, an image of water and/or an image of fat is reconstructed from echo signals, while contributions from water and fat to echo signals are separated using the two-point Dixon method in the first area of k space and the single-point Dixon method in the second area of k space, wherein the first area differs from the second area. In other words, the invention proposes adaptive switching between the two-point Dixon method for the water/fat separation applied to both the first and the second echo signals, and the single-point Dixon method applied to one of two echo signals, i.e., data of the first echo signal or data of the second echo signal depending on the position in k space.

EFFECT: providing the reliable water/fat separation in their imaging.

18 cl, 5 dwg

Description

The technical field to which the invention belongs

The invention relates to the field of magnetic resonance (MR) imaging. It concerns a method of MR imaging of an object placed in the examination volume of an MR device. The invention also relates to an MP device and a computer program executable on an MP device.

State of the art

MR imaging techniques that use the interaction between magnetic fields and nuclear spins to generate 2D or 3D images are currently widely used, in particular in the field of medical diagnostics, since they are superior to other imaging techniques in many respects for soft tissue imaging, do not require applications of ionizing radiation and are generally non-invasive.

In general, according to the MR method, the body of the examined patient is placed in a strong uniform magnetic field B ₀ , the direction of which simultaneously determines the axis (usually the z-axis) of the coordinate system on which the measurement is based. The magnetic field B ₀ creates different levels of energy for individual nuclear spins depending on the strength of the magnetic field, which can be excited (spin resonance) by applying an electromagnetic alternating field (radio frequency (RF) field) of a certain frequency (the so-called Larmor frequency, or MR frequency ). From a macroscopic point of view, the distribution of individual nuclear spins creates an overall magnetization that can deviate from equilibrium by applying an electromagnetic pulse of the appropriate frequency (RF pulse) perpendicular to the z-axis, so that the magnetization undergoes a precessional motion around the z-axis. The precessional motion describes the surface of a cone, the opening angle of which is called the angle of deflection of the magnetization vector. The magnitude of the angle of deviation of the magnetization vector depends on the strength and duration of the applied electromagnetic pulse. In the case of the so-called 90°-momentum, the spins deviate from the z-axis to the transverse plane (the angle of deviation of the magnetization vector (the so-called flip angle) is 90°).

After the end of the RF pulse, the magnetization returns to its original equilibrium state, in which the magnetization in the z direction is created again with the first time constant T1 (spin-lattice or longitudinal relaxation time), while the magnetization in the direction perpendicular to the z direction relaxes with the second time constant T2 (time of spin-spin or transverse relaxation). The change in magnetization can be detected by using RF receiving coils positioned and oriented in the volume of interest of the MR device so that the change in magnetization is measured in the direction perpendicular to the z-axis. The damping of the transverse magnetization is accompanied, after the application of, for example, a 90° pulse, by the transition of nuclear spins (caused by local inhomogeneities of the magnetic field) from an ordered state with the same phase to a state in which all phase angles are uniformly distributed (dephasing). Dephasing can be compensated for with a refocusing pulse (eg 180° pulse). This creates an echo in the receiver coils.

To implement spatial resolution in the body, constant magnetic field gradients along the three main axes are superimposed on a uniform magnetic field B ₀ , which leads to a linear spatial dependence of the spin resonance frequency. In this case, the signal received at the receiver coils contains components of different frequencies, which may be associated with different locations in the body. The signal data received by the pickup coils corresponds to the spatial frequency domain and is referred to as k-space data. The k-space data typically includes multiple lines received with different phase encodings. Each row in k-space is digitized by collecting a number of samples. The k-space data set is converted into an MR image, for example by a Fourier transform.

