WO1987006699A1

WO1987006699A1 - A method for performing volume-selected nmr spectroscopy

Info

Publication number: WO1987006699A1
Application number: PCT/AU1987/000112
Authority: WO
Inventors: David Michael Doddrell; William Miles Brooks
Original assignee: University Of Queensland
Priority date: 1986-04-24
Filing date: 1987-04-24
Publication date: 1987-11-05
Also published as: ZA872921B; JPS63503122A

Abstract

A method for performing volume-selected NMR spectroscopy where all unwanted magnetization is eliminated from the selected volume element of the body by using shaped rf pulses to excite all spin outside the volume element, dephasing the coherence of these spins by application of a field gradient and then observing the spin of interest with a single read pulse.

Description

Title: "A METHOD FOR PERFORMING VOLUME-SELECTED NMR SPECTROSCOPY" BACKGROUND OF THE INVENTION

(1) Field of the Invention THIS INVENTION relates to a method for per¬ forming volume-selected NMR spectroscopy and NMR imaging.

(2) Prior Art

NMR spectroscopy and NMR imaging both provide spatially encoded information about the sample of interest. One way of determining this information is by initially making a slice of magnetization. This slice can then be used as the basis for volume-selected spectroscopy or imaging. A number of objects can be prescribed for effective slice selection. It is important that the development of any technique at least addresses these objects and preferably satisfies the demands therein. These objects include the ability to be able to obtain slices of variable thickness with high sensitivity. The technique should use as little radio-frequency pulse power as possible. Because of rapid spin-spin (T₂ ) relaxation of signals compared with spin-lattice (T_x) relaxation, magnetization of interest in the slice of interest should remain in the transverse plane for as little time as possible so that firstly, strong signals result and secondly, to ensure that T₂ contrast between various components does not take place. Slice formation must facilitate incorporation into effective volume selection and imaging schemes.

In spectroscopy, three orthogonal and inter¬ secting slices are made and the chemical-shift signal acquired from the volume element centred about this intersection point. In imaging, where chemical shift is not normally resolved, initial slice selection is followed by some mechanism to encode spatial information in a second direction and a gradient applied during acquisition ensures encoding in the final direction. Fundamentally however, the initial slice selection is important as it, to a great extent, determines the final signal to noise obtained. Slicing is usually done by applying a frequency selective pulse (e.g. sine or gaussian) in the presence of some magnetic field gradient . If this is done to a collection of spins at equilibrium in a magnetic field, a slice, with width dependent upon the gradient strength and frequency bandwidth of the pulse applied, will be excited. It is the object of volume-selection and imaging schemes to treat such slices in such a manner that they efficiently yield localized information of interest.

A number of methods have been previously described to carry out slice selection for spectroscopy. These include the methods known as DRESS, [Depth- resolved surfεce-coil spectroscopy (DRESS) for in vivo *H, "p, and "c NMR, P. A. Bottomley, T. H. Foster and R. D. Darrow, J. Magn. Reson. 5j>_, 338, (1984)]; ISIS, [Image-selected in vivo Spectroscopy (ISIS). A new technique for spatially selective NMR Spectroscopy. R. J. Ordidge, A. Connelly and J. A. B. Lohman. J. Magn. Reson. 6_, 283, (1986)]; SPARS, [Solvent-suppressed spatially resolved spectroscopy. An approach to high resolution NMR on a whole-body MR system. P. R. Luyten, A. J. H. Marrien B. Sijtsma and J. A. den Hollander, J. Magn. Reson. 67_, 148 (1986)]; and SPACE [Spatial and chemical-shift-encoded excitation. SPACE, a new technique for volume-selected NMR spectroscopy. D. M. Doddrell, W. M. Brooks, J. M. Bulsing, J. Field, M. G. Irving and H. Baddeley. J. Magn. Reson. 6_8_, 367, (1986)]. The known methods can be summarised as follows:-

(1) DRESS uses a selective rf pulse in the presence of an applied field gradient to excite a slice of magnetization. Gradient reversal is then used to form a spin-echo which refocusses the off-resonance effects of the pulse. The major drawback of this method is that the magnetization of interest lies in the trans¬ verse plane during refocussing and so is T₂ con¬ trasted. Furthermore the acquired signals are either gradient broadened or reduced in intensity or both because of the need to allow gradient fall to occur before switching the receiver on. Volume selection is only one dimensional and the finite dimensions of the receiver coil are used to limit the volume in the plane of the slice.

