GB2614034A - Method and apparatus for magnetic resonance imaging - Google Patents

Abstract

Method of generating magnetic resonance image data includes selecting slices to image, perpendicular to a slice direction, transmitting a pulse perpendicular to the main B0 field direction in the presence of a slice gradient which enables separation along the slice direction. An initial signal spectrum is acquired including applying an additional gradient along the slice direction which enables separation along a frequency direction (frequency encoding), and processing the initial signal spectrum to give, for a plurality of positions along the slice direction, a frequency projection along the proposed phase encoding direction. A plurality of signal strength indicator values,each associated with a phase encode gradient value, are derived from each frequency projection. An ordering is created based on these, defining a time-varying sequence of phase encode gradient values. Subsequently imaging is carried out in the order defined by this time-varying sequence, and the acquired data is processed to produce an image-data array. Preferably, the signal strength indicator uses signal to noise ratios. Also disclosed is a magnetic resonance imaging system.

Description

Method and Apparatus for Magnetic Resonance Imaging

BACKGROUND OF THE INVENTION

This invention relates to generating magnetic resonance image data.

Magnetic Resonance Imaging (MRI) techniques are widely used for medical imaging. MRI techniques rely on the fact that many nuclei have nuclear spin and all nuclei are electrically charged. Protons in these nuclei (e.g. in a hydrogen nucleus, which is abundant in water which makes up most of the human body) align when a strong external magnetic field is applied (using a superconducting magnet for example).

After applying a magnetic field, short bursts of radio waves are sent into a target area of the body, knocking the protons out of alignment. This causes an energy transfer for some protons from a base energy to a higher energy level. When the radio waves are turned off, the protons realign i.e. the spins return to their base level. This causes energy of radio frequency to be emitted. These radio frequency signals when properly encoded using gradient magnetic fields provide information about the exact location of the protons in the body. They also help to distinguish between the various types of tissue in the body, because the protons in different types of tissue realign at different speeds and produce distinct signals. The signals can therefore be measured and processed in order to yield an image (or multiple images) of a target. Such an image shows different tissue types in different contrast, thereby providing medically useful information.

One of the major drawbacks to MRI is the time taken to acquire the data. An MRI scan can entail a patient spending an extended period inside an MRI machine (i.e. within a magnetic core) which can be cramped and distressing for a patient. Successful imaging requires a patient to stay very still throughout this period, which can prove challenging and uncomfortable. Certain patients e.g. neonates or those with movement disorders are unable to stay still for a sufficiently long time to obtain useful MRI images. -2 -

It would therefore be desirable to provide a method of producing magnetic resonance imaging data which can produce usable (i.e. sufficiently high quality) images, in a shorter period of time than conventional MRI.

SUMMARY OF THE INVENTION

From a first aspect, the invention provides a method of generating magnetic resonance image data corresponding to a target, in the presence of a magnetic field defining a main Bo field direction, comprising: selecting a plurality of slices through the target to image, wherein the plurality of slices are each perpendicular to a slice direction; transmitting an initial multi-band radio frequency pulse perpendicular to the main Bo field direction, in the presence of a slice gradient which enables separation of the initial signal spectrum along the slice direction; acquiring, using at least one radio-frequency receiver, an initial signal spectrum comprising signals produced by at least a part of the target, as a result of excitation by the initial multi-band radio frequency pulse, comprising applying an additional gradient along the slice direction which enables separation of the initial signal spectrum along a frequency direction, perpendicular to the proposed phase encoding direction and the slice direction; processing the initial signal spectrum to give, for each of a plurality of positions along the slice direction, a frequency projection along a proposed phase encoding direction, perpendicular to the slice direction; deriving from each frequency projection a plurality of signal strength indicator values, each associated with a phase encode gradient value; creating an ordering of the phase encode gradient values based on the plurality of signal strength indicator values derived from the plurality of frequency projections, so as to define a time-varying sequence of phase encode gradient values; subsequently acquiring an image-data array by: transmitting a multi-band radio frequency pulse sequence into the target, perpendicular to the main Bo field direction, using the time-varying sequence of phase encode gradient values applied along the phase encoding direction; -3 -acquiring, using the at least one radio-frequency receiver, an imaging signal spectrum comprising signals produced by at least a part of the target, as a result of excitation by the multi-band radio frequency pulse sequence, comprising applying an additional gradient along the slice direction which enables separation of the imaging signal spectrum along the frequency direction; and processing the imaging signal spectrum to produce the image-data array According to a second aspect, there is provided a magnetic resonance imaging system, comprising a magnet, arranged to produce a magnetic field defining a main Bo field direction, a radio-frequency transmitter, and at least one radio-frequency receiver, and a processor, the system configured to: select a plurality of slices through a target to image, the target being within the magnetic field, wherein the plurality of slices are each perpendicular to a slice direction; transmit, using the radio-frequency transmitter, an initial multi-band radio frequency pulse perpendicular to the main Bo field direction" in the presence of a slice gradient which enables separation of the initial signal spectrum along the slice direction; acquire, using the at least one radio-frequency receiver, an initial signal spectrum comprising signals produced by at least a part of the target, as a result of excitation by the initial multi-band radio frequency pulse, comprising applying an additional gradient along the slice direction which enables separation of the initial signal spectrum along a frequency direction, perpendicular to the proposed phase encoding direction and the slice direction; process, using the processor, the initial signal spectrum to give, for each of a plurality of positions along the slice direction, a frequency projection along a proposed phase encoding direction, perpendicular to the slice direction; derive from each frequency projection a plurality of signal strength indicator values, each associated with a phase encode gradient value; create an ordering of the phase encode gradient values based on the plurality of signal strength indicator values derived from the plurality of frequency projections, so as to define a time-varying sequence of phase encode gradient values; -4 -transmit, using the transmitter, a multi-band radio frequency pulse sequence into the target, perpendicular to the main Bo field direction, using the time-varying sequence of phase encode gradient values applied along the phase encoding direction; acquire, using the at least one radio-frequency receiver, an imaging signal spectrum comprising signals produced by at least a part of the target, as a result of excitation by the multi-band radio frequency pulse sequence, comprising applying an additional gradient along the slice direction which enables separation of the imaging signal spectrum along the frequency direction; and process the imaging signal spectrum to produce the image-data array.