MR imaging often requires information about the relative contributions of water and fat to the overall signal, either to suppress the contribution of one of them, or to separately or jointly analyze the contribution of each of them. These contributions can be calculated by combining information from two or more corresponding echoes obtained with different echo times. This can be thought of as chemical shift coding, in which an additional value, the chemical shift value, is determined and encoded by obtaining two or more MR images with slightly different echo times. For water/fat separation, these types of experiments are often referred to as Dixon measurements. Dixon MR imaging, or Dixon water-fat MR imaging, achieves a water/fat separation by calculating the contribution of water and fat based on two or more corresponding echoes obtained with different echo times. In general, such a separation is possible due to the presence of a known precessional difference in the frequencies of hydrogen in fat and in water. In its simplest form, water and fat images are created by adding or subtracting "in phase" and "out of phase" datasets.

In recent years, several methods of MR imaging using the Dixon method have been proposed. In addition to various water/fat separation strategies, known methods are generally characterized by the number of echoes (or "points") obtained and the limitations they impose on the echo times used. Conventional so-called two- and three-point methods require in-phase and anti-phase echo times, in which the water and fat signals are parallel and anti-parallel in the complex plane, respectively. Three-point methods have been gradually generalized to provide flexible echo times. Thus, they no longer limit the angle or phase between the water and fat signals at echo times to specific values. Thus, they provide more freedom in the construction of the imaging sequence and allow, in particular, to provide a trade-off between signal-to-noise ratio gains (SNR, signal-to-noise ratio) from data acquisition and SNR losses during separation. On the other hand, sampling only two instead of three echoes is desirable to reduce scan time. Eggers et al. (Eggers et al., Magn. Reson. Med., 65: 96-107, 2011) have proposed a flexible double echo Dixon MR imaging technique. Using such Dixon MR imaging techniques with flexible echo times, in-phase and anti-phase images are no longer necessary, but are instead synthesized from water and fat images if necessary.

In Dixon imaging, various strategies can be used to collect echoes with two different echo times, including: (i) two-pass strategies in which each echo is obtained separately after an RF excitation pulse using a positive amplitude magnetic field gradient read, ( ii) flyback strategies in which both echoes are acquired after the same RF excitation pulse using a positive amplitude read magnetic field gradient in combination with a negative amplitude cleanup magnetic field gradient, and (iii) bipolar strategies in which both echoes are generated after the same RF excitation pulse, with one echo generated using a positive amplitude read magnetic field gradient and the other echo using a negative amplitude read magnetic field gradient.

Bipolar strategies offer many attractive advantages such as shorter scan times, higher signal-to-noise ratio (SNR) performance, more reliable field map estimation, reduction of motion-induced artifacts, and less sensitivity to short T ₂ *. However, variable readout magnetic field gradients present a number of problems, including delay effects and image misregistration, which limit the direct applicability of existing Dixon water/fat separation methods. Lu et al. (Lu et al., Water-fat separation with bipolar multi-echo sequences), Magn. Reson. Med. 2008, 60, 198-209) proposed the use of post-processing of collected echoes prior to water/fat separation. This post-processing includes k-space data shifting to correct k-space echo misregistration and image warping based on a low resolution field map to eliminate misalignment caused by field non-uniformity. The actual water/fat separation is then performed in k-space, thereby removing chemical shift artifacts from the separate water and fat images.

According to the chosen echo times, the inverse problem of the Dixon method used for water/fat separation is usually well conditioned in the central part of k-space. However, due to the finite sampling rate, the inverse problem can be ill-conditioned in more peripheral regions of the k-space. To this end, Lu et al. additionally proposed diagonal loading regularization to prevent excessive noise amplification in these regions of k-space. However, this approach introduces distortions in the assessment of the contribution of signals from water and fat.

Disclosure of the essence of the invention

Thus, it is an object of the invention to provide a method for efficient and reliable Dixon water/fat separation using multiple echo acquisition methods such as TSE double echo or GRASE double echo acquisition or bipolar acquisition strategies.

In accordance with the invention, a method for MR imaging of an object (eg, a patient's body) placed in an MR device's test volume is disclosed. The method includes the following steps:

exposing the object to an imaging sequence comprising at least one RF excitation pulse and switchable magnetic field gradients, wherein two echoes, a first echo and a second echo, are generated with different echo times,

receive echoes from the object, and

reconstructing the water image and/or the fat image from the echo signals, wherein the contributions from water and fat to the echo signals are separated using the two-point Dixon method in the first k-space region and the one-point Dixon method in the second k-space region, the first region different from the second area.

