WO2021123384A1

WO2021123384A1 - A method and controller for magnetic resonance imaging

Info

Publication number: WO2021123384A1
Application number: PCT/EP2020/087324
Authority: WO
Inventors: Johan BERGLUND; Henric RYDÉN
Original assignee: Berglund Johan; Ryden Henric
Priority date: 2019-12-20
Filing date: 2020-12-18
Publication date: 2021-06-24
Also published as: GB201918935D0; GB2590492A

Abstract

A method of obtaining image data from a sample in a magnetic resonance imaging system, the method comprising the steps of: exciting the sample with a radio frequency, RF, pulse; acquiring image data by sampling an RF-signal from the excited sample over a sampling-time-window; and applying a readout-gradient-waveform, with an amplitude- profile within the sampling-time-window that is asymmetric about a fixed-time-axis corresponding to a temporal midpoint of the sampling-time-window.

Description

A Method and Controller for Magnetic Resonance Imaging

Field

The present disclosure relates to a method and controller for magnetic resonance imaging (MRI) and in particular to a method of applying a pulse sequence to obtain image data from a sample in MRI.

Background

In medical applications of Magnetic Resonance Imaging (MRI), it is of special interest to encode the chemical shift between fat and water. The rationale may be to locate fat, to get information about fat content, or to suppress the fat signal in order to improve visualization of pathology or contrast enhanced tissue [1]

The simplest form of chemical shift encoding is opposed phased imaging, where the signal readout happens when fat and water are 180 degrees out of phase. The chemical shift encoding will then manifest as signal dropout in voxels containing both fat and water. Signal readout at two or more time points with varying degrees of fat/water dephasing enables subsequent separation of fat and water signals. Thus, a fat-only and a water-only image may be formed in the reconstruction step. Such fat/water separation is also known as the Dixon method [2].

Chemical shift encoding can be performed in gradient echo sequences as the degree of fat/water dephasing is proportional to the echo time. However, in conventional spin echo (SE) or fast spin echo (FSE) sequences, there is no dephasing between fat and water, because the spin-echo coincides with the centre of a readout-gradient-waveform and a corresponding sampling-time-window, resulting in in-phase data. Nonetheless, chemical shift encoding can be achieved by shifting the readout gradient relative to the spin echo [2].

Often, certain degrees of fat/water dephasing are preferable over others. In opposed phase imaging, the signal dropout effect is maximised if fat and water are exactly 180° out of phase. In the two-point Dixon method, the signal-to-noise ratio (SNR) will be optimal if one point is sampled exactly opposed phase and the other exactly in phase [3]. Attending to such a prescription often prolongs the acquisition time or introduces a dead time in the sequence, lowering the SNR efficiency. This is especially problematic for FSE Dixon imaging, where the shifting of readout gradients will require longer echo spacing, resulting in both prolonged acquisition time and lower image quality due to aggravated T2 blurring.

For example, with 320 readout samples acquired with a receiver bandwidth (rBW) of ±25 kHz, the readout window duration is 6.4 ms. To achieve a 180-degree-phase shift (a TT- phase shift), dead times of 4.6 / 2.3 ms are required at a main magnetic field strength of 1.5 / 3 T (to fulfil CP G (Carr-Purcell-Meiboom-Gill) conditions). This results in a sampling occupancy of 58 / 74 %. In addition to this inefficient sampling, the echo spacing is prolonged, resulting in stronger T2 blurring.

Summary

According to a first aspect of the present disclosure there is provided a method of obtaining image data from a sample in a magnetic resonance imaging system, the method comprising the steps of: exciting the sample with a radio frequency, RF, pulse; acquiring image data by sampling an RF-signal from the excited sample over a sampling-time-window; and applying a readout-gradient-waveform, with an amplitude-profile within the sampling-time-window that is asymmetric about a fixed-time-axis corresponding to a temporal midpoint of the sampling-time-window.

Applying a readout-gradient-waveform with an asymmetric amplitude profile can control a gradient-echo-time within the sampling time window. Therefore, applying an asymmetrical gradient waveform can advantageously be used to eliminate or reduce sequence dead time or improve sampling efficiency through the more flexible control of the gradient-echotime relative to the temporal footprint of the readout-gradient-waveform.

Applying a readout-gradient-waveform may comprise applying a readout-gradient- waveform along a single linear readout-axis.

Applying a readout-gradient-waveform during the sampling-time-window may provide chemical shift encoding to the image data. In this way, the method may be a method of obtaining chemical shift encoded image data.

Applying the readout-gradient-waveform may control a gradient-echo-time within the sampling time window. A gradient-echo-time may be offset from the temporal midpoint of the sampling-time- window,

A gradient-echo-time may coincide with a central moment of the amplitude-profile within the sampling-time-window.

The sampling-time-window may comprise a samp!ing-start-time and a sampling-stop-time. The readout-gradient-waveform may comprise a readout-gradient-on-time and a readout- gradient-off-time. The sampling-start-time may be cotemporaneous with, or occur after, the readout-gradient-on-time and the sampling-stop-time may be cotemporaneous with, or occur before, the readout-gradient-off-time.

The temporal midpoint of the sampling-time-window is a time half-way between the sampling-start-time and the sampling-stop-time. The readout-gradient-on-time may be a time at which a magnitude of the amplitude-profile increases from zero. The readout- gradient-off-time may be a time at which the magnitude of the amplitude-profile decreases, or returns, to zero.

A gradient-echo-time may be offset from a temporal midpoint of the readout-gradient- waveform.

Applying a readout-gradient-waveform during the sampling-time-window may comprise; applying a first part of the readout-gradient-waveform from the start time to the temporal midpoint of the sampling-time-window; and applying a second part of the readout-gradient-waveform from the temporal midpoint of the sampling-time-window to the stop-time, wherein an integral of the amplitude-profile of the first part differs from an integral of the amplitude-profile of the second part.

Applying a readout-gradient-waveform during the sampling-time-window may comprise applying a readout-gradient-waveform with a spline-based amplitude-profile.

Applying a readout-gradient-waveform during the sampling-time-window may comprise applying a readout-gradient-waveform with a triangular-amplitude-profile. An apex of the triangular-ampiitude-profile may be temporally offset from the temporal midpoint of the sampling-time-window. Applying a readout-gradient-waveform during the sampling-time-window may comprise applying a readout-gradient-waveform with a varying amplitude-profile that is skewed relative to the temporal midpoint of the sampling-time-window.

Applying a readout-gradient-waveform during the sampling-time-window may comprise applying a readout-gradient-waveform with a single polarity.

The readout-gradient-waveform may be substantially non-zero during the sampling-time- window.

The method may further comprise the step of resampling the image data on a Cartesian grid.

The method may further comprise the step removing coloured noise from the image data based on the amplitude-profile. Removing the coloured noise may comprise applying a noise whitening filter to the image data.

The method may further comprise the steps of: acquiring second image data by sampling a second-RF-signal from the excited sample during a second-sampling-time-window; and applying a second-readout-gradient-waveform during the second-sampling-time- window with a second-amplitude-profile that is symmetric about a fixed-time-axis corresponding to a temporal midpoint of the second-sampling-time-window.

The image data may comprise out-of-phase image data and the second image data may comprise in-phase data. In one or more embodiments the image data may comprise opposed-phase image data.

The method may further comprise the steps of: acquiring one or more further image data by sampling corresponding further-echo- signals during corresponding further-sampling-time-windows; and for each further-sampling-time-window, applying a corresponding further-readout- gradient-waveform during the further-sampling-time-window with a further-amplitude- profile that is asymmetric about a fixed-time-axis corresponding to a temporal midpoint of the further-sampling-time-window. Each of the further-readout-gradient-waveforms may have different further-amplitude- profiles to each other and to the readout-gradient-profile.

The further image data may comprise further out-of-phase image data.

The method may further comprise the steps of: deriving water content image data based on the out-of-phase image data; and deriving fat content image data based on the out-of-phase image data.

The method may further comprise applying a RF refocussing pulse before the sampling- time-window to provide a method of conventional spin echo or fast spin echo magnetic resonance imaging.

The method may further comprise applying a dephasing-readout-waveform before the sampling-time-window with a dephasing-polarity which is opposite to a readout-polarity of the readout-gradient-waveform during the sampling-time-window to provide a method of gradient echo magnetic resonance imaging.

The method may provide a method of gradient and spin echo, GRASE, magnetic resonance imaging, wherein one or more gradient-echoes and one or more spin-echoes are acquired between successive pairs of RF refocussing pulses.

According to a second aspect of the present disclosure there is provided a controller for controlling a magnetic resonance imaging system to perform any method disclosed herein.

According to a further aspect of the present disclosure there is provided a magnetic resonance imaging system comprising any controller disclosed herein or configured to perform any method disclosed herein.