Thus it will be seen that, in accordance with the invention, by combining a plurality of signal strength indicator values, from a plurality of slices, and using these values, from multiple slices, to define an optimised order for the phase encoding steps of the later imaging process, the quality of the images ultimately produced may be maximised overall as the data acquisition steps carried out first are chosen according to their signal strength indicator value. The use of a slice gradient allows the frequency projections corresponding to each slice to be separated out which in turn allows an overall imaging (or encoding) order to be defined based on data from all of the slices.

The arrangement may conveniently be considered with reference to Cartesian axes, for example where the x-axis defines a frequency (encoding) direction, the y-axis defines the (proposed) phase encoding direction, and the z-axis defines the slice direction, to which each image slice is perpendicular. It will be understood that the phase encoding direction can be referred to as a "proposed" phase encoding direction until an acquisition using such phase encoding is carried out. It will be understood that these terms serve as labels equivalent to labelling the axes as "x", "y" and "z", or "first", "second" and "third".

It will be understood that the slices need not correspond to adjacent spatial locations i.e. slices which are adjacent along the slice direction (z-axis) need not correspond to slices through the target which are adjacent or spatially continuous. -5 -

Thus it will be understood that the purpose of the initial multi-band radio frequency pulse is that this allows a frequency projection to be acquired for each slice, from which in turn an overall ordering of phase encode gradient values can be defined. The ordering which is created lists phase encode gradient values, in a particular order defined by their signal strength indicator value. In some embodiments, the signal strength indicator value for a particular phase encode gradient value, for a particular frequency projection (i.e. relating to a frequency projection of a particular slice), is representative of a signal-to-noise ratio for data of that frequency projection at that particular phase encode gradient value. The signal strength indicator value may be the signal-to-noise ratio value, or may be a value related to or indicative of this value, or providing an alternative measure of relative signal strength. This advantageously provides a useful measure which is indicative of the quality of data which might be obtained from each particular slice at that particular phase encode gradient value (i.e. in that particular phase encoding step), thus indicating which are likely to be the most useful phase encoding gradient values to apply first.

The ordering is used to define a single series of phase encoding steps, i.e. each individual phase encoding gradient value is applied only once in the time-varying sequence, but the ordering is defined based on a plurality of sets of signal strength indicator values, each set corresponding to a particular slice i.e. the defined ordering thus includes one defined ranking of phase encoding steps, even though the same phase encode gradient value may have a different signal strength indicator value for each slice. Possible methods for deriving definitive ordering of these phase encode gradient values based on the different signal strength indicator values from different slices include: for each phase encode gradient value, taking an average of its respective signal strength indicator value for each frequency projection, and then placing the phase encode gradient values in order based on their average signal strength indicator values across the slices (e.g. from highest to lowest signal strength indicator value), or alternatively, selecting each phase encode gradient value in turn which has the next highest signal strength indicator value in any of the frequency projections.

Thus, in some embodiments creating an ordering of phase encode gradient values comprises summing the signal strength indicator values for each phase encode gradient value, to give an averaged signal strength indicator value for each phase -6 -encode gradient value, and then placing the phase encode gradient values in an order based on the averaged signal strength indicator values (e.g. in order from highest to lowest average signal strength indicator value). This provides a convenient and effective method by which to derive a single phase encoding order based on frequency projections corresponding to each slice.

In other embodiments, creating an ordering of phase encode gradient values comprises interleaving sequentially the order of the phase encode gradient values based on an ordering of signal strength indicator values of the phase encode gradient values in any of the frequency projections for the different slices. By sequentially interleaving, it is meant that the phase encode gradient value having e.g. the highest signal strength indicator value for any of the frequency projections is identified, and that phase encode gradient value is placed first in the ordering. Then the phase encode gradient value having e.g. the next highest signal strength indicator value across any of the frequency projections is identified, and that phase encode gradient value is placed next in the ordering (provided that it does not already appear in the ordering). Thus the phase encode gradient value having the highest signal strength indicator value (or a signal strength indicator value having a different property e.g. lowest, depending on the chosen ordering) across all of the slices is chosen as first in the ordering, followed by the phase encode gradient value having the next highest signal strength indicator value (in any of the slices) etc., unless that phase encode gradient value has already appeared in the ranking in which case it is omitted so that no phase encode gradient value appears in the ordering more than once. Or, considered differently, this is the ordering that would be achieved if all phase encode gradient values for all the slices' frequency projections were placed in a single ordering based on signal strength indicator values (so that each phase encode gradient value would appear the same number of times as there are frequency projections) and then all duplicate phase encode gradient values other than the first appearance of that phase encode gradient value in the ordering were removed.

