surface coil techniques for in vivo nmr - weizmann.ac.il · surface coil techniques for in vivo nmr...

Surface Coil Techniques for in Vivo NMR

M. Robin Bendall*

School of ScienceGriffith University

Nathan, QueenslandAustralia 4111

(* On leave in the Department of Applied Sciences in Medicine,University of Alberta, Edmonton, Canada.)

PAGEI. INTRODUCTION 18

II. INVERSION-RECOVERY AND SPIN-ECHO SEQUENCES ININHOMOGENEOUS R.F. FIELDS 19

III. THE R.F. FIELD OF THE SURFACE COIL 21

IV. MAXIMIZATION OF SIGNAL-TO-NOISE 23

V. LOCALIZATION USING PULSED FIELD GRADIENTS 24A. Selective Pulses in Pulsed Field Gradients 24B. Two-Dimensional Fourier Transform Localization 26C. Sensitive Point Steady-State Free Precession 27

VI. LOCALIZATION USING RADIOFREQUENCY INHOMOGENEITY 27A. Rotating Frame Zeugmatography 27B. Simple Depth Pulse Sequences 27C. Composite Depth Pulses 30D. Fourier Series Windows 30

VII. LOCALIZATION BY COMBINING THE METHODS OF PULSED FIELDGRADIENTS AND RADIOFREQUENCY INHOMOGENEITY 31

VIII. LOCALIZATION USING MULTIPLE R.F. COILS 32A. Elimination of Multiple Coil Coupling 32B. Separate Transmit/Receive Coils 33C. Separate Transmit Coils 34D. Other Coils 35

IX. HETERONUCLEAR METHODS WITH SURFACE COILS.. 36

Vol. 8, No. 1/2 17

A. Polarization Transfer 36B. Heteronuclear Spin-Echo Techniques 37

X. HOMONUCLEAR 'H SPECTRAL SIMPLIFICATIONSWITH SURFACE COILS , 38A. Water Suppression 38B. 'H Spectral Editing 38

XI. CONCLUSION 39

REFERENCES 39

/. INTRODUCTION

The application of nuclear magnetic resonanceto the study of live animal and human subjects is arapidly growing field of interest of establishedimportance for biochemical and clinical studies.NMR imaging is now a routine modality in manyciinical settings and in vivo NMR spectroscopy isa widely-accepted research method. Imaging isnormally restricted to the detection of theabundant and NMR-sensitive *H nuclei of water andfat, and commonly, circumscribing r.f. coils whichsurround the head or body are used for r.f.irradiation and detection of the NMR signals.Generally, circumscribing coils do not havesufficient sensitivity to detect other 'Hcontaining metabolites, or other magneticallyactive nuclei, in localized regions of a human oranimal, and almost all in vivo spectrocopicstudies have utilized surface coils placed close tothe region of interest. Surface coils of varioustypes will probably remain the major tool of the invivo spectroscopist. In the short term it isreasonable to invest considerable research effortin their development which is the most promisingmeans of expanding routine clinical applications.

This review is addressed entirely to surfacecoil techniques, especially pulse sequences usefulfor optimization of signal-to-noise, localizationand spectral simplification. The already myriad ofapplication papers, especially utilizing 31P NMR,are not discussed.

Maximum sensitivity is of paramountimportance in in vivo spectroscopy. A small flatsurface coil, placed on the surface of an animal orhuman subject, is the most sensitive way ofobtaining spectra from tissues close to thesurface and, in consequence, most physiologicaldata from intact animals or humans have beenobtained in this way. Optimization of signal in thepresence of the spatially varying r.f. field of thesurface coil is discussed briefly in section IV, andconsidered further in relation to methodsdiscussed in later sections.

The variation of sensitivity throughout thesample is directly related to the spatial variationof the r.f. field produced by the surface coil. Thisproperty of r.f. inhomogeneity displayed by allsurface-type coils must be taken into account forany sequence of r.f. pulses applied with the coil.The problem can be simplified, since most usefulpulse sequences can be broken down into the well-known inversion-recovery or spin-echo sequences,and such an analysis is used below to describevarious localization methods, as well as methodsfor simplifying or editing in vivo NMR spectra.The simple phase cycling schemes required forinversion-recovery or spin-echo sequences whenusing surface coils are described in section II.

The rapid decrease of the sensitivity of thesurface coil to regions at increasing distance fromthe coil provides a crude form of samplelocalization. However, the sensitive volume doesnot have hard boundaries and its detailed shape

18 Bulletin of Magnetic Resonance

(discussed in section III) is quite complex anddependent on the experimental variables selected.It is always dangerous to assume that this crudelocalization is sufficient, though it may be insome cases, and many published studies whichhave ignored this problem are questionable. Acommon problem is the intense signals from theintervening tissue between the surface coil andthe tissue of interest. For example, surface coil3 'P spectra of the brain will include extracranialmuscle and scalp tissue. Conversely 'HNMRshowing lactate in abnormal brain tissue wouldprobably not be compromised by 'H signals fromthe scalp. The reverse situation applies to thehuman limbs, where 'H lactate signals from withinischaemic muscle would be swamped by lipidsignals from the intervening fat layer, but the fatlayer produces no significant3 'P signals toinvalidate the 3*P spectrum. For the liver, theintervening muscle wall and fat layer wouldcompromise both the 31P and the 'H spectrumrespectively. Another common problem is thediffuse boundary of the sensitive volume whichmay for example extend into non-exercised musclein the case of human limb studies, or into normalbrain tissue in the case of surface brain tumor orstroke investigations. Clearly, in general,additional means of sensitive-volume localizationare required so that it can be stated withcertainty that say 80 to 90% of the signal for aparticular metabolite originates from the regionof interest. Such a criterion could be relaxed ifchanges in the level of a particular metabolitewere measured in a set of routine experiments andif it were known that such changes were onlyoccurring in the tissue of interest, but this wouldhave to be proven by complete localization prior tothe set of routine experiments.

The major part of this review is devoted toadditional techniques for localizing the sensitivevolume. These fall into two main classes:methods which require magnetic field gradients(section V) and methods which utilize r.f.inhomogeneity (section VI). A third class combinesthese two methods (section VII) and a fourth classuses multiple r.f. coils, one of which will be asurface coil (section VIII). Throughout, mostemphasis will be placed on surface-type coils, butthe generalization of some methods tothree-dimensional localization as might also be

appropriate for a circumscribing coil will also bediscussed.

In traditional NMR spectroscopy, thesimplification of spectra by relaying informationbetween two J-coupled nuclei has been a veryimportant area of research for the last decade.There are also important applications in in vivospectroscopy and this is explored in sections IXand X.

//. INVERSION-RECOVERY ANDSPIN-ECHO SEQUENCES IN

INHOMOGENEOUS R.F. FIELDS

For a homogeneous r.f. coil theinversion-recovery and spin-echo pulse sequencesmay be written respectively as

180° - x - 90° ; acquire signal,

and 90° - x -180° - x - acquire,

(1)

(2)

where r is a delay period between the pulses. Forinhomogeneous r.f. coils such as surface coils, the180° and 90° pulses have little meaning becausepulse angles vary continuously and rapidlythrough-out sample space, and variable 29 and 9pulses are appropriate. When 9 differs markedlyfrom 90°, intense anomolous signals arise, and ithas been widely believed that these sequencescannot be used with a surface coil (1-3). However,it has been shown using simple vector diagrams(4,5) or rotation matrices (6), that these artifactscan be entirely eliminated for any value of 9 bysuitable phase cycling of the pulses during aseries of transients, which for sequences (1) and(2) respectively is

29[±x] - x - 9; acquire, (3)

and

9 - x - 29[+x,±y] - x - acquire, receiver[+,-] (4)

The phase-cycled inversion pulse, 29[+x], signifiesthat the pulse phase is alternated between +x and-x for alternate transients, and this is a moregeneral variation (5-8) of the method introducedby Demco et al (9) and initially used with surface

Vol. 8, No. 1/2 19

coils (4,10,11). The phase cycling of the 29[±x,±y]refocusing pulse is as introduced by Bodenhausenet al (12) for two-dimensional high-resolutionNMR, and receiver [+,-] means that the receiverphase is inverted (i.e. the signal is subtracted)when the 29 pulse is cycled through the ±y phases.

We have shown (5,6) that the three types ofpulses in sequences (3) and (4) are basic buildingblocks of surface-coil sequences. Each sequencemust have one excitation pulse 9, but depending onthe purpose of the sequence, any number of 29[+x]or 29[±x,±y] type pulses may be used (provided thephase cycling for each pulse is independent of allothers), with or without x delays.

To quantify the effect of these variable r.f.pulses, abstract mathematics can be used (6) or abetter physical feel for the various pulsesequences may be obtained using vector diagrams(4,5). For example, Figure 1a shows the effect of a

IgCOSa

Figurei. (a) An r.f. pulse of angle (or length) 9 andof x phase rotates an initial z magnetization Mo

through an angle 9 about the x axis of the rotatingreference frame and generates a detectabletransverse component of y phase(absorption-mode) proportional to sin9. (b) Thesame pulse applied off resonance rotates M0aboutthe tilted x' axis. This rotation is equivalent tothe resultant of a stationary component Mosincxand rotation of M0cos<x in the tilted yz plane. Thefinal y component is M0coscx sin9" and an additionalcomponent of x phase (dispersion mode) isgenerated equal to M0sin2cx sin2(e72).

variable 9 pulse on initial z magnetization. Thusit can be readily shown that a 29[±x]pulse before the 9 excitation pulse reduces initialz magnetization to an amount given by cos29. The9 pulse converts unit z magnetization todetectable transverse (xy) magnetization equal tosin9 (as shown in Figure 1a), and a 29[±x,±y] pulsereduces the latter to an amount given by sin29. Formultiple pulse sequences, these varioustrigonometric factors are multiplied together: eg.if two 29[+x] pulses were used prior to 9 and one29(±x,+y] pulse after 9, signal intensity would beproportional to cos229 sin39.

