MR imaging using motion compensated image reconstruction

Abstract

The invention relates to a method for MR imaging of an object (10). The object of the invention is to achieve MR imaging in the presence of motion of the imaged object, wherein the acquired MR signals are fully utilized and a high-quality MR image substantially free of motion artifacts is obtained. The method of the invention comprises the following steps: generating MR signals by subjecting the object (10) to an imaging sequence comprising RF pulses and switched magnetic field gradients; acquiring the MR signals as signal data within a given time period (T); subdividing the time period into a plurality of consecutive time segments (S0, S1, S2, ..., Sn); deriving a geometric transformation (DVF1, DVF2, ..., DVFn) in image space for each pair of consecutive time segments (S0, S1, S2, ..., Sn), the geometric transformation (DVF1, DVF2, ..., DVFn) reflecting the motion occurring between the two time segments of the corresponding pair; and reconstructing the MR image from the signal data, wherein motion compensation is applied according to the derived geometric transformation (DVF1, DVF2, ..., DVFn). Furthermore, the invention relates to an MR device (1) and a computer program for an MR device (1).

Description

Technical Field

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

[0002] Today, MR imaging methods that utilize the interaction between magnetic fields and nuclear spins to form two-dimensional or three-dimensional images are widely used, especially in the field of medical diagnosis, because they are superior to other imaging methods in many aspects for imaging soft tissues, do not require ionizing radiation and are generally non-invasive.

[0003] Typically, according to the MR method, the body of the patient to be examined is arranged in a strong and uniform magnetic field B0, the direction of which simultaneously defines the axis of the coordinate system relevant for the measurement (usually the z-axis). The magnetic field B0 generates different energy levels for individual nuclear spins depending on the magnetic field strength, which can be excited (spin resonance) by applying an electromagnetic alternating field (RF field) of a defined frequency (the so-called Larmor frequency or MR frequency). From a macroscopic point of view, the distribution of the individual nuclear spins produces an overall magnetization (which can be deviated from the equilibrium state by applying an electromagnetic pulse (RF pulse) of appropriate frequency), while the corresponding magnetic field B1 of the RF pulse extends perpendicular to the z-axis, so that the magnetization performs a precession around the z-axis. The precession describes the surface of a cone, the aperture angle of which is called the flip angle. The size of the flip angle depends on the intensity and duration of the applied electromagnetic pulse. In the case of a so-called 90° pulse, the magnetization is deflected from the z-axis to the transverse plane (flip angle 90°).

[0004] After the RF pulse is terminated, the magnetization relaxes back to the original equilibrium state, wherein the magnetization in the z direction is reestablished with a first time constant T1 (spin lattice or longitudinal relaxation time) and the magnetization in the direction perpendicular to the z direction relaxes with a second shorter time constant T2 (spin-spin or transverse relaxation time). The transverse magnetization and its changes can be detected with the aid of a receiving RF coil, which is arranged and oriented in the examination volume of the MR device so that the changes in the magnetization are measured in a direction perpendicular to the z axis. The decay of the transverse magnetization is accompanied by dephasing after the RF excitation due to local magnetic field inhomogeneities, which promotes the transition from an ordered state with the same signal phase to a state in which all phase angles are uniformly distributed. The dephasing can be compensated with the aid of a refocused RF pulse (e.g., a 180° pulse). This generates an echo signal (spin echo) in the receiving coil.

[0005] In order to achieve spatial resolution in the body, time-varying magnetic field gradients extending along the three main axes are superimposed on the uniform magnetic field B0, resulting in a linear spatial dependence of the spin resonance frequency. The signal picked up in the receiving coil then contains components of different frequencies, which can be associated with different locations in the body. The signal data obtained via the receiving coil corresponds to the spatial frequency domain and is called k-space data. K-space data typically includes multiple lines of different phase encodings acquired. Each line is digitized by collecting multiple samples. A set of k-space data is converted into an MR image with the aid of Fourier transform.

[0006] Motion during MR signal data acquisition is a source of various image artifacts. One way to reduce motion artifacts is to modify the acquisition in such a way that data is acquired under the same motion conditions, either by gating or triggering or by requiring the patient to hold their breath. These methods all have the disadvantage that they reduce the scan rate and require patient cooperation.

