MR imaging for radiotherapy planning

Abstract

The invention relates to a method of MR imaging of a body (10) of a patient located in an examination volume of an MR device (1). It is an object of the invention to provide a method which enables a geometrically corrected MR-only radiotherapy planning with a minimum number of scans. The method of the invention comprises the following steps: acquiring first MR imaging data representing at least one region of the body (10); analyzing the first MR imaging data to delineate at least one anatomical structure within the body region; acquiring second MR imaging data of the body region using a multi-point Dixon sequence; deriving a B0 map from the second MR imaging data; analyzing the B0 map to determine at least one low-fidelity region of the B0 map; performing a B0 mapping using a multi-shot gradient echo sequence limited to at least one region which fully or partially overlaps with the delineated anatomical structure and the low-fidelity region to refine the B0 map; and, correcting geometric distortions in the first and / or second MR imaging data using the refined B0 map. Furthermore, the invention relates to an MR device (1) for performing the method and to a computer program to be executed on an MR device (1).

Description

TECHNICAL FIELD

[0001] The present invention relates to the field of magnetic resonance (MR) imaging. The invention concerns a method of MR imaging of an object located in an examination volume of an MR device. The invention further relates to an MR device and to a computer program for execution on an MR device. BACKGROUND

[0002] Image-forming MR methods which utilize the interaction of magnetic fields with nuclear spins are now widely used, particularly in medical diagnostics, for soft tissue imaging since they are superior to other imaging methods in many respects, they do not require ionizing radiation and are usually not invasive.

[0003] Generally, according to the MR method, the body of the patient to be examined is arranged in a strong, uniform magnetic field Bo whose direction at the same time defines the axis of the coordinate system on which the measurement is based (usually the z-axis). The magnetic field Bo causes the individual nuclear spins which have magnetic moments of their own due to their magnetic field strength to take on different energy levels in dependence on the electromagnetic alternating field (RF field) of defined frequency which can be excited (spin resonance). From a macroscopic point of view the distribution of the individual nuclear spins produces an overall magnetization which can be tilted out of the equilibrium state by means of suitable RF pulses, causing the magnetization to perform precessional motion about the z-axis. The precessional motion describes a surface of a cone whose opening angle is referred to as flip angle. The magnitude of the flip angle is dependent on the strength and the duration of the applied electromagnetic pulse. In the case of a so-called 90° pulse, the spins are tilted from the z-axis to the transverse plane (flip angle 90°).

[0004] After the RF pulse has terminated, the magnetization relaxes back to the initial equilibrium state where it establishes with the first time constant Ti (spin lattice or longitudinal relaxation time) in the direction of the z axis and with the second time constant T2 (spin-spin or transverse relaxation time) in the direction perpendicular to it, i.e. in the transverse plane. The relaxation in the direction perpendicular to the z axis can be detected by means of receive coils arranged and oriented within the examination volume of the MR device. After a 90° pulse, for example, the decay of the transverse magnetization occurs with an accompanying shift (dephasing) of the nuclear spins from an ordered state with the same phase to a state in which all phase angles are uniformly distributed (phase shift). This can be compensated by means of a refocusing pulse (e.g. 180° pulse). This results in an echo signal (spin echo) in the receive coils.

[0005] To achieve spatial resolution in vivo, linear magnetic field gradients extending along the three main axes are superimposed on the uniform magnetic field Bo, resulting in a linear spatial dependence of the spin resonance frequency. The signal picked up in the receiver coils then contains components of different frequencies, which can be associated with different locations within the body. The signal data obtained via the receiver coils correspond to the spatial frequency domain and are referred to as k-space data. The k-space data usually comprise a plurality of lines acquired with different phase encoding. Each line is digitized by collecting a number of samples. A set of k-space data is converted into an MR image by means of a Fourier transformation.