These trigonometric factors are exact onresonance, but for in vivo spectroscopy r.f. pulseswill often be so weak that the factors no longerhold true off resonance across a normal spectralwidth. Weak r.f. pulses act along axes which aretilted by an angle ex towards the main field (z)axis where <x is related to the resonance offset

AH (Hz) and 90° pulse time on-resonance t90(s) by

tancx = 4 AH t90. (5)

For any on-resonance pulse 9, the rotation angleoff resonance is increased to 9' given by

91 = 9 seccx. (6)

Thus the vector diagram of Figure 1 a translates tothat of Figure 1b off resonance. The cos29 factorfor a 29 [±x] pulse converts to (1-2cos2cx sin29')off resonance. The sin9 factor for the 9 excitationpulse becomes cos<x sin9', but in addition to thisnormal absorption-mode signal, an unwanteddispersion-mode signal proportional to sin29sin2(972) arises (see Figure 1 b). Lastly, theoff-resonance factor for a 29[+x,±y] pulse is+cos2cx sin29', where the minus sign signifies thatdispersion signal is inverted by the pulse. Again,these various factors must be multiplied togetherfor each pulse in a sequence, and although theresult may be complicated, calculations as afunction of 9 and AH need to be carried out foreach prospective method to check its extent ofvalidity off resonance. For example, it is foundthat apart from eventual loss of signal intensityoff resonance, accurate Tn values can be measured


using sequence (3), and any spin-echo method canbe employed with sequence (4).

A fourth and final building block for our pulsesequences, 29[±x,0], is derived by subtracting theresult of two transients, obtained using just9; acquire, from sequence (3) giving

29[±x,0] - r - 9; acquire (7)

where 0 signifies that the pulse is not applied andthe receiver phasejs inverted. The trigonometricfactor for a 29[±x,0] pulse applied before the 9excitation pulse is sin29 on resonance andcos2o< sin29" off resonance, and so it has the samequantitative effect as a 29[±x,±y] pulse used after9 except that dispersion signals are not inverted.This equivalence between 29[±x,0] and 29[+x,±y]holds true no matter whether 29 is a hard, shapedor composite pulse (13).

///. THE R.F. FIELD OF THE SURFACECOIL

The basic surface coil is one or more circularturns of wire connected to a tune and matchnetwork. During an r.f. pulse a large oscillatingcurrent flows in the wire and this induces anoscillating magnetic field (or r.f. field) along thesame field lines as the magnetic field produced bya direct current in the coil. Clearly the r.f. field isstrongest closest to the coil wire as depicted inFigure 2.

As described in the early report by Ackerman etal (14), the r.f. field produced by any coilconsisting of continuous wire may be calculatedby using the Biot Savart Law to determine themagnetic field produced by direct current in thecoil. Detailed three-dimensional calculations havebeen published (4,7,15,16) and these areillustrated in Figure 3a and d. The calculated fieldin Figure 3a is given in terms of pulse anglecontours assuming 9 = 90° at 1.0 radius depth. Onehalf of the xy plane through the center of the coilis shown with the vertical axis being the coil axisor x axis. After a single 9 pulse the signalresponse from each volume element is given by9sin9, where the 9 term accounts for thedecreasing sensitivity of the coil to volumeelements further away from it. Thus signal will

nodeline

»ory

Figure 2. The r.f. field lines produced by a surfacecoil. If the plane displayed is the xz plane, therewill be node lines where the r.f. field lines areparallel to z, the external magnetic field direction.

be detected from all points in sample space exceptwhere 9 = 180, 360° . . . , and the signal will bepositive for 9 ~ 90°, negative for 9 ~ 270° and soon as indicated.

The shape of the r.f. field may also be revealedby experiment. The images in Figure 3b and c wereobtained by adding pulsed field gradients tosequence (4) and imaging the signal response froma slice phantom of H2O [eg. see Ref. (17)]. Thesignal response is consequently given by 9sin39,yielding similar banded results to that shown inFigure 3a but with more diffuse boundaries. ForFigure 3b, 9 was set at 90° at about 1.0 radiusdepth to mimic Figure 3a, but in Figure 3c thepulse lengths have been more than halved toillustrate that these banded areas can be pushed tovariable depth into the sample. Note the lowersensitivity to the 90° signal region in Figure 3bcompared to the 270° region ( ~ 3 times asintense) and the 450° region ( ~ 5 times asintense) as determined by the 9 sensitivity term.However, because the volumes of these "high flux"regions become smaller with increasing pulseangle, their total contribution is of a similarmagnitude to the 90° region, and for simplicity wegenerally ignore the 9 sensitivity term in

Vol. 8, No. 1/2 21

(a)9 0 TO MAGNETIC FIELD

405°

positive signals

node

90° region

positive signals

1 negative signals

Figure 3. (a) Plot of pulse angle contours for asurface coil for the xy plane (transverse to theexternal magnetic field). The shaded areascorrespond to regions from which signal would bedetected using an appropriate depth pulse sequence(e.g. signal given by cos229 sin3e for the depthpulse (20[±x])2;e; 29[±x,±y];acquire as described insection VI). A 90° pulse is assumed at 1.0 radiusdepth, but any other pulse angle can be accountedfor by scaling the values given. The pulse anglecontours also correspond to sensitivity contours,(b) Experimental image of the sensitive volume inthe xy plane of a surface coil obtained by Ordidge(18). (c) As for (b) with all pulse lengths dividedby 2.5. (d) As for (a) for the xz plane (parallel tothe external magnetic field).


theoretical expressions such as 9sin39.In an NMR experiment, the active component of

the r.f. field is the oscillating component at rightangles to the external field axis (z). For the xyplane displayed in Figs 3a to c the r.f. fieldis always transverse to the z axis. However, fromFigure 2 it can be seen that in the xz plane there isa node line where the r.f. field is parallel to z andno NMR signal can be excited or detected. Thecalculated result for the xz plane is shown inFigure 3d, which demonstrates that this node linefurther complicates the shape of the sensitivevolume.

The normal situation for a surface coil asillustrated in Figure 3 is clearly unsatisfactory.When trying to sample regions at a modest depthinto the sample, intense high flux signals aregenerated near the surface, and even if these canbe suppressed, the 90° region curves back into thesurface outside the circumference of the coil.Obviously, additional means of sensitive volumelocalization are required whenever it is necessaryto study a region beneath a surface layer.

IV. MAXIMIZATION OFSIGNAL-TO-NOISE

Haase et al (15) and Evelhoch et al (16) haveextended the theoretical calculation ofdistribution of surface coil signal intensity toinclude the relaxation time T r and the time foreach transient, TR. At high repetition rates, sayTR = 0.05^, and reduced pulse lengths, thedistribution of signal intensity is significantlyaltered with less high flux signals near thesurface and more signal intensity at one radiusdepth, ie. a more uniform distribution of signal,and total signal intensity is doubled for optimumpulse lengths at these high repetition rates.However, despite the improvements at higherrepetition rate, the signals from an interveningsurface layer would still swamp an inner organ bya factor of 3 to 9 times, ie.^dditional methods aredefinitely required in such cases.

In maximizing total signal-to-noise, theapproach of Evelhoch et al (16) is valid forexperiments where a thick surface layer isto be detected, or where an underlying layer isdetected but the intervening surface layer does not

contribute any interfering signals. Improvementscan also be obtained in signal-to-noise ratiowithout increasing the repetition rate of theexperiment (and disturbing the normal ratio ofspectral peak lines), by using a compositeexcitation pulse 9 which provides for uniformexcitation of transverse magnetization across arange of pulse angles around 9 = 90° . However,contrary to some published claims (3), thesimplest composite 90° pulse introduced byFreeman et al (19), 9 = 9[x];9[y], does not increasesignal-to-noise, since the additional signalcomponents which are generated are entirelydispersion-mode, have different phases either sideof 9 = 90°, and so mutually cancel. Nevertheless,Tycko et al (20) have described pulses whichprovide uniform excitation across a factor of threevariation in pulse angle and which have a muchreduced phase problem. These should prove useful.

Hetherington et al (21) have used the 9[y];9[x];9[-y];9[x] composite pulse of Levitt and Ernst (22)and have achieved a 20% increase insignal-to-noise in surface coil studies of the ratbrain. Although this composite 90° pulse doesproduce dispersive components, Hetherington et aleliminated these by using the phase-cycledequivalent 9[-y];9[x];9[y];9[x] for alternatetransients.

Hetherington et al have also extended the use ofcomposite pulses to spin-echo sequences andsurface coils, with both the 9 excitationpulse and the 29 refocusing pulse replaced bycomposite pulses, and achieved a 40% increase insignal-to-noise over simple pulses. This howeverrequires a more complex but nevertheless generalphase cycling scheme of the 29 refocusing pulse,than that shown in sequence (4), as determined byHetherington and Rothman (23).

Recently we have developed a complexexcitation pulse where amplitude and frequencyare continuously modulated during the pulse assine and cosine functions (eg. from sin0° tosin90°). This "sin/cos" pulse operates underconditions known as "adiabatic half passage" intraditional NMR and our results (24) show that itcan be used under practical conditions withsurface coils and that it will excite more than 90%of z magnetization to give absorption-mode signalacross a variation of at least a factor of 10 in the

Vol. 8, No. 1/2 23

strength of the r.f. field. Further preliminarycalculations (25) indicate that if the length of the"sin/cos" pulse is doubled (e.g. to extend from sin0° to sin 180°), a 180° inversion pulse of similarefficiency is produced. This sin/cos pulse appearsto be somewhat shorter but gives similar resultsto the "sech/tanh" pulse previously described bySilver et al(26). But both these adiabatic pulsesintroduce very bad phase problems if utilized inspin-echo sequences. However our most recentcalculations indicate that suitable 180°refocusing pulses can be formed by reversing theamplitude and frequency modulation to givecos/sin and tanh/sech pulses, with the formeragain having a length advantage over the latter.These pulses have not been exploited in in vivoapplications, but there appears to be goodprospects of being able to maximizesignal-to-noise even with inversion-recovery andspin-echo sequences.