In some embodiments, the ordering is created by placing the phase encode gradient values in order of highest to lowest signal strength indicator value (or averaged signal strength indicator value), where any suitable signal strength indicator value may be chosen e.g. signal-to-noise ratio. This advantageously provides a simple choice or ordering which allows production of a high quality image, since those views having the -7 -highest signal strength indicator value are acquired first, according to a chosen signal strength indicator value which is relevant for the given situation e.g. if the signal strength indicator of the data obtained declines over time.

In some embodiments the plurality of slices are selected by adjusting the timing of the initial multi-band radio frequency pulse, and the multi-band radio frequency pulse sequence. In some embodiments the multi-band radio frequency pulse sequence is frequency modulated, for example as in a sine modulated sinc function. These may provide simple and easy to use encoding techniques for MRI.

It will be understood that the term "gradient" (e.g. phase encode gradient), refers to an encoding gradient waveform applied along a particular direction such that signals received from regions, or pixels, a different distance along that direction are phase-shifted by a different amount, i.e. so that if a phase encode gradient has been applied a distance along that direction can be determined based on the phase of the received signal. Thus, for example, the time-varying sequence of phase encode gradient values enable separation of the further imaging signal spectrum along the phase encoding direction.

A slice (selection) gradient is a gradient applied along the slice axis in the presence of a multi-band radio frequency pulse to simultaneously select multiple slices in the slice direction. In some embodiments the step of transmitting a multi-band radio frequency pulse sequence (i.e. the pulse sequence used to acquire an image-data array after a phase encoding order has been determined) into the target is carried out in the presence of an imaging slice gradient which enables separation of the imaging signal spectrum along the slice direction. This allows the slices to be separated for the final image-data which is collected, just as they are for the initial projection. It will be understood that the slice gradient will advantageously be chosen to select, or separate, the same slices as the imaging slice gradient i.e. the slice gradient used in the initial "projection" which provides the signal strength indicator values may be the same as the slice gradient used in the later imaging process, so that in each case the imaged slices are separable by the gradient in the same way. It will be understood that in such embodiments a complete imaging sequence including slice selection and frequency encoding is used with the modified phase table (i.e. phase encode ordering) to generate a complete data set. -8 -

In some embodiments, the system comprises a plurality of radio-frequency receivers e.g. coils. The method may comprise acquiring the initial signal spectrum, using the plurality of radio-frequency receivers. The method may also comprise acquiring the imaging signal spectrum using the plurality of radio-frequency receivers.

In a preferred set of embodiments the method comprises phase undersampling to further reduce scan time. Thus, in some embodiments the time-varying sequence of phase encode gradient values comprises only a sub-set of the phase encode gradient values (i.e. less than the total number of phase encode gradient values, increasing the distance of sampling positions in k-space). For example, equal to or less than half of the phase encoding steps may be applied. This allows the imaging process to be carried out much more quickly (e.g. twice as quickly if data for half as many phase encoding steps is acquired) and thus reduces scan time. In some embodiments, the data for the rest of the phase encode gradient values can be inferred, interpolated, extrapolated or reconstructed based on parameters of the plurality of radio-frequency receivers, optionally the sensitivity profiles of the plurality of radio-frequency receivers. This allows the amount of data required to be reduced (i.e. the number of phase-encoding steps is reduced), reducing scan time, without a significant or noticeable (or preferably any) impact on the quality of the data or image produced. The image-data array may be further processed to remove the effects of phase undersampling.

It will be understood that the purpose of the subsequent multi-band radio frequency pulse sequence is to allow an image-data array to be acquired, from which images can be produced. Thus in some embodiments, the image data array is processed to produce an image (e.g. a two-dimensional image) corresponding to each of the plurality of slices. It will be understood that the imaging directions, i.e. the x and y directions in the plane of these images, are perpendicular to the slice direction. Thus in some embodiments the image-data array is processed to produce a plurality of slices separated in both the slice and frequency directions into individual images. This advantageously produces from the collected data a series of separate images representing different slices of a target e.g. part of a patient, which can provide medically useful information. -9 -

Although the method described above, and detailed further below with reference to the Figures, produces two-dimensional images, it will be appreciated that this method could equally be carried out so as to acquire an image-data array suitable for producing three-dimensional images. Thus in some embodiments an ordering of phase encode gradient values is further derived for a second phase encoding direction, perpendicular to the first phase encoding direction. The method may further comprise processing the image-data array to produce a three-dimensional image corresponding to each of the plurality of slices.

Features of any aspect or embodiment described herein may, wherever appropriate, be applied to any other aspect or embodiment described herein. Where reference is made to different embodiments or sets of embodiments, it should be understood that these are not necessarily distinct but may overlap.

BRIEF DESCRIPTION OF THE DRAWINGS

Certain preferred embodiments of the invention will now be described, by way of example only, with reference to the accompanying drawings, in which: Figure 1 is a schematic drawing showing a magnetic resonance imaging system; Figure 2 is a schematic drawing from above representing schematically an MRI image of a brain, showing the frequency and phase encoding axes; Figure 3 is schematic diagram representing a frequency encoding axis, as shown in Figure 2, and a slice direction; Figure 4 is a schematic diagram representing the spin echo MR pulse sequence along each of the axes of Figure 3; Figure 5 shows four images of a human lower limb acquired using various MRI techniques; Figure 6 is a graph showing an example of signals received as a result of a projection made along the phase encoding direction; Figure 7 is a graph showing the projection of Figure 6, rearranged in ascending signal-to-noise ratio order; Figure 8 is a graph representing the position within a phase-encode order, defined using the graph of Figures 6 and 7, of each pixel; and Figure 9 is a flow chart representing a method according to an embodiment of the present invention.