(2) ISIS is based on a phase inversion between signals in the slice of interest and those arising from outside. This is carried out by applying an inversion pulse tailored to the appropriate bandwidth in the presence of a gradient. The computer memory is then used to eliminate the signals by addition and sub¬ traction of appropriate signals. The. primary dis¬ advantage of ISIS is that it relies on the ability of the computer to distinguish between the small signals within the volume of interest and the residual signal which will often be several orders of magnitude stronger. This, of course, is reliant upon the dynamic range of the computer memory as well as the spectrometer preamplifier and receiver systems. (3) SPARS uses a refocussing pulse to form a spin-echo in the presence of an applied field gradient . A selective pulse of appropriate bandwidth is then applied to rotate the magnetization in the slice of interest back to the applied field direction following which the gradient is collapsed. This is carried out in all directions to yield the desired volume of interest which can then be read out with a single pulse. In this method the magnetization of interest suffers the effect of three gradient rises or falls while it is inthe trans- verse plane for each direction of slicing which means any irrecoverable losses associated with gradient rises or falls are extreme. The time for refocussing is extended because gradient changes require finite time whereupon T₂ relaxation becomes important. The sensitivity of this technique is also dependent on efficient refocuss¬ ing and on applying the selective pulse at precisely the correct moment. It appears that the phase evolution of the signal during this pulse also contributes to the weakening of signal strength. (4) SPACE uses a selective pulse in the presence of a gradient to rotate a slice of z-magnetiza- tion into the transverse plane and then uses a high power refocussing pulse to refocus the phase roll. After an appropriate time the slice of magnetization is pulsed back to the field axis and the gradient collapsed. As in other techniques this procedure is repeated in all directions. Finally a read pulse is used to tip the magnetization of interest into the transverse plane for acquisition in the absence of applied gradients. The major drawback of

SPACE is that a hard pulse with bandwidth at least as wide as the selective pulse must be used to ensure accurate refocussing of phase coherence. If this is not effective loss of signal can result. One major problem encountered when carrying out slice selection is that there is generally a large signal phase distortion across the slice associated with the pulses used. This is because of off resonance evolution during the extended duration of selective pulses compared with short non-selective pulses. The result, if not carefully accounted for, can be serious loss of signal. This phase roll can be eliminated in a number of ways including the use of gradient reversal or a refocussing pulse to form a spin-echo. Because these methods involve the use of the transverse plane for this refocussing, the coherence of interest suffers spin-spin relaxation which in some cases may lead to a serious change in the relative intensity of signals, and if extreme, complete loss of those signals with short T_a values.

SUMMARY OF THE PRESENT INVENTION The broad object of the present invention is to obtain nuclear magnetic resonance data from a well defined region in space, without the need to change or move either the apparatus used to carry out the techni¬ que or the sample body.

Further preferred objects are to obtain this information with (1) high signal to noise, (2) little or no T₂ distortion, (3) low rf pulse power, (4) high spectral resolution and (5) complete volume-selection within a single pass through the pulse sequence.

Just as in previously known slicing methods for volume-selection and imaging, the present invention takes advantage of the relation between nuclear magnetic Larmour frequency and magnetic field strength. By apply¬ ing a suitable radio-frequency pulse to the sample body in the presence of a field gradient superimposed upon the initial homogeneous static magnetic field, certain spins can be excited. If the frequency of this excita- tion pulse and the strength of the gradient are known, the position and size of this volume element can be accurately determined.

Consider, as an example, a sample container filled with water. Suppose we wish to acquire NMR signals from a 1 cm volume of spins located at the centre of this sample. In other known volume-selection experiments, traditional wisdom demands selective excitation of the spins of interest and to ignore all other spins. However, it has been demonstrated that selective excitation followed by the application of a field gradient (acting as a homospoil) can achieve quite remarkable signal suppression. [Water signal elimina¬ tion in vivo, using "Suppression by Mistimed Echo and Repetitive Gradient Episodes". D. M. Doddrell, G. J. Galloway, W. M. Brooks, J. Field, J. M. Bulsing, M. G. Irving, and H. Baddeley. J. Magn. Reson. 7_, 176, (1986)].