V. LOCALIZATION USING PULSEDFIELD GRADIENTS

A. Selective Pulses in Pulsed Field Gradients

From Figure 3 it is clear that when using asurface coil to detect a region below the surfaceit is necessary to discriminate against the surfacelayer. This may be achieved by applying aselective pulse in a pulsed field gradient, wherethe gradient axis is the surface coil or x axis, sogenerating a selected slice which is parallel to theplane of the coil. The center of the slice shouldclosely coincide with 9 = 90° on the coil's axis, inwhich case the r.f. inhomogeneity of the coil willlimit the lateral dimensions of the selected sliceto about the diameter of the surface coil. Thesensitive volume would be disc-shaped, centeredon the coil's axis and flattened along the zdimension (because of the r.f. node line). Thisgeneral method has been proved using aphase-cycled pulse (a depth pulse (7,27,28)], butusually the shaped pulses commonly employed inNMR imaging will be more convenient. There arenumerous types of shaped pulses that may be used,and this is a rapidly developing area of research.The interested reader is referred to recent

articles (29-33) and numerous preliminary reportsin the abstracts of recent internationalconferences.

A major problem with pulsed field gradients isthat eddy currents are generated in thesurrounding metal of the probe and magnet whichdie away exponentially, thus slowing down thegradient switching. These eddy currents can becompensated electronically but a short delay ofabout 10 ms is normally required before signalacquisition, otherwise the resolution of the finalspectrum will be reduced. During this delay, themagnetization in the selected slice must bepreserved by some means, and this is discussedbelow in more detail. A second limitation ofselective pulse/field gradient methods is that thespatial frequency shifts are mixed with thenatural chemical shifts, thus generating a spreadof localized slices depending on the magnitude ofthe chemical shift. This problem is reduced byusing maximum field gradient strength, butthis increases eddy currents. The effect is mostsignificant for larger bore, higher field strengthmagnets and is a serious problem for 13C NMR,whilst being partially limiting for 31P NMR and tolesser extent for 'H NMR. Both these problemswill increase with the progressive move to higherfield whole-body magnets.

There are three possible ways of using selectivepulses:

1. Selective excitation pulses . Denoting aselective pulse in a field gradient along the x axisas Gsx, a straightforward experiment would appearto be

9sx[x];SL[y]; acquire, [8]

where the 9SX pulse has x phase. SL[y] is aspin-lock pulse of y phase, and is included to lockand so preserve the selected magnetization duringthe eddy current decay period. Although the use ofa spin-lock pulse in this way has been proved byexperiment (7,27,28), to retain its simplicity thismethod requires a shaped excitation pulse whichdoes not permit the nuclear spins to dephase in thetransverse plane during the pulse as a result oftheir natural chemical shift and the imposedspatial frequency shift. Unfortunately, suchshaped pulses which do not show the dephasing


problem have yet to be devised.Bottomley et at (34,35) have obtained good

results from the human brain by applying asine 90° pulse in a field gradient then acquiringthe NMR signal immediately after (or slightlybefore) the nuclear spins are rephased by a reversegradient. Thus no x delay period was used to alloweddy currents to decay, but the eddy currentproblem was avoided by minimizing the usualcoupling between the gradient coils and thesurrounding metal magnet by using 0.6 meterdiameter field gradient coils in their 1.5T/1 meterbore magnet. This is a proper solution but is notgenerally applicable to smaller bore magnets asused for animal systems, or higher field systems.Note that the effects of natural chemical shiftoccurring during the sine 90° pulse and thereversed gradient are not refocused and the finalspectrum will have a large phase roll which whencorrected can lead to considerable spectraldistortions. Bottomley et al used a cylindrical NMRtransmit coil and a surface receive coil, and thisaspect of their work is discussed further insection VIII, B, 2. However, as can be appreciatedfrom Figure 3, the extent of the slices parallel tothe surface coil depends only on the sensitivity ofthe surface coil as a receiver, and this sensitivitydecreases gradually in the y direction, giving poorlocalization in this direction.

2. Selective inversion pulses. Ordidge andco-workers (36) have proposed the use ofselective inversion pulses, 29S, in place of 29 insequence (7) to select a slice parallel to a surfacecoil and have obtained 31P brain spectra in thisway. The x delay of sequence (7) becomes the eddycurrent decay period, and this method has theadvantage that the selected magnetization is lostvia T, relaxation during x, the slowest possiblerelaxation mechanism. Additional 20S [±x,0) pulsescan be added with field gradients in the y and zdirection to select slices in each of the threespatial dimensions and thus a cuboid sensitivevolume:

The ±x phase alternations remove unwantedtransverse signals, although these are partiallyeliminated by dephasing in the pulsed fieldgradients. For ideal 180° and 90° pulses, assupplied by a homogeneous r.f. coil, thesetransverse signals will be small and so thedephasing is sufficient to allow the ±xalternations to be omitted. Thus the 64-transientcycle of sequence (9) becomes an 8-transientcycle:

180°sx[x,0];180oSY[x.°];

180osz[x,0J - x -90°; acquire. (10)

29sx[±x,0];2eSYt±x, 0];

29sz[±x,0] - x - 9; acquire. (9)

This is the method recently used by Ordidge et al(37) for localized in vivo 'H NMR using acyclindrical r.f. coil.

A disadvantage of these methods is that spinsoutside the selected volumes are excited by the 9or 90° pulse, and the resulting large signals areonly eliminated by the alternate addition andsubtraction of the detected signal. Instrumentalinstabilities and movement of the live samplemeans that such subtractions can never be perfectand random errors, or "subtraction noise" results.For the applications with a homogeneous r.f. coil,these large signals outside the selected region canbe removed by using 90°; 90°s[±x] or 90°s[±x]; 90°in place of each 180°s[x,0] in sequence (10) forexample. The hard 90° pulse would generatetransverse magnetization outside the selectedslice which would be substantially dephased andso eliminated by the pulsed field gradient appliedduring the pulse. However this combination ofhard and shaped pulses as proposed by Young (38)is presently prevented by the dephasing of spinsduring the shaped pulse (noted above for selectiveexcitation pulses).

Aue and coworkers (39,40) have developed apartial solution to this dephasing problem by usinga sandwich of soft and hard pulses, 45°s[x];90°[-x];45°s[x] (first introduced by Post et al (41)for imaging), which is equivalent to 90°s[x];90°[-x]. This sandwich corresponds to the zeropart of 180°s[x,0] and the 180° inversion part isobtained using the second sandwich, 45°s[-x];

Vol. 8, No. 1/2 25

90°[-x]; 45°s[-x]. However, although the dephasingof spins during the 45°s pulses is refocused andeliminated in the first sandwich, the problemremains for the second sandwich. Calculationsalso show that significant signal is generatedoutside the selected slice when the 90° pulse inthese sandwiches is not very hard (say >50psec)(42). Note also that Aue and coworkers (40)mistakenly used the sandwich, 45°s[x];90°[x];45°s[x], as the second variation and that thiscannot function to properly cancel signals outsidethe selected slice. Furthermore, they used only atwo transient cycle for the three dimensionallocalization experiment. This must be at least an8 transient cycle as in sequence (10), and shouldprobably be extended to a 64 transient cycle as insequence (9) when using surface coils. Finally,when using surface coils the sandwiches are of theform es/2;9;es/2 and the hard 9 pulse on averageconverts only half of the detectable zmagnetization outside the selected slice totransverse magnetization. Thus subtraction noiseis also only reduced by half compared to using acomplete 29S pulse.

3. Selective refocusing pulses. Having noted theequivalence between 29[±x,0] and 29(±x,±y], a sliceparallel to a surface coil can be generated using29SX in sequence (4), and the equivalent method tosequence (9) is

9-r-29sx[±x,±y];29SY[±x,iy];

29sz[±x,±y] - x - acquire (11)

Although the selected spins relax via the morerapid Hahn T2 time and the total x delay periodsare twice as long as the equivalent delay in theselective inversion methods, these spin-echomethods are applicable to small mobile 'Hmetabolites for example, especially as thespin-echo helps to eliminate large fat and watersignals having a shorter T2. A second disadvantagecompared to the inversion methods is that the 29S

pulses must be accurately timed to be at theeffective center of the pulsed field gradients toachieve refocusing of the selected spins. An

advantage is that the large signals excited outsidethe selected volume by the initial 9 pulse aredephased and reduced by the pulsed field gradients,so subtraction noise is less serious. Again, the[±x,+y] alternations are included for surface coilstudies to remove transverse error signals, butsince these are small enough for homogeneous r.f.coils to be efficiently dephased by the pulsed fieldgradients, sequence (11) reduces to

90o-r-180°s x[x,y] ;180°S Y;

180°sz - r - acquire, receiver[+,-]. (12)

One phase alternation is retained to eliminate anyresidual signal excited by the 90° pulse and notdephased by the field gradients, giving a final2-transient cycle. Though previously suggested(43) these methods have yet to beexploited, although Ordidge (36) has combinedsimilar sequences with selective excitationpulses, a technique that has now been superceded.

B. Two-Dimensional FourierTransform Localization

1. 2DFT localization. As introduced byMaudsley and coworkers (44) and further describedby Haselgrove et al (45), an incremented pulsedfield gradient can be employed in the first xperiod of a spin-echo (sequence (4) for surfacecoils), and a second Fourier transform providesseparate spectra from a series of slicesperpendicular to the gradient axis (x axis forsurface coils). In general, the method hassignificant advantages over slice-selective pulsesbecause spatial frequency shifts and chemicalshifts are not mixed, and because weakergradients can be used leading to lessproblematical eddy currents. For a surface coil,the transaxial extent of the slices is limited bythe usual 9sin39 dependence of signal intensity forsequence (4). Thus only a field gradient in onedimension (along the coil's axis) is required, soavoiding the usual long acquisition timerequirements of multi-dimensional chemicalshift imaging. The use of one incrementedgradient in this way is the same as the use of anincremented phase-encoding gradient in normal


imaging, so this procedure is well established.

