Parallel MR imaging with Nyquist ghost correction for EPI

Abstract

A method of parallel MR imaging includes subjecting the portion of the body (10) to an imaging sequence of at least one RF pulse and a plurality of switched magnetic field gradients. The MR signals are acquired in parallel via a plurality of RF coils (11, 12, 13) having different spatial sensitivity profiles within the examination volume. The method further includes deriving an estimated ghost level map from the acquired MR signals and from spatial sensitivity maps of the RF coils (11, 12, 13), and reconstructing a MR image from the acquired MR signals, the spatial sensitivity maps, and the estimated ghost level map.

Claims

1. A method of parallel MR imaging of at least a portion of a body placed within the examination volume of a MR device, the method comprising the steps of: a) subjecting the portion of the body to a single-shot or a multi-shot EPI imaging sequence of at least one RF pulse and a plurality of switched magnetic field gradients, wherein MR signals are acquired in parallel via a plurality of RF coils having different spatial sensitivity profiles with a FOV matrix size doubled to separate Nyquist-ghosts from image information within the examination volume, b) deriving an estimated Nyquist-ghost level map from the acquired MR signals and from spatial sensitivity maps of the RF coils based on solving the following system of equations:
m.sub.j(x)=.sub.iS.sub.j(x.sub.i)p.sub.obj(x.sub.i)+S.sub.j(x.sub.i+FOV/2)p.sub.ghost(x.sub.i), wherein m.sub.j(x) is the MR image (prior to unfolding) reconstructed from MR signals acquired via one of the RF coils 11, 12, 13, with j indicating the respective RF coil, S.sub.j(x) is the spatial sensitivity map of RF coil j, p.sub.obj(x) is the signal contribution from the object (the body 10), and p.sub.ghost(x) is the signal contribution from the ghost image, FOV is the size of the field of view, and the estimated ghost level map g(x) is then calculated as:
g(x)=p.sub.ghost(x)/p.sub.obj(x) and spatially smoothing the estimated Nyquist-ghost level map c) reconstructing a MR image from the same acquired MR signals from which the Nyquist-ghost level map is derived, the spatial sensitivity maps, and the estimated Nyquist-ghost level map and the reconstruction involving regularisation using the regularisation constraint g(x)p.sub.obj(x)p.sub.ghost(x)=0, that is derived from the smoothed estimated Nyqiust-ghost level map; and iteratively repeating steps b) and c) with the same MR signals to successively increase the accuracy of the ghost level map.

2. The method of claim 1, wherein the estimated ghost level map and/or the MR image is reconstructed in steps b) and/or c) using compressed sensing.

3. The method of claim 1, wherein each of steps b) and c) comprises the computation of a MR image and a MR ghost image using a linear inversion method.

4. The method of claim 3, wherein a regularisation parameter determining the weighting of the regularisation constraint in the linear inversion is tuned depending on a signal-to-noise ratio of the MR image reconstructed in step c).

5. The method of claim 1, wherein the MR signals are acquired in step a) with undersampling of k-space.

6. A magnetic resonance (MR) device comprising at least one main magnet coil configured for generating a uniform, study magnetic field within an examination volume, a number of gradient coils configured for generating switched magnetic field gradients in different spatial directions within the examination volume, at least one RF coil configured for generating RF pulses in the examination volume, a plurality of RF coils having different spatial sensitivity profiles within the examination volume configured for receiving MR coils from at least a portion of a body of a patient positioned in the examination volume, a control unit configured for controlling the temporal succession of RF pulses and switched magnetic field gradients, and a reconstruction unit configured for reconstructing MR images from the received MR signals, wherein the MR device is arranged to perform the following steps: a) subjecting the portion of the body to a single-shot or multi-shot EPI imaging sequence of at least one RF pulse and a plurality of switched magnetic field gradients, wherein MR signals are acquired in parallel via the RF coils, b) deriving an estimated Nyquist-ghost level map from the acquired MR signals and from spatial sensitivity maps of the RF coils and spatially smoothing the estimated ghost level map, wherein the Nyquist-ghose level map is derived from the following system of equations:
m.sub.j(x)=.sub.iS.sub.j(x.sub.i)p.sub.obj(x.sub.i)+S.sub.j(x.sub.i+FOV/2)p.sub.ghost(x.sub.i), wherein m.sub.j(x) is the MR image (prior to unfolding) reconstructed from MR signals acquired via one of the RF coils 11, 12, 13, with j indicating the respective RF coil, S.sub.j(x) is the spatial sensitivity map of RF coil j, p.sub.obj(x) is the signal contribution from the object (the body 10), and p.sub.ghost(x) is the signal contribution from the ghost image, FOV is the size of the field of view, and the estimated ghost level map g(x) is then calculated as:
g(x)=p.sub.ghost(x)/p.sub.obj(x) and spatially smoothing the estimated Nyquist-ghost level map, c) reconstructing a MR image from the acquired MR signals, the spatial sensitivity maps, and the smoothed estimated Nyquist-ghost level map from step b), and the reconstruction involving regularisation using a regularisation constraint g(x)p.sub.obj(x)p.sub.ghost(x)=0, that is derived from the smoothed estimated Nyquist-ghost level map.