Thus, a more satisfactory strategy, provided by the method of the present invention, is to firstly eliminate all unwanted magnetization and then acquire the signal from the remaining volume of interest by applying a single read pulse. This may be done by using appropriate shaped rf pulses to excite all spins outside the volume of interest, dephase the coherence of these spins by application of a field gradient (which prefer¬ ably will be the slice selection gradient) and then to observe the spins of interest with a single read pulse. During such an experiment, the magnetization of interest is preferably never in the transverse plane and so is not influenced by short T. values. As well, there are preferably no dynamic range problems as the spins out¬ side the volume of interest preferably have no phase coherence in the transverse plane and hence yield no net signal during data acquisition. To perform this experi- ment, the preferred requirement is an rf pulse or pulses which do not excite the body within the slice of interest but do excite for a wide band outside the slice. That is, the rf pulse must preferably have a hole of controllable width in its excitation profile. BRIEF DESCRIPTION OF THE PREFERRED DRAWINGS Such an rf pulse combination can be developed in two different ways, which will now be described in preferred embodiments of the present invention with reference to the accompanying drawings, in which: FIG. 1(a) shows the definition of the unit step function H(x);

FIGS. Kb) to 1(f) show the functions f_x (x) and f₂ (x) used to define the possible rf pulse excita¬ tion profiles F^t), used for the {ssc} pulse, and F_a (T);

FIG. 2(a) shows the calculated z-magnetization profile following a 2.048ms {ssc} pulse, where the shape determination constants a and β were set to 7.0 and 7.5 respectively and two cycles (-2 to +2 π ) of the rf pulse were employed (the rotation angle was set as hereinafter described in the text ) and was adjusted to yield the minimum z-magnetization in the desired excita¬ tion profile;

FIG. 2(b) shows the calculated z-magnetization profile if the {ssc} pulse is applied for a second time, where the assumption of zero transverse magnetization is made before application of the second {ssc} pulse;

FIG. 2(c) shows the experimentally determined z-magnetization profile determined following one appli- cation of the {ssc} pulse, pulse sequence (A), where the _τ time was set at 10ms for convenience, and the 2 hard pulse time was 70ms and the spectral width shown corresponds to 15kHz;

FIG. 2(d) shows the experimentally determined z-magnetization following two applications of the {ssc} pulse and this experiment corresponds to the result expected using pulse sequence (B);

FIG. 3 shows the shape of the sin-sine rf pulse. (The rf phase alternates +,- starting from one end of the pulse to the other, there being 56 lobes in the pulse. In these embodiments this pulse is generated using 256 data points in the waveform memory of the computer and two cycles of the functions F_x(t) = sin(7xt)sin(7.5κt)/*t for -2s to 2 τ . This is the curve used to generate the {ssc} pulses shown in FIG. 2(a);

1

FIG. 4(a) shows normal H spectrum obtained from the arrangement of bottles as hereinafter discussed in the text. FIGS. 4(b) to 4(d) show slicing in the x and z directions using pulse train (C) set to acquire signals from FIG. 4(b) the bottle containing cyclo- hexane, FIG. 4(c) the bottle containing methanol and FIG. 4(d) the bottle containing benzene. (Each spectrum represents the average of 8 scans obtained with a recycle time of 15s.); FIG. 5 shows the pulse and gradient sequence used for imaging using the sin-since pulse for slice selection;

1

FIG. 6 shows a H image obtained at 100MHz using the sequence described in FIG. 5. The sample was the phantom hereinafter described in the text.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS METHOD 1

Noting that there is an approximate relation- ship between the Fourier transform of the pulse shape and the pulse excitation profile, possible pulse shapes cna be derived by Fourier transformation of f_x* (x) or f₂(x), as shown in FIG. 1. Noting that: f ₍x₎ _ H(x-g) - H(x+α) •^{■ ■}<_ _j

¹ β B f ₍x₎ -H⁽x-g⁾ + H⁽x+p_t) ₍₂₎

² β B where H(x) is the step function as shown. Possible pulse shapes of use are obtained as follows:

/e^{i tx} _fl ( x ) _dx = _Fl ( t ) = 4isin «tsinBt _{( 3 )} 4coso_ tsinβ t (4) e^{i tx} _a ( x ) dx = F₂ ( t ) =