2. Selective 2DFT localization. Referring backto Figure 3, it would seem that only a fewincrementations of a gradient along the coil's axiswould be necessary to generate a convenient thickslice at some depth (centered on e = 90°) anddiscard the slices closer to the surface.Unfortunately, normal slices (or pixels) havea sinx/x signal dependence relative to the xgradient dimension and adjacent slices are notwell resolved (46). More gradient incrementationscould be used to generate thinner slices and onewell resolved thick slice could be obtained bysumming N thinner slices. But this leads to asignal-to-noise penalty of N1/2 (47,48). Mareci andcoworkers (49-52) have solved this problem byshowing that if a variable number of transients,weighted according to a Gaussian function, areobtained at each setting, the final slices obtainedafter the second Fourier transform will beGaussian shaped. Maximum signal-to-noise ismaintained at the center of such slices and only afew gradient incrementations are required togenerate a thick slice, at some depth relative to asurface coil, uncontaminated by surface signals.There appears to be no clear disadvantages forthis selective 2DFT method and it is likely to bewidely used in its own right or in conjunction withdepth pulses (section VI) or even uniformexcitation methods (section IV) to generate aseries of slices with maximum signal-to-noise.

C. Sensitive Point Steady-State FreePrecession Method

This technique, which utilizes r.f. pulses with arapid repetition rate applied in the presence ofslowly alternating field gradients (53-58),requires further development before its value canbe assessed. However it would seem thatcomplete localization could be obtained using thesensitive point method with a surface coil bygenerating a selected slice perpendicular to thesingle alternating field gradient along the coil'saxis.

VI. LOCALIZATION USINGRADIOFREQUENCYINHOMOGENEITY

A: Rotating Frame Zeugmatography

By incrementing a single excitation pulse lengthduring a series of transients, and applying atwo-dimensional Fourier transform, a series ofspectra can be obtained from curved slices whoseboundaries are determined by pulse angle contoursas in Figure 3. Hoult (59) described the basis ofthis method which was further developed byCox and Styles (60) and recently discovered byBolton (61). Haase etal (62) introduced thetechnique to surface coils and Garwood et al (63)used it to obtain 31P metabolite maps from abovine eye. This imaging method is analogous tonormal 2DFT imaging (section V, B) with thedifference that the incremented magnetic fieldgradient is replaced with an incremented r.f. fieldgradient. Extending this analogy, each slice (orpixel) will show poor resolution and theirsummation leads to much less signal-to-noise

than can be obtained by other methods (eg. depthpulses - next section). However, a weightedsummation can be used to optimizesignal-to-noise, as in the selective 2DFT method,and this leads to the Fourier series windowtechnique described below (section VI, D).

B. Simple Depth Pulse Sequences

1. Design. Depth pulse sequences may be used toobtain spectra with maximum signal-to-noisefrom single curved slices whose boundaries aredetermined by pulse angle contours as in Figure 3(4-7,11, 27, 28, 64-67). These sequences havebeen reviewed (5) and improved (6) and are formedfrom the phase-cyled pulses of sequences (3)and (4), giving in general

(m*2e[±x])n; {X«xe}; 2e[±x,±y]; acquire (13)

A 29[+x] type pulse may be used as a fraction ormultiple, m*2e (eg. e/3[+x] or 40[±]), and so, forexample, the on resonance factor is cos(m><2e).Often, several of these pulses are useful (withdifferent values of m) as signified by thesubscript n. The excitation pulse 9 may also be

Vol. 8, No. 1/2 27

Figure 4. Signal magnitude versus pulse angle forvarious depth pulses. Some examples show the270° signal region suppressed, but in all casesboth 270° and 450° signals can be readilyeliminated, by modification of the depth pulse,with a small reduction (up to 20%) in signalintensity at 9=90°. The continuous curve with270° and 450° signals retained corresponds to theshaded signal regions in Figure 3.

used as a fraction or multiple, as indicated by ?X9,and several values of I may be used with theresults of the various transients summed as alinear combination using calculated coefficients(hence X) . Lastly, if significant unwanteddispersion signals occur for any sequence, thesecan be entirely removed by repeating sequence (13)with 29[±x,0] at the beginning, in place of29[±x,±y] at the end, and summing the results. Aselection of the range of possibilities isillustrated in Figure 4. A single 9 pulse would berepresented by a sin9 curve, so compared to this,the limits of signal intensity around 9 = 90° can bepushed in to a variable degree, and this can be

achieved with or without the complete removal of270° and 450° high flux signals. In terms of Figure3, the high flux signal regions can be eliminatedand the 90° signal region can be narrowed orexpanded at will. T1 measurements can be made byinserting an inversion-recovery r delay after aninitial 29[+x] pulse, and any spin-echo method canbe applied by inserting v delays on either side ofthe 29[±x,±y] pulse in sequence (13).

2. Off-resonance effects. For in vivo work wehave noted that a consideration of off-resonanceeffects is very important. These have beencalculated for all depth pulses using thetrigonometric factors listed in section II, and atypical result is shown in Figure 5a.This has been calculated by substituting te*90/9

for tgo in equation (5) to allow for the variationof tg0 throughout sample space. We have foundsuch plots to be more informative than previousversions where the resonance offset axis, AH, wasgiven in terms of reciprocal tgo (68), In Figure 5,the frequency offsets can be obtained bysubstituting for the length of the 9 pulse, te(s),used. In all cases it is found that r.f.discrimination is retained off resonance, but thereis a gradual loss of signal intensity , with thedepth pulses useable to where signal drops to say50% of maximum, at about 0.1 t g " l

Recently, Shaka and Freeman have also noted theutility of mx29[±x] pulses which they calledprepulses (69,70). Shaka and Freeman (70) didhowever introduce a new type of phase cycled


pulse angle, 8

0

0-1 t"e'-

aI)

90

/ml /m

Figure 5. Contour plots of signal intensity (totaltransverse magnetization) against pulse angle 9and frequency offset for (a) depth pulse [I] of refs.(5, 6) (the off-resonance characteristics aretypical of simple depth pulses), (b) a sequenceequivalent to depth pulse [I] on resonance buthaving all simple phase-cycled pulses replaced bycomposite phase-cycled pulses as according toShaka and Freeman (70) and (c) a 9 term FSW withdispersion signals removed. In some cases theoff-resonance proporties of simple depth pulsescan be modestly improved to give similar behaviorto that shown for the FSW in (c).

prepulse which can be written as49[±x(2/3), 0(1/3)], where 2/3 and 1/3 indicatesthat the pulse is not applied for one third of alltransients. This type of pulse can be generalizedto include established pulses such as 29[±x,0] andnew pulses such as 49[±x,0] and 69[+x(2/3),0(1/3)](68). In this way many new depth pulses can bedevised but these are generally similar toestablished sequences or give identical results onresonance (68). However, 29[+x] pulses can beeliminated from some depth pulses using the newpulses and this may lead to a modest improvementoff resonance. In such favorable cases, signalintensity extends a little further off resonancesimilar to Figure 5c, but the useful limits are stillrestricted to about ± 0.11@~1 because of the

pronounced curvature of the pulse angle window tolower 9 values (68).

3. Short recycle times. The localizationprovided by depth pulses (or any other methoddescribed in this section VI) may fail at shortrecycle times, TR, though we have experienced noexperimental difficulty at T R > 2 T r Garwoodetal (63) eliminated this problem for rotating framezeugmatograghy using presaturation. Decorps et al(71) have shown presaturation to be generallyapplicable and that maximum signal is obtained atTR = T1 although there is little gain over TR =(2 to 3)xT1. However, presaturation would allowfor a large range of T1 values as occurs for in vivo

Vol. 8, No. 1/2 29

3 'P species for example.*

4. Applications. Ail the localization methodsin this section VI will provide sensitive volumesin the form of pulse angle slices. From Figure 3 itis clear that these slices will always intersectthe sample surface and may not be convenientlyshaped. Combined methods of localization thateliminate surface signals are described insections VII and VIII. Nevertheless, there areimportant applications for depth pulses alone,especially when a modest reduction in surfacesignals is sufficient and when it is important toprevent detection of regions beyond a certaindepth. In vivo applications along these lines havebeen described (11,66,72,73). Even with completeelimination of surface signals, the curved natureof depth pulse sensitive volumes may not seemvery satisfying from a geometrical point of viewcompared to the more easily visualized cuboidshapes that can be generated using pulsed fieldgradients. But live animals do not fit well tocuboid shapes and there will be many instanceswhere curved slices fit better, eg. surface infarctsand trauma to the brain.

Note that signal-to-noise will be a commonlimiting factor, especially in these eariy stages ofthe development of in vivo spectroscopy. Forexample, a simple depth pulse may easily limitsignal detection to one-third of the originalvolume and this would require nine times moretransients to acquire a reasonable spectrum. Sucha penalty can often not be afforded and it has beenfound that depth pulses which localize to a smalldegree (eg. elimination of high flux signals withminimal reduction of the 90° signal region) areimportant (72, 74). Along these lines, Gadian et al(75) and Decorps et al (76) have utilized theminimal extra localization provided by a 29[±x,±y]pulse in spin-echo sequences.

C. Composite Depth Pulses

1. Narrowband inversion pulses. Shaka andcoworkers (77,78) and Tycko and Pines (79) havedescribed some depth pulses in which one or more

*Recent results using one complete depth pulselocalization method (Bendall, Foxall, Nicols, andSchmidt, J. Magn. Reson., in press, 1986) showed noloss of localization at short recycle times.