The invention provides adaptive switching between a two-point Dixon method for water/fat separation applied to both the first and second echoes, and a one-point Dixon method applied to one of the two echoes, i.e. the first echo data or data second echo, depending on the position in k-space. Alternatively, Dixon's single point method can be applied to both echoes separately, and the results can then be combined to reduce noise.

Preferably, two echoes are obtained using the bipolar readout magnetic field gradients according to the invention. Two echoes are obtained using a pair of time-contiguous reading magnetic field gradients having opposite polarities. The first echo is obtained at a first echo time using a positive amplitude read magnetic field gradient, and the second echo is obtained at a second echo time using a negative amplitude read magnetic field gradient (or vice versa).

Due to delay effects and image misalignment that occurs with bipolar data acquisition, the two echoes are preferably corrected for misalignment in k-space prior to reconstruction of the water and fat images. The two echoes need to be further corrected for phase errors and alignment inaccuracies in image space due to base field non-uniformity prior to reconstruction of water and fat images. For details of these correction steps and the performance of water/fat separation in k-space, refer to the above article by Lu et al.

As mentioned earlier, imaging parameters are usually chosen such that the inverse problem of the two-point Dixon method used for water/fat separation is well conditioned in the central part of k-space. However, under the same visualization parameters, the inverse Dixon two-point separation problem can be ill-conditioned in more peripheral parts of k-space. Therefore, the first region of k-space in which Dixon's two-point method is applied should be defined so that it covers the central part of k-space (where the corresponding inverse problem is well conditioned). The invention provides insight that although the two-point Dixon method becomes ill-conditioned, in this example towards the periphery of k-space, the inverse of the one-point Dixon method applied to the first and/or second echo data becomes (or remains) fairly well conditioned. Therefore, the second region of k-space where the one-point Dixon method is applied is preferably defined as covering the peripheral part of the k-space according to the invention.

In general, the invention proposes to define the first and second regions of k-space in accordance with the predictive conditioning of the inverse problems of the two-point and one-point Dixon methods in the respective regions of k-space. The predicted water/fat distortion of the corresponding Dixon separation can also be taken into account. Finally, the water image and/or the fat image is reconstructed using the k-space data after separating the water and fat contributions.

In accordance with a preferred embodiment of the invention, the contributions of water and fat in k-space are assumed to be Hermitian in those regions of k-space where the inverse problem of the corresponding two-point or one-point Dixon method is ill-conditioned. As noted above, the conditionality of the inverse problem of Dixon's method can be improved by regularization. However, this regularization introduces distortions into the assessment of the contribution of water and fat signals. The invention provides insight that inverse conditioning can be improved (without corrupting the estimate of water and fat contributions due to regularization) by limiting that the initial phases of the water and fat signal contributions in image space are equal whenever the (unrestricted) inverse the task is ill-conditioned. In this case, the water and fat contributions are assumed to be real in image space and hence Hermitian in k-space. The conditionality of the inverse problem can be determined by amplifying the noise of the corresponding water/fat separation algorithm. The inverse problem can be considered as ill-conditioned, for example, if the noise amplification level is above a given threshold value.

The approach of the invention, in which Dixon's algorithm assumes that the signal contributions from water and fat are Hermitian in k-space, provides one-point and two-point water/fat separation in k-space, which operates on two echoes. It can be used locally for selected regions of k-space (according to the definition of the corresponding inverse problem) or even globally for all regions of k-space. Unlike the usual complex water/fat separation, it often remains well-conditioned without regularization.

The method disclosed above according to the invention can be implemented by means of an MR device, including at least one main magnetic coil for generating a substantially uniform static magnetic field B ₀ in the investigated volume, a number of gradient coils for generating switchable magnetic field gradients in different spatial directions in the test volume, at least one RF body coil for generating RF pulses in the test volume and/or for receiving MR signals from the patient's body located in the test volume, a control unit for controlling the time sequence of RF pulses and switchable magnetic field gradients, and a reconstruction unit for reconstructing the MR images from the received MR signals. The method according to the present invention can be implemented by appropriate programming of the reconstruction unit and/or the control unit of the MP device.