There may be provided a computer program, which when run on a computer, causes the computer to configure any apparatus, including a controller, system, or device disclosed herein or perform any method disclosed herein. The computer program may be a software implementation, and the computer may be considered as any appropriate hardware, including a digital signal processor, a microcontroller, and an implementation in read only memory (ROM), erasable programmable read only memory (EPROM) or electronically erasable programmable read only memory (EEPROM), as non-limiting examples. The software may be an assembly program. The computer program may be provided on a computer readable medium, which may be a physical computer readable medium such as a disc or a memory device, or may be embodied as a transient signal. Such a transient signal may be a network download, including an internet download.

Brief Description of the Drawings

One or more embodiments will now be described by way of example only with reference to the accompanying drawings in which:

Figure 1 shows an example MR! system;

Figure 2 illustrates a shifted-readout-pulse-sequence for use in spin-echo MRI; Figure 3 shows fat and water separated spin-echo MRI images;

Figure 4 illustrates an asymmetric-readout-pulse-sequence for use in a spin-echo MRI method according to an embodiment;

Figure 5A illustrates example asymmetric-readout-gradient-waveforms for use in a method according to an embodiment;

Figure 5B illustrates a further example asymmetric-readout-gradient-waveforms for use in a method according to an embodiment;

Figure 6 illustrates an asymmetric-readout-pulse-sequence for use in a gradient- echo MRI method according to an embodiment;

Figure 7 shows fat and water spin-echo MRI images obtained by a method according to an embodiment; and

Figure 8 illustrates a flow chart depicting a method according to an embodiment. Detailed Description

One or more of the examples disclosed herein can address the deficiencies of the above described methods by providing a method and controller for acquiring MRI data with chemical shift encoding that avoids or reduces dead time and is compatible with many MRI sequences including spin-echo and gradient-echo techniques.

Figure 1 illustrates a schematic cross-section of an example MRI system 100. The MRI system 100 comprises a main magnet 102, gradient coils 104 and an RF coil 106 surrounding a sample 108. The main magnet 102, gradient coils 104 and RF coil 106 may be arranged in a tubular structure surrounding the sample 108. The main magnet 102 applies a static magnetic field Bo across the sample 108. The gradient coils 104 are configured to superimpose spatially distributed magnetic field gradients upon the static magnetic field and the sample 108. In some examples the magnetic field gradients generated by the gradient coils 104 may be along three orthogonal axes and may comprise: a slice gradient for selecting a slice through the sample; a phase gradient for phase encoding nuclear spins (or nuclear magnetic moments) within a selected slice of the sample 108; and a readout gradient for frequency encoding the nuclear spins within the selected slice. Encoding the nuclear spins by phase and frequency along two orthogonal axes within a slice can provide spatial resolution when detecting a RF signal from the spins within the slice.

The RF coil 106 is configured to provide an RF pulse to excite the sample 108. The RF coil is also configured to receive RF signals emitted from the nuclear spins within the sample in response to the RF excitation pulse. In some examples the excitation and receiving functions may be provided by separate RF coils.

A controller 110 is configured to control the MR! system 100 for example by controlling the operation of the gradient coils 104 and the RF coil 106. The controller 110 may control the MR! system 100 by providing a sequence of waveforms (or a pulse sequence) to each of the gradient coils 104 and the RF coil 106. The controller 110 may also receive signals from the RF coil 106 based on the response of the RF coil 106 to RF signals received from the sample 108. The MR! system 100 may comprise other intermediate components not shown, such as amplifiers, detectors, drivers etc as known in the art.

Figure 2 illustrates a spin-echo MRI pulse sequence 211 with temporally shifted readout gradient waveforms that can be applied to a MRI system to obtain image data with chemical shift encoding. The shifted readout pulse sequence 211 is plotted against a time axis 212. The pulse sequence may be implemented by the controller of Figure 1 to acquire image data from the MRI system.

Throughout this disclosure, spin-echo MRI may refer to either or both of fast spin-echo MRI or conventional spin-echo MRI and the term is not intended to be limiting unless explicitly state otherwise.

An RF excitation pulse sequence comprises an initial RF excitation pulse 214 followed by a series of RF refocussing pulses 216A-216D. The RF excitation pulse sequence can be applied using a RF coil, such as the one illustrated in Figure 1. In this example, the initial RF excitation pulse 214 is a 90-degree RF excitation pulse. In other examples, the intial RF excitation pulse 214 may be less than a 90-degree RF excitation pulse. In this example the series of RF refocussing pulses 216A-216D are 180-degree RF refocussing pulses. In other examples, the RF refocussing pulses 216A-216D may be less than 180-degree RF refocussing pulses to reduce heating of the sample.

Under the sole influence of the static magnetic field, B₀, the nuclear spins of a sample align along the same axis as the static magnetic field. The magnetic moments of the nuclear spins precess around the static magnetic field at the Larmor Frequency. In MRI, the referred to nuclear species is typically a proton. Applying a substantially 90-degree initial RF excitation pulse 214 excites protons in the sample by flipping their proton magnetic moments (or proton moments) into a plane transverse to the primary magnetic field Bo. Immediately following the initial RF excitation pulse 214, the excited proton moments are in phase and aligned along the same axis in the transverse plane and continue to precess around the static magnetic field axis.

The ensemble of protons in the sample will precess around the static magnetic field axis at different frequencies due to a variation in their local environments. As a result, dephasing of the proton moments (that is dephasing of the proton precessions) occurs and corresponding RF signals emitted by the protons, gradually fall out-of-phase. By applying a 180-degree RF refocussing pulse 216A, the proton moments are rotated 180° about an axis in the transverse plane. As a result, slower varying (lower frequency) proton moments, that lagged behind faster varying (higher frequency) proton moments, now lead the faster varying proton moments. After a further period of time (equal to the period of time between the 90-degree RF pulse 214 and the first 180-degree RF pulse 216A) the fast varying proton moments catch back up with the slower varying proton moments and the ensemble of proton moments are in phase again. The refocussed phases of the proton moments results in a RF pulse echo, known as a spin-echo. Following the spin- echo, the proton moments dephase again. Following this, the whole process can be repeated again in conventional spin-echo MRI. A further approach is to apply further RF refocussing pulses 216B, 216C, 216D to obtain further spin-echoes before the signal decays due to T₂ relaxation. Applying a plurality of RF refocussing pulses 216 following a single initial RF pulse 214 is known as fast-spin-echo (FSE). The first spin-echo occurs at a time ESP (echo spacing) after the initial RF pulse 214 for a first refocussing pulse 216A applied at ESP/2. To a first approximation, the further refocussing pulses 216B-216D and further spin-echoes occur at the same subsequent intervals. In this example, the first spin echo occurs at a time ESP that is approximately 10.2 milliseconds after the initial RF excitation pulse 214.

The pulse sequence 211 may also comprise a slice gradient sequence. The slice gradient sequence may comprise slice gradient waveforms 218A-218D that may be applied during the RF pulses 214, 216A-216D. Each slice-gradient-waveform 218A-218D can be identified in the Figure as comprising three trapezoid shaped sub-waveforms surrounding and coinciding with the RF pulses 214, 216A-216D. The slice gradient waveforms 218A- 218D may be applied by the gradient coils of Figure 1. Applying a slice gradient waveform applies a magnetic field gradient along a slice-selection-axis of the sample such that the RF excitation pulses 214, 216 (applied at a specific excitation frequency) only excite protons in a particular planar slice of the sample perpendicular to the slice-selection-axis. The slice-selection-axis may be co-axial with the axis of the static magnetic field.

A readout gradient sequence 220 can be applied in the intervening period between successive RF pulses 214, 216 (and corresponding slice selection gradient waveforms). The readout gradient sequence 220 comprises a series of readout gradient waveforms 222, 224. The readout gradient waveforms 222, 224 may be applied by a gradient coil, such as the gradient coils of Figure 1. The readout gradient waveforms 222, 224 may be applied along a readout-axis which may be perpendicular to the slice-selection-axis.

The readout gradient sequence 220 is applied to measure the RF response of the sample to the RF excitation 214, 216. Application of the readout gradient sequence itself can generate additional dephasing and rephasing of the proton spins along the readout-axis.

A dephasing readout gradient waveform 222 is applied between the initial RF excitation pulse 214 and the first RF refocussing pulse 216A. Applying this initial readout waveform produces known dephasing of the proton spins along the readout-axis, prior to the first refocussing pulse 216A.

Subsequent readout gradient waveforms 224A-224D occur between the RF refocussing pulses 216A-216D. These readout-gradient-waveforms 224A-224D are applied during corresponding sampling-time-windows 226A-226D. The sampling-time-windows 226A- 226D are illustrated in the figure as vertically striped patterned regions. The sampling- time-windows 226A-226D define a time over which the RF signal is sampled. The readout-gradient-waveforms 224A-224D rephase the proton spins along the readout- axis. An area of the readout-gradient-waveforms 224A-224D may be equal to twice the area of the dephasing gradient waveform 222. In this way, a gradient echo 217A-217D occurs at a centre of the readout-gradient-waveform 224A-224D (or equivalently a centre of the sampling-time-window 226A-226D) corresponding to realignment of the proton spins in the sample along the readout-axis. The second half of the readout-gradient-waveform 224A-224D then dephases the proton spins in the same manner as the dephasing readout gradient waveform 224A-224D prior to the next RF refocussing pulse 216A-216D.