[0006] Radiation therapy or radiotherapy is a treatment using ionizing radiation, generally as part of cancer treatment, to control or kill malignant cells. Ionizing radiation is usually delivered in precisely adjusted doses and by a linear accelerator to selectively treat diseased tissue. Several years ago, dose calculations performed based on MR imaging have been reported, in which MR imaging is used as a complementary modality to computed tomography. In any case, since MR imaging offers excellent soft tissue contrast and high accuracy in delineating the anatomy to be irradiated (compared to computed tomography), the concept of performing all steps of a radiotherapy planning based on MR imaging as the only modality (so-called MR-only radiotherapy) becomes increasingly important.

[0007] In MR-only radiotherapy treatment planning, geometrically corrected MR imaging is of great importance, as the target tumor and organs at risk are delineated based on MR images, and patient positioning and dose planning for treatment are performed based on these delineations. There is strong interest in using MR-only variants to eliminate the redundant CT scan, and more importantly, to avoid the geometric uncertainty during CT-MR registration, which has been shown to be on the order of a few millimeters. The geometric accuracy of MR images is limited by various types of distortions. System-level distortions stem from inherent gradient nonlinearities and B0 field inhomogeneity of the MR equipment used. Both effects are largely corrected during image reconstruction based on per-system reference data. However, the chemical shift and susceptibility effects of fat cause distortions, which depend on the patient’s anatomy, and thus typically require additional data acquisition to perform correction. Spatial variations in susceptibility, most pronounced at interfaces with large susceptibility differences, such as air / tissue or bone / metal, make the B0 field inhomogeneity in the human head as high as about 4 ppm, which can make the warping at the venous sinus / tissue interfaces in the brain as high as 4 mm. The chemical shift and patient-specific distortions due to variations in susceptibility are typically addressed by scanning with high readout bandwidth. Furthermore, precise B0 mapping using multi-shot gradient echo (GRE) sequences is increasingly used to quantify and correct such distortions related to B0 field inhomogeneity. Multi-shot gradient echo sequences include multiple radio frequency (RF) excitation pulses. One or more gradient echoes are acquired after each RF excitation pulse, while the one or more gradient echoes are acquired at different echo times after the respective RF excitations. A very practical example of a multi-shot gradient echo sequence is a two-shot gradient echo sequence. Such a two-shot gradient echo sequence includes two consecutive gradient echo acquisitions after the respective RF excitation pulses, and the gradient echoes of the respective shots have different echo times. In such an imaging sequence, gradient-recalled echo signal data is acquired at two different echo times, and a B0 map is derived from the difference in signal phase at the two different echo times. The signal phase is proportional to the local field strength. However, such a sequence requires additional scan time and image processing steps (e.g., phase unwrapping) to apply the correction.

[0008] Recently, a novel and fast method for assessing image distortions in MR-only radiotherapy planning has been proposed, which uses B0 maps derived from multi-point Dixon (mDIXON) imaging (see Weiss et al. “A novel and rapid approach to estimate patient-specific distortions based on mDIXON MRI” Phys. Med. Biol. 2019, 64(15): 155002). The use of mDIXON imaging is advantageous because it is usually performed for the purpose of generating synthetic CT images in MR-only radiotherapy planning workflows (see Berker et al. “MRI-Based attenuation correction for hybrid PET / MRI systems: a 4-class tissue segmentation technique using a combined ultrashort-echo-time / Dixon MRI sequence” J. Nucl. Med. 2012, 53: 796-804; Tyagi et al. “Dosimetric and workflow evaluation of first commercial synthetic CT software for clinical use in pelvis” Phys. Med. Biol. 2017, 62(8): 2961-2975). This secondary use of mDIXON MR imaging data also leads to the inherent simultaneous acquisition of B0 maps and radiotherapy planning data, which minimizes the risk of misalignment due to patient motion. As for distortion correction, the mDIXON method allows for the correction of most patient-specific distortions using only the mDIXON-based B0 maps. Since this field map is estimated in the mDIXON reconstruction with the purpose of water / fat separation according to the requirements of the CT simulation, the corresponding correction saves the full scan time of a dedicated field mapping step, which is typically 2-4 minutes, and thus saves more time than the scan time of the mDIXON scan itself.