-10 -

DETAILED DESCRIPTION

Figure 1 is a schematic drawing showing a typical magnetic resonance imaging system 100.

The magnetic resonance imaging system 101 includes a magnet 104, which surrounds a gradient coil 106 and a radio-frequency coil 108, the functioning of which is described below. Within the radio-frequency coil 108 there is a cavity 102. A part of a patient who is to be imaged will be placed inside this cavity e.g. the patient may lay on a bed which is then slid into the cavity 102.

The magnet 104 has a static field and causes a small portion of the water molecules in a patient's body, within the cavity 102, to align along one direction, being polarised to align either with the magnetic field, or anti-parallel to the magnetic field, so that the spins are respectively in quantum states up' and 'down'. It is specifically the nucleus of the hydrogen atoms of the water molecules, i.e. the proton, which is aligned by the magnetic field. Additional energy is then added to these protons by applying a radio wave to the magnetic field applied by magnet 104 using RF coil 108. The radio-frequency signal applied by the RF coil 108 is then switched on and off in a series of quick pulses; these pulses are controlled by a spectrometer 110, as represented by signal arrow 112. These pulses cause the proton of each hydrogen atom within the imaged part of the patient's body to alter its alignment.

When the RF coil 108 is switched off, the aggregate magnetic-moment vector returns to its resting state, and the protons of the water molecules emit an RF signal i.e. the hydrogen atoms produce a signal when they return to their relaxed states. This signal is received by receiver coils 120, which are spaced around the patient, who is located in cavity 102. The receiver coils 120 may be the same as the RF coil 108, or may be separate coils. This signal is then sent to the spectrometer 110, as represented by arrow 116. This is the signal that creates raw data for images.

The signal produced by the hydrogen atoms depends on the strength of the magnetic field. The signal produced will not be spatially localised e.g. a signal will just be received from the target overall. It is thus necessary to alter the signal produced across the target such that the signal received by the receiver coils 120 can be separated out to correspond spatially to the target.

To achieve this, the signal applied by the RF coil 108 is frequency encoded along a "frequency encoding" direction by applying a frequency "gradient" along the frequency encoding direction, such that the received signals will vary according to that frequency. This allows a received signal to be "separated out" along that frequency encoding direction, to identify which spatial locations contributed to the received signal.

The MRI system 101 also includes a gradient coil 106. This applies another magnetic field which is varied along another direction, the "phase encoding direction", perpendicular to the "frequency encoding" direction. This is also controlled by the spectrometer 110, as represented by arrow 114. This gradient coil 106 therefore produces a signal which causes the phase of the received signal to vary along the "phase encoding" direction, allowing signals received along that direction to be separated into those corresponding to different spatial locations along that direction which contribute to the signal.

Figure 2 is a drawing representing (schematically) an MRI image 1 which may be produced using the MRI system 101 described with reference to Figure 1, showing a single cross-sectional slice through a human brain 2 from above. The x-axis 4 is the frequency encoding axis. Signals are transmitted into the human brain 2, which include a frequency gradient along the frequency encoding axis, as described above. This frequency gradient effectively causes each line, or row, of protons in the human brain 2 perpendicular to the frequency encoding axis 4, to resonate at different selected encoding frequencies. Thus when an imaging signal spectrum is received from the entire plane, as a result of this excitement, the signal is encoded by the frequency encoding gradient as a range of frequencies in a time dependent signal. The frequencies are separated out after Fourier transformation of the time dependent signal in the frequency direction.

The y-axis 6 is the phase encoding axis. A phase gradient which is applied by gradient coils 106, as described above, will be applied along the y-direction, causing the resonance frequency of each successive location (e.g. pixel) in a given vertical column to be phase shifted by a successively larger amount.

-12 - Thus the x-coordinate of a chosen location (e.g. pixel) can be identified based on its frequency, and the phase shift of the signal at a given point depends on the y-coordinate of that point. When no gradient is applied to the signal, all points in the same column will have the same phase. Therefore in order to create an image of the entire human brain 2, a 2D matrix of received signal values is required. This 2D matrix is acquired by applying many phase encode steps, typically 128-256 different phase values, while collecting 256 time points of frequency information. The frequencies are separated out after 2D Fourier transformation of the entire matrix of, so called, k-space data.

Each of these phase encoding steps involves the application of a phase gradient, along the phase encoding direction, where each step corresponds to a gradient of different magnitude, such that overall each step is separated by an equal increment, and lasts for the same period of time. Various phase encoding orders are known. For example linear phase encoding goes from either the most negative gradient to the most positive (or vice versa) or centric order phase encoding goes from the central gradient (i.e. 0) outwards alternating in the positive and negative directions. A different phase encode ordering technique, according to the present invention, is discussed below.

The method according to the particular embodiment described is explained by reference to three techniques, as laid out below.

Phase encode undersamolina Phase encode undersampling is a class of known techniques using multiple receivers (sometimes referred to as receiver coils), where generally the placement and sensitivities of the receiver coils are used to assist spatial localization of the MR signal i.e. reconstruction based on receiver coil sensitivity profiles. This spatial localization allows the number of phase-encoding steps required during image acquisition to be reduced (i.e. an image can be reconstructed using fewer samples than would otherwise be needed), improving speed. This is achieved by increasing the distance of sampling positions in k-space while maintaining the maximum k-values. Scan time is therefore reduced whilst preserving spatial resolution. The factor by which the number of k-space samples is reduced is referred to as the reduction factor R. -13 -Phase undersampled sensitivity encoding, referred to herein as SENSE, is one such "undersampling" method. It will be described below for exemplary purposes.

