GB2206970A - Improvements in or relating to nmr spectroscopy and nmr imaging - Google Patents

Description

220C970 IMPROVEMENTS IN OR RELATING TO NMR SPECTROSCOPY AND NMR IMAGING

The present invention relates to NMR spectroscopy and NMR imaging and more particularly to the selection of a small volume of material from within a much larger sample.

In order to select a small volume of material from a much larger sample, a succes-sful spatial localization technique must be capable of eliminating the unwanted signal from a large volume of the surrounding material.

Selection of a cubic volume of tis.sue within a -specimen which is ten times. larger along' all three axes, necessitates. that the s-ignal contribution from materia 1 outside the cube can be reduced by at least a factor of 10000. This s.uppres.sion ratio would only ensure that the measured signal from within the cube is. ten times. larger than the error s - ignal. In practice, a suppress-ion ratio of at least 20000 is.desirable to reduce the error signal below the nois.e level usually found in in-vivo phosphorous spectroscopy.

Known mecbanisms.for signal suppres-s.ion vary between techniques. Several methods. however rely upon signal cancellation following the application of a series- of experiments-. The ISIS R.J. Ordidge, A. Connely and J.A.B. Lohman; J. Magn. Reson. 66, 283 (1986) and VSE W.P. Aue, S. Mueller, T.A. Cros-s and J. Seelig; J. Magn. Reson. 56, 350 (1984) techniques. both rely on this principle, and require accurate cancellation of large s.ignal amplitudes. in order to detect relatively small s.ignals. originating from the volume of interes.t. The cancellation is. prone to instabilities. in the main magnetic field, the magnetic field gradients., the NMR spectrometer system and motion of the sample.

D.D. Doddrell, G.G. Galloway, W.M. Brooks, J.M. Bulsing, J.C. Field, M.G. Irving and H. Baddeley; Magn.

Reson. Med. 3, 970 (1986) have des-cribed a radio frequency (RF) pre-pulse which destroys_longitudinal spin magnetization in the unwanted regions of the sample by application of a selective Tr/2 spin nutation. The transverse magnetization is- then allowed to decay prior to the application of a suitable spatial localization technique. The method relies- on the application of a high power selective RF puls-e us-ing a homogeneous transmitter coil. This ensures that most of the unwanted spin magnetization undergoes-an accurately determined7r/2 nutation. In practice, the combination of high RF power, and a homogeneous_ RF coil which completely surrounds. the sample, is_ only found in small and medium bore in-vivo spectrometer sys.tems. For whole-body in-vivo spect-ros.copy studies, the RF power requirement would be considerable and may well exceed the current safety guidelines. for such equipment.

A similar technique called LOCUS spectros-copy has. been proposed by A. Haase; Magn. Reson. Med. 3, 963 (1986). In this- method the unwanted magnetization iseffectively removed by a sequence of s.elective IT/2 pulses. The LOCUS technique requires-a highly homogeneous_ RF coil, and a s-econd experiment is - needed to remove the effect of spin lattice relaxation.

It is an object of the present invention to provide a method of eliminating unwanted s.ignals to assist in enhancing wanted signals_.

According to the present invention there is_provided a method of eliminating signals_ from unwanted frequency bands- in NMR systems by using a shaped RF puls-e comprising at least two regions of random frequency components- between which at least one region of zero components is situated.

It is a further object of the present invention to provide a method for selection of a small W' volume of material from within a much larger sample which is- highly efficient in the use of RF power and which provides.good selectivity.

According to the present invention there is. also provided a method of more accurate definition of a small volume of material in NMR systems ' within a larger sample by removal of unwanted signals through the application of a shaped RF pulse in conjunction with a linear magnetic field gradient which effectively randomises. the net longitudinal spin magnetization in all volumes- outs.ide the selected volume.