[0009] However, it turned out that the reduction of the scan time when using the mDIXON based B0 mapping method comes at the cost of a reduced fidelity of the B0 map (see the article of Weiss et al. cited above). Although the difference between the distortions derived from the mDIXON imaging on the one hand and the distortions derived from the dedicated B0 mapping on the other hand is overall much smaller than the total distortions estimated by the pure B0 mapping, almost all differences observed between the two methods which are larger than 50% are very locally present in regions with large B0 spatial variations. If these local errors in the distortion estimation are present in locations to be treated by radiotherapy, they become very important. The tumor can not receive the full dose as planned or a risk organ can receive a higher dose than planned. SUMMARY

[0010] From the foregoing, it can be readily appreciated that there is a need for an improved MR imaging technique. It is therefore an object of the present invention to provide a method enabling a geometrically corrected MR-only radiotherapy planning with a minimum of scan time.

[0011] According to the present invention, a method of MR imaging a body of a patient located in an examination volume of a MR device is disclosed. The method comprises the following steps:

[0012] - acquiring first MR imaging data representing at least one region of the body;

[0013] - analyzing the first MR imaging data to delineate at least one anatomical structure within the body region;

[0014] - acquiring second MR imaging data of the body region using a multi-point Dixon sequence;

[0015] - deriving a B0 map from the second MR imaging data;

[0016] - analyzing the B0 map to determine at least one low-fidelity region of the B0 map;

[0017] - performing a B0 mapping to refine the B0 map using a multiple-acquisition gradient echo sequence limited to at least one region where the delineated anatomical structure and the low-fidelity region fully or partially overlap; and

[0018] - correcting geometric distortions in the first and / or second MR imaging data using the refined B0 map.

[0019] It is proposed according to the invention to acquire the first MR imaging data, for example using a T2-weighted MR imaging sequence, which provides a suitable contrast for depicting anatomical structures such as tumors or organs at risk. The second MR imaging data is acquired using a multi-point Dixon method. This mDIXON data is used for the purpose of generating a synthetic CT image in a MR-only radiotherapy planning workflow. Since the first and second MR imaging data are acquired in the same reference frame by the same modality, the depiction of the anatomical structures can be transferred to the second MR imaging data (and the resulting synthetic CT image) without loss of geometric accuracy.

[0020] According to known mDIXON techniques, the spectral difference between fat and water protons is exploited to separate MR signals emitted from water-containing tissue and MR signals emitted from fat tissue. In mDIXON, multiple acquisitions of k-space are repeated with different echo times. The simplest mDIXON technique, 2-point Dixon, acquires two complete k-space data sets, where the fat magnetization in the second acquisition is out of phase with respect to the first acquisition at the respective echo times. By simple addition or subtraction of the complex MR signal data sets, separate and distinct water and fat images are obtained. In general, a B0 map, a water map and a fat map can be obtained by mDIXON techniques. A specific advantage is that B0 mapping using mDIXON is very fast and provides, in addition to the B0 map, useful information about the water distribution and fat distribution within the field of view in the form of water and fat maps. According to the present invention, the B0 map is utilized to determine and correct geometric image distortions caused by the spatial distribution of B0.

[0021] A basic step of the inventive method is the step of analyzing the mDIXON B0 map to determine one or more low-fidelity regions. As indicated above, the mDIXON B0 map lacks fidelity only in very limited regions, so that the accuracy of the B0 map is fully sufficient to correct distortions with sufficient accuracy in a wide range of regions. It is thus proposed to automatically identify those regions in the mDIXON-based B0 mapping and related distortion estimation which are very erroneous, so that the dose planning is compromised. It is proposed to use a gradient echo sequence limited to at least one region overlapping the depicted anatomical structure and the low-fidelity region to automatically perform a dedicated B0 mapping to refine the B0 map. In this way, the scan time consumption for B0 mapping is limited to a minimum by acquiring a conventional B0 map (using a multi-acquisition gradient echo sequence, in particular a two-acquisition gradient echo sequence) only in those regions relevant for radiotherapy planning, i.e. in those regions where the mDIXON B0 accuracy can be assumed to be insufficient, but only in those low-fidelity regions overlapping the regions to be subjected to radiotherapy, i.e. in regions covered by the depicted anatomical structure.