Most of the rephasing/dephasing of the protons along the readout-axis may be imposed by the readout-gradient-waveform.

During the sampling-time-windows 226A-226D, only a readout-gradient-waveform 224A- 224D should be applied. In other words, the sampling-time-windows 226A-226D should not coincide with the application of the slice-gradient-waveforms 228A-228D or a phase encoding gradient waveform (not shown).

Each sampling-time-window 226A-226D has a sampling-start-time and a sampling-stop time. Each readout-gradient-waveform 224A-224D has a readout-gradient-on-time and a readout-gradient-off-time. The readout-gradient-on-time may be cotemporaneous with, or occur before, the sampling-start-time of the corresponding sampling-time-window. The readout-gradient-off-time may be cotemporaneous with, or occur after, the sampling-stoptime of the corresponding sampling-time-window. In this way, readout-gradient- waveforms 224A-224D are substantially non-zero during, or over, the corresponding sampling-time-windows 226A-226D. The sampling-time-window 226A-226D may be considered as a period of time within a duration of the readout-gradient-waveform 224A- 224D in which the RF signal including the gradient echo 217A-217D is sampled.

In the example of Figure 2, the readout-gradient-on-time occurs prior to the sampling-start- time, and the readout-gradient-off-time occurs after the sampling-stop-time, of the corresponding sampiing-time-window. In this way, an amplitude of the readout-gradient- waveform can rise from zero to a value required for sampling the RF signal in the time between the readout-gradient-on-time and the sampling-start-time. Similarly, the amplitude of the readout-gradient-waveform can decrease to zero from the sampling value in the time between the sampling-stop-time and the readout-gradient-off-time. As a result, the readout gradient waveforms 224A-224D have an amplitude-profile with a slight trapezoid shape with: a ramp-on-portion between the readout-gradient-on-time and the sampling-start-time; a ramp-off-portion between the sampling-stop-time and the readout- gradient-off-time; and a fixed (or DC) amplitude-profile between the sampling-start-time and the sampling-stop-time.

The readout gradient waveforms 224A-224D have a fixed amplitude-profile during the sampling time windows 226A-226D. In other words, the sampling-time-windows 226A- 226D coincide with the fixed amplitude-profile portion of the corresponding readout- gradient-waveforms 224A-224D, The amplitude of the readout-gradient-waveforms is measured in T/m, For a fixed amplitude in a sampling-time-window duration, the sampling will occur at a constant rate or bandwidth,

The RF signal including the gradient echo 217A-217D may be sampled over the corresponding sampling-time-windows 226A-226D by measuring the RF signal with an RF coil, such as the RF coil of Figure 1. Applying the readout gradient waveform 224A-224D over the sampling-time-window 226A-226D, may comprise applying a magnetic field gradient along the readout-axis. The magnetic field gradient may apply a linear frequency gradient to the proton moments along the readout-axis. This can frequency encode the proton moments producing the RF signal and provide spatial resolution along the readout- axis. This frequency encoding can provide spatial resolution of the protons along the readout axis. Decoding the measured RF signals according to frequency can be achieved by Fourier transforming the measured temporal response during the sampling-time-window 226A-226D. Applying a phase gradient waveform (not shown) before each echo, can provide spatial resolution along a phase-axis perpendicular to the read-out axis and the slice-selection axis. In this way, spatial resolution of the protons may be provided for a slice of the sample. Similarly, a second phase encoding axis can be applied along the slice-selection axis to provide spatial resolution along all three spatial dimension, i.e. 3D imaging.

In this example, two of the sampling-time-windows 226B, 226D are centred on corresponding spin-echoes at times 2ESP and 4ESP. Equivalently, the temporal midpoints of the sampling-time-windows 226B, 226D are at a time ( 2ESP , 4ESP ) equidistant between the two corresponding consecutive RF refocussing pulses 216. In this way, the gradient echoes 217B, 217D are cotemporaneous with the corresponding spin echoes at times 2ESP, 4ESP. Applying corresponding readout-gradient-waveforms 224B, 224D over these sampling-time-windows 226B, 226D will acquire in-phase RF signals and therefore in-phase image data from the corresponding gradient echo 217B, 217D. The other two samp!ing-time-windows 226A, 226C are offset from the centre of the corresponding spin-echoes at times ESP and 3ESP. Equivalently, the temporal midpoints of the sampling-time-windows 226B, 226D are at a time that is not equidistant between the two corresponding consecutive RF refocussing pulses 216. In this way, the gradient echoes 217A, 217C are not cotemporaneous with the corresponding spin echoes at times ESP, 3ESP . Applying corresponding readout-gradient-waveforms 224A, 224C over these sampling-time-windows 226A, 226C will acquire out-of-phase RF signals and therefore out-of-phase image data from the corresponding gradient echo 217A, 217C. These out- of-phase sampling-time-windows 226A, 226C and their associated out-of-phase waveforms are examples of shifted readout gradient waveforms and sampling-time- windows.

It will be understood that throughout this disclosure, “out-of-phase” refers to any phase relationship that is not exactly or substantially in-phase. “Opposed-phase” refers to a specific example of “out-of-phase” in which the phase relationship corresponds exactly or substantially to a 180 degree phase shift

As mentioned above, following the spin-echo, the proton moments in the sample begin to dephase again due to variations in their local environments. The magnetic moments of protons in water-based portions of the sample will have a higher precession frequency than proton moments of fat-based portions. As a result, the RF signals of the water-based portions and the fat-based portions move in and out of phase with time. Acquiring both in- phase and out-of-phase image data can enable the isolation of water-based images and fat-based images. Applying a shifted readout pulse sequence 211 to acquire out-of-phase RF echo signals in MRI is an example of chemical shift encoding.

Figure 3 illustrates MRI image data acquired using a chemical shift encoding spin-echo MRI pulse sequence, such as that illustrated in Figure 2. In-phase image data 330 and out-of-phase image data 332 are illustrated in the top two images. In this example, the out-of-phase image data 332 is opposed-phase image data with a phase shift of 180 degrees between the fat and water signals. A black boundary 334 known as an India ink artefact can be seen in the out-of-phase image data 332. The boundary 334 corresponds to boundaries of fat tissue and water tissue. As both types of proton are present and both are 180 degrees out of phase, the RF signals cancel and no image data is detected in the boundary 334. To a first approximation: a water image 336 can be obtained by summing the in-phase image data 330 and the opposed-phase image data 332; and a fat image 338 can be obtained by subtracting the opposed-phase image data 332 from the in-phase image data 330. Some complementary features of the water image 336 and the fat image 338 can be seen in the Figure.

Returning to Figure 2, the echo spacing, ESP, or equivalently the time between the RF refocussing pulses 216, should be kept to a minimum to reduce an overall acquisition time (or scan time) of the MRI image acquisition process and reduce any signal loss due to T2 relaxation. As described above, only a readout-gradient-waveform 224 (and not the slice- gradient 228 or phase-gradient) should be applied during the sampling-time-windows 226. Therefore, the minimum possible time, ESP, between the RF refocussing pulses 216 is determined by a duration, T_s, of the sampling-time-windows 226. Each sampling-time- window 226 should have the same duration, T_s, which may be based on a bandwidth of the RF receiver and a sampling size. However, to allow the shifted sampling-time-windows 226A, 226C and the shifted readout-gradient-waveforms 224A, 224C to sample the out- of-phase RF signals without overlapping the slice gradient 228, the time, ESP, between the RF refocussing pulses 216 must be larger than the duration of the sampling-time- window, T_s, As a result, there is a dead-time To between each refocussing pulse 216, and before (and / or after) the sampling-time-window 226, during which no gradient fields are applied and the RF signal is not sampled. Therefore, for a shifted readout/sampling-time- window approach to chemical shift encoding, a dead-time T_D is required, which increases the echo spacing ESP which in turn increases the acquisition time of the MRI image acquisition. The dead-time To may be twice the desired temporal shift. For example, to achieve opposed-phase image data at a static magnetic field of 1.5 T, a temporal shift of 2.3 ms is required resulting in a dead-time of 4.6 ms added to each echo spacing. In the example of Figure 2, the sampling duration of the sampling-time-window is approximately 4 ms and the dead time is approximately 2 ms. An increased echo spacing ESP also results in increased T2 relaxation for each refocused echo which can reduce the signal level of the RF echoes and result in image blurring.