The selective pre-puls-e can be followed by a suitable spatial localization technique such as ' ISIS, R.J. Ordidge, A. Connely and J.A.B. Lohman; J. Magn. Reson. 66, 283 (1986) or VSE W.P. Aue, S. Mueller, T.A. Cross and J. Seelig; J. Magn. Reson. 56, 350 (1984).

Emoodiments of the present invention will now be described, by way of example with reference to the accompanying drawings in which:- Figure 1A shows- the distribution of components. in the real frequency domain which form the irradiation spectrum for a selective noise pulse. Twenty points. in the centre of the 512 point spectrum are set to zero in both real and imaginary domains., Figure 1B shows the selective RF pulse shape obtained by Fourier trans. formation of the data in Figure 1A. A similar function is obtained in the imaginary time domain; Figure 1C shows. a computer s-imulation of the effect of the selective pulse shown in Figure 1B applied in combination with a magnetic field gradient upon a uniform spin distribution. The vertical axis represents the residual longitudinal magnetization following the pulse and is plotted as a function of pos.ition in the linear magnetic field gradient. The RF amplitude was chosen to give a minimum net longitudinal magnetization outside the unperturbed slice;

Figure 2 shows a plot showing the percentage of net residual longitudinal magnetization outside the unperturbed slice, plotted as. a function of the normalized RF amplitude of a quadrature selective pulse with two independent random frequency components_. An RF amplitude of onecorresponds to a non-s-elective Tr/2 RF pulse of equal duration; Figure 3 shows_ a plot showing the percentage of net res-idual longitudinal magnetization outside the unperturbed slice, plotted as - a function of th e normalized RF amplitude of a quadrature selective pulse with frequency components.defined in polar co-ordinates..

The- frequency components were specified by a constant amplitude and a random phase angle which could vary between + TF. An RF amplitude of one corresponds to a non-selective Ir/2 RF pulse of equal duration.

Figure 4 shows sequence timing diagrams illustrating the principles_ of the present invention for spatial localization and elimination in three dimensions (X, Y and Z); and Figure 5 shows the principles_ of the present invention adopted for an elimination technique to remove dominant peaks_ for more accurate assessment of minor peaks.

The selective RF pre-pulse is- constructed from noise. A random noise generation routine can be found for most computers_, which usually creates- a sequence of random numbers with an equal probability ofoccurrence over a pre-determined range. For the case cons-idered in this discuss.ion, the noise sequence cons.isted of 512 complex points_, with both the real and imaginary components randomly chosen to have a distribution centred around zero. This. sequence was- used to represent an initial selective irradiation spectrum. If the NMR spins_ in a central slice of the object are to be left unperturbed, the corresponding frequency components in Q 1 the irradiation spectrum must be removed. This- is achieved by setting a consecutive number of points to zero in the centre of the irradiation spectrum. A nonselected s.lice of corresponding frequency width is. thus left unperturbed when the pulse is.applied in conjunction with a field gradient. A suitable noise distribution is shown in Figure 1A, with the central 20 points. set to zero in both the real and imaginary frequency. domains -.

The selective RF puls - e shape is generated by Fourier transformation of the irradiation spectrum and is. shown in Figure!B. After application of this RF pulse shape in conjunction with a magnetic field gradient, the spin magnetization has undergone varying degrees of nutation as a function of pos.ition along the selection axis. For RF power levels greater than a minimum value, the longitudinal magnetization outs.ide the unperturbed slice is randomized in size and polarity, and for objects with reasonable spatial homegeneity, will average to zero. The transverse magnetization excited by this. pulse can then be allowed to decay before application of subsequent experiments. Since the NMR spins of interest remain unperturbed, the NMR signal measured by a suitable spatial localization technique will contain a full signal contribution from the selected spins, and a greatly reduced error signal originating from these outer regions.