The method according to the present invention can be successfully implemented on most MR devices in clinical use at the present time. This simply requires the use of a computer program by which the MP device is controlled to carry out the above-described steps of the method according to the invention. The computer program may either be present on a storage medium or be present on a data network by being downloaded for installation in the control unit of the MP device.

Brief description of the drawings

The accompanying drawings disclose preferred embodiments of the present invention. However, it should be understood that the drawings are for illustrative purposes only and not for delimiting the invention. On the drawings:

in fig. 1 shows an MP device for carrying out the method of the present invention;

in fig. 2 shows a schematic (simplified) pulse train plot for a Dixon imaging sequence using bipolar readout magnetic field gradients according to the invention;

in fig. 3 is a graph illustrating noise amplification by Dixon water/fat separation in k-space as a function of position in k-space.

Implementation of the invention

Consider Fig. 1, where the MP device 1 is shown in block diagram form. The device contains superconducting or resistive main magnetic coils 2 so that a substantially uniform, time-constant main magnetic field B ₀ along the z-axis is created, passing through the volume under study. The device further comprises a set of (1st, 2nd and, where applicable, 3rd order) shim coils 2', wherein the current flowing through the individual shim coils of the set 2' is controlled to minimize deviations B ₀ in within the investigated volume.

The magnetic resonance generation and manipulation system applies a series of RF pulses and switchable magnetic field gradients to invert or excite nuclear magnetic spins, induce magnetic resonance, refocus magnetic resonance, manipulate magnetic resonance, spatially and otherwise encode magnetic resonance, saturate spins, and the like. to perform MR imaging.

In particular, the gradient pulse amplifier 3 delivers current pulses to selected coils from the whole body gradient coils 4, 5 and 6 along the x, y and z axes of the volume to be examined. The digital RF transmitter 7, through the transmit/receive switch 8, transmits the RF pulses or pulse bursts to the RF body coil 9 for transmitting the RF pulses into the volume of interest. A typical MR imaging sequence consists of a burst of short duration RF pulse segments which, together with any applied magnetic field gradients, provide selective nuclear magnetic resonance manipulation. The RF pulses are used to saturate the resonance, excite the resonance, invert the magnetization, refocus the resonance, or manipulate the resonance and select the part of the body 10 located in the volume of interest. The MR signals are also picked up by the RF body coil 9.

To generate MR images of limited areas of the body 10, a set of local array RF coils 11, 12, 13 are placed adjacent to the area selected for imaging. Matrix coils 11, 12, 13 can be used to receive MR signals induced by RF transmissions via the RF body coil.

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

The host computer 15 controls the shim coils 2' as well as the gradient switching amplifier 3 and transmitter 7 to generate imaging sequences according to the invention. For the selected sequence, receiver 14 receives one or more lines of MR data in quick succession after each RF excitation pulse. The acquisition system 16 performs A/D conversion of the received signals and converts each line of MP data into a digital format suitable for further processing. In modern MR devices, the data acquisition system 16 is a separate computer dedicated to collecting raw image data.

Finally, the original digital image data is reconstructed into an image representation by a reconstruction processor 17 applying a Fourier transform or other appropriate reconstruction algorithms such as SENSE. The MR image may represent a flat section of the patient's body, an array of parallel flat slices, a three-dimensional volume, and the like. 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 imaging. , for example, with a video monitor 18 providing a human-readable display for the acquired MR image.

Host computer 15 and reconstruction processor 17 are programmed to perform the method of the invention as discussed above and below.