[0007] Another way to reduce motion artifacts is to use dedicated reconstruction algorithms that account for motion as part of the reconstruction model.

[0008] In addition, gating techniques have been developed that only accept MR signal data acquired within a certain predetermined respiratory gating window. A multi-gating window approach (called PAWS, see US7039451B1) has been proposed that uses multiple separate motion states (bins) instead of a predetermined gating window. Each motion state corresponds to one of multiple continuous ranges of displacement caused by the motion of the examined body. The final MR image in PAWS is reconstructed based on the MR signal data attributed to the motion state, for which a complete set of MR signal samples is first acquired. The motion binning can be controlled by one or more navigator signals, which can be derived in many different ways, such as from a central k-space navigator, a pencil beam navigator, from an ECG signal, from a respiratory belt, a camera, etc.

[0009] One disadvantage of the binning approach described above is that attributing the MR signal data to a certain motion state is a difficult signal processing task. Furthermore, binning the data into different motion states often fails in practice because the actual motion does not correspond to the model of quasi-periodic motion. Confounding factors are, for example, offsets that may not be well picked up by the navigator signal, unexpected significant motions, and transitions between different breathing patterns.

[0010] Another disadvantage of the binning approach is that some bins always contain more data than others. That is, the signal-to-noise ratio / artifact level from different bins may differ to a large extent. Many bins are often filled to such a small extent that no useful image can be reconstructed for these bins at all. Therefore, a considerable part of the MR signal data is essentially discarded in the reconstruction, which renders the approach ineffective. Due to this problem, it is practically impossible to fully image the continuous time course of the contrast agent dynamics in a DCE ("dynamic contrast enhancement") scan.

[0011] The article “An optimal radial profile order based on the golden ratio for time-resolved MRI” by S. Winkelmann et al. (IEEE Trans. Med. Im. 26 (2007) 68-76) mentions reconstructing highly undersampled data sets for motion analysis and retrospectively adding motion-corrected data to obtain morphological images with higher spatial resolution. Summary of the invention

[0012] From the foregoing, it is readily appreciated that there is a need for MR imaging techniques with improved motion compensation. It is therefore an object of the present invention to enable MR imaging in the presence of motion of an imaged object, wherein the acquired MR signals are fully utilized and high-quality MR images substantially free of motion artifacts are obtained.

[0013] According to the present invention, a method for MR imaging of an object positioned in an examination volume of an MR device is disclosed. The method comprises the following steps:

[0014] generating MR signals by subjecting the object to an imaging sequence comprising RF pulses and switched magnetic field gradients;

[0015] Acquiring the MR signal as signal data within a given period of time;

[0016] subdividing the time period into a plurality of consecutive time segments;

[0017] deriving, for each pair of consecutive time segments, a geometric transformation in image space that reflects motion occurring between the two time segments of the corresponding pair; and

[0018] An MR image is reconstructed from the signal data, wherein motion compensation is applied according to the derived geometric transformation.

[0019] The present invention provides a solution for the above purpose by subdividing the time period of acquiring signal data into a series of short time segments so that no motion occurs or only negligible motion occurs in each time segment. The motion is determined as a geometric transformation. This is based on the following assumption: the difference in the patient position in adjacent time segments can be described by a geometric transformation (e.g., a displacement vector field) in the image space. This geometric transformation between consecutive time segments is used in the reconstruction according to the present invention by aligning the different motion states that occur over the entire duration of the scan.

[0020] Thus, a high quality MR image is obtained that contains contributions from all signal data.

[0021] In addition to the MR images, the method of the invention generates a geometric transformation that reflects the motion that occurred during the scan. This allows the MR images to be transformed into all motion states at different time segments.

[0022] In addition, reconstruction can be performed by corresponding application of the determined geometric transformation in the reconstruction procedure so that an MR image is reconstructed for any one of the time segments. Therefore, a high-quality MR image can be reconstructed for each time segment and thus the motion state. Therefore, the method of the present invention can be used to reconstruct a dynamic series of MR images.