Samples are taken with a lower density than in known techniques, which causes "aliasing" in the image i.e. overlap in the image of spatially separated parts of the target, since signals from a number of different spatial positions in the field of view contribute to the same pixel in the aliased image. In SENSE imaging, one such "aliased" image is created for each receiver element (using a discrete Fourier transform (DFT)). These aliased images (one for each receiver) are then separated out, to give a single image showing the full field of view i.e. target. The separation of the aliased images is possible because for each image i.e. corresponding to each receiver, the relative weight of each spatial location contributing to a given pixel is different, because it depends on the sensitivity of the particular coil, as explained further below.

SENSE imaging is described in "SENSE: Sensitivity Encoding for Fast MR/" Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P.. Magn Reson Med 1999;42(5): 952-62 with reference to a single slice image. np denotes the number of pixels superimposed to form a single pixel in an aliased image, and no denotes the number of coils used. Vector a contains the complex image values the chosen pixel has in the aliased images. The complex coil sensitivities at each of the np superimposed positions together form a sensitivity matrix S, which has dimensions no x np: Sy0 = sy (re) where the subscripts y, p count the coils and the superimposed pixels, respectively, rp denotes the position of the pixel p, and sy is the spatial sensitivity of the coil y. The sensitivity matrix is used to calculate the unfolding matrix: U = (.00-1-sylsHip-1 where the superscript H indicates the transposed complex conjugate, and 111 is receiver noise matrix (having dimensions no x no (see, which describes the levels and correlation of noise in the receiver channels. Using the unfolding matrix, signal separation is performed by: -14 -v = U a where the resulting vector v has length np and lists separated pixel values for the originally superimposed positions. By repeating this procedure for each pixel in the reduced FOV a non-aliased full-FOV image is obtained.

Although in the described example phase encode undersampling is used to reduce the number of required samples (and therefore to increase imaging speed), it will be understood that this is not essential in the method of the present invention, and instead the "standard" number of samples i.e. as in known previous techniques, could be used e.g. one for each location or pixel/voxel in k-space. However, inclusion of this technique may help to further improve the scanning speed time which can be achieved in accordance with the invention.

Simultaneous slice excitation According to the simultaneous slice excitation approach, multiple physical slices of the target are excited at the same time, using wide-and or ultra-wide-band pulses e.g. dual band RF pulses. A received signal corresponding to a particular part of the target will therefore contain data corresponding to more than one of these slices at the same time, and this data will therefore need to be separated out into data corresponding to each slice. This can be achieved using a slice gradient Gz (e.g. the same principle as described above with reference to Figure 2) in combination with phase encode undersampling described above, but modifying the encoding function to include the sensitivity of the coils at multiple excited slice locations. This given an encoding function of: VP =1Fmtvidmrdc, ydc where vp is the nv image values, my,k is the magnetization from the yth coil at the Kth k-space location, p counts the aliased voxels, and F is the reconstruction matrix of the size nyncnk, where nk is the reduced number of k-space sampling points and tic is the number of array coils used.

The net image encoding function: is a product of the standard phase encoding function: etk K -15 -and the individual coil sensitivity function: sye (r) which, in the case of parallel slice encoding, denotes the sensitivity functions of the y coils at the different slice locations. In the case of two slices, these would be separated in frequency by half the readout bandwidth. Multiple excited slices can be acquired without any loss of SNR if frequency oversampling (increased number of acquired points, equivalently decreased sample duration, same pixel resolution) is used to increase the size of the FOV without changing the strength or length of the readout gradient.

Various gradients are used to separate out the aliased slice images. These gradients are represented in Figure 3, which shows a graph like that of Figure 2, but showing a view along the axis 6 of Figure 2 i.e. with the phase encoding axis directed into the page. The y-axis 6' (phase encoding axis) is into the page, represented by a circle in Figure 3, and is thus along the phase-encoding direction. A phase gradient Gp is applied along y-axis 6', to enable image extraction, as described above with reference to Figure 2.

X-axis 4' represents the frequency encoding direction, along which frequency encoding is used to enable image extraction, as described above with reference to Figure 2. This frequency encoding is referred to below as frequency gradient Gr. A particular slice 10' is aligned with the x-axis 4' in this example, and additional slices are selected at regular intervals along the slice direction, z-axis 8'. An additional "slice gradient" Gs is applied along the z-axis 8'.

By frequency modulating a standard RF pulse, the simultaneously excited slices have similar slice properties (slice thickness, phase profile) to the slice produced by the original unmodulated pulse. As the sequence can be operated in conventional interleaved multislice mode, full volume coverage is thus possible in half of the time required for a conventional sequence. The separation of the slices Oz is determined by the frequency modulation of the dual excitation pulse (which can be freely adjusted) and the slice gradient strength, that is, 67 =6,1y0z.

The additional gradient during acquisition, "slice gradient" Gs', separates the slices along the frequency direction. The ratio of the additional slice gradient strength Gs' to -16 -the read gradient strength Gz determines the separation of the slices in the frequency direction in a linear fashion and independently from the separation along the slice axis.

When Gs'=Gz, the centers of the two slices will be shifted by a frequency equivalent to that occupied by the original FOV. Also, during frequency encoding, a projection of each selected slice is rotated from the frequency toward the slice direction, the angle being determined by the relative size of the additional slice gradient and the frequency encoding gradient, that is, B = tan-' (Gs'/Gz) as shown in Figure 3.