In FIG. 2 is a schematic diagram of a pulse train for a Dixon imaging sequence using bipolar readout magnetic field gradients according to the invention. The graph shows switchable magnetic field gradients in the frequency encoding direction (M), the phase encoding direction (P), and the slice selection direction (S). In addition, the graph shows the RF excitation pulse as well as the time intervals labeled ACQ1 and ACQ2 during which the echoes are received. The graph covers the acquisition of one pair of echoes. A number of such pairs of echoes is obtained by repeating the depicted sequence many times using different phase codes (P) to completely cover the required region of k-space. Each pair of echoes is obtained using a respective pair of magnetic field readout gradients (M) having opposite polarities. The time intervals and amplitudes of the bipolar reading gradients are chosen so as to shift the windows ACQ1, ACQ2 of receiving echo signals in such a way as to provide different echo times TE1 and TE2 and, accordingly, different phase shifts of the contributions of signals from water protons and fat protons. Dixon separation of these signal contributions is based on these phase shifts in the final stage of water and/or fat image reconstruction.

Prior to reconstruction, the two echoes are corrected for k-space misalignment and phase errors, as well as misalignment in image space due to main field non-uniformity (see, e.g., Lu et al. , Water-fat separation with bipolar multi-echo sequences, Magn Reson Med 2008, 60, 198-209).

The subsequent reconstruction involves separating the contributions from water and fat to the k-space echoes. For this purpose, the two-point Dixon method (see, for example, Eggers et al. (Eggers et al., Dual-echo Dixon imaging with flexible choice of echo times), Magn. Reson Med. 2011, 65, 96-107) is applied in the first region of k-space, and the single-point Dixon technique (see, for example, Ma, A single-point Dixon technique for fat-suppressed fast 3D gradient-echo imaging with a flexible echo time (Dixon's single-point method for imaging with fast 3D gradient echo in fat-suppression mode with flexible echo time, J. Magn. Reson. Imaging 2008, 27, 881-890) is applied in the second k-space region. to one of the two echoes, i.e. the data of the first echo or the data of the second echo Alternatively, the single point Dixon method can be applied to both echoes separately, and then the results can be combined to reduce noise.

Unlike Dixon's two-point method, Dixon's one-point method models water and fat signals in image space as real variables. In k-space, this corresponds to conjugate symmetric water and fat signals. To exploit this property, the water/fat separation is advantageously performed in k-space jointly for the respective negative and positive frequencies. Alternatively, k-space echoes to which one-point Dixon is to be applied can be converted to image space for water/fat separation since fat shift is unipolar in single echo data and can be easily corrected after water/fat separation. .

The invention proposes to determine the first and second regions of k-space in accordance with the predictive conditioning of inverse problems of two-point and one-point Dixon methods in the respective regions of k-space. This is shown in Figure 3. The graph shows noise amplification with the appropriate k-space water/fat separation algorithm as a function of position in k-space (sample size), in one case using Dixon's two-point method in the absence (dashed portion of the curve). 31) and in the presence (solid curve 31) of adaptive regularization, as in the aforementioned article by Lu et al. echo signal (dashed curve 32), as well as both individual and averaged results (curve 33). This example assumes a Dixon turbo spin echo (TSE) sequence with echo offsets of -0.5 ms and 1.5 ms at 1.5 T and a partial echo coefficient of 0.75. Near the center of k-space (sample size 256), the noise amplification in the applied two-point Dixon method is close to the optimal value of 0.5, which corresponds to a two-fold averaging. However, towards the periphery of the k-space, the noise amplification sharply increases (unless regularization is applied). As shown by solid curve 31, in this example, the noise gain is capped at 2.0. However, this regularization introduces distortions into the estimation of the contributions of signals from water and fat. While noise enhancement becomes unfavorable in k-space periphery for the two-point Dixon method (indicating that the inverse problem becomes ill-conditioned), it becomes more favorable for the one-point Dixon method applied to the first or second echo. This can be seen from curves 32, 33. Wherever one of the two curves 32, 33 remains below curve 31, it is preferable to rely on the one-point Dixon method instead of the two-point Dixon method. This characterizes the second region of k-space in the sense of the present invention. The central k-space region where the noise amplification is minimal along curve 31 forms the first k-space region in the sense of the present invention. At certain positions in k-space (also within the second region), averaging the results of the one-point Dixon method applied separately to the first and second echo data provides further noise gain reduction.