Figure 4 illustrates a spin-echo pulse sequence 411 that can be applied to a MRI system by a method according to an embodiment of the present disclosure. The pulse sequence 411 , shown in the lower half of Figure 4, termed an asymmetric readout pulse sequence 411 may be applied to obtain image data with chemical shift encoding, while avoiding or reducing a deadtime in the readout gradient sequence 420. The shifted readout pulse sequence 211 of Figure 2 is also repeated in the upper half of Figure 4 for comparison. Features of the asymmetric readout pulse sequence 411 which are also present in the shifted readout pulse sequence 211 have been given corresponding numbers in the 400 series and will not necessarily be described again here.

The asymmetric readout pulse sequence 411 comprises a RF excitation pulse sequence comprising an initial RF excitation pulse 414 followed by a series of RF refocussing pulses 416A-416D. The RF excitation pulse sequence 411 of the asymmetric readout pulse sequence 411 is substantially the same as that of the shifted readout pulse sequence 211 with the exception that the RF excitation pulse sequence of the asymmetric readout pulse sequence 411 has a reduced echo spacing, ESP. As explained further below, this can arise because the asymmetric readout pulse sequence 411 may not require a deadtime. The RF excitation pulse sequence may be applied to the MR! system by a RF coil, such as the RF coil of Figure 1.

When the RF excitation pulse sequence is applied to the RF coil of a MRI system a series of RF spin-echoes is produced from the sample (at times corresponding to integer multiples of the echo spacing ESP ) by the same process as described above in relation to Figure 2 (the only difference being the shorter echo spacing ESP).

A slice gradient sequence of the asymmetric readout pulse sequence 411 comprises slice gradient waveforms 418A-418D that are substantially the same as the slice gradient waveforms 218A-218D of the shifted readout pulse sequence 211.

A readout gradient sequence 420 can be applied in the intervening period between successive RF pulses 414, 416 (and corresponding slice selection gradient waveforms). The readout gradient sequence 420 comprises a series of readout-gradient-waveforms 422, 424A-424D. The readout gradient waveforms 422, 424A-424D may be applied by a gradient coil, such as the gradient coils of Figure 1. The readout-gradient-waveforms 422, 424A-424D may be applied along a readout-axis which may be perpendicular to the slice- selection-axis.

A dephasing-readout-gradient-waveform 422 is applied between the initial RF pulse 414 and the first RF refocussing pulse 416A in a similar manner to that described above in relation to Figure 2. The dephasing-readout-gradient-waveform 422 is shorter in duration relative to the corresponding waveform 222 of Figure 2, due to the reduced echo spacing ESP. Subsequent readout-gradient-waveforms 424A-424D occur between the RF refocussing pulses 416A-416D. These readout-gradient-waveforms 424A-424D are applied during corresponding sampling-time-windows 426A-426D. As described above in relation to Figure 2, the sampling-time-windows 426A-426D define a time over which the RF signal is sampled. During the sampling-time-windows 426A-426D, only the readout-gradient- waveform 424A-424D should be applied. In other words, the sampling-time-windows 426A-426D should not coincide with the application of the slice selection gradient 428 or a phase encoding gradient waveform (not shown).

The RF signal may be sampled over the sampling-time-windows 426A-426D by measuring an RF signal with an RF coil, such as the RF coil of Figure 1 , Applying the readout gradient waveform 424A-424D over the sampling-time-window 426A-426D, may comprise applying a magnetic field gradient along the readout-axis. The magnetic field gradient may apply a linear frequency gradient to the proton moments along the readout-axis as described above in relation to Figure 2.

In contrast to the shifted readout sequence of Figure 2, all of the sampling-time-windows 426A-426D in the asymmetric-readout-pulse-sequence 411 are centred on corresponding spin-echoes at times corresponding to integer multiples of the echo spacing ESP. In other words, temporal midpoints of the sampling-time-windows 426A-426D correspond to temporal midpoints of the corresponding echo signals 417A-417D. Equivalently, the temporal midpoints of the sampling-time-windows 426A-426D are at a time equidistant between the two corresponding consecutive RF refocussing pulses 416.

In the same way as described for Figure 2, each sampling-time-window 426A-426D has a sampling-start-time and a sampling-stop-time. Each readout-gradient-waveform 424A- 424D has a readout-gradient-on-time, at which a magnitude of the amplitude-profile increases from zero, and a readout-gradient-off-time, at which the magnitude of the amplitude-profile decreases, or returns, to zero. The readout-gradient-on-time may be cotemporaneous with, or occur before, the sampling-start-time of the corresponding sampling-time-window. The readout-gradient-off-time may be cotemporaneous with, or occur after, the sampling-stop-time of the corresponding sampling-time-window. In this way, readout-gradient-waveforms 424A-424D may be substantially non-zero during, or over, the corresponding sampling-time-windows 426A-426D, The sampling-time-window 426A-426D may be considered as a period of time within the duration of the readout- gradient-waveform 424A-424D in which the RF signal is sampled. Each of the readout-gradient-waveforms 424A-424D have a ramp-on-portion between the readout-gradient-on-time and the sampling-start-time and a ramp-off-portion between the sampling-stop-time and the readout-gradient-off-time. As a result, the ramp-on-portion and the ramp-off-portion occur outside the sampling-time-windows 426A-426D. In this way, the amplitude-profile of the readout-gradient-waveforms 424A-424D may be substantially non-zero over the sampling-time-window 426A-426D.

To obtain in-phase RF echo signals (or in-phase image data) from the sample, two in- phase readout-gradient-waveforms 424B, 424D have the same amplitude-profile as the read-out-gradient-waveforms 224A-224D of Figure 2. That is, the in-phase readout- gradient-waveforms 424B, 424D have a fixed amplitude-profile over the corresponding sampling-time-windows 426B, 426D. For each of these in-phase readout-gradient- waveforms 424B, 424D, the fixed amplitude-profile is symmetric, within the sampling-time- window, about a fixed-time-axis corresponding to a temporal midpoint of the corresponding sampling-time-window 426B, 426D. In this way, the in-phase readout-gradient-waveforms 424B, 424D will rephase the proton spins along the readout axis and produce a corresponding gradient echo 417B, 417D at the centre of the readout-gradient-waveform. That is the gradient echo 417B, 417D coincides with a corresponding spin echo at time 2ESP, 4ESP.

Applying these in-phase readout-gradient-waveforms 424B, 424D with an amplitude- profile within the sampling-time-windows 426B, 426D that is symmetric about a fixed-time- axis corresponding to a temporal midpoint (2ESP, 4ESP in this example) of the sampling- time-window 426B, 426D will acquire in-phase RF signals and therefore in-phase image data from the corresponding gradient echo 417B, 417D.

To obtain out-of-phase RF gradient echo signals from the sample, two asymmetric- readout-gradient-waveforms 424A, 424C have a variable amplitude-profile during the corresponding sampling-time-windows 426A, 426C. In particular, these asymmetric- readout-gradient-waveforms 424A, 424C applied over the sampling-time-windows 426A, 426C have an amplitude-profile that is asymmetric about a fixed-time-axis corresponding to a temporal midpoint (ESP, 3ESP in this example) of the corresponding sampling-time- windows 426A, 426C.

Applying an asymmetric-readout-gradient-waveform 424A, 424C will produce a gradient echo 417A, 417C in the RF signal that is temporally offset from the temporal midpoint of the corresponding sampling-time-window 426A, 426C. The gradient echo 417A, 417C will occur at a time corresponding to a central moment of the amplitude-profile of the asymmetric-readout-gradient-waveform 424A, 424C. The central moment (or centre of mass) is a time within the sampling-time-window 426A, 426C at which the time integral of the amplitude-profile (from the sampling-start-time) is half of the total time integral of the amplitude-profile over the sampling-time-window 426A, 428C (from sampling start-time to sampling-stop-time). The central moment corresponds to a time at which the readout- gradient-waveform has rephased the proton spins along the readout-axis. In this way, applying an asymmetric-readout-gradient-waveform 424A, 424C can be considered as controlling a gradient-echo-time (or a time of the gradient-echo) in the sampling-time- window 426A, 426C. In this example, the gradient echo 417A, 417C is temporally offset from the corresponding spin echo at time ESP, 3ESP. Therefore, the effect of temporally (chemically) shifting a gradient echo 417 A, 417C can be achieved without a corresponding shift in the sampling-time-window 426A, 426C / readout-gradient-waveform 424A, 424C.

In this example, the asymmetric-readout-gradient-waveforms 424A, 424C have a higher amplitude value after the temporal midpoint of the corresponding sampling-time-window 426A, 426C, than before it. In this way, the gradient echo 417A, 417C occurs after the temporal midpoint of the sampling-time-window 426A, 426C. In other examples, the asymmetric-readout-gradient-waveforms 424A, 424C may have a lower amplitude value after the temporal midpoint of the corresponding sampling-time-window 426A, 426C, than before it. In further examples, the asymmetric amplitude-profile may vary in a more complex manner with peaks and troughs on either side of the temporal midpoint of the sampling-time-window 426A, 426C. In some examples, the asymmetric-readout-gradient- waveform 424A, 424C may have a single polarity. In other examples, the asymmetric- readout-gradient-waveform 424A, 424C amplitude-profile may comprise both positive and negative amplitude-values. It will be appreciated that the variations discussed are not specific to spin-echo MRI pulse sequences and can apply to gradient-echo MRI pulse sequences (described further below) or other MRI pulse sequences, including 3D sequences.