[0023] While the previously mentioned binning methods (eg PAWS) restrict the use to repetitive motions (respiratory or cardiac), the method of the present invention is generally applicable to all types of motion, independent of spatial or temporal characteristics.

[0024] The present invention acquires data portions of magnetic resonance signals for consecutive time segments (typically sampled from different portions of k-space). Within each individual time segment, motion can be ignored, and motion between consecutive time segments is represented by geometric transformations. For each moment of an individual time segment, the geometric transformation from each of the (other) time segments can be formed by synthesizing the geometric transformations between several pairs of consecutive time segments. For each moment, a motion-corrected data portion can be formed from the acquired data portion; the correction follows directly from the applicable synthesized geometric transformation to the moment at the problem of the data portion from other moments. Finally, for each moment, a spatially high-resolution magnetic resonance image can be reconstructed from the appropriately corrected data portion for that moment. Thus, a dynamic series of spatially high-resolution diagnostic quality magnetic resonance images is obtained over the entire time period formed by the common consecutive time segments. Thus, for each moment, a spatially high-resolution magnetic resonance image corresponds to a frame reconstructed from the data portion acquired at that moment and the corrected data portion of the other time segments. At an individual time segment, only a portion of the k-space is sampled, and a spatially high-resolution magnetic resonance image is reconstructed from the entire k-space sampled from the set of motion-corrected data from the corresponding time segment. That is, the spatially high-resolution magnetic resonance image is corrected based on the inter-frame motion occurring within the sampling of the k-space data for reconstructing the spatially high-resolution magnetic resonance image.

[0025] In a preferred embodiment of the present invention, a low-resolution MR image is reconstructed for each time segment based on the signal data belonging to the corresponding time segment, and the geometric transformation is derived as a displacement vector field by registering the low-resolution MR images of the corresponding pair with each other. Preferably, a radial or spiral acquisition strategy is used so that signal data from the central k-space can be used for each time segment, thereby achieving reconstruction of the low-resolution MR image.

[0026] In a preferred embodiment, an iterative reconstruction technique is applied to reconstruct MR images, preferably in combination with parallel imaging. In the parallel imaging technique, a set of RF coils with different spatial sensitivity profiles are used for MR signal acquisition. Parallel imaging allows a reduction in phase encoding steps and, therefore, a reduction in scanning time. In general, the essence of parallel imaging reconstruction is to solve a system of linear equations representing the encoding scheme, or more specifically, to invert a generalized encoding matrix determined by the applied magnetic field gradient modulation and the coil sensitivity profile. The size of the encoding matrix is ​​quite large, and a direct inversion is numerically prohibitive. An efficient way is to perform reconstruction iteratively, as proposed, for example, by Pruessmann et al. (Magn. Reson. Med., 2001, vol. 46, pp. 638-651). In this technique, the large system of linear equations on which the image reconstruction is based is solved as a least squares minimization problem (optionally with a suitable regularization scheme) using a conjugate gradient iteration approach. The present invention proposes an adaptation of this known iterative reconstruction scheme. Reconstruction iteratively refines the reconstructed MR image, wherein sequentially for each time segment, the geometric transformations belonging to the respective time segment are applied and the MR image is caused to conform to the signal data belonging to the respective time segment. By passing through all time segments in this way, all signal data are used to iteratively reconstruct the MR image with inherent compensation for motion occurring during acquisition.

[0027] In a preferred embodiment of the method of the invention, an inconsistency of the MR signal attributable to each time segment is determined, wherein if an inconsistency is detected, the time segment is subdivided into a greater number of shorter time segments.

[0028] Furthermore, if at least one of the derived geometric transformations indicates that too fast or too strong a motion occurs between two time segments of the corresponding pair, the time period may be subdivided into a larger number of shorter time segments. On the other hand, if the derived geometric transformation indicates that there is no motion occurring between consecutive time segments or that there is only insignificant motion occurring between consecutive time segments, the time period may be subdivided into a smaller number of longer time segments. In this way, the geometric transformation information may be advantageously used to refine the initial subdivision of the time segments, optionally in combination with a data consistency analysis. Depending on the motion occurring during the MR signal acquisition, the acquired time period may be subdivided into finer or coarser time segments, as indicated by the corresponding derived geometric transformation.