7. The magnetic resonance device of claim 6 wherein the MR device is further arranged to: iteratively repeat steps b) and c) with the same MR signals to successively increase the accuracy of the Nyquist-ghost level map in the reconstructed MR image.

8. A non-transitory computer-readable medium carrying a computer program which when executed on a computer controls an MR device to: a) generate a single or a multi-shot EPI imaging sequence of at least one RF pulse and a plurality of switched magnetic field gradients, and acquiring MR signals in parallel via a plurality of RF coils having different spatial sensitivity profiles, b) derive an estimated Nyquist-ghost level map from the MR signals acquired in step a) and from spatial sensitivity maps of the RF coils, and spatially smoothing the estimated ghost level map, wherein the Nyquist-ghose level map is derived from the following system of equations:
m.sub.j(x)=.sub.iS.sub.j(x.sub.i)p.sub.obj(x.sub.i)+S.sub.j(x.sub.i+FOV/2)p.sub.ghost(x.sub.i), wherein m.sub.j(x) is the MR image (prior to unfolding) reconstructed from MR signals acquired via one of the RF coils 11, 12, 13, with j indicating the respective RF coil, S.sub.j(x) is the spatial sensitivity map of RF coil j, p.sub.obj(x) is the signal contribution from the object (the body 10), and p.sub.ghost(x) is the signal contribution from the ghost image, FOV is the size of the field of view, and the estimated ghost level map g(x) is then calculated as:
g(x)=p.sub.ghost(x)/p.sub.obj(x) and spatially smoothing the estimated Nyquist-ghost level map, and c) reconstruct a MR image from the acquired MR signals, the spatial sensitivity maps, and the smoothed estimated Nyquist-ghost level map and the reconstruction involving regularisation using a regularisation constraint g(x)p.sub.obj(x)p.sub.ghost(x)=0, that is derived from the smoothed estimated Nyquist-ghost level map.

9. The non-transitory computer-readable medium of claim 8, wherein which, when the computer program is executed on the computer, further controls the MR device to: iteratively repeat steps b) and c) with the same MR signals to successively increase the accuracy of the Nyquist-ghost level map in the reconstructed MR image.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The enclosed drawings disclose preferred embodiments of the present invention. It should be understood, however, that the drawings are designed for the purpose of illustration only and not as a definition of the limits of the invention. In the drawings:

(2) FIG. 1 schematically shows a MR device for carrying out the method of the invention;

(3) FIG. 2 shows EPI images of a phantom reconstructed conventionally and according to the method of the invention;

(4) FIG. 3 shows noise amplification maps for different weightings of the regularisation constraint applied according to the method of the invention;

(5) FIG. 4 shows a ghost level map computed according to the invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

(6) With reference to FIG. 1, a MR device 1 is shown. The device comprises superconducting or resistive main magnet coils 2 such that a substantially uniform, temporally constant main magnetic field B.sub.0 is created along a z-axis through an examination volume. The device further comprises a set of (1.sup.st, 2.sup.nd, andwhere applicable3.sup.rd order) shimming coils 2, wherein the current flow through the individual shimming coils of the set 2 is controllable for the purpose of minimizing B.sub.0 deviations within the examination volume.