It should be understood that the symmetry of the in-phase readout-gradient-waveforms 426B, 426D and the asymmetry of the asymmetric-readout-gradient-waveforms 424A, 424C is defined within the corresponding sampling-time-window 426A-426D and about the fixed-time-axis defined by the temporal midpoint of the sampling-time-window 426A-426D. The asymmetric-readout-gradient-waveforms can be considered as having a varying bandwidth, as the sampled spatial frequencies are non-equispaced. However, the physical sampling rate is constant.

Applying the asymmetric-readout-gradient-waveform 424A, 424C with an amplitude-profile within (or over) the sampling-time-window 426A, 426C that is asymmetric about a fixed- time-axis corresponding to a temporal midpoint (or temporal centre) of the sampling-time- window 426A, 426C will acquire out-of-phase RF signals and therefore out-of-phase image data from the corresponding gradient-echo 417A, 417C.

Therefore, applying asymmetric-readout-gradient-waveforms 424A, 424C that have asymmetric amplitude-profiles within the sampling-time-windows 426A, 426C can achieve chemical-shift-encoding. Advantageously, no temporal shifting of the asymmetric-readout- gradient-waveforms 424A, 424C and corresponding sampling-time-windows 426A, 426C is required. All sampling-time-windows 426A-426D are centred on the corresponding RF spin-echo at times corresponding to integer multiples of the echo spacing. As a result, it may be possible to maintain a constant echo spacing, ESP, between consecutive RF refocussing pulses 416 without requiring a deadtime prior to, or subsequent to, the sampling-time-window 426A-426D. An asymmetric pulse sequence without a deadtime can have shorter echo spacings resulting in reduced T2 relaxation between echoes and the associated image blurring.

In the illustrated example, the readout-gradient-on-time of the asymmetric-readout- gradient-waveform 424A, 424C occurs before the corresponding sampling-start-time of the sampling-time-window 426A, 426C. Similarly, the sampling-stop-time occurs before the readout-gradient-off-time. In other embodiments, the sampling-start-time may be cotemporaneous with the readout-gradient-on-time and / or the sampling-stop-time may be cotemporaneous with the readout-gradient-off-time. In the example, during the ramp- on-time and the ramp-off time, both a readout-gradient-waveform 424A, 424C and a corresponding slice-gradient-waveform 418A, 418C are present. However, during the sampling-time-window 426A, 426C only a readout-gradient-waveform 424A, 424C is applied, as described above. However, in some examples, more than one readout- gradient-waveform may be applied at the same time.

To summarise, a method of obtaining image data from a sample in a MRI system comprises; exciting the sample with a RF pulse 418; acquiring image data by sampling an RF signal from the excited sample over a sampling-time-window 426A; and applying a readout-gradient-waveform 424A with an amplitude-profile within the sampling-time- window 426A that is asymmetric about a fixed-time-axis corresponding to a temporal midpoint of the sampling-window 426A. Such a method may be performed by applying the asymmetric-readout-pulse-sequence 411 of Figure 4 to an MRI system.

As indicated above, applying the asymmetric-readout-gradient-waveform 424A, 424C controls a gradient-echo-time within the sampling time window. The gradient-echo-time is offset from the temporal midpoint of the sampling-time-window 426A, 426C. The control is provided as the gradient-echo-time coincides with the central moment of the amplitude- profile of the asymmetric-readout-gradient-waveform 424A, 424C. Therefore, the amplitude-profile can be selected to achieve a specific gradient-echo-time. For example, the gradient-echo-time can be shifted to a prescribed degree of fat/water dephasing without changing a temporal footprint of the readout-gradient-waveform.

In one or more examples, applying an asymmetric-readout-gradient-waveform may comprise applying the asymmetric-readout-gradient-waveform along a single linear readout-axis. In other words, the asymmetric-readout-gradient-waveform is applied along one, and only one, linear readout-axis. The single linear readout-axis may correspond to an axis of a gradient coil or may correspond to a linear combination of two or more axes each corresponding to a respective gradient coil. By applying the asymmetric-readout- gradient-waveform along a single linear readout-axis, Cartesian trajectories can be used for the image acquisition and reconstruction. In other examples, non-cartesian trajectories (for example, spiral, radial or propeller) can be used by applying two or more asymmetric- readout-gradient-waveforms along two or more corresponding gradient axes at the same time (within the sampling time window).

In one or more examples, the method may further comprise: applying an in-phase readout- gradient-waveform 424B, 424D having a symmetric amplitude-profile within the sampling- time-window 426B, 426 D for a second echo. Water and fat MR image data can then be obtained from the resulting in-phase and out-of-phase data. In some examples, water and/or fat MR data can be obtained from the out-of-phase data alone.

In one or more examples, the method may further comprise applying one or more further asymmetric-readout-gradient-waveform 424A, 424C having a corresponding further asymmetric amplitude-profile within the sampling-time-window 426A, 426C for one or more corresponding further echos. The further asymmetric amplitude-profiles may differ from each other and the asymmetric amplitude-profile of the asymmetric-waveform applied for the first echo. In this way, the one or more further asymmetric-readout-gradient- waveforms 424A, 424C can correspond to different chemical shifts, or phase shifts, to the first asymmetric-readout-gradient-waveform 424A, 424C. Applying one or more further asymmetric-readout-gradient-waveforms 424A, 424C may or may not be performed in combination with the applying of an in-phase readout-gradient-waveform 424B, 424C.

The asymmetric-readout-gradient-waveforms can be used to achieve any degree of dephasing (or chemical shift) between fat and water, for example generally out-of-phase, opposed phase or quadrature phase. In a three-point Dixon method, sampling one point in quadrature phase can be beneficial [5],

It will be appreciated that these methods of applying one or more asymmetric readout- gradient-waveforms and / or symmetric readout-gradient-waveforms is not specific to spin- echo MR! pulse sequences and is also applicable to other MRI pulse sequences such as gradient-echo pulse sequences, discussed further below in relation to Figure 6.

Applying an asymmetric-readout-gradient-waveform 424A, 424C with an amplitude-profile within the sampling-time-window 426A, 426C that is asymmetric about a fixed-time-axis corresponding to a temporal midpoint of the sampling-time-window 426A, 426C can also be described in the following corresponding ways:

• Applying a first part of the asymmetric-readout-gradient-waveform 424A, 424C from a start time of the sampling-time-window 426A, 426C to the temporal midpoint of the sampling-time-window 428A, 426C; and applying a second part of the asymmetric-readout-gradient-waveform 424A, 424C from the temporal midpoint of the sampling-time-window 426A, 428C to a stop time of the sampling-time-window 426A, 426C, wherein an integral of (or an area under) the amplitude-profile of the first part differs from an integral of (or area under) the amplitude-profile of the second part. It should be noted that this definition does not limit the asymmetric- readout-gradient-waveform 424A, 424C to comprise two distinct parts and the asymmetric-readout-gradient-waveform 424A, 424C may have a continuous profile, including at a transition from the first part to the second part. The artificial division into a first temporal part and a second temporal part is merely to aid the description of the amplitude-profile asymmetry within the sampling-time-window 426A, 426C. • Applying an asymmetric-readout-gradient-waveform 424A_» 424C with a varying amplitude-profile within the sampling-time-window 426A, 426C that is skewed relative to the temporal-midpoint of the sampling-time-window 426A, 426C.

• Applying an asymmetric-readout-gradient-waveform 424A, 424C with an amplitude-profile within the sampling-time-window 426A, 426C having a central moment corresponding to a time that is offset from a temporal midpoint of the sampling-time-window 426A, 426C. The central moment is a time within the sampling-time-window 426A, 426C at which the time integral of the amplitude- profile of the readout-gradient-waveform 424A, 424C equals half of its total value integrated across the whole sampling-time-window 426A, 426C. The sampling- time-window 426A, 426C may be considered as comprising a plurality of individual sampling-time-steps. The central moment of the amplitude-profile may also be considered as a weighted average sampling-time-step within the sampling-time- window 426A, 426C, wherein the weighting is according to the amplitude-profile at a corresponding sampling-time-step.