[0029] In another preferred embodiment of the present invention, MR signals are acquired according to a star stacking scheme. In the known so-called three-dimensional (3D) star stack (stack-of-stars) acquisition scheme (for example, see WO2013 / 159044A1), a number of spatially non-selective or slat-selective RF excitations are applied, and one or more MR signals (for example, gradient echo signals) are acquired after each excitation, wherein each MR signal represents a k-space profile. MR signals are acquired from multiple parallel k-space planes as a radial k-space profile. The slices are arranged at different positions along a direction perpendicular to the direction of the radial k-space profile. In this direction (usually, k z Standard Cartesian phase encoding is performed in the k direction, while the MR signal is encoded in each single slice along the direction around the center (k x =k y =0). This results in a cylindrical k-space coverage consisting of stacked disks (a "star stack"). Technically, this is achieved by simultaneously generating magnetic field gradients in the in-plane direction of the k-space plane and modulating their amplitudes. Different schemes can be used to select the temporal order of the k-space profile acquisition steps. For example, all phase encoding steps along the slice direction can be acquired sequentially before acquiring k-space profiles at different angular positions (rotation angles). This ensures that the period of Cartesian sampling remains short, resulting in high data consistency within the slice stack and retaining the general motion robustness of radial sampling for the star stack approach. The Cartesian phase encoding step can be performed from the center slice towards the periphery of k-space (from the center to the outside), or from -k z,max To+k z,max The imaging sequences are performed in a linear order. Optionally, a Cartesian phase encoding scheme can be selected to acquire the central 3D volume of k-space more frequently than the peripheral k-space parts. For angular ordering, the imaging sequence can use equidistant angular sampling with multiple overlaps, or the so-called golden angle scheme. In the equidistant scheme, the angular distance (i.e., the increment of the rotation angle of the radial k-space profile) is calculated according to ΔΦ=180° / n. total To calculate, where n total is the total number of spokes. Acquiring spokes using multiple overlaps (or "rotations") can be beneficial because overlaps reduce temporal coherence in k-space. As a result, motion inconsistencies are spread out in k-space and artifacts are attenuated. In the golden angle scheme, the rotation angle of the k-space profile is increased each time by ΔΦ=111.25°, which corresponds to 180° multiplied by the golden ratio. Therefore, subsequently sampled spokes always add supplementary information while filling the largest gap within a previously sampled set of spokes. As a result, any set of successively acquired spokes covers k-space approximately uniformly. Specifically in combination with sampling the central volume of k-space more densely in time than the peripheral portions in k zCartesian phase encoding in the direction, which facilitates the reconstruction of a low-resolution MR image for each time segment according to the invention, is used to derive a geometric transformation for each pair of consecutive time segments.

[0030] Similarly, in the also known spiral stack acquisition scheme, after each non-selective or slat-selective RF excitation, one or more MR signals representing the spiral k-space profile are acquired. As in the star stack method, the planar spirals are arranged at different positions in k-space along a direction perpendicular to the plane of the spiral, wherein standard Cartesian phase encoding is performed in the direction perpendicular to the plane, while within each single plane along the source point at the center of k-space (k x =k y =0) to acquire MR signals.

[0031] The aforementioned 3D radial star stacking scheme and spiral stacking scheme offer several advantages for clinical 3D and 4D MR imaging, such as high intrinsic motion robustness and benign aliasing artifacts.