30

of the phase-cycled pulses are compositenarrowband inversion pulses. The basic method isdescribed by sequence (4) or (7) (with r = 0)where 29 is now the composite pulse. Ourexperiments and calculations (13,68) have shownthat these composite depth pulses give similarresults on resonance to some established simpledepth pulses, but r.f. discrimination is lost atresonance offsets which are much less than theuseful range for simple depth pulses. Thesecomposite sequences require a reduced phase cyclecompared to the equivalent simple depth pulse,but this small advantage is out-weighed by thepoor off-resonance characteristics of the presentnarrowband inversion composite pulses andconsequently they are not not competitive (13).

2. Broadbanded composites. Recently, Shakaand Freeman have extended the useful range ofoffset frequencies of a depth pulse bysubstituting a new type of composite pulse foreach simple pulse (except the excitation pulse) inthe simple depth pulse sequence. In favourablecases (68) the offset range can be increased to0.25te~' and an example is shown in Figure 5b.However, these composite pulses greatly increasethe length of the depth pulse and it is easily shownthat simple depth pulses retain the advantage interms of length and/or deposition of r.f. power inlive tissues provided there is sufficient r.f. powerto cover the required spectrum (68). This willoften be the case given the large r.f. amplifiersnow being employed on in vivoimagers/spectrometers.

D. Fourier Series Windows

1. Design. Garwood and colleagues (80) haveintroduced a method whereby a single pulse isincremented during a series of transients (likerotating frame zeugmatography) but the transientsare summed using pre-calculated coefficientswhich are determined by the Fourier componentsof a square window. Our most recent calculations(81) show that the range of possibilities indicatedin Figure 4 for depth pulses can be mimicked withthese Fourier series windows (FSW's) and some ofthese features have also been described by Metzand Briggs (82). The ideal square shape of thewindow is lost because of truncation of the

Bulletin of Magnetic Resonance

i

Fourier series, but the "wobbles" of signalintensity on either side of the selectedpulse-angle slice (caused by truncation) can beminimized to <4% as for depth pulses.Alternatively, similar windows can be generatedusing Gaussian coefficients (as described for the2DFT method). The FSW's show a small loss insignal-to-noise over the equivalent depth pulse(10-15%), but they may be experimentally moreconvenient. In particular, there is the abilityafter data collection to "zoom" in from a broadwindow to a narrower window by discardingspectra using the longest r.f. pulses to generatethe broader windows. This aspect will be usefulfor determining when the degree of localization issufficient to eliminate signals from surroundingtissues.

2. Off-resonance effects. Resonance offsetcalculations for FSW's show that r.f.discrimination is lost beyond about 0.05 \Q~ ]

because of dispersion error signals. These can beremoved using

{B*9};29[±x,±y];

acquire + 29[±x,0J;{Xl!x9}; acquire (14)

as for depth pulses, where {I£X9} signifies theFSW and an FSN, with dispersion signals removed,is illustrated in Figure 5c indicating that theusefulness of FSN's is limited to about 0.1tgo"'.by the curvature to lower pulse angles. Analternative to sequence (14) is to use a spin-lockpulse to dephase the unwanted dispersion signals,

{I«x9}[x];SL[y]; acquire, (15)

a procedure previously described for depth pulses(7).

3. Related methods. An early extension ofrotating frame zeugmatography to FSW's wasdescribed by Pekar et al (83), but unfortunatelythey chose to mimic the most inefficient depthpulse sequences. The new "very fast" methods ofPekar and coworkers (84,85) also appear to beimpractical because they require a priorknowledge of 1% times, which must be uniform

across the sample and the same for each chemicalspecies.

As in sequence (14), FSW's can take the place ofsome of the phase-cycled pulses in the depth pulsemethods described in the following sections. It ismerely necessary to replace the excitation pulseby {Ll x&] and then omit whichever phase-cycledpulses are no longer necessary.

VII. LOCALIZATION BY COMBINING THEMETHODS OF PULSED FIELD GRADIENTSAND RADIOFREQUENCY INHOMOGENEITY

1. Selective depth pulses in pulsed fieldgradients. The localization that can be achievedusing a depth pulse may be good enough for aparticular metabolic time-course study, but unlessthis can be proved, the data obtained ismeaningless. It would be valuable to filter out theintervening surface signals using a pulsed fieldgradient method and if this showed that theproposed experiment was prejudiced by thesesurface signals then a time-course study couldproceed at maximum efficiency without the fieldgradient. All simple depth pulses show goodfrequency selectivity, being similar in this regardto a Gaussian shaped pulse, and so the whole depthpulse, DP(9), can be treated as a selectiveexcitation pulse and applied in a pulsed fieldgradient. That is, DP(9) takes the place 9SX inequation (8). This combined localization methodhas been proven by phantom studies (7,27,28).

2. Selective 2DFT with depth pulses. Theselective 2DFT method of Mareci and coworkers(49-52, section V, B, 2) can easily be combinedwith any depth pulse which contains a 29[±x,±y]pulse, and these authors have demonstrated someof the possible variations (51,52). Writing a depthpulse as DP(9); 29[±x,±y], a simple spin-echo isappropriate as in sequence (4) with DP(9) in placeof 9 and the incremented gradient inserted in thefirst r period. It will be common to filter out thesurface signals using an incremented x gradient,but it is also possible to divide up the 90° signalregion into voxels using incremented gradients inboth the y and z dimensions (5,7). Again, thegradients can be discarded if the depth pulse is

Vol. 8, No. 1/2 31

tune

HI-' match

receiver, R -

transmitter, T

Hh-

Figure 6. Circuits useful for eliminating multiplecoil coupling.

proven to be sufficient.

VIII. LOCALIZATION USING MULTIPLER.F. COILS

For a simple surface coil, the 90° region ofsignal intensity always curves back into thesurface outside the circumference of the coil.This is true of any planar coil (including the "fluxconcentrator" (86, 87)). Complete localizationwithout field gradients requires at least a secondinhornogeneous r.f. coil, whose r.f. field differsmarkedly in shape from the first coil. The secondcoil may be used as a receive coil because of itsdifferent sensitivity to sample space, or the twocoils may both be used as transmit coils so thatthe localized region is the region of overlapbetween the sensitive volumes of the two coils.The basis of these two methods was described inone of the original depth pulse articles (64).

A. Elimination of Multiple Coil Coupling.

1. Separate transmit/receive coils. When twor.f. coils are tuned to the same frequency, theyinteract and modify each other's r.f. field. Fortransmit/receive coils, this unwanted coupling hasbeen removed by using crossed diodes in serieswith the transmit coil and in parallel with thereceive coil (64). Recently we have improved onthis using the circuitry in Figure 6 (88). In

comparison to the normal tuned circuit depicted inFigure 6a, a 7J4 cable and crossed diodes toground as in Figure 6b deactivates a receive coilduring pulses from a transmit coil. A cable ofvariable length £ in the transmit line as in Figure6c, or the equivalent circuit in Figure 6d,adequately detunes a larger transmit coil duringsignal acquisition (89). We have found that thiscoil circuitry is efficient even for the worst caseof a surface receive coil inside and parallel to acylindrical transmit coil. Haase (91) has alsoproposed a circuit of the type shown in Figure 6c,but unlike Haase we find that the length I shouldspecifically not be equal to nV4 (n an integar). Inthe case of transmit/receive coaxial coplanarsurface coils it is usually unnecessary to achievecomplete decoupling of the coils, because the r.f.fields are already inhomogeneous, and Styles et alhave found a less efficient circuit to be adequate(90). However, the circuits of Figures 6b,c and dare general and can be used for this applicationtoo.

2. Separate transmit coils. The abovetransmit/receive decoupling methods are allpassive - they rely on the shorting of crosseddiodes during r.f. pulses. For separate transmitcoils a passive method cannot be used but we havesuccessfully employed the actively switchedvariant of Figure 6b as in Figure 6e (17). It wasnecessary to use rapid reed relays as the


switches, so this method is restricted to lowfrequency nuclei in small animal magnets toenable the \IA cables to extend outside the mainmagnetic field. Present research is focused on themore general circuits in Figure 6c and d withactively switched PIN diodes in place of thenormal crossed diodes.* Hedges and Hoult (92)have also proposed this solution to the problem.

3. Separate coils for imaging and spectroscopy.It is extremely useful to obtain a proton image ofthe head or body using a circumscribing coil andthen obtain a localized spectrum with a surfacecoil already in place. The position of the surfacecoil relative to the image can be calibrated usingsmall H2O phantoms adjacent to the coil, or betterstill, the sensitive volume of the surface coil(including depth pulse localization) can bedisplayed on the image by generating a secondimage using the surface coil but the same fieldgradient settings as for the first image (74).However, care must be taken to minimize couplingbetween the two coils. Normally there will be noproblem if the surface coil is much smaller thanthe cylindrical coil and if they are orthogonal andtuned to different nuclei. If tuned to the samenucleus, one should be detuned mechanically (bymissetting the tuning capacitance or insert-ing a switch) whilst using the other. Switcheswould also be necessary for non-orthogonal coils.Figure 6e would be appropriate for the surface coilor the switch should be directly in the circuit inplace of the A/4 cable. Switches substituted forthe crossed diodes in Figure 6c or d should besufficient for the larger imaging coil.