[0022] The B0 map refined in this way in the respective overlapping regions is finally used to correct the first and / or second MR imaging data, thereby achieving a high level of geometric accuracy at a minimum scan time.

[0023] In a preferred embodiment, the B0 map is determined to be of low fidelity at locations where both the amplitude of the B0 gradient and the amplitude of the geometric distortion associated with B0 are above respective predetermined thresholds. In this way, only those regions will be subject to the refined B0 map which can lead to large errors in dose planning. These regions are characterized by large geometric distortions (i.e. high B0 bias) and at the same time high spatial gradients of B0 (i.e. strong spatial variation of B0). The respective thresholds can be selected by a user to achieve an optimal trade-off between scan time and geometric correction.

[0024] The reduced fidelity of the B0 map is of particular relevance for single- or two-point mDixon imaging with flexible echo spacing (i.e. not in-phase and out-of-phase for water and fat components of the magnetic resonance signal). Therefore, in another embodiment of the method of the present invention, such single- or two-point mDixon sequence is combined with at least a three-point mDixon sequence to refine the B0 map, wherein the delineated anatomical structure and the low-fidelity region overlap completely or partially.

[0025] In another preferred embodiment, a simulated CT image is computed from the corrected first and / or second MR imaging data by assigning Hounsfield unit values to each pixel or voxel of the second MR imaging data. A simulated CT image is required for actual dose calculation in generating a radiotherapy plan, because a CT image with Hounsfield unit values assigned to each image location provides the required imaging tissue radiation attenuation properties. It is known that mDIXON imaging providing water / fat separation is effective for the purpose of generating a synthetic CT image in a MR-only radiotherapy planning workflow (see reference cited above). It is also possible to compute the simulated CT image based on the first MR imaging data, any further available MR imaging data or any combination of these data. Any additional MR contrast information of the same region can be used to determine the Hounsfield unit for each image location.

[0026] In a possible embodiment of the present invention, the step of analyzing the first MR imaging data involves manual or automatic segmentation of the relevant anatomical structure (e.g. a tumor or an organ at risk). When using automatic segmentation, the method of the present invention can be performed in a fully automatic manner.

[0027] In yet another preferred embodiment, the overlay of the depicted anatomical structure with the determined low-fidelity region of the B0 map is displayed to the user. This informs the user of potential errors in the dose planning. The user can react by adjusting the dose plan. This can specifically be advisable in regions where the low-fidelity region (part) intersects with the depicted tumor or risk organ.

[0028] The method of the application as thus described can be performed by an MR device comprising at least one main magnet coil for generating a uniform, steady magnetic field Bo within an examination volume, a number of gradient coils for generating switched magnetic field gradients in different spatial directions within the examination volume, at least one RF coil for generating RF pulses within the examination volume, one or more receiver coils for receiving MR signals from an object located in the examination volume, a control unit for controlling the temporal succession of RF pulses and switched magnetic field gradients, and a reconstruction unit. The method of the application is implemented by a corresponding programming of the reconstruction unit and / or the control unit of the MR device.

[0029] The method of the application can advantageously be implemented in most MR devices currently in clinical use. For this purpose, it is only necessary to utilize a computer program running on a computer by which the MR device is controlled such that it performs the above-mentioned method steps of the application. The computer program can be present on a data carrier or in a data network in order to be downloaded for installation in the control unit of the MR device. BRIEF DESCRIPTION OF DRAWINGS

[0030] The drawings disclose preferred embodiments of the application. It is to be understood, however, that the drawings are designed solely for purposes of illustration and are not intended to limit the definition of the present application. In the drawings:

[0031] Figure 1 An MR device for performing the method of the application is shown;

[0032] Figure 2 The method of the application is shown schematically in a flow chart;

[0033] Figure 3 The process of analyzing the B0 map to determine the low-fidelity region is shown schematically in a flow chart;

[0034] Figure 4 MR image data illustrating the background of the method of the application is shown. DETAILED DESCRIPTION

[0035] REFERENCE Figure 1, a MR device 1 is shown. The device includes a superconducting or resistive main magnet coil 2 creating a substantially uniform, time invariant main magnetic field Bo along the z axis through an examination volume. The device also includes a set of shim coils 2' in which the current through each shim coil of the set 2' is controllable for minimizing Bo inhomogeneity within the examination volume.