[0032] The MR signals may also be sampled according to the so-called "koosh ball" sampling scheme. This technique provides nearly silent MR imaging, in which the RF excitation and the acquisition of the MR signals are performed in the presence of a magnetic field gradient. The magnetic field gradient is applied for purely frequency-encoded, radial center-out k-space encoding. The spatially non-selective excitation must uniformly cover the full frequency bandwidth spanned by the readout magnetic field gradient, which is usually accomplished by radiating short, hard RF pulses. The acquisition of the free induction decay (FID) signal begins immediately after the radiating of the RF pulse. After the FID readout, only a minimum time is required to set the next readout magnetic field gradient before the next RF pulse can be applied, thus achieving a very short repetition time (TR). The magnetic field gradient vector that determines the readout direction changes incrementally from repetition to repetition until the spherical volume in k-space is sampled to the desired extent. Such a radial center-out k-space scanning technique is called "koosh ball" scanning, in which the radial k-space "spokes" and their arrangement in k-space constitute the filaments (chords) of the known toy ball design. Without the need to turn off the readout magnetic field gradients during the entire scan, MR imaging can be performed almost silently (see, for example, Weiger et al., Magnetic Resonance in Medicine, vol. 70, p. 328-332, 2013).

[0033] In another possible embodiment of the present invention, the imaging sequence adopted by the method of the present invention can be, for example, a fast field echo (TFE) sequence or a balanced (fast) field echo sequence or echo planar imaging (EPI) or a fast spin echo (TSE) sequence or a GRASE sequence.

[0034] The method according to the invention described herein can be performed with the aid of an MR device comprising at least one main magnet coil for generating a uniform, stable 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 body of a patient positioned in the examination volume, the control unit for controlling the temporal succession of the RF pulses and the switched magnetic field gradients, and the reconstruction unit for reconstructing an MR image from the received MR signals. The method according to the invention can be implemented by corresponding programming of the reconstruction unit and / or the control unit of the MR device.

[0035] The method of the present invention can be advantageously performed on most MR devices currently used clinically. To this end, it is only necessary to use a computer program that controls the MR device to make the MR device perform the above-mentioned method steps of the present invention. The computer program can exist on a data carrier, or can exist in a data network so as to be downloaded and installed in the control unit of the MR device. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings disclose preferred embodiments of the present invention. However, it should be understood that the accompanying drawings are designed for illustrative purposes only and are not intended to be a definition of the limitations of the present invention. In the accompanying drawings:

[0037] Figure 1 A block diagram of an MR device for performing the method of the present invention is shown;

[0038] Figure 2 Schematically illustrates the subdivision of the acquisition time period into time segments according to the present invention;

[0039] Figure 3 The method of the present invention is shown in the form of a flow chart. DETAILED DESCRIPTION

[0040] refer to Figure 1, shows a MR device 1 in a block diagram. The device comprises a superconducting or resistive main magnet coil 2, so that a substantially uniform, temporally constant main magnetic field B0 is created along a z-axis through an examination volume. The device further comprises a set of (first-order, second-order, and, where applicable, third-order) shim coils 2', wherein the currents flowing through the individual shim coils of the set of shim coils 2' are controllable for minimizing B0 deviations within the examination volume.

[0041] Magnetic resonance generation and manipulation systems apply 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 and otherwise encode magnetic resonance, saturate spins, etc., thereby performing MR imaging.

[0042] More specifically, the gradient amplifier 3 applies current pulses or waveforms to selected coils of the whole body gradient coils 4, 5 and 6 along the x-axis, y-axis and z-axis of the examination volume. A digital RF frequency transmitter 7 transmits RF pulses or pulse packets to the body RF coil 9 via a transmit / receive switch 8, thereby transmitting RF pulses into the examination volume. A typical MR imaging sequence includes a packet of RF pulse segments of short duration, which together with any applied magnetic field gradients achieves a selected manipulation of the nuclear magnetic resonance signals. The RF pulses are used to saturate resonance, excite resonance, invert magnetization, refocus resonance or manipulate resonance and select a portion of the body 10 positioned in the examination volume. The MR signals are also picked up by the body RF coil 9.

[0043] To generate MR images of limited regions of the body 10 or for scan acceleration by means of parallel imaging, a set of local array RF coils 11, 12, 13 are placed adjacent to the region selected for imaging. The array coils 11, 12, 13 can be used to receive MR signals induced by body coil RF transmissions.

[0044] The resulting MR signals are picked up by the body RF coil 9 and / or the array RF coils 11,12,13 and demodulated by a receiver 14 which preferably includes a preamplifier (not shown).