Figure 1C shows a computer s-imulation of the effect of this selective noise pulse on the longitudinal magnetization immediately following the pulse period. A uniform spin distribution along the selection axis. has been assumed. Nutation of the NMR spins. during application of the pulse has - been calculated by application of successive matrix operators according to boundary conditions specified by the Bloch equations.tst-c, R.J. Ordidge, i i Pb.D. Thesis, Nottingham Univers-ity, England (1981). For non-uniform samples_, the residual net longitudinal magnetization should average to zero because of the high frequency components. in the nois.e excitation spectrum. Minor variations in the applied RF pulse power can be used to effectively ensure that this condition isreached. For a ten millisecond s- elective pulse, the frequency bandwidth of the unperturbed slice is approximately 2 kHz, compared to a total frequency bandwidth of 51.2 kHz, within which the remaining magnetization is randomized.

Three selective pre-pulses_ are required in order to achieve complete randomization of longitudinal magnetization along all three spatial axes-. These pulses. however may be applied consecutively as_ shown in Figure 4.

The effect of the selective nois.e pulse can be compared with a s. elective-rr/2 puls-e of the same frequency bandwidth. A selective -W12 pulse nutates- all magnetization into the transverse plane, and the NMR signal is then allowed to dephas.e under the influence of a field gradient, as described by D.D. Doddrell, G.G. Galloway, W.M. Brooks, J.M. Bulsing, J.C. Field, M.G. Irving and H. Baddeley; Magn. Reson. Med. 3, 970 (1986). This randomization process is- different from the action of a selective noise pulse, which randomizes.the res.idual longitudinal magnetization. However, in both cases the net effect is. to remove signal contributions from unwanted regions of the sample during the subsequent acquisition period. An advantage of the noise pulse compared with the 17/2 selective pulse of equal frequency bandwidth is. the considerable saving in RF power, which amounts to at least a factor of 50 in the present case. The noise pulse is- more efficient because of the randomness of the irradiation spectrum. This_ generates a

1 selective pulse with very little coherence between frequency components.. Therefore, the resultant RF modulation function does-not contain any intervals during which a large RF amplitude is. required s ince the individual frequency components_do not become coherent in phase.

A rectangular non-selective RF pulse is. the most efficient means of nutating NMR magnetization, and can be used as the standard by which all other puls ' es ' are judged. In general we can cons.ider a selective noise pulse as a random sequence of RF am litude variations P applied as a function of time. Application of this waveform causes NMR spins to experience the net effect of N individual RF pulses of random sign and size, where N is r-he number of time intervals. used to repres.ent the selective puls.e waveform. Since any average random noise level increases. by the s.quare root of the number in the average, then the average effect of the puls ' e will be to cause a nutation which is, J, times. larger than that occurring during 'any s.ingle time interval in the pulse waveform. However, compared with a rectangular RF pulse of equal duration, the nois.e pulse is IN- times less efficient. This is-compensated by the fact that the noise pulse has an excitation frequency bandwidth which is N times.larger than the rectangular pulse. A rectangular IT/2 puls.e with the same excitation bandwidth would require a much larger RF power than the noise. pulse. The noise pulse however, mus.t produce spin nutations.of up to + Tr, and therefore is only N-/2 times. more ef f icient in RF amplitude than the equivalent frequency selective IT/2 pulse, or N/4 times more efficient in RF power.

The most convenient way of expres.s_ing the efficiency of the noise pulse is to compare its performance with a rectangular non-selective Tr/2 pulse. If unity representsthe amplitude of a rectangular non-selectiveIT/2 pulse of fixed duration, then in the cas.e considered here, an RF amplitude of 492 would be required to achieve the desired effect using a s.ingle sPlective puls-e as. des.cribed by D.D. Doddrell, G.G. Galloway, W.M. Brooks., J.M. Buls.ing, J.C. Field, M.G._ Irving and H. Baddeley; Magn. Reson.

Med. 3, 970 (1986). This assumes. that both puls.es, are of equal duration.