Noise from the two slice locations is included within the encoded data sets, and if the frequency separation of the two fields of view is not complete, then the noise as well as the signal will add in areas of overlap. The noise contribution for each slice will depend on the details of the detection system. For example, if a single volumetric coil was used to acquire data, then the noise across the entire readout frequency band of the two slices would result from the entire volume included within the coil. However, if multiple surface coils with possibly different sizes and sensitivity profiles were used to acquire data localized to the different slices, then only the noise measured by the local coil would appear in the relevant frequency band for the slice under consideration.

Given the measured coil sensitivity reference functions acquired with dual-slice excitation and the individual images from each coil, it is possible to reconstruct the unaliased images and then separate the multiple excited slices by cutting the images along the frequency direction at the appropriate positions.

Figure 4 is a schematic diagram representing the spin echo MR pulse sequence along z-axis 8', that uses multiple excitation band 90 and 180 RF pulses 12, 14, and an additional gradient 16 applied along the slice axis during readout to remove frequency direction aliasing. Figure 4 also represents the phase encoding gradient 18 applied along y-axis 6' to assist with phase encoding undersampling, and the frequency gradient 20 applied along x-axis 4' along with the multiple channel parallel data acquisition 22 carried out by coils 1-4.

Figure 5 shows four images acquired using various MRI imaging techniques to image the lower limb of a human. Image A of Figure 5 shows a full-phase sampled image without the additional slice gradient G,' applied. It can be seen that since no additional -17 -slice gradient is applied, the projections of the two slices overlap within the field of view (compared to e.g. image B).

Image B of Figure 5 shows the full-phase sampled image with the additional slice gradient applied, as described above, which completely resolves the aliasing seen in Image A. Images C and D were produced by combining the simultaneous slice excitation method with the phase encode undersampling technique described above. Image C was produced at x4 the speed of a standard image, because undersampling produced a factor of two increase (half as many samples taken) and parallel slice excitation produced a factor of two increase.

Image D shows the x8 image (x4 phase undersampled, x2 slice excitation). It can be seen that relative to images A-C there has been a decrease in SNR and an increase in artifacts, which is to be expected. A masking process is used for image reconstruction which has reduced the noise to zero outside the limbs (i.e. in the dark areas) in images C and D. SNR phase order k-space encoding In accordance with the present invention, simultaneous slice excitation, and optionally phase encode undersampling, may be combined with phase order encoding, to provide a fast imaging technique capable of producing multiple high quality images in a relatively short time.

According to the SNR phase order k-space encoding technique generally, a phase-encoding order (i.e. an order of the phase encoding steps) is selected, based on signal-to-noise ratio information, producing an image having improved quality, particularly in the case of samples with signals that change during data acquisition.

Examples are in hyperpolarized helium gas imaging of the lungs where polarization is lost with each RF pulse or the signal changes observed in rapid dynamic studies with T1 or T2* contrast agents when mixing is taking place.

Phase-encode order is a major determinant of both contrast-to-noise ratio and artifact-to-noise ratio in MR imaging. It is the object structure in the phase encoding direction -18 -that is the major determinant of k-space SNR along that axis. Such an approach relies on acquisition of a single 'pre-projection' using a frequency encode gradient applied in the intended phase-encoding direction. The power spectrum of the projection is sorted from maximum to minimum (or vice versa for a signal that increases during image acquisition), and the indices of the sorted array are used to create a phase-encode table i.e. ordering associating each position in k-space (corresponding to a phase encode gradient value) with a position in the phase encode order, thus ensuring that the phase encoding steps required to obtain "views" with highest SNR are carried out first before the signal decays. The technique is described in greater detail below.

The phase encoding direction 6, 6' has been described above. With 2D Fourier encoding, it is possible to perform an in-plane rotation so that the phase encoding direction is aligned along any direction within the plane to be imaged. This can be used to advantage in minimizing scan time by using asymmetric fields of view (FOVs) or to avoid phase ghosting over structures of interest. However, once a phase-encode direction has been chosen, the structure of the object in this direction will determine the detailed k-space structure.

The k-space represents the spatial frequency information in two or three dimensions of an object and is defined by the space covered by the phase and frequency encoding data. The data acquisition matrix contains raw data before image processing.

The relationship between k-space data and image data is the Fourier transformation. In 2-dimensional (2D) Fourier transform imaging, a line of data corresponds to the digitised MR signal at a particular phase encoding level (i.e. a particular magnitude of phase encoding gradient). The position in k-space is directly related to the gradient across the object being imaged. By changing the gradient over time, the k-space data are sampled in a trajectory through Fourier space.

Every point in the raw data matrix contains part of the information for the complete image, but this point in the raw data does not correspond to a physical part of the image. The outer rows of the raw data matrix, the high spatial frequencies, provide information regarding the borders and contours of the image, the detail of the structures. The inner rows of the matrix, the low spatial frequencies, provide information on the general contrast of the image.

-19 -The average of this k-space structure can be seen as a function of k value by taking a frequency encoded projection in the proposed phase encoding direction, which can be achieved in a single acquisition. This projection data can then be sorted into SNR order and used to guide the order of phase encoding to maximize SNR at the start of the scan, which is important if a signal decays during image acquisition.

This technique is illustrated with reference to Figures 6, 7 and 8. Figure 6 shows an example of signals received as a result of a projection which has been made along the phase encoding direction. The x-axis represents an index value of a particular "phase encoding step" i.e. a particular phase encode gradient value. In this example 256 such phase encoding steps are carried out, as seen from the scale on the x-axis, where each value from 0-255 corresponds to a different value of phase encode gradient. For example, these indices might initially be assigned starting with the most negative gradient and numbering sequentially towards the most positive gradient value. The y-axis represents the magnitude of the received signal.