B. Separate Transmit/Receive Coils.

1. Coaxial coplanar surface coils. Experimentalresults obtained using a depth pulse and coaxialcoplanar transmit/receive surface coils areillustrated in Figure 7 (7). High flux signals near

large transmit small receivecoil coil

*This research showed that the circuits inFigure 6c and 6d were inadequate, but that ofFigure 6e with PIN diodes in place of the reedrelays has been very successful (Bendall, Foxall,Nicols and Schmidt, J. Magn. Reson., in press,1986).

surface

Figure 7. Signal intensity contours for coaxialcoplanar surface coils. The contours have beenextrapolated from the experimental results givenin ref. (7). The plane displayed is the xy plane.

the surface were eliminated with the depth pulse,and the small receive coil discriminates againstthe 90° surface signals (outside thecircumference of the transmit coil) on the basis ofsensitivity, since sensitivity decreases morerapidly outside the circumference of a coil than itdoes along the coil's axis (see Figure 3). It isnow clear that in our introductory experiment, weattempted to sample a region which was too farfrom the probe. The depth should be limited toless than the diameter of the small coil. This willgreatly increase the sensitivity of the receive coilto the region near the coil's axis without muchincrease in the surface signals, thus gaining betterdiscrimination overall and restricting signaldetection to a region where the small receiver coilis reasonably efficient. Styles et al (93) have usedsimilar dimensions to this with rotating framezeugmatography and have obtained human liverspectra separate from the intervening musclewall.

2. Homogeneous-transmit/surface-receivecoils. When using separate transmit and receivecoils there is a potential phase problem wheneverthe r.f. field lines of force from the two coils takedifferent relative directions in the laboratory xyplane at different points in sample space. Thesedifferent relative directions are directlytransferred to the NMR experiment as phase shiftsin the detected signal from the different samplepoints, and so signal for each chemical species

Vol. 8, No. 1/2 33

cannot add coherently. This problem does notoccur in the xz plane because only the r.f. fieldcomponent in the x direction is active (see Figure2). We have determined that for the coaxialcoplanar surface coils of Figure 7 the phase shiftis not large across the sensitive volume in the xyplane and so the problem is not serious (5,7,64).However the problem is severe for a largecylindrical transmit coil used with a smallsurface receive coil. The r.f. lines of force of thehomogeneous r.f. coil can be superimposed asparallel lines on those of the surface coil in Figure2, and so in the xy plane identical signals will

cancel from two sample points where the surfacecoil lines of force are pointing in oppositedirections (94). It has been argued that thecancellation of signal because of phase differencedoes itself enhance localization, but this is onlytrue for a homogeneous sample (which is notinteresting to an in vivo spectroscopist).

These problems represent a further limitation ofthe method of Bottomley et al (34,35), which canbe summarized as uncertain transaxiallocalization depending on the homogeneity of thesample, and provide a powerful argument againstusing homogeneous-transmit/surface-receivecoils.

C. Separate Transmit Coils

1. Coaxial coplanar surface coils. If somephase-cycled pulses in a depth pulse sequence areapplied with one r.f. coil and some with another,signals will only be acquired from the overlapregion of the two sensitive volumes (because thetrigonometric factors discussed in section II forthe various pulses are multiplied together). Suchan overlap region is illustrated by theexperimental image in Figure 8c, which comparesto the images for the large surface coil (Figure 8a)

surfqcj

large0 coil

small$ coil

broodband coil

Figure 8. Images of the sensitive volumes in thexy plane obtained using a slice phantom of H3P04and the coaxial coplanar surface coils depicted(17). The small signal regions close to the coilwire in (a) are 270° regions and there is also acomplicated high flux region in (b) closer to thesmall coil. The broad band coil is a continuouscircle of wire which couples with the transmitcoils and enhances the difference in shape of thetwo overlapped sensitive volumes.


and the small coaxial coplanar surface coil (Figure8b) when used individually (17). An improvedversion of the pulse sequences used in the originalstudy (17) is

(2<j>[±x,0])2 - r - DP(9); acquire with 9 coil, (16)

where the <j> pulses are applied with the small coil<j>; the tuning of the coils is switched during r;DP(0) is a depth pulse applied with the large coil 0which eliminates the 270° signals shown in Figure8a; and the signal is acquired with the large coil.The phase problem referred to in the previoussection does not affect the application of aninversion pulse (2<{>) by the second coil as insequence (16), and the phase error was alsoavoided in the original method (17) by using doublespin-echo. Note that if the degree of localizationprovided by the 2<j>[±x,0] pulses is insufficient,this can be increased using narrowband inversionpulses (section (Vi, C, 1) such as the Q1 pulse ofShaka and Freeman (77). Increased pulse powerwill be necessary because such pulses operateover a reduced bandwidth, and they should beapplied as Q ± 1 or Q + ' ; Q"1 to remove deleteriousoff-resonance effects as recently described (13,68, 78).*

2. Homogeneous and surface transmit coils. Itwould seem contradictory to attempt additionallocalization using a homogeneous r.f. coil, but thishas been proposed by Doddrell and coworkers (95).These authors went to considerable trouble toobtain "phase coherence" between a homogeneousr.f. coil (saddle) and a surface coil but as explainedin section VIII, B, 2 this is not possible. Second,the effectiveness of the achieved "phasecoherence" was demonstrated using aphase-alternated inversion recovery sequence likesequence (3) except that the inversion pulse wasapplied with the homogeneous r.f. coil and theexcitation pulse with the surface coil (like insequence (16)). However, as mentioned for

'Better control of the shape of the localizedvolume has been recently obtained using furtherimproved versions of sequence (16) (Bendall,Foxall, Nicols and Schmidt, J. Magn. Reson., inpress, 1986).

sequence (16), the phase of inversion pulsesrelative to the excitation pulse in such alternatedsequences is immaterial since no transversemagnetization is generated by them, yet,impossibly, these authors generated a substantial"out-of-phase" error in the final signal. Third, the"volume-selection capabilities" of their methodwas achieved primarily by matching the surfacecoil 6 pulse to average 180° over one of two smallphantom samples, the first and crudestsurface-coil localization procedure ever used (96).Fourth, "excellent volume selection" was achievedby adding a "purging" 90° pulse applied with thehomogeneous r.f. coil. Such a pulse can have noeffect in the xz plane (as in Figure 3d), butimposes a node line in the xy plane of Figure 3alike that in Figure 3d. This improvement does notappear to be particularly worthwhile.

D. Other Coils

These multiple coil techniques are generalmethods and there are good prospects for improvedmodelling of localized regions. For example, twosurface coils could be used on opposite sides ofthe head with the overlapped sensitive volumes ata variable distance in between, or one coil couldproduce a curved sensitive volume just inside theskull on the opposite side of the head close to asmall moveable surface coil. There is also a vastrange of possiblities for using coils which aredifferently shaped to the flat surface coil.Preliminary results have been described for thesaddle surface coil (97) and "floppy" surface coils(98), and several groups are calculating r.f. fieldprofiles for various shapes (97, 99). The sectorialloop-gap resonator looks promising (100). As afurther example, an array of four mutuallyorthogonal surface coils (with a commonintersection point like an open four-petaledflower) can be used to produce a discrete sensitivevolume which curves away from the surface (101).R.f. fields may also be shaped by introducingpassive coils which couple with the transmitcoils. An example is the circle of wire (broad bandcoil) indicated in Figure 8, and Holcomb and Gore(98) have also made use of "satellite" coils tomodify r.f. fields.

Vol. 8, No. 1/2 35

IX. HETERONUCLEAR METHODS WITHSURFACE COILS

A sequence of r.f. pulses applied to twoJ-coupled heteronuclei can be used to transferinformation or properties from one nucleus to itsJ-coupled neighbor. For in vivo NMR, the mostimportant applications are to ' H-' 3C systems,especially when using 13C-enriched compounds, buthere are other possibilities, e.g. 'H - ^N systems,and 1H-19F systems as exist in anaesthetics. Avital aspect of the 'H-^C and }H-]5H methods isthat the weak 13C signal can be enhanced, or theproperties of the 13C nucleus can be transferred tothe 'H nucleus and the 'H signal can be detectedwith much greater sensitivity. The various pulsesequences fall into one of two classes,polarization transfer and spin-echo methods and arecent finding, additional to our knowledge ofthese heteronuclear methods within mainstreamNMR, is that both the polarization transfer and thespin-echo sequences may at the same time be usedfor localization using depth pulse principles (102,103).

A. Polarization Transfer

1. Localization using r.f. inhomogeneity plussignal enhancement. Polarization transfer from 'Hto 13C leads to a factor of four signalenhancement for the I3C signal. The mostconservative way in terms of pulses to achievethis is to use the DEPT sequence (104) which forsurface coils is (102,105)

redundant as shown (105). If applied with adouble-tuned surface coil with <j> = 9 for maximumsignal, the spatial distribution of signal will besimilar to that depictedin Figure 3a and d. If applied with two coaxialcoplanar surface coils, where the larger 9 coil isthe 13C transmit/receive coil and the <|> coilis the 'H transmit coil, complete localization canbe achieved like that shown in Figure 8 by addingadditional phase-cycled pulses to sequence (17)using normal depth pulse principles. For example a2<j>[+x] pulse and 9/3[±x];(29[±x])2 pulses in frontof the <{> and 9 pulses respectively of sequence (17)would be sufficient. This general method has beenproven experimentally (102).

2. Spectral editing. DEPT may be used to edit invivo 13C spectra into CH, CH2 and CH3 spectra(106,107), but there is a considerablesignal-to-noise penalty if this is done withsurface coils (108), so the spin-echo method ofsection B below is preferable. By reversing the13C and 'H labels in sequence (17) to give inverseDEPT (109), 'H spectra of only 13C enrichedmetabolites can be obtained with a factor of up to16 enhancement in sensitivity over normal 13CNMR. This is useful for 'H resonances close to H2O,but otherwise the spin-echo method of section Bgives a larger sensitivity gain.