-The graph of Figure 2 shows the variation of residual longitudinal magnetization outside the unperturbed s-lice as. a function of the applied RF amplitude of the noise puls-e. The firs-t zero-crossing occurs. at a normalized amplitude of 69. The afore mentioned theory predicts - that the nois-e puls-e should be IN-12 times_more efficient than a s-elective pulse of equal duration and excitation bandwidth. For a pulse with 512 time intervals_, this predicts a zero net longitudinal magnetization at a relative RF amplitude of 45.

Therefore, with this particular method of noise generation a selective puls-e is. produced with a relatively poor efficiency in randomizing the spin system. At higher RF amplitudes-the residual longitudinal magnetization oscillates_ in polarity with a decaying amplitude, however the profile of unperturbed longitudinal magnetization begins to deteriorate at the edges of the non-selected slice. These os.cillations.could prove useful in the optimization of RF amplitude for maximum cancellation of unwanted signal with non-uniform samples. The dependence of net res.idual magnetization upon RF amplitude is greatly reduced at higher levels. of RF amplitude.

Unfortunately, the nois.e puls.e does- not caus.e an even range of spin nutation angles across the complete excitation spectrum of the pulse. This feature, which is common to all selective pulses, is caused by the application of a discrete RF pulse shape rather than a k 1 continuous.function. The effect of dis-creteness-in the RF waveform may be calculated by cons.idering the excitation spectrum of a -IT/2 RF pulse with a time duration which is equal to the individual intervals. in the waveform. The nutation angle for off-res.onance spins. may therefore be calculated, and the corresponding frequency components in the selective puls-e adjusted to compensate this. effect. The required correction neces. s.itates multiplying the relevant frequency components by a factor which is. equal to I-t/2 divided by the calcuated nutation angle in radians. The correction function C( w) is- des-cribed by the equation:

C( W) = TT/ 2'8 where e - 0-rr- 'Los 111 2 Wis the angular frequency offs ' et for the initial irradiation spectrum, and Iris. the time duration of the discrete intervals in the s-elective puls.e shape.

The nois.e spectrum must therefore be substantially increased in amplitude for large frequency offsets. in order to maintain an even - frequency response over the excited bandwidth of the puls-e. This. modification has.. been included in the calculation of previous results.

Further refinement of the sPlective puls.e can be achieved by modification of the nois-e characteristics- in the irradiation spectrum. Instead of using two independent noise amplitudes to represent the quadrature components of the irradiation spectrum, a random noise can be generated in the polar coordinate system using a constant amplitude with a random phase angle. The resultant excitation pulse requires. less RF power to achieve a zero in net residual magnetization, however the oscillations. in residual mangetization at high RF amplitudes are much larger. Figure 3 shows- the net residual longitudinal magnetization plotted against RF amplitude for a uniform sample. The first zero-crossing occurs. at an RF amplitude of approximately 50, which is_ much closer to the theoretical value of 45. The saving in RF power compared to a 7/2 s-elective puls.e of equal duration and similar effect, is- now approximately a factor of 95.

other forms. of noise modulation can be us-ed. The dependence of net longitudinal magnetization upon RF amplitude can be varied by changing the noise characteris-tics.

A selective pre-puls-e has. been described which caus.es- des-truction and cancellation of the unwanted spin magnetization from regions- outs- ide the sPlected slice. Three such puls-es may be applied consecutively in order to leave an unperturbed volume of material that can subsequently be inves.tigated us-ing a spatial localization technique such as- ISIS, as, shown in Figure 4. With reference now to Figure 4 the pulse sequence is_sbown for gradients Gx, Gy and Gz with the shaped RF pulse shown for each gradient pulse. The gradient and RF pulses are followed in known manner by a spatial localization experiment or alternatively an NMR imaging -experiment. Application of these pre-pulses-should greatly reduce the error signals from unwanted volumes- of tissues, and will thus- extend the us-efulnes.s-of the ISIS technique.

A similar pulse might be us-ed to remove broad resonance lines. in standard NMR spectros.copy, and may also be useful in the production of "zoomed" images_ in NMR imaging. This. technique also makes- the spatial 1 localization procedure much les.s. sensitive to subject motion during the NMR s.tudy.