(7) A magnetic resonance generation and manipulation system applies a series of RF pulses and switched magnetic field gradients (also referred to as gradient pulses) to invert or excite nuclear magnetic spins, induce magnetic resonance, refocus magnetic resonance, manipulate magnetic resonance, spatially and otherwise encode the magnetic resonance, saturate spins, and the like to perform MR imaging.

(8) More specifically, a gradient pulse amplifier 3 applies current pulses to selected ones of whole-body gradient coils 4, 5 and 6 along x, y and z-axes of the examination volume. A digital RF frequency transmitter 7 transmits RF pulses or pulse packets, via a send-/receive switch 8, to abody RF coil 9 to transmit RF pulses into the examination volume. A typical MR imaging sequence is composed of a packet of RF pulse segments of short duration which taken together with each other and any applied magnetic field gradients achieve a selected manipulation of nuclear magnetic resonance. The RF pulses are used to saturate, excite magnetic resonance, invert magnetization, refocus resonance, or manipulate resonance and select a portion of a body 10 positioned in the examination volume.

(9) For generation of MR images of regions of the body 10 by means of parallel imaging, a set of local array RF coils 11, 12, 13 having different spatial sensitivity profiles are placed contiguous to the region selected for imaging. The array RF coils 11, 12, 13 are used to receive MR signals induced by body-coil RF transmissions.

(10) The resultant MR signals are picked up by the body RF coil 9 and by the array RF coils 11, 12, 13 and demodulated by a receiver 14 preferably including a pre-amplifier (not shown). The receiver 14 is connected to the RF coils 9, 11, 12 and 13 via send-/receive switch 8.

(11) A host computer 15 controls the current flow through the shimming coils 2 as well as the gradient pulse amplifier 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, and the like. For the selected sequence, the receiver 14 receives a single or a plurality of MR data lines in rapid succession following 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 suitable for further processing. In modern MR devices the data acquisition system 16 is a separate computer which is specialized in acquisition of raw image data.

(12) Ultimately, the digital raw image data is reconstructed into an image representation by a reconstruction processor 17 which applies appropriate reconstruction algorithms, such like SENSE. The MR image may represent a planar slice through the patient, an array of parallel planar slices, a three-dimensional volume, or the like. The image is then stored in an image memory where it may be accessed for converting slices, projections, or other portions of the image representation into appropriate format for visualization, for example via a video monitor 18 which provides a man-readable display of the resultant MR image.

(13) According to the invention, the body 10 is subjected to a single-shot or multi-shot EPI sequence, wherein MR signals are acquired in parallel via the array of RF coils 11, 12, 13 having different spatial sensitivity profiles within the examination volume. As in conventional SENSE imaging, spatial sensitivity maps of the RF coils are determined from low-resolution reference data obtained by a SENSE reference scan. An estimated ghost level map indicating the level of ghosting for each image position is derived from the acquired MR signals and from the spatial sensitivity maps of the RF coils 11, 12, 13 by a first SENSE unfolding step based on the following system of equations:
m.sub.j(x)=.sub.iS.sub.j(x.sub.i)p.sub.obj(x.sub.i)+S.sub.j(x.sub.i+FOV/2)p.sub.ghost(x.sub.i),
wherein m.sub.j(x) is the MR image (prior to unfolding) reconstructed from MR signals acquired via one of the RF coils 11, 12, 13, with j indicating the respective RF coil. S.sub.j(x) is the spatial sensitivity map of RF coil j, p.sub.obj(x) is the signal contribution from the object (the body 10), and p.sub.ghost(x) is the signal contribution from the ghost image. FOV is the size of the field of view (i.e. the folded field of view equal to the planned FOV divided by the acceleration factor of the parallel imaging technique). The formula takes into account that the signal contribution from the Nyquist ghost is weighted by the spatial sensitivity map shifted by one-half field of view. Tikhonov regularisation may be used in this unfolding step as in common SENSE reconstruction. This kind of regularisation provides information about the expected level of output signal per image position. This includes information as to the image area in which the object is positioned and the background. Using this information improves the SNR. The standard Tikhonov regularisation is also applied for the ghost image p.sub.ghost(x.sub.i). It is based on the same data, with the only difference that the ghost image is shifted by one-half FOV to account for the different spatial location of the ghost image. The estimated ghost level map g(x) is then calculated as:
g(x)=p.sub.ghost(x)/p.sub.obj(x)