[0045] The host computer 15 controls the shim coils 2' as well as the gradient pulse amplifiers 3 and the transmitter 7 to generate any of a plurality of MR imaging sequences, such as echo planar imaging (EPI), echo volume imaging, gradient and spin echo imaging, fast spin echo imaging, etc. For the selected sequence, the receiver 14 receives single or multiple MR data lines in rapid succession after each RF excitation pulse. The data acquisition system 16 performs analog-to-digital conversion on the received signals and converts each MR data line into a digital format suitable for further processing. In modern MR devices, the data acquisition system 16 is a separate computer dedicated to acquiring raw image data.

[0046] Ultimately, the digital raw image data is reconstructed into an image representation by a reconstruction processor 17, which applies a Fourier transform or other suitable reconstruction algorithm, such as SENSE or GRAPPA. 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 a slice, projection, or other portion of the image representation into a suitable format for visualization, such as via a video monitor 18, which provides a human-readable display of the resulting MR image.

[0047] The host computer 15 is programmed to execute the method of the present invention as described above and below.

[0048] According to the present invention, in step 19 (see Figure 3 ), for example in a golden angle star stacking scheme with overlapping overview sequences as described above, MR signals are acquired within a given time period. In step 20, the time period is subdivided into a plurality of consecutive time segments. This Figure 2 The time period for acquiring MR signals is Figure 2Indicated by T in. The time period T is subdivided into a plurality of shorter time segments S0, S1, S2, ..., Sn, each having a duration ΔT. For example, the duration of each time segment ΔT may be 0.5s. This is usually short enough in the case of normal respiratory motion so that the motion within a time segment can be ignored. In step 21, a low-resolution three-dimensional MR image is reconstructed for each time segment based on the signal data of each time segment. This is possible because approximately 10-15 radial k-space spokes are typically available within 0.5s due to the overlapping overview sequence. In addition, step 21 includes the derivation of a geometric transformation in the image space for each pair of consecutive time segments S0, S1, S2, ..., Sn, which reflects the motion occurring between the two time segments of the corresponding pair. The geometric transformation is derived by estimating the deformation vector field DVF1, DVF2, ..., DVFn for each pair of consecutive low-resolution MR images. From the deformation vector fields DVF1, DVF2, ..., DVFn, a geometric transformation reflecting the motion between arbitrary time segments can be calculated by summing all deformation vector fields across the segments:

[0049] DVF(i->j)=DVFi+1+DVFi+2+…+DVFj

[0050] In step 22, the inconsistency of the signal data is calculated for all time segments. If a high inconsistency within a certain time segment Si is associated with a large amplitude of the corresponding deformation vector field DVFi, a possible cause of this situation is rapid motion. In this case, the subdivision of the time segment in step 20 is repeated with a shorter duration ΔT. Alternatively, the corresponding time segment Si is divided into a plurality of shorter time segments. The geometric transformation between the shorter time segments can be obtained by interpolation between the original subdivided deformation vector fields DVF1, DVF2, ..., DVFn. In this way, the duration of the time segment Si is shortened and the inconsistency within the time segment Si is reduced. On the contrary, if the data consistency is high and the amplitude of the deformation vector field is small, the time segment Si can be combined. This means negligible motion, and the combination of time segments increases the amount of available data for iterative reconstruction and also increases the convergence speed. In step 23, the MR image is reconstructed from the signal data by an iterative reconstruction technique, wherein motion compensation is applied based on the derived geometric transformation. Iterative reconstruction is initialized with an MR image reconstructed from the entire set of MR signal data collected in a time period T. The MR image may have low quality because it is blurred by motion. The initial MR image is then iteratively refined. For this purpose, as the next step, the index of the time segment S0, S1, S2, ..., Sn is selected. Different strategies can be used to select the processing order of the time segment. In the case of a scan using the application of a contrast agent, a linear order will be preferred because this takes into account the temporal continuity of the contrast agent concentration change. In the case of a free breathing scan without a contrast agent, it may be advantageous to select the processing order based on the similarity of the low-resolution MR images belonging to the time segment. Similarity can be quantified using the derived deformation vector field DVF1, DVF2, ..., DVFn. The MR image is then transformed according to the geometric transformation of the motion state corresponding to the corresponding selected index. The MR image is then updated and made to conform to the signal data of the time segment belonging to the selected index, for example, by applying multiple iterations of the standard iterative SENSE algorithm. After iterating through the indices of all time segments in this manner, a high quality MR image containing contributions from the entire set of signal data is obtained. In addition to the MR image, the method generates a motion transformation that allows the MR image to be transformed into all motion states of different time segments. Furthermore, the reconstruction process can be repeated through all time segments after the first pass, wherein the geometric transformation is applied in each iteration in such a way that a high quality MR image is reconstructed for each time segment and thus motion state.