Erroneous. s.ignal can be caused by non-cancellation of s.ignal responses. from tis-sue which moves. during the experimental sequence. The noise pulse tends.to randomize longitudinal magnetization within these moving organs_ which provides-more effective s.ignal cancellation.

The technique can be added to any NMR s,equence in order to minimize the s-ignal response from volumes of tis's.ue outside the region of interest. Alternatively s.ignal frequency components. (e.g. lines. in the chemical shift spectrum) may be removed by pre-excitation with the selective noise pulse as_ shown in Figure 5. With reference to Figure 5 by application of one shaped pulse with spectral noise bands. identified by SP1 and SP2, major peaks P1 and P2 in a spectrum can be eliminated using the techniques. of the present invention thereby enabling the receiver gain to be increased and thereby enhancing the minor peak P3 by better use of the k to D converter (see Figure 5c).

The shape of the noise pulse is_ generated from an initial irradiation spectrum cons.isting of an array of 2 N real points.and 2 N imaging points- where N is- an integer. Any number or dis.tribution of these points may be set to zero, and the remainder is. filled with a random noise pattern. A distribution of gaps_ in the noise spectrum is thus generated and the shape of the selective RF pulse isproduced by Fourier transformation of the irradiation spectrum. Signal within these frequency gaps. therefore remains- unperturbed after application of this. pulse, whereas parts. of the sample which have experienced the effect of the noise pulse produce a markedly reduced signal through randomization of the longitudinal magnetization.

The technique can be combined with NMR imaging to produce images of zoomed regions- of the sample. The technique can also be combined with chemical shift imaging to produce chemical shift images.of inner volumes within the sample. Finally, the method may be applied with any of the existing spatial localization procedures,.

r 7

Claims (11)

1. A method of eliminating signals_ from unwanted frequency bands,in NMR systemsby us-ing a shaped RF puls.e compris.ing at least two regions_ of random frequency components. between which at least one region of zero components_is,s.ituated.

2. A method of more accurate definition of a small volume of material in NMR systems. within a larger sample by removal of unwanted s.ignal through the application of a shaped RF puls-e applied in conjunction with a linear magnetic field gradient which effectively randomizes. the net longitudinal spin magnetization in all volumes. outside the selected volume.

3. A method as-claimed in Claim 2 wherein one shaped RF pulse and gradient is,used to leave unperturbed a s.lice.

4. A method as-claimed in Claim 2 wherein two shaped RF puls-es_ and gradients- are used to leave unperturbed a column.

5. A method as. claimed in Claim 2 wherein three shaped RF puls-es. and gradients- are used to leave unperturbed a rectanguloid volume.

6. A method as. claimed in Claim 2, 3, 4 or 5 in which each shaped RF pulse has_ spectral components- which are random over a defined period and has. no spectral components in a specified frequency band.

7. A method as- claimed in Claim 6 in which the RF pulses. and gradients provide NMR or chemical shift imaging.

8. A method as. claimed in any one of Claims- 2 to 7 in which each shaped RF pulse has- spectral components which are random in phas.e only.

9. A method as- claimed in any one of Claims- 2 to 7 in which noise s. ignals of any characteristic are used to generate the random spectral components. of the shaped RF pulse.

10. A method as claimed in any one of Claims 2 to 9 in which the application of the shaped RF pulse is_ used in conjunction with a form of localized spatial techniques_ to provide an accurately defined localized volume.

11. A method of more accurate definition of a small volume subs.tantially as. described with reference to the accompanying drawings, t i Published 1988 at The Patent Office. State House. 66 71 High Holborn. London WC1R 4TP. Further copies may be obtained from The Patent Office. Sales Branch. St. Mary Cray, Orpington, Kent BR5 3RD. Printed by Multiplex techniques ltd, St Mary Cray, Kent. Con. 1187.