Figure 7 shows the projection of Figure 6, rearranged in ascending signal-to-noise ratio (SNR) order. The x-axis represents a value in a new ordering or list of phase encoding steps, i.e. gradient values. The y-axis represents a signal-to-noise ratio value.

Figure 8 represents the phase-encode order of pixels, calculated from the indices of the sorted data set. The x-axis represents the index initially assigned to the particular phase encoding step (i.e. phase encode gradient value), the y-axis represents the corresponding position of that same pixel in the new ranking/ordering shown in Figure 7, thus giving a "table" assigning to each phase encode gradient value a corresponding signal-to-noise ratio ranking.

In accordance with the present invention, a similar principle to the phase order encoding described above, may be combined with simultaneous slice excitation, in such a way that a beneficial phase-encoding order is selected both within each slice, and across different slices, thus giving a new and advantageous technique. This method will be described below with reference to Figure 9.

-20 -Figure 9 is a flow chart showing schematically the stages in a method of MRI imaging according to an embodiment of the present invention.

First, in a planning stage 80, slice positions through a target are selected by choice of timing and slice gradient strength of a multi-band radiofrequency (RF) pulse.

Next, in a projection stage 82, a single, frequency encoded projection is projected along the selected phase encoding direction, together with an additional slice gradient applied in order to allow later separation of the projections for each slice during processing. The Applicant has appreciated that the use of an additional gradient together with the pre-projection allows the SNR values for each slice to be extracted separately.

These SNR values for the different slices are then ordered, in sorting stage 84, to give one overall SNR ranking for all phase encoding steps, i.e. independent of slice. This enables SNR to be maximised across the multiple slices.

In some embodiments, the SNR ranking is optimized for all slices by summing all the projections and then sorting into descending SNR phase order and using the indices calculated from the sort to order the phase encode table.

In other embodiments, the phase encode order is determined by sequentially interleaving the highest SNR views (i.e. phase encoding steps) from each slice assuming they corresponded to different phase views, until the entire phase table is filled i.e. selecting each remaining highest value, unless the value corresponds to a phase encoding step which already appears in the ranking, in which case the value is discarded. In other words, a phase encoding step having the highest signal-to-noise ratio in any of the slices will be identified and will be selected as the phase encoding step to be carried out first. Then the phase encoding step having the next highest signal-to-noise ratio value across any of the slices will be identified, and selected as the next step in the phase encoding order, unless this phase encoding step already appears in the ordering in which case the ordering process will simply move on.

This sorting stage 84 results in a phase encode order, which lists phase encode coordinates (in k-space) in order of highest to lowest signal strength indicator value -21 -e.g. SNR. During the final data acquisition stage 86, as described below, repeated transmissions are made into the target. During these data transmissions and corresponding acquisition steps, the phase encoding gradient is selectively applied (in addition to the other gradients described above) in the ordering defined in the sorting stage 84.

Thus, in an acquiring stage 86, magnetic resonance data is acquired in the order derived in stage 84. Images produced from this collected data are then separated along the frequency direction to produce a plurality of simultaneously acquired images.

Such images can be acquired more quickly than using known techniques, and are capable of providing high quality images. For example, imaging two slices at once, and undersampling by x4 gives a total scan time which is eight times faster than a standard MRI scan, whilst the use of re-ordering across these slices ensures that the images across all the slices are of high quality. This represents a significant advancement as it makes scanning of certain patients e.g. neonates, or patients with motor conditions, feasible with a practical system. Such an approach improves image quality obtainable using time-decaying signals to the point where it is clinically useful.

In summary, embodiments of the invention reduce scan time while optimising SNR for imaging dynamic or rapidly decaying signals provides significant advantages for improved patient diagnostic examinations as well as for therapy monitoring of interventional MRI procedures.

It will be appreciated by those skilled in the art that the invention has been illustrated by describing one or more specific embodiments thereof, but is not limited to these embodiments; many variations and modifications are possible, within the scope of the accompanying claims.

Claims (1)