3. Localization using pulsed field gradients.Pulsed field gradient localization methods can bereadily combined with sequence (17). For example,incremented field gradients can be inserted intothe first (2J)"1 period (or if necessary an extended

13 C

2<f>[±x,±y] - (2J ) " 1 - <J>/2[y] - (2J)I I9 29[x,y]

- |dec.I

The *H decoupling is optional. For simplicity, thefirst two 1H pulses and the '3C pulses can beregarded as comprising spin-echo sequences as insequence (4), hence the phase cycling of the 2<j>and 29 pulses, though we have recently found thathalf the phase cycling for the 29 pulse is

first delay period (110)) to enable the selective2DFT method (section V, B, 2). Alternatively, Aueet al (111) have shown that if a selectiveinversion pulse in a pulsed gradient (section V, A,2) is applied to the 'H spins prior to DEPT, thesmall spread of *H chemical shifts does not cause


a significant spread of localized regions as wouldresult from selective 13C methods (see section V,A). Thus for example, using either method, awell-localized 13C spectrum can be obtained withmaximum signal-to-noise using a homogeneouscylindrical 'H coil (in place for ]H imaging) and a13C surface coil, with a single pulsed fieldgradient along the surface coil's axis. There are asusual several other permutations.

B. Heteronuclear Spin-Echo Techniques.

1. General methods. Without exception, any ofthe pulse sequences described in this review arealready spin-echo sequences, or can be convertedto spin-echo sequences suitable for inhomogeneousr.f. coils, by the addition of - r - 20[+x,+y] - r -(4). For a surface coil, the gated on version of thewell known gated-decoupled spin-echo experimentfor a coupled AX system can be written as

2. Editing of 13C spectra. If A=13Cand X^ 'Hinsequence (18) and alternate transients areaccumulated into separate computer memoryblocks, then addition of the two blocks yields amethylene/quartemary subspectrum andsubtraction yields a methyl/methine subspectrum.This valuable simplification of the 13C spectrum isobtained in the same time as a normal spectrum(112-114) and the benefit to in vivo spectroscopyhas recently been established in a study of13C-enriched organs (107).

3. Editing of 'H spectra. If A ^ H and X=13C, aspectrum of only those protons attached to 13Cnuclei can be obtained by subtraction of alternatetransients. This has obvious applications in13C-enrichment studies and benefits from a largepotential gain in signal-to-noise over conventional13C spectroscopy. The maximum possible gains arefactors of 64,128 and 192 for CH, CH2 and CH3

S(9) — X—26[±x,±y]- [acquire

I I| decouple for || J-1for || alternate |decoupleI transients I

(18)

Decoupling is optional during signal acquisition.The alternative gated off version may also be used.S(e) is the rest of the pulse sequence, and may bejust 9 or any depth pulse (phase-cycled,composite or FSW), or a selective pulse sequenceutilizing a pulsed field gradient. The 20[±x, ±y]refocusing pulse may be any sort of compositepulse or selective pulse in a pulsed field gradient.The spin-echo may be part of an imaging sequenceusing incremented field gradients for the selective2DFT method. The pulsed field gradients can beused to achieve complete localization in thevarious normal ways (section V). The r period maybe much longer than J " 1 to facilitate suppressionof water and fat signals by T2 relaxation.

moieties, though experimentally there are variouslosses, including subtraction inaccuracies (or"noise"), which lead to much lower gains.Nevertheless, Rothman et al (115) obtained an11-fold increase in sensitivity for the CH3 oflactate and a 6-fold increase for the CH2 ofglutamate in the rat brain in vivo during acontinuous infusion with 13C-labelled glucose. Bycomparing the 1H resonance of just the13CHnspectrum to the ordinary spectrum (12CHn +13CHn), the fractional enrichment was alsodetermined. Modem stable instrumentation shouldgive further sensitivity gains.

Vol. 8, No. 1/2 37

4. Editing of 'H spectra plus localization usingr.f. inhomogeneity. The detection of only thoseprotons attached to )3C nuclei was firstintroduced using a pulse on the 13C nucleus ratherthan gated decoupling (116,117). Including alarger x period suitable for various purposes, thepulse sequence for surface coils is

'H DP(9) x 20[±x,±y] x |acquire

I _ I13C | - J - 1 - 20c [x ,0 ] |decouple

(19)The 2$c [x ,0] pulse is a composite narrowbandinversion pulse of the type describe by Shaka et al(77) or Tycko and Pines (79), or simpler versionsgiven by us (103), and is applied for alternatetransients with receiver phase alternation. DP(9)is any depth pulse sequence, including a Fourierseries. We recently proved, using phantomsamples, that sequence (19) allows completesensitive volume localization like that shown inFigure 8 when using separate ]H{&) and 13C(<|>)surface coils (103).

X. HOMONUCLEAR W SPECTRALSIMPLIFICATIONS WITH SURFACE

COILS

A. Water Suppression

The water resonance is so massive compared tomillimolar metabolites that its broad base coversthe whole ]H spectrum. Presaturation with lowpower single frequency r.f. is commonly used, butwhilst this is efficient at the center of the H2Osignal, a significant proportion of the broad wings,which result from field inhomogeneity, remains.These wings can be removed with reasonableefficiency if the method in use includes receiverphase alternation (115) (as in sequence (19) or(20) for example), but this will lead tosubtraction noise. Williams et al have used aCarr-Purcell-Meiboom-Gill (CPMG) pulse train butthis is not suitable for surface coils and a simpleHahn echo does not provide sufficient H2Osuppression (118). Binomial water suppression

(119-121) functions by not exciting the H2Oresonance but excites resonances within anadjustable window on either side of the H2Oresonance. Brindle et al (122) have used binomialwater suppression for both pulses in a spin-echosequence and Hetherington et al (123) have shownthat this can be applied with a surface coil. Indeedit can be shown using rotation matrices (6) thatevery pulse in a depth pulse sequence may be splitinto a binomial sequence (124) so that H2O is neverexcited, and recent experience shows that thesebinomial methods are easily efficient enough forin vivo ]H NMR with surface coils (74,123).

B. 'H Spectral Editing

1. Selective decoupling. Despite the ability tosuppress the H2O resonance, the 'H spectrum isstill mostly intractable because of large broadmulticomponent fat resonances and numerousoverlapping 'H metabolites. However, Rothman etal (125) have used an editing technique ofCampbell and Dobson (126) to individually detectthe resonances of alanine, lactate, glutamine andglutamate in excised leg muscles and heart of arat and to observe alanine, £-hydroxybutyrate,glutamate and glutamine in a perfused mouse liver.Williams et al (118) have since used the samemethod to detect the build up of lactate in a ratleg during ischemia. Although these studies weredone usino homogeneous r.f. coils. Rothman et alhave also used the method to reveal lactate in thein vivo rat brain using a surface coil, for whichthe editing experiment is

'H 9 — (2J ) - '— 29[±x,±y] — (2J)

IiH x | selective decoupling for

I alternate transients

- i i acquireI

(20)

The method relies on the homonuclear coupling (J)between two adjacent protons in the metabolite,HA and Hx, e.g. between the lactate CH3 and CHprotons. Single frequency decoupling is appliedto Hx for alternate transients which aresubtracted.


2. Selective inversion. If binomial watersuppression is used, and Hx lies in the unexcitedwindow (as will be common when Hx is a CHproton and HA are CH2 or CH3 protons), the methodbecomes exactly analogous to the heteronuclearcase. Writing the binomial pulses as 9W, themethod becomes

'H 9W x 2ew[+x,+y] x acquire

' H x | — J " 1 —20s[x,0] (21)

2$ s is a chemical shift selective 180° pulseapplied just to Hx, and since both HA and Hx areseparately irradiated in the sequence, it functionsin the same way as sequence (19). For 2$ s to beselective, it will be of tens of millisecondsduration, so x is set geater than J " 1 to permitthis. Hetherington etal (123) have used thismethod to detect lactate and alanine in the ratbrain postmortem using a surface coil. We havesince been thwarted by significant errors arisingfrom intervening fat layers in other applications,but have theoretically proposed that the use of theequivalent sequence

>H w — x —29w[±x, ±y] • | acquire(22)

24Ux,0]—

in addition to (21) will eliminate this (128).Recent results appear to confirm this (74). Notethat by analogy with sequence (19), completelocalization can be achieved with sequences (21)and (22) using appropriate binomial depth pulses inplace of 9W and by applying the 0 and $ pulseswith separate coils. Alternatively the pulsed fieldgradient methods of section V could be combinedwith these editing methods.

XI. CONCLUSION

Given the reality of large-bore high fieldstrength magnets, and spectrometers that willdeliver any sequence of hard or soft pulses with or

without pulsed field gradients, the challenge isto optimize localization, sensitivity and spectralediting for in vivo NMR spectroscopy. A plethoraof different methods has resulted and there arenow real solutions to the problem ofsensitive-volume localization, though none are inroutine use - insufficient time has passed.Although we may now have a reasonable idea ofthe range of possible solutions, no doubt newsolutions will be devised for some time to come,and the task of optimizing the present methods forroutine application has hardly begun. It isimpossible to conclude that one method is bestoverall - that conclusion depends on theparticular investigation in mind. However, in anyone case it is possible to assess the sources oferrors and judge whether a localization method isrequired or sufficient. Perhaps the biggest worryis whether there will be sufficient intrinsicsignal-to-noise to permit the routine use oflocalization procedures. Certainly, the need formaximum sensitivity ensures the futureimportance of inhomogeneous coils.

REFERENCES

'E.A. Shoubridge, R.W. Briggs and G.K. Radda, FEBSLett. 140,288(1982).

2 J.J.H. Ackerman, J.L. Evelhoch, B.A. Berkowitzand G.M. Kichura, R.K. Devel and K.S. Lown, J. Magn.Reson. 56, 318 (1984).

3M.G. Crowley, J.L. Evelhoch and J.J.H. Ackerman,Proceedings of the Third Annual Meeting of theSociety of Magnetic Resonance in Medicine (NewYork, August 1984), p.173.

4M.R. Bendall and R.E. Gordon, J. Magn. Reson. 53,365 (1983); M.R. Bendall, U.S. Patent No.4,486,709(1984).

5M.R. Bendall in Biomedical Magnetic Resonance(T.L. James and A.R. Margulis, Eds.), p. 99,Radiology Research and Education Foundation, SanFrancisco, 1984.