[0051] Optionally, the assumption that the motion within each time segment is negligible can be quantified and tested by calculating the inconsistency of all signal data within the time segment. If the inconsistency is low, the assumption that the motion during the corresponding time segment can be negligible is valid, if on the other hand it is high, the corresponding time segment can be excluded from the reconstruction. This situation may occur as an exception, for example in case of unexpected significant motion, swallowing, coughing, etc.

[0052] Yet another option is to gradually increase the resolution of the reconstructed MR images during the iterative reconstruction process from coarse to fine as the number of iterations increases. Many iterations may be required to progress from a blurry initial MR image to a final high-quality MR image. In this case, computational time can be saved by performing the initial iterations on a lower resolution (fewer voxels) MR image. A small number of voxels may be sufficient to represent all the information of the blurry initial image.

Claims

1. A method for MR imaging of an object (10) positioned in an examination volume of an MR device (1), the method comprising the following steps: generating MR signals by subjecting the object (10) to an imaging sequence comprising RF pulses and switched magnetic field gradients; acquiring the MR signals as signal data during a given time period (T), in which data portions are sampled from different portions of k-space of the magnetic resonance signals for consecutive time segments; subdividing the time period into a plurality of consecutive time segments (S0, S1, S2, ..., Sn) such that no movement or only negligible movement occurs in each time segment; deriving, for each pair of consecutive time segments (S0, S1, S2, ..., Sn), a geometric transformation (DVF1, DVF2, ..., DVFn) in image space, the geometric transformation (DVF1, DVF2, ..., DVFn) reflecting the motion occurring between the two time segments of the corresponding pair; reconstructing respective MR images for respective moments in time of individual time segments from the signal data of the plurality of time segments, wherein motion compensation is applied for each moment in time to the data portion from the respective time segment according to the derived respective geometric transformation (DVF1, DVF2, ..., DVFn) between the individual time segment and the moment in time, and An inconsistency of the MR signal attributable to each time segment (S0, S1, S2, ..., Sn) is determined, wherein if an inconsistency is detected, the time period (T) is subdivided into a greater number of shorter time segments.

2. The method according to claim 1, wherein: A low-resolution MR image is reconstructed for each time segment (S0, S1, S2, ..., Sn) based on the signal data belonging to the corresponding time segment (S0, S1, S2, ..., Sn).

3. The method according to claim 2, wherein: Each of the geometric transformations (DVF1, DVF2, ..., DVFn) is derived as a displacement vector field by registering the corresponding pair of the low-resolution MR images with each other.

4. The method according to any one of claims 1 to 3, wherein: A motion state is attributed to each time segment (S0, S1, S2, ..., Sn), and an MR image is reconstructed for each motion state.

5. The method according to claim 1, wherein: Iterative reconstruction techniques are applied to reconstruct the MR images.

6. The method according to claim 5, wherein: The reconstruction iteratively improves the reconstructed MR image, wherein, for each time segment (S0, S1, S2, ..., Sn), the geometric transformation (DVF1, DVF2, ..., DVFn) belonging to the corresponding time segment (S0, S1, S2, ..., Sn) is applied and the MR image is made to conform to the signal data belonging to the corresponding time segment (S0, S1, S2, ..., Sn).

7. The method according to any one of claims 1 to 3, wherein: If at least one of the derived geometric transformations (DVF1, DVF2, ..., DVFn) indicates that too fast or too strong a motion occurred between two time segments of the corresponding pair, the time period is subdivided into a greater number of shorter time segments (S0, S1, S2, ..., Sn).