By way of summary of Figures 2 and 4, in conventional spin echo or fast spin echo imaging, chemical shift encoding can be performed by shifting the readout gradient, such that the gradient echo does not coincide with the spin echo [4]. This introduces a dead time on the "other side" of the spin echo. The resulting increase in echo spacing can lead to unsharp images (increased T2 blurring). Additionally, a lower number of slices can be sampled in the repetition time (TR). Increasing TR to sample the same number of slices results in a longer acquisition time. The asymmetrical-readout-gradient-waveforms of Figure 4 can advantageously perform chemical shift encoding without temporally shifting the sampling- time-windows or readout-gradient-waveforms. In this way, the echo spacing does not need to increase relative to the echo spacing used for in-phase acquisition. Therefore, the problems of increased blurring and a decreased number of slices / increased acquisition time can be reduced or avoided. The use of asymmetrical-readout-gradient-waveforms can beneficially result in less T2 blurring and higher SNR efficiency compared to the use of shifted gradients.

In the example of Figure 4, every other readout-gradient-waveform is chemical shift encoded. In other examples, acquisitions may be implemented in other ways, such as "sequential" rather than "interleaved" acquisition of the two echoes. Regardless, asymmetrical gradient waveforms can be used to eliminate or reduce sequence dead time or improve sampling efficiency through the more flexible choice of chemical shift encoding relative to the gradient temporal footprint. Figure SA illustrates two asymmetric-readout-gradient-waveforms 524-1, 524-2 that may be applied to a MRI system during a sampling-time-window in a method according to an embodiment of the present disclosure. For reference, a shifted-readout-gradient- waveform 524-3 is also illustrated. The asymmetric-readout-gradient-waveforms 524-1, 524-2 can each enable out-of-phase image data to be acquired from a MRI system as part of chemical shift encoding.

For the asymmetric waveforms 524-1, 524-2 and the shifted waveform 524-3, corresponding segments of a RF excitation fast spin echo pulse sequence comprising two RF refocussing pulses 516A, 516B are illustrated. Slice-gradient-waveforms 518A, 518B associated with the corresponding 180-degree RF pulses 516A, 516B are also shown. A time ESPx equidistant between the two RF refocussing pulses 516A, 516B, corresponding to a time of a spin-echo (at an integer multiple of the echo spacing ESP after an initial RF pulse) is also illustrated.

For each readout-gradient-waveform 524-1, 524-2, 542-3 the corresponding sampling- time-windows 526-1, 526-2, 526-3 have a sampling-start-time cotemporaneous with the readout-gradient-on-time and corresponding to a time t=0. The sampling-stop-time is cotemporaneous with the readout-gradient-off-time and corresponds to a time t_s.

A time-to-centre (time to central moment of the readout-gradient-waveform), t_c, corresponds to a desired gradient-echo-time at which the gradient echo will occur. In this FSB example, the desired gradient-echo-time corresponds to a desired phase shift (or chemical shift) for acquiring out-of-phase image data. That is an offset time, D = t_c- ESP, corresponds to a desired amount of dephasing between fat proton spins and water proton spins. In other examples, such as gradient echo MRI (discussed below in relation to Figure 6), the desired gradient-echo-time may correspond to in-phase or out-of-phase image data.

The time-to-centre, t_c, may be defined such that an integral or sum of the amplitude-profile, G_read(t), of the readout-gradient-waveform from t=0 to t₀ corresponds to 50% of the integral or sum of the amplitude-profile from t=0 to fe

In other words, the time-to-centre, t_c, is the time at which a cumulative-amplitude-value of the readout-gradient-waveform is 50% of a total-amplitude-value. The time-to-centre, t_c, can be considered as a time corresponding to the central moment of the amplitude-profile of the readout-gradient-waveform within the sampling-time-window.

To satisfy the above equation for a non-zero offset time (D = t_c - ESP ) and corresponding phase-shift, the shifted-readout-gradient-waveform 524-3 requires a deadtime, To, prior (or subsequent) to the sampling-time-window 526-3, as described above. To achieve the desired phase shift, the shifted-readout-gradient-waveform 524-3 and corresponding sampling-time-window 526-3 are offset from ESP such that the shifted-readout-gradient- waveform 524-3 is centred on t_c. In other words, tJ2 = t_c and a deadtime, To, is required to obtain a chemical shift

The two asymmetric-readout-gradient-waveforms 524-1 , 524-2 have an amplitude-profile that is asymmetric about a fixed-time-axis corresponding to the temporal midpoint of the sampling-time-window 526. In this way, the asymmetric-readout-gradient-waveforms 524 1, 524-2 can satisfy the above equation, for a non-zero offset time (D = t_c - ESP ) and corresponding phase-shift, due to their varying amplitude-profiles. As a result, the corresponding sampling-time-windows 526-1 , 526-2 can run between the end of the first slice-gradient-waveform 518A and the beginning of the second slice-gradient-waveform 518B and no deadtime is required. As a result, the echo spacing, ESP, and corresponding acquisition time is less for the asymmetric-readout-gradient-waveform approach relative to the shifted-readout-gradient-waveform approach.

The two asymmetric-readout-gradient-waveforms 524-1, 524-2 illustrate two different example amplitude-profiles that can be used to obtain a desired chemical shift corresponding to the offset time (D = t_c - ESP),

A first asymmetric-readout-gradient-waveform 524-1 comprises an asymmetric triangle comprising a first segment 540 and a second segment 542. The first-segment 540 can be defined by a first duration, t_a, and a peak amplitude, Go. A first-slew-rate, s_a, can be defined as Go/t_a. The second-segment 542 can be defined by a second duration, t_g, and the amplitude, Go- A second-slew-rate, s_a, can be defined as Go/fc.

The time-to-centre, t_c, duration, t_s = t_a + U, and area, M, of the asymmetric triangle can be determined according to the desired chemical shift, a duration of the sampling-time- window 526 and a desired resolution, respectively. For a positive phase-shift, the first segment duration is given by: For a negative phase-shift, the asymmetric triangle waveform is mirrored about the midpoint of the sampling-time-window 526. The first asymmetric-readout-gradient-waveform 524-1 can achieve a maximum positive offset time (A=t_c-TE) and corresponding phase-shift when fe = 0 (or t_a = 0 for a negative phase shift). This results in a maximum positive offset time of:

A second asymmetric-readout-gradient-waveform 524-2 comprises a spline-based amplitude-profile. In this example the asymmetric-readout-gradient-waveform 524-2 comprises a cubic-spline-based amplitude-profile, but in other examples it may comprise a quadratic or other spline-based profile. The waveform may be described by the following two interval polynomials, with three knots may be located at t=0, t_c, and t_s:

In total, these two polynomials have eight degrees of freedom. Two degrees of freedom may be eliminated by letting:

G_s( 0) — f¾(t_s) — 0

Further, two degrees of freedom may be eliminated by letting these polynomials satisfy the time-to-centre integral condition defined above, that:

Further, three degrees of freedom may be eliminated by letting the values, the first derivatives, and the second derivatives of the polynomials to be equal at t_c. The final degree of freedom may be eliminated by letting the second derivative be zero at t_s in line with the concept of “natural splines”.

The above conditions can be satisfied and provide a maximum positive phase shift for the asymmetric-readout-gradient-waveform 524-2 that is larger than the asymmetric triangular waveform 524-1. The cubic-spline based asymmetric-readout-gradient-waveform 524-2 can provide a larger range of phase-shift (larger offset time, A=i_c-TE ) than the asymmetric triangular waveform 524-1. The first derivative of the cubic-spline based asymmetric- readout-gradient-waveform 524-2 is continuous which can improve gradient fidelity during playout.

Figure 5A also illustrates cumulative-amplitude-profiles 544 for each of the three waveforms. The cumulative-amplitude-profiles can be considered as the integral of the amplitude-profile over the sampling-time-window.

The cumulative-amplitude-profile 544-3 for the shifted-readout-gradient-waveform 524-3 increases linearly because of the fixed amplitude-profile of the shifted-readout-gradient- waveform 524-3. As a result, the time-to-centre, t_c, occurs at tJ2. Therefore, measuring out-of-phase image data (or producing an offset gradient echo) requires a deadtime prior or subsequent to the sampling-time-window.

The cumulative-amplitude-profiles 544-1, 544-2 for the asymmetric-readout-gradient- waveforms 524-1 , 524-2 are non-linear because of the varying / asymmetric amplitude- profiles of the asymmetric-readout-gradient-waveforms 524-1 , 524-2. As a result, the time- to-centre, t_c, is offset from the temporal midpoint of the sampling-time-window (f/2). Therefore, out-of-phase RF echo signals can be acquired without a temporal offset or deadtime.

Figure 5B illustrates a further example asymmetric-readout-gradient-waveforms for use in a method according to an embodiment. Figure 5B is substantially identical to Figure 5A with the exception that the echo spacing is the same for the asymmetric-readout-gradient- waveforms 524-4, 524-5 and the shifted-readout-gradient-waveform 524-6.

The asymmetrical-readout-gradient-waveforms 524-4, 524-5 can achieve the same chemical shift encoding as the s h ifted -reado ut-g rad ient-wa veto rm 524-6 while avoiding any dead time. In this way, the asymmetric-readout-gradient-waveforms 524-4, 524-5 have a longer sampling-time-window than the shifted-readout-gradient-waveform 524-6 and the SNR efficiency is increased.