(14) As a next step, spatial smoothing is applied to g(x) in order to obtain the low frequency information about the ghost level. In this way, the assumption that the ghost level is spatially smoothly varying is introduced into the reconstruction method. A further SENSE unfolding step is then performed based on the above system of equations and using the following regularisation constraint:
g(x)p.sub.obj(x)p.sub.ghost(x)=0,
wherein g(x) is the smoothed estimated ghost level map. The finally reconstructed MR image (without Nyquist ghost artifacts) then corresponds to p.sub.obj(x). A regularisation parameter (Lagrange factor) determining the weighting of the regularisation constraint in the second unfolding step is tuned depending on the signal-to-noise ratio of the MR image. If the constraint is strongly weighted this results in p.sub.ghost(x)=g(x) p.sub.obj(x) and the increase of the noise level of the reconstructed MR image is minimal. If the weighting is low, in contrast, the level of Nyquist ghosting is minimized (since any errors in the computation of g(x) are shifted into p.sub.ghost(x)). However, SNR is decreased in the final image.

(15) Optionally, further iterations over the above two SENSE unfolding steps can be performed, wherein the weighting of the regularisation constraint may be increased from iteration to iteration. In this way, the estimated ghost level map can be determined more and more accurately. If the inversion problem is too ill-conditioned in the first step, which may be the case if the SENSE acceleration factor is large, the iteration may start with a pre-set ghost level map, for example assuming a fixed ghost level or a ghost level map obtained from an initial reference measurement.

(16) The total reconstruction time required by the method of the invention is only marginally increased vis-a-vis conventional SENSE, since the unfolding procedure constitutes only a small part of the total reconstruction. Moreover, the modified unfolding of the invention typically requires only a limited number of iterations.

(17) FIG. 2 shows EPI images of a phantom acquired with SENSE factor 2. The upper left image is reconstructed without Nyquist ghost correction. The image shows strong ghost artifacts. Conventional one-dimensional phase correction based on a reference measurement is applied in the upper middle image. The level of Nyquist ghosting is substantially reduced. The upper right image is reconstructed according to the method of the invention with a low weighting of the regularisation constraint. An intermediate regularisation weight is applied in the bottom left image, the bottom right image shows the reconstructed MR image using a strong regularisation weight. The images reconstructed according to the invention show different levels of Nyquist ghost suppression.

(18) FIG. 3 shows corresponding noise amplification maps. The noise amplification maps are computed according to:

(19) $\mathrm{NAM}\left(x,y\right)=\sqrt{\frac{H_{\mathrm{SENSE}}\left(x,y\right)H_{\mathrm{SENSE}}^h\left(x,y\right)}{H_{\mathrm{CLEAR}}\left(x,y\right)H_{\mathrm{CLEAR}}^h\left(x,y\right)}},$
wherein H.sub.SENSE(x,y) is the combination factor matrix obtained during the SENSE unfolding for image location x, y (size 1number of RF coils) and H.sub.clear(x,y) is the combination factor matrix obtained during standard SENSE unfolding with SENSE factor 1 (CLEAR operation). NAM(x,y) can never be less than 1 and only provides valid information inside the imaged anatomy. The upper left image in FIG. 3 shows NAM(x,y) for conventional SENSE reconstruction. The upper right image corresponds to the modified SENSE unfolding with a low weighting of the regularisation constraint. An intermediate regularisation weight is applied in the bottom left image, the bottom right image shows NAM(x,y) using a strong regularisation weight.

(20) As becomes evident from FIGS. 2 and 3, Nyquist ghosting can be successfully removed by the method of the invention with a minimal/negligible increase of the noise level. The SNR only drops very slightly as compared to standard SENSE reconstruction.

(21) FIG. 4 shows the ghost level map g(x,y) estimated according to the method of the invention. FIG. 4 illustrates that the assumption of a constant ghost level in the phase encoding direction is not made according to the invention. Hence, the technique is superior to one-dimensional phase correction approaches used conventionally for Nyquist ghost removal and thus generally provides an improved image quality.

(22) The invention has been described with reference to the preferred embodiments. Modifications and alterations may occur to others upon reading and understanding the preceding detailed description. It is intended that the invention be constructed as including all such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.