8. The method according to any one of claims 1 to 3, wherein: If the derived geometric transformation (DVF1, DVF2, ..., DVFn) indicates that there is no motion or only insignificant motion between consecutive time segments, the time period (T) is subdivided into a smaller number of longer time segments (S0, S1, S2, ..., Sn).

9. The method according to any one of claims 1 to 3, wherein: The MR signals are acquired according to a star stack or spiral stack scheme or according to a koosh ball scanning scheme.

10. The method according to any one of claims 1-3, further comprising reconstruction of a dynamic series of MR images.

11. The method according to claim 1, wherein: The MR signals are acquired by parallel imaging using multiple RF receive coils (11, 12, 13) with different spatial sensitivity profiles.

12. The method according to any one of claims 6 and 11, wherein: The MR images are reconstructed by iterative SENSE reconstruction.

13. An MR device, comprising at least one main magnet coil (2), a plurality of gradient coils (4, 5, 6), at least one RF coil (9), a control unit (15) and a reconstruction unit (17), wherein the at least one main magnet coil is used to generate a uniform and stable magnetic field B0 in an examination volume, the plurality of gradient coils are used to generate switched magnetic field gradients in different spatial directions in the examination volume, the at least one RF coil is used to generate RF pulses in the examination volume and to receive MR signals from an object (10) positioned in the examination volume, the control unit is used to control the temporal continuity of the RF pulses and the switched magnetic field gradients, and the reconstruction unit is used to reconstruct an MR image based on the received MR signals, wherein The MR device (1) is arranged to perform the following steps: generating MR signals by subjecting the object (10) to an imaging sequence comprising RF pulses and switched magnetic field gradients; acquiring the MR signals as signal data during a given time period (T), in which data portions are sampled from different portions of k-space of the magnetic resonance signals for consecutive time segments; c) subdividing the time period into a plurality of consecutive time segments (S0, S1, S2, ..., Sn) such that no movement or only negligible movement occurs in each time segment; deriving, for each pair of consecutive time segments (S0, S1, S2, ..., Sn), a geometric transformation (DVF1, DVF2, ..., DVFn) in image space, the geometric transformation (DVF1, DVF2, ..., DVFn) reflecting the motion occurring between the two time segments of the corresponding pair; reconstructing respective MR images for respective moments in time of the individual time segments from the signal data, wherein motion compensation is applied to the data portion from the respective time segment for each moment in time according to the derived respective geometric transformation (DVF1, DVF2, ..., DVFn) between the individual time segment and the moment in time, and An inconsistency of the MR signal attributable to each time segment (S0, S1, S2, ..., Sn) is determined, wherein if an inconsistency is detected, the time period (T) is subdivided into a greater number of shorter time segments.

14. A computer program product storing a computer program to be run on an MR device, the computer program comprising instructions for: generating an imaging sequence comprising RF pulses and switched magnetic field gradients; acquiring MR signals as signal data within a given time period (T), in which data portions are sampled from different portions of k-space of the magnetic resonance signal for consecutive time segments; subdividing the time period into a plurality of consecutive time segments (S0, S1, S2, ..., Sn) such that no movement or only negligible movement occurs in each time segment; deriving, for each pair of consecutive time segments (S0, S1, S2, ..., Sn), a geometric transformation (DVF1, DVF2, ..., DVFn) in image space, the geometric transformation (DVF1, DVF2, ..., DVFn) reflecting the motion occurring between the two time segments of the corresponding pair; A corresponding MR image is reconstructed for a corresponding moment of an individual time segment according to the signal data, wherein: Motion compensation is applied for each time instant to the data portion from the corresponding time segment according to the derived corresponding geometric transformation (DVF1, DVF2, ..., DVFn) between the individual time segment and the time instant, and An inconsistency of the MR signal attributable to each time segment (S0, S1, S2, ..., Sn) is determined, wherein if an inconsistency is detected, the time period (T) is subdivided into a greater number of shorter time segments.

Images