Excitation profiles of radio frequencey pulses having the shapes given by F_j(t), a sin-sine pulse, hereafter labelled {ssc}, and F₂ (t) have been calculated by methods discussed previously. F₂(t) failed to give the desired response while a {ssc} pulse gave the residual z-magnetization profile dependent upon reson- ance off-set as shown in FIG. 2(a). If the {ssc} pulse is repeated again, the z-magnetization profile is as shown in FIG. 2(b). In this instance, the values of o. and B ere set to 7.0 and 7.5 respectively. In practice however, these can be tailored to suit particular experimental needs without comprising results. METHOD II

A second method to produce a hole in the z- magnetization profile can be developed as follows. It is known that a sine pulse, the pulse shape being (sinπt)/πt, produces a rectangular excitation profile. Thus, if one preferably sine pulse is applied at a frequency, shifted by an amount

from the natural excitation frequency, such that the pulse does not excite on-resonance and a second preferably sine pulse is applied at -^)ω a hole will be generated in the z- magnetization profile at the natural resonance frequency provided the values of_lω are set as noted below. In this method, as in the first method proposed above, a field gradient is employed to modify the resonance frequency of the spins dependent upon spatial location. In this way, a z-magnetization hole is generated at a known spatial location. A satisfactory pulse sequence would be

G_Z / \ rf (x,01) (x,02) 01 and 02 refer to two different frequencies. A delay may need to be inserted between the two sine pulses. To increase the frequency bandwidth of the excitation, it is possible to use more than one excitation pulse on either side of the slice of interest. These pulses are preferably set with off-set frequency so that their excitation band abuts but does not overlap that of the neighbouring pulse. Thus if the slice of interest and the excitation pulse each have frequency widths of pHz and qHz respectively, then the off-set frequency of each pulse is given by ±(p₊nq)/2 where n = 1, 2, 3... and is limited by prevailing experimental constraints. It may be possible to use other pulse shapes which yield useful excitation profiles. In what follows, experimental tests have been performed using the first pulse generation method (Method I) proposed above. These tests are equally applicable to the second procedure proposed. The tests were performed upon a Bruker MSL-100 spectroscope system with a 2.35T 40cm bore magnet.

To test the response of a {ssc} pulse a 5.0cm cube of water spins was used as a test sample and the

1

H signals acquired. A gradient sufficient to place some spins outside the excitation profile of the { ssc} pulse was applied and the following pulse experiments were performed. In each case, the gradient is left on during data acquisition and the phase cycling is set to sample the residual z-magnetization remaining following the {ssc} pulses. {ssc,±x}-τ - (x), receiver, +, + (A)

{ssc, i}- τ- {ssc, j}-τ- (x) receiver -. (B)

In pulse sequence (B), i and j refer to the appropriate combinations of {ssc } pulse phase required to eliminate spin-echos formed during data acquisition in the presence of a gradient. The experimental results for pulse trains (A) and (B) are shown in FIGS. 2(c) and 2(d), showing excellent agreement between theory and experiment, the bandwidth of the hole in the z-magneti- zation being measured as approximately 1.4kHz in this instance. This value can be varied as a function of the total pulse duration and by varying the constants α and β discussed previously. The {ssc! pulse time in these experiments was 2.048ms and 256 data points of the pulse shape were placed into the waveform memory of the computer. The {ssc} pulse shape is shown in FIG. 3.

For the first volume-selected experiment, three bottles (internal diameter 2.0cm external dia¬ meter 2.5cm) were arranged in the form of an equi- lateral triangle and placed into a 250ml beaker filled with water. The long axes (6cm) of these bottles were parallel with the y-axis. Two bottles stood on the z- axis and one on the x-axis of the magnet. This latter bottle was filled with benzene and the other two with ethanol and cyclohexane. FIG. 4(a) shows the normal

¹ H spectrum obtained with a hard «it pulse (pulse time 70μs using 5.0kW of pulse power). In FIGS. 4(b), 4(c)

1 and 4(d) are shown H volume selected spectra obtained using the pulse train ^Gsl * if \ rf {ssc} - ti - {ssc} {ss -τ,. - {ssc} -τ₂ -^- [13

Acquire (C) Here, i refers to a quadrature phase cycle which is also followed by the receiver. The small residual signals arise from our inability to apply the {ssc} pulse more than twice. It is, of course, feasible to apply each {ssc } pulse more than twice to better define the slice of interest and to reduce the contribution from residual signals arising from outside the slice. Careful atten- tion must be paid so that unwanted echo signals are not reformed by the refocussing effect of rf pulses. This is preferably done by the use of appropriate delays between each pulse.