[0036] Magnetic resonance generation and manipulation system applies a series of RF pulses and switched magnetic field gradients to invert or excite nuclear magnetic spins, induce magnetic resonance, refocus magnetic resonance, manipulate magnetic resonance, spatially or otherwise encode magnetic resonance, saturate spins, etc. to perform MR imaging.

[0037] More specifically, a gradient pulse amplifier 3 applies 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 a body RF coil 9 for transmitting RF pulses into the examination volume. A typical MR imaging sequence consists of a package of short duration RF pulse segments combined together with any applied magnetic field gradients to selectably manipulate nuclear magnetic resonance. The RF pulses serve to saturate, excite resonance, invert magnetization, refocus resonance, or manipulate magnetic resonance and select a portion of the body 10 located in the examination volume. MR signals are also picked up by the body RF coil 9.

[0038] To generate MR images of a limited region of the body 10, a set of local array RF coils 11, 12, 13 is placed adjacent the region selected for imaging. The array coils 11, 12, 13 can be used as receive coils to receive MR signals induced by body-coil RF transmission.

[0039] The resulting MR signals are picked up by the body RF coil 9 and / or the array RF coils 11, 12, and 13 and demodulated by a receiver 14, which preferably includes preamplifiers (not shown). The receiver 14 is connected via the transmit / receive switch 8 to the RF coils 9, 11, 12, and 13.

[0040] A host computer 15 controls the shim coils 2' as well as the gradient pulse amplifier 3 and the transmitter 7 to generate any one of a number of MR imaging sequences, such as echo planar imaging (EPI). For the selected sequence, the receiver 14 receives single or multiple MR data lines in rapid succession after each RF excitation pulse. A data acquisition system 16 performs analog-to-digital conversion of the received signals and converts each MR data line to a digital format for further processing. In modern MR devices, the data acquisition system 16 is a separate computer dedicated to acquiring raw image data.

[0041] Ultimately, the digital raw image data is reconstructed into an image representation by a reconstruction processor 17 which applies a Fourier transform or other appropriate reconstruction algorithm, such as SENSE. The MR image can 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, which can be accessed, e.g., via a video monitor 18 providing a human readable display of the resulting MR image, to convert slices, projections, or other portions of the image representation into an appropriate format for visualization.

[0042] Embodiments of the method of the present application are described with reference to Figures 2-4 and further with reference to Figure 1 are described as follows:

[0043] After positioning the body 10 in the examination volume of the main magnet coil 2, a first MR imaging scan is started in step 21 using T2-weighted scanning, for example, to acquire first MR imaging data. The first MR imaging data represents a region of the anatomical body of the body 10.

[0044] In step 22, a delineation of at least one anatomical structure, e.g., a tumor to be treated or a risk organ which shall be protected from irradiation in a radiation therapy, is performed manually, e.g., by a radiologist interactively analyzing the first MR imaging data displayed on the video monitor 18, or by a suitable automatic segmentation technique known per se in the art. The result of the delineation can be a map (referred to as TRB map in the following) covering only the border of the delineated anatomical structure. The width of the border can be a preset parameter depending on the type of tumor or risk organ.

[0045] In step 23, second MR imaging data is acquired covering the same region of the anatomical body of the body 10. For this purpose, a multi-point Dixon technique is employed. Step 23 comprises deriving a fat map, a water map and a B0 map from the acquired mDIXON data.