Finally, the water image and/or the fat image is reconstructed using the k-space data after the above-described separation of the water and fat contributions.

The two combined echoes S ₁ and S ₂ (after phase estimation and demodulation caused by the main field non-uniformity) can be modeled as

,

,

wherein

,

,

where W and F denote the water and fat signals in k-space, t is the acquisition time, and w and Δf are the relative amplitude and peak frequency of the fat spectral model used. The solution for W and F has the form

,

,

wherein

,

,

where λ denotes the regularization parameter.

As explained above with reference to FIG. 1-3, a suitable choice of λ allows the noise gain to be limited to the desired value. However, the diagonal loading of the C ^H C matrix introduces a deviation in the water/fat separation. To avoid this, the invention proposes to introduce a constraint so that the initial phases of the water and fat signals in image space are equal if the C ^H C matrix inversion is ill-conditioned. In this case, the water and fat contributions are assumed to be real in image space and hence Hermitian in k-space. When considering only one echo, S is described as

,

,

where indices_R and_I denote the real and imaginary parts, andk ₂ = - k _one. Turning on the second echo also results in

,

,

which forms an overdetermined linear system of eight equations. W and F in both cases are obtained as above, but without regularization. This limited Dixon method provides one-point and two-point k-space separation of water and fat, which works on sample pairs in k-space. It can be used locally for selected regions of k-space, or globally for all regions of k-space. As a measure for conditioning the inverse problem, the graph in FIG. 4 shows the noise amplification by the corresponding k-space water/fat separation algorithm as a function of position in k-space (sample size), in one case using the conventional complex two-point Dixon method in the absence (dashed portion of curve 41) and in the presence of ( the solid portion of regularization curve 41) and using the proposed limited water/fat separation without regularization, taking into account only the first echo (curve 42) and both echoes (curve 43). As shown in FIG. 4, in contrast to the conventional complex separation (curve 41), the separation limited by the real contributions of the water and fat signals (curves 42, 43) remains well-conditioned without regularization. Its noise gain exhibits only minor changes in k-space and remains within reasonable limits.

Finally, a proposed limited Dixon water/fat separation is shown as an example in FIG. 5. The pelvis of a volunteer was visualized by a 3D T ₁ -weighted double gradient echo sequence with clearing on a 3 Tesla MR device. Field of observation 350 (AP) × 350 (RL) × 200 (FH) mm ³ , resolution 0.7 × 0.7 × 2.5 mm ³ , TE ₁ /TE ₂ /TR 2.0/3.7 were chosen /5.5 ms and the deflection angle of the magnetization vector is 10°. The pixel pixel bandwidth was 650 Hz, which corresponds to a signal shift from the dominant fat spectrum peak of 0.65 pixels in single echo images. The phase caused by the main field non-uniformity and the initial phase of the water and fat signals were first determined and excluded from the single echo imaging, ignoring any alignment inaccuracy. Then complex separation was applied in k-space according to the usual Dixon algorithm, with and without regularization, as well as the proposed "hybrid complex-real" restricted separation without regularization. The results for one slice are shown in Fig. 5. In FIG. Figure 5 shows water images reconstructed using the conventional Dixon complex water/fat separation (top row), with no (left) and with (right) regularization, and using the proposed hybrid complex-real separation (bottom) without regularization. While regularization is essential in conventional complex separation to prevent high frequency noise breakthrough, it is clear that it is not necessary in the approach of the invention. With the method of the present invention, the loss of image sharpness commonly seen as an adverse consequence of regularization can be avoided without being affected by excessive noise amplification.