Therefore, applying asymmetric-gradient-waveforms can eliminate a requirement for a dead time providing: (i) an improved SNR for the same acquisition time as the sh ifted - readout-gradient-waveform, as illustrated in Figure 5B; or (ii) a reduced acquisition time relative to the shifted-readout-gradient-waveform, as illustrated in Figure 5A. These advantages of improved SNR or reduced acquisition time are not exclusive to fast-spin- echo and also apply to gradient echo MRI. Figure 6 illustrates a gradient echo pulse sequence 611 that can be applied to a MRI system by a method according to an embodiment of the present disclosure. The asymmetric readout pulse sequence 611 may be applied to obtain image data of any phase, while avoiding or reducing a dead time in the sequence. A shifted readout-gradient- sequence 620-1 is also shown for comparison.

The asymmetric readout pulse sequence 611 comprises a RF excitation pulse 614, The RF excitation pulse 614 may be a 90-degree RF excitation pulse or may be less than a 90- degree excitation pulse. In gradient echo MRI there are no RF refocusing pulses. Instead, the readout gradient can be applied to de-phase and re-phase the proton spins, as described below.

The asymmetric readout pulse sequence 611 comprises a slice gradient waveform 618. The slice gradient waveform 618 may select a slice through the sample in the same way as described above in relation to figures 2 and 4.

The Figure illustrates a shifted readout gradient sequence 620-1 and an asymmetric readout gradient sequence 620-2. A first de-phasing part 646-1, 646-2 of the readout- gradient-sequences 620-1 , 620-2 can be applied by gradient coils to induce dephasing between the proton spins in the sample along the readout-axis.

Following the de-phasing part 646-1 , 646-2 of the readout-gradient-sequences 620-1 , 620- 2, a readout-gradient-waveform 624-1 , 624-2, having opposite polarity to the dephasing part 646-1 , 646-2, can be applied to rephase the proton spins and produce a gradient- echo. The readout-gradient-waveform 624-1 , 624-2 may be applied over a sampling-time- window 626-1 , 626-2 to measure the RF gradient-echo.

The Figure illustrates a desired echo time, TE, for producing a gradient echo and obtaining image data. To produce a gradient echo at TE using the shifted-readout-gradient- waveform 624-1 , the corresponding sampling-time-window 626-1 is centred at TE. A deadtime, 7b, is required to achieve this because the amplitude-profile of the shifted- readout-gradient-waveform 624-1 has a fixed value during the sampling-time-window 626- 1. The shifted-readout-gradient-waveform 624-1 has an amplitude-profile within the sampling-time-window 626-1 that is symmetric about a fixed-time-axis (TE) corresponding to a temporal midpoint of the sampling-time-window 626-1. In contrast, the sampling-time-window 626-2 for the asymmetric-readout-gradient- waveform 624-2 can begin immediately following the de-phasing part 646-2 of the readout- gradient-sequence 620-2. A gradient echo can be produced at the desired echo time, TE, by applying a readout-gradient-waveform 624-2 with an amplitude-profile within the sampling-time-window 626-2 that is asymmetric about a fixed-time-axis corresponding to a temporal midpoint of the sampling-time-window 626-2. In this way, image data can be captured using a gradient-echo-pulse-sequence 611 at any desired echo time without requiring a deadtime in the readout-gradient-sequence 620-2.

The asymmetric readout pulse sequence 611 is not limited to obtaining out-of-phase data in gradient echo MRI. The benefit of the asymmetric amplitude-profile applies to any desired echo time for which a symmetric waveform would require a dead time. For example, an asymmetric-readout-gradient-waveform may be applied to acquire in-phase image data.

One benefit of the asymmetric-readout-gradient-waveform 624-2 is added sampling time. The shifted-readout-gradient-waveform 624-1 has a fixed sampling duration (of the sampling-time-window 626-1) determined by the receiver bandwidth and number of samples. As a result, the sampling-start-time / readout-gradient-on-time of the shifted- readout-gradient-waveform 624-1 is determined by the desired echo time and a dead time is required. In this way, a shifted-readout-gradient-waveform 624-1 cannot utilize all the available sampling time. The asymmetric-readout-gradient-waveform 624-2 can replace the dead time with sampling. As a result, the signal-to-noise ratio is increased for an asymmetric waveform as more time is spent acquiring image data.

In one or more examples of gradient echo MRI, an opposed phase image may be desired. The desired echo time, TE, can be set such that fat and water will be in opposed phase. For a conventional trapezoid shifted-readout-gradient-waveform 624-1 with a fixed sampling duration (corresponding to the desired image resolution), obtaining opposed phase image data may require a dead time before the sampling-time-window. This can be avoided by lowering the gradient amplitude, but only at the expense of prolonging the gradient duration, thus increasing the total image acquisition time. Applying an asymmetric-readout-gradient-waveform 624-2 provides a higher degree of freedom in sequence flexibility. For example, the dead time can be replaced by signal sampling, while not increasing the total acquisition time. Gradient echo MRI pulse sequences for fat/water imaging are often implemented as a train of at least two gradient echoes after the RF excitation pulse 614. Asymmetrical readout gradient waveforms allow a more flexible choice of gradient echo times relative to the readout gradient on and off times / sampling start and stop times. An increase in SNR efficiency can be realised by avoiding sequence dead time. The improved flexibility of asymmetrical readout gradients can also provide further image quality improvement as certain combinations of gradient echo times can be better than others in terms of SNR in the calculated images [5].

A gradient echo MR! pulse sequence may comprise one or more asymmetric-readout- gradient-waveforms 624-2. In one or more examples, a two-point gradient echo acquisition may produce a first gradient echo that is exactly in-phase and a second gradient echo that is exactly opposed phase. This can maximise SNR in the calculated images. A first-asymmetrical-readout-gradient-waveform could be provided to produce the in-phase gradient echo at the desired echo time without requiring a dead-time. A second asymmetrical-readout-gradient-waveform could produce the out-of-phase gradient echo with an amplitude profile of equal area but shorter duration to shorten the image acquisition time. Such short acquisition times could be applied in medical applications such as abdominal breath-hold imaging, flow imaging, time-of-flight angiography, etc.

In summary, the method of obtaining image data with an asymmetric-readout-gradient- waveform from a sample in a MRI system can be applied to both gradient echo sequences and spin-echo sequences. In yet further embodiments, the method of applying an asymmetric-readout-gradient-waveform over the sampling-time-window can be applied to other MRI sequences, for example, Gradient And Spin Echo (GRASE) MRI.

One or more of the examples disclosed herein can also be used for application using chemical shift encoding to separate other chemical species, such as ¹H fat/water/silicone separation [6] or species separation for other isotopes, such as ¹³G [7]. Such examples can also be advantageous in applications encoding-resonance due to BO inhomogeneity, such as BO-mapping, susceptibility weighted imaging (SWl) [8], and quantitative susceptibility mapping (QSM) [9], Furthermore, examples disclosed herein can also be used for chemical shift encoding based temperature mapping [10].

The amplitude-profile of the asymmetric-gradient-waveforms varies during readout. Therefore, a constant sampling rate can result in a varying k-space sampling density in the readout-axis. However, a desired field-of-view in the reconstructed images may correspond to a specific constant sampling density. To obtain the prescribed field-of-view, the acquired image data can be resampled on a Cartesian grid. This resampling is restricted to the readout dimension. The monotonic trajectories can enable resampling by sine interpolation [11]

Another effect of the varying amplitude-profile during signal readout is that the amount of noise can vary between samples. In other words, coloured noise may be introduced, which may not be desired. For opposed phase imaging, the coloured noise may be counteracted by a noise whitening filter. For fat/water imaging, the conditionality of the model matrix may also vary, since the chemical shift encoding can vary between sample pairs. Some spatial frequencies may therefore be more susceptible to noise amplification. This can be adjusted by using a Tikhonov regularized pseudo-inverse of the model matrix when solving the inverse problem. The amount of regularization required for each spatial frequency to achieve white noise can be calculated as described elsewhere [12]

In both cases, the noise whitening can alter the modulation transfer function to some extent. Nonetheless, the advantages of using asymmetric-readout-gradient-waveforms can outweigh the cost of any additional processing required to remove noise.

Figure 7 illustrates fast spin-echo MRI images of a human cervical spine obtained by an asymmetric readout pulse sequence method according to an embodiment of the present disclosure. Similar images obtained by a shifted readout pulse sequence are also illustrated for reference.

Chemical shift encoded image-data was acquired with the asymmetric readout pulse sequence method in a time of 3.0 minutes. Similar image-data was acquired with the shifted readout pulse sequence in 3.7 minutes. The longer acquisition time of the shifted- readout-method results from the deadtime requirement as illustrated in Figures 2 and 4.