1. -22 -Claims 1. A method of generating magnetic resonance image data corresponding to a target, in the presence of a magnetic field defining a main Bo field direction, comprising: selecting a plurality of slices through the target to image, wherein the plurality of slices are each perpendicular to a slice direction; transmitting an initial multi-band radio frequency pulse perpendicular to the main Bo field direction, in the presence of a slice gradient which enables separation of the initial signal spectrum along the slice direction; acquiring, using at least one radio-frequency receiver, an initial signal spectrum comprising signals produced by at least a part of the target, as a result of excitation by the initial multi-band radio frequency pulse, comprising applying an additional gradient along the slice direction which enables separation of the initial signal spectrum along a frequency direction, perpendicular to the proposed phase encoding direction and the slice direction; processing the initial signal spectrum to give, for each of a plurality of positions along the slice direction, a frequency projection along a proposed phase encoding direction, perpendicular to the slice direction; deriving from each frequency projection a plurality of signal strength indicator values, each associated with a phase encode gradient value; creating an ordering of the phase encode gradient values based on the plurality of signal strength indicator values derived from the plurality of frequency projections, so as to define a time-varying sequence of phase encode gradient values; subsequently acquiring an image-data array by: transmitting a multi-band radio frequency pulse sequence into the target, perpendicular to the main Bo field direction, using the time-varying sequence of phase encode gradient values applied along the phase encoding direction; acquiring, using the at least one radio-frequency receiver, an imaging signal spectrum comprising signals produced by at least a part of the target, as a result of excitation by the multi-band radio frequency pulse sequence, comprising applying an additional gradient along the slice direction which -23 -enables separation of the imaging signal spectrum along the frequency direction; and processing the imaging signal spectrum to produce the image-data array 2. The method of claim 1, wherein the signal strength indicator value for a particular phase encode gradient value, for a particular frequency projection, is representative of a signal-to-noise ratio for data of said particular frequency projection at said particular phase encode gradient value.3. The method of claim 1 or 2, wherein creating an ordering of phase encode gradient values comprises summing the signal strength indicator values for each phase encode gradient value, to give an averaged signal strength indicator value for each phase encode gradient value, and then placing the phase encode gradient values in an order based on the averaged signal strength indicator values.4. The method of claim 1 or 2, wherein creating an ordering of phase encode gradient values comprises interleaving sequentially the order of the phase encode gradient values based on an ordering of signal strength indicator values of the phase encode gradient values in any of the frequency projections.5. The method of any preceding claim, further comprising creating the ordering by placing the phase encode gradient values in order of highest to lowest signal strength indicator value, or averaged signal strength indicator value.6. The method of any preceding claim, wherein the plurality of slices are selected by adjusting a timing of the initial multi-band radio frequency pulse and the multi-band radio frequency pulse sequence.7. The method of any preceding claim, wherein the multi-band radio frequency pulse sequence is frequency modulated.8. The method of any preceding claim, further comprising phase undersampling.-24 - 9. The method of any preceding claim, comprising carrying out said step of acquiring the initial signal spectrum using a plurality of radio-frequency receivers and wherein the time-varying sequence of phase encode gradient values comprises only a sub-set of the phase encode gradient values.10. The method of claim 9, further comprising inferring, interpolating, extrapolating or reconstructing the data for the rest of the phase encode gradient values based on the sensitivity profiles of the plurality of radio-frequency receivers.11. The method of any preceding claim, further comprising processing the image data array to produce an image corresponding to each of the plurality of slices.12. A magnetic resonance imaging system, comprising a magnet, arranged to produce a magnetic field defining a main Bo field direction, a radio-frequency transmitter, and at least one radio-frequency receiver, and a processor, the system configured to: select a plurality of slices through a target to image, the target being within the magnetic field, wherein the plurality of slices are each perpendicular to a slice direction; transmit, using the radio-frequency transmitter, an initial multi-band radio frequency pulse perpendicular to the main Bo field direction" in the presence of a slice gradient which enables separation of the initial signal spectrum along the slice direction; acquire, using the at least one radio-frequency receiver, an initial signal spectrum comprising signals produced by at least a part of the target, as a result of excitation by the initial multi-band radio frequency pulse, comprising applying an additional gradient along the slice direction which enables separation of the initial signal spectrum along a frequency direction, perpendicular to the proposed phase encoding direction and the slice direction; process, using the processor, the initial signal spectrum to give, for each of a plurality of positions along the slice direction, a frequency projection along a proposed phase encoding direction, perpendicular to the slice direction; derive from each frequency projection a plurality of signal strength indicator values, each associated with a phase encode gradient value; -25 -create an ordering of the phase encode gradient values based on the plurality of signal strength indicator values derived from the plurality of frequency projections, so as to define a time-varying sequence of phase encode gradient values; transmit, using the transmitter, a multi-band radio frequency pulse sequence into the target, perpendicular to the main Bo field direction, using the time-varying sequence of phase encode gradient values applied along the phase encoding direction; acquire, using the at least one radio-frequency receiver, an imaging signal spectrum comprising signals produced by at least a part of the target, as a result of excitation by the multi-band radio frequency pulse sequence, comprising applying an additional gradient along the slice direction which enables separation of the imaging signal spectrum along the frequency direction; and process the imaging signal spectrum to produce the image-data array.13. The system of claim 12, wherein the signal strength indicator value for a particular phase encode gradient value, for a particular frequency projection, is representative of a signal-to-noise ratio for data of said particular frequency projection at said particular phase encode gradient value.14. The system of claim 12 or 13, configured to create an ordering of phase encode gradient values by summing the signal strength indicator values for each phase encode gradient value, to give an averaged signal strength indicator value for each phase encode gradient value, and then place the phase encode gradient values in an order based on the averaged signal strength indicator values.15. The system of claim 12 or 13 configured to create an ordering of phase encode gradient values by interleaving sequentially the order of the phase encode gradient values based on an ordering of signal strength indicator values of the phase encode gradient values in any of the frequency projections.16. The system of any of claims 12 to 15, configured to create the ordering by placing the phase encode gradient values in order of highest to lowest signal strength indicator value, or averaged signal strength indicator value.-26 - 17. The system of any of claims 12 to 16, configured to select the plurality of slices by adjusting a timing of the initial multi-band radio frequency pulse and the multi-band radio frequency pulse sequence.18. The system of any of claims 12 to 17, wherein the multi-band radio frequency pulse sequence is frequency modulated.19. The system of any of claims 12 to 18, configured to use phase undersampling.20. The system of any of claims 12 to 19, comprising a plurality of radio-frequency receivers, and wherein the time-varying sequence of phase encode gradient values comprises only a sub-set of the phase encode gradient values.21. The system of claim 20, configured to infer, interpolate, extrapolate or reconstruct the data for the rest of the phase encode gradient values based on the sensitivity profiles of the plurality of radio-frequency receivers.22. The system of any of claims 12 to 21, further configured to process the image data array to produce an image corresponding to each of the plurality of slices.