In the second pulse method proposed, two sine pulses applied at two different frequencies, could be applied to perform volume-selection in the following manner. A wait time has been inserted and the frequen¬ cies of the two sine pulses are set at ±Δo . A possible pulse sequence would be c_{z /} \ rf ι(+^)- _Tl- |(- ω₂)-τ₂- |( +Δi )-**-τr \( ) , Acq.

The bandwidth of the | sine pulse is set at 2Δnx yielding a hole in the z-magnetization of width 2(_/|ω₁ -_/|ω) , clearly_< ω₁><d_ω.

The use of the sin-sine pulse for efficient slice selection may also be demonstrated for use in magnetic resonance imaging. In particular, it can be used in the field of fast imaging. The strong signal strength associated with slices obtained using the method of the present invention means that complete image information may be obtained following a single slice selection procedure. Because at the end of slice selection the magnetization from the slice of interest is "stored" along the static magnetic field axis, a small rotation angle pulse may be used to tip a small amount of magnetization into the transverse plane for each projection. If all projections for a particular image are acquired in a time short compared with the spin-lattice relaxation times of the sample, the image may be constructed from a single slice selection process.

Other known methods which use the magnetic field axis for "storage" of magnetization generally involve the use of the so-called stimulated echo. However, this approach is inherently less sensitive because the transverse coherence is allowed to dephase before it is pulsed back to the field axis and so the signal strength is reduced by a factor of 1/2. Further- more, the use of the transverse plane results in signal attenuation as well as T₂ constrast being present in images obtained this way. The use of a sin-sine pulse eliminates signal loss from either of these sources. One possible pulse and gradient sequence incorporating a sin-sine pulse for slice selection in an imaging experi¬ ment is described in FIG. 5 while FIG. 6 shows an image of the phantom described herein obtained using this sequence. In this instance, the angle was kept constant during the experiment. However, it is preferable to vary this angle so that the transverse component of the coherence acquired for each projection is constant throughout the experiment. This may be done by varying θ so that _^■<_ ^θn ^{= tan " θ}n-l and where the rotation angle for the final projection is 90°.

Experiments were performed at 100.16MHz on a Bruker MSL-100 spectrometer equipped with a 2.35T 40cm bore horizontal magnet system. Modification was made to the hardware to ensure that the waveform memory, rf modulators, and amplifier system faithfully reproduced the desired pulse shape. A digital resolution of 256 data points defining the shape of the {ssc} pulse appears to be adequate for the first excitation method proposed.

The embodiments described are by way of illus¬ trative examples only and various changes and modifica¬ tions may be made thereto without departing from the scope of the present invention defined in the appended claims.

Claims

1. A method for performing volume-selected NMR spectroscopy including the steps of: eliminating all unwanted magnetization outside the volume of interest; and then acquiring the signal from the volume of interest by applying a single read pulse.

2. A method according to Claim 1 wherein: the unwanted magnetization outside of the volume of interest is eliminated by using shaped rf pulses to excite all spins outside the volume of interest; and dephasing the coherence of these spins by application of a field gradient.

3. A method according to Claim 2 wherein: the field gradient is the slice selection gradient.

4. A method according to any one of Claims 1 to 3 wherein: the signal is acquired by observing the spins in the volume of interest with a single read pulse.

5. A method according to any one of Claims 1 to 4 wherein: the magnetization of the volume of interest is in the transverse plane for as little time as possible so ^'that it is not influenced by short T₂ value.

6. A method according to any one of Claims 1 to 5 wherein: the spins outside the volume of interest have no phase coherence in the transverse plane and so yield no net signal during data acquisition.

7. A method according to any one of Claims 1 to 6 wherein: the selective rf pulse or pulses do not excite the body within the slice of interest but do excite for a wide band outside the slice.

8. A method according to Claim 7 wherein: the rf pulse or pulses have a hole of controllable width in their excitation profile.