[0046] In step 24, the derived B0 map is analyzed to determine one or more low-fidelity regions. This involves Figure 3The steps depicted in figure 2. In step 33, the magnitude of the spatial gradient of the B0 map is computed for each image location. In step 34, a map G is generated that covers all regions (i.e. indicates all image locations) for which the magnitude is greater than a predetermined threshold. In step 35, a map D is computed that covers all regions for which the geometric distortion caused by the B0 map at the respective image locations is greater than a further predetermined threshold. The measure of local geometric distortion can be, for example, the magnitude of the pixel / voxel shift caused by the respective local B0 value. Finally, in step 36, a map E = G D is computed that covers all regions covered by both maps G and D. E marks the image locations at which low fidelity of the B0 map can be expected. E only encompasses those regions in which the mDIXON B0 map is known to be prone to errors. It can be expected that a geometric correction based on the mDIXON B0 map only in these low-fidelity regions will lead to significant errors in the dose planning for radiotherapy treatment, as the geometric distortion itself and its spatial variation is large.

[0047] In step 25, the superposition of the low-fidelity map E with the result of the delineation of the anatomical structure in step 22 is displayed on the video monitor 18. This informs the user about potential errors in the dose planning. The user can react by adjusting the dose planning, which can be advisable in regions where E and TRB coincide.

[0048] In step 26, a map R = E TRB is computed that indicates the image regions where the delineated anatomical structure and the low-fidelity regions of the B0 map (fully or partially) overlap, i.e. the error-prone regions of the B0 map at the tumor or risk organ boundaries.

[0049] In step 27, a field of view is determined that covers the map R. The field of view can be subdivided into several different regions. In step 28, an automatic dedicated B0 mapping scan using a multi-shot gradient echo imaging sequence is performed on the determined field of view. The B0 mapping scan can consist of several sub-scans, each sub-scan being directed at a different region of the field of view.

[0050] Then in step 29, the mDIXON B0 map is updated accordingly in the regions indicated by the map R by replacing the B0 values of the mDIXON B0 map with the corresponding values obtained from the dedicated B0 mapping scan in these regions. The result is a refined, i.e. higher-fidelity, B0 map.

[0051] In step 30, the refined B0 map is then used to correct the geometric distortions in the first and second MR imaging data.

[0052] In step 31, a simulated CT image is computed from the distortion-corrected second MR imaging data. This involves assigning Hounsfield unit values to each pixel or voxel of the second MR imaging data.

[0053] In step 32, the user / radiologist uses the simulated CT image and the geometry-corrected first MR imaging data to perform dose planning for radiotherapy. If the tumor and the organs at risk are delineated using automatic segmentation in step 22, all steps 21-31 can be performed completely automatically. In particular, the automatic acquisition of the B0 map can be updated directly after the mDIXON scan. This means that the patient does not have to stay in the examination volume of the MR device 1. Patient throughput can thus be maximized.

[0054] It must also be noted that, in typical cases, the map R will contain only very few or even no regions at all. This means that the additional acquisition time required for the dedicated B0 mapping scan is either zero or at least very small. Thus, the approach proposed by the present invention combines high precision of distortion correction and very short acquisition time if compared to conventional B0 mapping for full field of view.

[0055] Figure 4 An example of a sagittal slice of MR image data of the head / neck region is shown. The image values of the two top images (a and 4b) indicate the local geometric image distortion (magnitude of the B0-induced voxel shift). The top image (a) shows the geometric distortion derived from the mDIXON B0 map. The lower image (b) shows the geometric distortion derived from the conventional dedicated B0 mapping. Figure 4 Figure 4 Figure 4 The third image of c shows the difference of the two images of a and b. Figure 4 Figure 4 The two images of a and b. Figure 4 Figure 4 The image of d is the mDIXON in-phase image shown as an anatomical reference. It can be seen that in regions with low spatial variation (e.g. the region indicated by the white circle 42), the mDIXON estimates the geometric distortion well. More than 50% of the difference occurs in regions of the B0 with high spatial variation (circle 41). As can be seen in the ​ The mDIXON B0 map with low fidelity mainly fills around the nasal cavity, the oral and ear cavities, the sphenoid sinus and dental fillings. If a tumor to be treated by radiotherapy or an organ at risk to be protected from irradiation is within one of these regions, there is a risk of severe misregistration and false dose calculation if the geometric correction is based on the mDIXON B0 map only. For this reason, the present invention proposes to refine the mDIXON B0 map in a targeted manner only in the relevant regions.​​​​