Claims (24)

1. The method of magnetic resonance imaging (MR) of visualization of an object (10) placed in the investigated volume of a magnetic resonance device (1), which includes the following steps: subjecting the object (10) to an imaging sequence comprising at least one radio frequency (RF) excitation pulse and switchable magnetic field gradients, wherein two echoes, a first echo and a second echo, are generated at different times (TE1, TE2) echo, receive echoes from the object (10), reconstructing the water image and/or the fat image from the echo signals, wherein the contributions from water and fat to the echo signals are separated using the two-point Dixon method in the first k-space region and the one-point Dixon method in the second k-space region, the first region different from the second area. 2. The method of claim 1, wherein two echoes are obtained using bipolar readout magnetic field gradients. 3. The method of claim 1 or 2, wherein the single point Dixon method is applied to the first echo data or to the second echo data. 4. The method of claim 1 or 2, wherein the one-point Dixon method is applied in two separate calculation steps to the first echo data and the second echo data, respectively, wherein the water and fat contributions are calculated by combining the results of the two separate calculation steps. . 5. The method according to any one of paragraphs. 1-4, in which the two echoes are corrected for k-space alignment inaccuracies prior to reconstruction of the water and fat images. 6. The method according to any one of paragraphs. 1-5, in which the two echoes are corrected for phase errors and alignment inaccuracies in image space due to base field non-uniformity prior to reconstruction of the water and fat images. 7. The method according to any one of paragraphs. 1-6, in which the first k-space region covers the central part of the k-space, while the second k-space region covers the peripheral part of the k-space. 8. The method according to any one of paragraphs. 1-7, in which the first and second regions of k-space are determined in accordance with the conditioning of the inverse problems of the two-point and one-point Dixon methods in the respective regions of k-space. 9. The method according to any one of paragraphs. 1-8, which assumes that the contributions from water and fat in k-space are Hermitian in those regions of k-space where the inverse problem of the corresponding two-point or one-point Dixon method is ill-conditioned. 10. The method according to any one of paragraphs. 1-9, in which the two-point Dixon method and/or the one-point Dixon method is applied without regularization. 11. The method of magnetic resonance (MR) imaging of an object (10) placed in the investigated volume of the magnetic resonance device (1), and the method includes the following steps: subjecting the object (10) to an imaging sequence comprising at least one radio frequency (RF) excitation pulse and switchable magnetic field gradients, wherein two echoes, a first echo and a second echo, are generated at different times (TE1, TE2) echo, receive echoes from the object (10), reconstructing the water image and/or the fat image from the echoes, wherein the contributions from water and fat to the echoes are separated using the two-point Dixon method and/or the one-point Dixon method, it being assumed that the contributions from water and fat in k-space are Hermitian in those regions of k-space where the inverse problem of the corresponding two-point or one-point Dixon method is ill-conditioned. 12. The method of claim 11 wherein two echoes are obtained using bipolar readout magnetic field gradients. 13. The method of claim 11 or 12, wherein the one-point Dixon method is applied in two separate calculation steps to the first echo data and the second echo data, respectively, wherein the contributions from water and fat are calculated by combining the results of the two separate calculation steps. 14. The method according to any one of paragraphs. 11-13, in which the two echoes are corrected for k-space alignment inaccuracies prior to reconstruction of the water and fat images. 15. The method according to any one of paragraphs. 11-14, in which the two echoes are corrected for phase errors and alignment inaccuracies in image space due to base field non-uniformity prior to reconstruction of the water and fat images. 16. The method according to any one of paragraphs. 11-15, in which the two-point Dixon method and/or the one-point Dixon method is applied without regularization. 17. Magnetic resonance (MR) device containing at least one main magnetic coil (2) for generating a uniform static magnetic field B0 in the volume under study, a number of gradient coils (4, 5, 6) for generating switchable magnetic field gradients in different spatial directions in the test volume, at least one radio frequency (RF) coil (9) for generating radio frequency pulses in the test volume and/or for receiving magnetic resonance signals from an object (10) located in the test volume, control unit (15) for controlling the time sequence of radio frequency pulses and switchable magnetic field gradients, and a reconstruction unit (17) for reconstructing magnetic resonance images from the received magnetic resonance signals, moreover, the magnetic resonance device (1) is configured to implement the steps of the method according to any one of paragraphs. 1-16. 18. A computer-readable medium containing a computer program executable on a magnetic resonance device and containing instructions for executing the method according to any one of paragraphs. 1-16.

Images