The upper two images comprise an asymmetric-readout-water-image 750 and a shifted- readout-water-image 752 constructed from the corresponding image data. The images are grayscale with whiter regions indicating higher water content. The asymmetric- readout-water-image 750 provides finer structural detail than the shifted-readout-water- image 752. This results from the reduced echo spacing and the resulting reduction in T2 relaxation between readouts and associated image blurring. For example, further detail can be seen at the top of the spinal column 758 in the asymmetric-readout-water-image 750 The lower two images comprise asymmetric-readout-fat-image 754 and a shifted-readout- fat-image 754 constructed from the corresponding image data. In a similar way to the water images 750, 752, the asymmetric-readout-fat-image 754 provides finer structural detail than the shifted-readout-fat-image 758. For example, a small tail 760 can be seen protruding from a vertebra in the asymmetric-readout-fat-image 754 which is not visible in the shifted-readout-fat-image 756.

Figure 8 shows a flowchart that summarises an embodiment of the method. The flowchart shows exciting (870) a sample with a radio frequency, RF, pulse; acquiring (880) image data by sampling an RF-signal from the excited sample over a sampling-time-window; and applying (890) a readout-gradient-waveform, with an amplitude-profile within the sampling-time-window that is asymmetric about a fixed-time-axis corresponding to a temporal midpoint of the sampling-time-window.

In summary, the disclosed methods and controller for implementing such can provide faster acquisition times in MR! with reduced T2 image blurring. The disclosed methods apply asymmetric-readout-waveforms to provide efficient out-of-phase sampling without a dead time or redundancy. The techniques allow for flexible echo shifts without introducing dead time while still fulfilling CPMG conditions. The concept can be applied to a range of pulse sequences including spin-echo sequences and gradient echo sequences.

References

[1] T. A. Bley, O. Wieben, C. J. Francois, J. H. Brittain, and S. B. Reeder. Fat and water magnetic resonance imaging. J. Magn. Reson. Imaging, 31(1 ):4-18, Jan. 2010.

[2] W. T. Dixon. Simple proton spectroscopic imaging. Radiology, 153(1 ):189-194, Oct. 1984.

[3] Q.-S. Xiang. Two-point water-fat imaging with partially-opposed-phase (POP) acquisition: An asymmetric dixon method. Magn. Reson. Med., 56(3):572-584, Sept. 2006.

[4] Q.-S. Xiang and L. An. Water-fat imaging with direct phase encoding (DPE), July 2000. [5] A. R. Pineda, S. B. Reeder, Z.Wen, and N. J. Pelc. Cramer-Rao bounds for three- point decomposition of water and fat. Magn. Reson. Med., 54(3):625-635, Sept. 2005.

[6] L. An and Q. S. Xiang. Chemical shift imaging with spectrum modeling. Magn. Reson. Med., 46(1): 126-130, July 2001.

[7] S. B. Reeder, J. H. Brittain, T. M. Grist, and Y.-F. Yen. Least-squares chemical shift separation for (13)c metabolic imaging. J. Magn. Reson. Imaging, 26(4): 1145-1152, Oct. 2007.

[8] E. M. Haacke, Y. Xu, Y.-C. N. Cheng, and J. R. Reichenbach. Susceptibility weighted imaging (SWI). Magn. Reson. Med., 52(3):612-618, Sept. 2004.

[9] L. de Rochefort, R. Brown, M. R. Prince, and Y. Wang. Quantitative MR susceptibility mapping using piece-wise constant regularized inversion of the magnetic field. Magn. Reson. Med., 60(4): 1003-1009, Oct. 2008.

[10] B. J. Soher, C. Wyatt, S. B. Reeder, and J. R. MacFall. Noninvasive temperature mapping with MRI using chemical shift water-fat separation. Magn. Reson. Med., 63(5): 1238-1246, May 2010.

[11] J. D. O'Sullivan. A fast sine function gridding algorithm for fourier inversion in computer tomography. IEEE Trans. Med. Imaging, 4(4):200-207, 1985.

[12] H. Ryden, J. Berglund, O. Norbeck, E. Avventi, and S. Skare. T1 weighted fat/water separated PROPELLER acquired with dual bandwidths. Magn. Reson. Med., 80(6):2501- 2513, Dec. 2018.

Claims

1 , A method of obtaining image data from a sample in a magnetic resonance imaging system, the method comprising the steps of: exciting the sample with a radio frequency, RF, pulse; acquiring image data by sampling an RF-signal from the excited sample over a sampling-time-window; and applying a readout-gradient-waveform, with an amplitude-profile within the sampling-time-window that is asymmetric about a fixed-time-axis corresponding to a temporal midpoint of the sampling-time-window.

2. The method of claim 1 , wherein applying a readout-gradient-waveform comprises applying a readout-gradient-waveform along a single linear readout-axis.

3. The method of claim 1 or claim 2, wherein applying a readout-gradient-waveform during the sampling-time-window provides chemical shift encoding to the image data.

4. The method of any preceding claim, wherein applying the readout-gradient- waveform controls a gradient-echo-time within the sampling time window.

5. The method of any preceding claim, wherein a gradient-echo-time is offset from the temporal midpoint of the sampling-time-window.

6. The method of any preceding claim, wherein a gradient-echo-time coincides with a central moment of the amplitude-profile within the sampling-time-window.

7. The method of any preceding claim, wherein; the sampling-time-window comprises a sampling-start-time and a sampling-stoptime; the readout-gradient-waveform comprises: a readout-gradient-on-time; and a readout-gradient-off-time; the sampling-start-time is cotemporaneous with, or occurs after, the readout- gradient-on-time; and the sampling-stop-time is cotemporaneous with, or occurs before, the readout- gradient-off-time.

8. The method of any preceding claim wherein a gradient-echo-time is offset from a temporal midpoint of the readout-gradient-waveform.

9. The method of any preceding claim, wherein applying a readout-gradient- waveform during the sampling-time-window comprises applying a readout-gradient- waveform with a spline-based amplitude-profile.

10. The method of any of claims 1 to 8, wherein applying a readout-gradient-waveform during the sampling-time-window comprises applying a readout-gradient-waveform with a triangular-amplitude-profile, wherein an apex of the triangular-amplitude-profile is temporally offset from the temporal midpoint of the sampling-time-window.

11. The method of any preceding claim, wherein applying a readout-gradient- waveform during the sampling-time-window comprises applying a readout-gradient- waveform with a varying amplitude-profile that is skewed relative to the temporal midpoint of the sampling-time-window.

12. The method of any preceding claim, wherein applying a readout-gradient- waveform during the sampling-time-window comprises applying a readout-gradient- waveform with a single polarity.

13. The method of any preceding claim further comprising the step of resampling the image data on a Cartesian grid.

14. The method of any preceding claim further comprising the step removing coloured noise from the image data based on the amplitude-profile.

15. The method of claim 14 wherein removing the coloured noise comprises applying a noise whitening filter to the image data.

16. The method of any preceding claim further comprising the steps of; acquiring second image data by sampling a second-RF-signal from the excited sample during a second-sampling-time-window; and applying a second-readout-gradient-waveform during the second-sampling-time- window with a second-amplitude-profile that is symmetric about a fixed-time-axis corresponding to a temporal midpoint of the second-sampling-time-window,

17. The method of claim 16, wherein: the image data comprises out-of-phase image data; and the second image data comprises in-phase data.

18. The method of any preceding claim further comprising the steps of: acquiring one or more further image data by sampling corresponding further-echo- signals during corresponding further-sampling-time-windows; and for each further-sampling-time-window, applying a corresponding further-readout- gradient-waveform during the further-sampling-time-window with a further-amplitude- profile that is asymmetric about a fixed-time-axis corresponding to a temporal midpoint of the further-sampling-time-window.

19. The method of claim 18, wherein the further image data comprises further out-ofphase image data.

20. The method of any preceding claim, wherein the method further comprises the steps of: deriving water content image data based on the out-of-phase image data; and deriving fat content image data based on the out-of-phase image data.

21. The method of any preceding claim, further comprising applying a RF refocussing pulse before the sampling-time-window to provide a method of conventional spin echo or fast spin echo magnetic resonance imaging.

22. The method of any of claims 1 to 20, further comprising applying a dephasing- readout-waveform before the sampling-time-window with a dephasing-polarity which is opposite to a readout-polarity of the readout-gradient-waveform during the sampling-time- window to provide a method of gradient echo magnetic resonance imaging.

23. The method of claim 22 providing a method of gradient and spin echo, GRASE, magnetic resonance imaging, wherein one or more gradient-echoes and one or more spin- echoes are acquired between successive pairs of RF refocussing pulses.

24. A controller for controlling a magnetic resonance imaging system to perform the method of any of claims 1 to 23.

25. A magnetic resonance imaging, MRI, system comprising the controller of claim 24 or configured to perform the method of claims 1 to 23.