Claims

1. A method of MR imaging of a body (10) of a patient located in an examination volume of an MR device (1), the method comprising the following steps: - acquiring first MR imaging data representing at least one region of the body (10); - analyzing the first MR imaging data to delineate at least one anatomical structure within the body region; - acquiring second MR imaging data of the body region using a multi-point Dixon sequence; - deriving a Bo map from the second MR imaging data; - analyzing the Bo map to determine at least one low-fidelity region of the Bo map; - performing Bo mapping using a multi-shot gradient echo sequence limited to at least one region fully or partially overlapping the delineated anatomical structure and the low-fidelity region, in particular using a two-shot gradient echo sequence, to refine the Bo map; and - correcting geometric distortions in the first and / or second MR imaging data using the refined Bo map.

2. The method of claim 1, wherein, The Bo map is determined to have low fidelity at locations where both the magnitude of the Bo gradient and the magnitude of the geometric distortion associated with Bo are above respective predetermined thresholds.

3. The method of claim 1 or 2, wherein, A simulated CT image is computed from the corrected first and / or second MR imaging data by assigning Hounsfield unit values to each pixel or voxel of the second MR imaging data.

4. The method of claim 3, wherein, The simulated CT image is used to generate a radiotherapy plan.

5. The method of claim 1 or 2, wherein, The analysis of the first MR imaging data involves automatic segmentation of the anatomical structure.

6. The method of claim 1 or 2, wherein, An overlay of the delineated anatomical structure and the low-fidelity region is displayed.

7. A method of MR imaging of a body (10) of a patient located in an examination volume of an MR device (1), the method comprising the following steps: - acquiring first MR imaging data representing at least one region of the body (10); - analyzing the first MR imaging data to delineate at least one anatomical structure within the body region; - acquiring second MR imaging data of the body region using a single-point or dual-point Dixon sequence; - deriving a Bo map from the second MR imaging data; - analyzing the Bo map to determine at least one low-fidelity region of the Bo map; - performing Bo mapping using at least a three-point Dixon sequence limited to at least one region fully or partially overlapping the delineated anatomical structure and the low-fidelity region to refine the Bo map; and - correcting geometric distortions in the first and / or second MR imaging data using the refined Bo map.

8. MR device, comprising: at least one main magnet coil (2) for generating a uniform, steady magnetic field Bo within an examination volume; a plurality of gradient coils (4, 5, 6) for generating switched magnetic field gradients in different spatial directions within the examination volume; at least one RF coil (9) for generating RF pulses within the examination volume; one or more receive coils (11, 12, 13) for receiving MR signals from a body (10) of a patient positioned in the examination volume; a control unit (15) for controlling the temporal succession of RF pulses and switched magnetic field gradients; and a reconstruction unit (17), wherein the MR device (1) is arranged to perform the following steps: acquiring first MR imaging data representing at least one region of the body (10); analyzing the first MR imaging data to delineate at least one anatomical structure within the body region; acquiring second MR imaging data of the body region using a multi-point Dixon sequence; deriving a Bo map from the second MR imaging data; analyzing the Bo map to determine at least one low-fidelity region of the Bo map; performing a Bo mapping using a multi-shot gradient echo sequence or an at least three-point Dixon sequence that is restricted to at least one region that overlaps the delineated anatomical structure and the low-fidelity region to refine the Bo map; and correcting geometric distortions in the first and / or second MR imaging data using the refined Bo map.

9. A computer program comprising program code for causing an MR device to perform the steps of the method according to any one of claims 1-7 when the computer program is executed on a computer controlling the MR device.

Images