6M.R. Bendall and D.T. Pegg, Magn. Reson. Med. 2,91 (1985).

7M.R. Bendall, J. Magn. Reson. 59,406 (1984).8W.P. Evanochko, T.C. Ng, T.T. Sakai, N.R. Krishna

and J.D. Glickson, Magn. Reson. Med. 1,149 (1984).9D.E. Demco, P. Van Hecke and J.S. Waugh, J. Magn.

Reson. 16, 467(1974).

Vol. 8, No. 1/2 39

10J.L Evelhoch and J.J.H. Ackerman, J. Magn.Reson. 53, 52 (1983).

' >T.C. Ng, J.D. Glickson and M.R. Bendall, Magn.Reson. Med. 1,450(1984).

12G. Bodenhausen, R. Freeman and D.L Turner, J.Magn. Reson. 27, 511 (1977).

13M.R. Bendall and D.T. Pegg, J. Magn. Reson. 63,494(1985).

H JJ .H . Ackerman, T.H. Grove, G.C. Wong, D.G.Gadian and G.K. Radda, Nature (London) 283,167(1980).

15A. Haase, W. Hanicke and J. Frahm, J. Magn.Reson. 56,401 (1984).

16J.L. Evelhoch, M.G. Crowley and J.J.H. Ackerman,J. Magn. Reson. 56,110 (1984).

17M.R. Bendall, J.M. McKendry, I.D. Cresshull andR.J. Ordidge, J. Magn. Reson. 60, 473 (1984).

18R.J. Ordidge, unpublished results.19R. Freeman, S.P. Kempsell and M.H. Levitt, J.

Magn. Reson. 38, 453 (1980).20R. Tycko, H.M. Cho, E. Schneider and A. Pines, J.

Magn. Reson. 61, 90 (1985).2tH.P. Hetherington, D. Wishart, S.M. Fitzpatrick,

P. Cole and R.G. Shulman, J. Magn. Reson., 66, 313(1986).

22M.H.. Levitt and R.R. Ernst, J. Magn. Reson. 55,247 (1983).

23H.P. Hetherington and D.L. Rothman, J. Magn.Reson., 65, 348(1985).

24M.R. Bendall and D.T. Pegg, J. Magn. Reson., 67,376(1986).

25M.R. Bendall and D.T. Pegg, unpublished results.26M.S. Silver, R.I. Joseph and D.I. Hoult, Phys. Rev.

A 31, 2753 (1985).2?M.R. Bendall, Bull. Magn. Reson. 5,191 (1983).28M.R. Bendall, Magn. Reson. Med, 1,105 (1984).29D.I. Hoult, J. Magn. Reson. 35, 69 (1979).30P.R. Locher, Phil. Trans. Roy. Soc. London Ser. B.

289,537(1980).31C. Bauer, R. Freeman, T. Frenkiel, J. Keeler and

A.J. Shaka, J. Magn. Reson. 58, 442 (1984).32M.S. Silver, R.I. Joseph and D.I. Hoult, J. Magn.

Reson. 59, 347(1984).3 3J . Frahm and W. Hanicke, J. Magn. Reson. 60, 320

(1984).34P.A. Bottomley, T.B. Foster and R.D. Darrow, J.

Magn. Reson. 59, 338 (1984).35P.A. Bottomley, W.A. Edelstein, T.H. Foster and

W.A. Adams, Proc. Nat. Acad. Sci., 82, 2148 (1985).

36R.J. Ordidge, M.R. Bendall, R.E. Gordon and A.Connelly, Proceedings of the Eleventh BiennialInternational Conference of Magnetic Resonance inBiological Systems (Goa, September 1984), p.387.

37R.J. Ordidge, A. Connelly and J.A.B. Lohman, J.Magn. Reson., 66, 283 (1986).

38I.R. Young, U.K. Patent application No. 2,122,753A.

39W.P. Aue, S. Muller, T.A. Cross and J. Seelig, J.Magn. Reson. 56, 350 (1984).

4 0S. Muller, W.P. Aue and J. Seelig, J. Magn. Reson.63,530(1985).

4 1 H. Post, P. Brunner and D. Ratzel, BrukerMedizintechnik.,Karlsruhe FRG, personalcommunication.

4 2M. Garwood, personal conmunication.43M.R. Bendall in Magnetic Resonance in Cancer

(P.S. Allen, D.P.J. Boisert and B.C. Lentle, Eds.), p.159, Pergamon, Toronto, 1986.

44A.A. Maudsley, S.K. Hilal, W.H. Perman and H.E.Simon, J. Magn. Reson. 51,147 (1983).

45J.C. Haselgrove, V.H. Subramanian, J.S. Leigh, L.Gyulai and B. Chance, Science 220,1170 (1983).

4BR. J. Ordidge, personal communication.47 W.A. Edelstein, Proceedings of the Third Annual

Meeting of the Society of Magnetic Resonance inMedicine (New York, August 1984), p. 202.

48D.L. Rothman, personal communication.49H.R. Brookerand T.H. Mareci, Magn. Reson. Med.,

1,118(1984).50T.H. Mareci and H.R. Brooker, J. Magn. Reson. 57,

157(1984).5 'T.H. Mareci, R.G. Thomas, K.N. Scott and H.R.

Brooker, Proceedings of the Third Annual Meetingof the Society of Magnetic Resonance in Medicine(New York, August 1984), p.493.

52T.H. Mareci, J.T. Mao. K.N. Scott and H.R. Brooker,Proceedings ofthe Fourth Annual Meeting of the Society ofMagnetic Resonance in Medicine (London, August1985), p. 1022.

53K.N. Scott, H.R. Brooker, J.R. Fitzsimmons, H.F.Bennett and R.C. Mick, J. Magn. Reson. 50, 339(1982).

54K.N. Scott, R.A. Schurehaus, J.R. Fitzsimmons,R.C. Mick, H.F. Bennett and H.R. Brooker, Magn.Reson. Med. 1,246(1984).

55R.C. Mick, J.R. Fitzsimmons, K.N. Scott and R.A.Schuerhaus, Magn. Reson. Med. 1, 203 (1984).


56K.N. Scott in Biomedical Magnetic Resonance(T.L. James and A.R. Margulis, Eds) p. 79, RadiologyResearch and Education Foundation, San Francisco,1984.

5?A. Macovski, Magn. Reson. Med. 2, 29 (1985).5 8 I . Shenberg and A. Macovski, Proceedings of the

Fourth Annual Meeting of the Society of MagneticResonance in Medicine (London, August 1985), p.1056.

59D.I. Hoult, J. Magn. Reson. 33,183 (1979).60S.J. Cox and P. Styles. J. Magn. Reson. 40,209

(1980).6 tP.H. Bolton, J. Magn. Reson. 63, 620 (1985).62A. Haase, C. Malloy and G.K. Radda, J. Magn.

Reson. 55,164(1983).5 3M. Garwood, T. Schleich, G.B. Matson and G.

Acosta, J. Magn. Reson. 60,268 (1984).64M.R. Bendall, Chem. Phys. Lett. 99, 310 (1984).65M.R. Bendall and W.P. Aue, J. Magn. Reson. 54,

149(1983).66T.C. Ng and M.R. Bendall, Magn. Reson. Med. 1,

216 (1984).57M.R. Bendall, Proceedings of the Third Annual

Meeting of the Society of Magnetic Resonance inMedicine (New York, August 1984), p.41.


69A.J. Shaka and R. Freeman, J. Magn. Reson. 62,340(1985).

70A.J. Shaka and R. Freeman, J. Magn. Reson. 64,145(1985).

7 1M. Decorps, M. Laval, A. Confort and J.-J.Chaillout, J. Mag. Reson. 61, 418 (1985).

72W.M. Chew, M.E. Moseley, M.C. Nishimura and T.L.James, Proceedings of the Fourth Annual Meetingof the Society of Magnetic Resonance in Medicine(London, August 1985), p.952.

73R. Gonzalez-Mendez, M.E. Moseley, J.Murphy-Boesch, W.M. Chew, L. Litt and T.L. James,Proceedings of the Fourth Annual Meeting of theSociety of Magnetic Resonance in Medicine(London, August 1985), p.971.

™C.C. Hanstock, D.P. Boisvert and M.R. Bendall,unpublished results.

75D.G. Gadian, E. Proctor, S.R. Williams, I.J. Coxand R.M. Gardiner, Proceedings of the FourthAnnual Meeting of the Society of MagneticResonance in Medicine (London, August 1985), p.785.

76M. Decorps, P. Blondet and J.P. Albrand,Proceedings of the Fourth Meeting of the Societyof Magnetic Resonance in Medicine (London, August1985), p.779.

77A.J. Shaka and R. Freeman, J. Magn. Reson. 59,169 (1984).

78A.J. Shaka, J. Keeler, M.B. Smith and R. Freeman,J. Magn. Reson. 61,175 (1985).

79R. Tycko and A. Pines, J. Magn, Reson, 60,156(1984).

8 0M. Garwood, T. Schleich, B.D. Ross, G.B. Matsonand W.D. Winters, J. Magn. Reson. 65, 239 (1985).

8 1M. Garwood, T. Schleich, M.R. Bendall and D.T.Pegg, J. Magn. Reson., 65, 510 (1985).

82K.R. Metz and R.W. Briggs, J. Magn. Reson. 64,172(1985).

8 3 J . Pekar, J.S. Leigh and B. Chance, J. Magn. Reson.64,115(1985).

8 4 J . Pekar, P.F. Renshaw and J.S. Leigh,Proceedings of the Fourth Annual Meeting of theSociety of Magnetic Resonance in Medicine(London, August 1985), p. 176.

8 5 J . Pekar and J.S. Leigh, Proceedings of theFourth Annual Meeting of the Society of MagneticResonance in Medicine (London, August 1985),P.178.

86G.K. Radda, Proceedings of the Third AnnualMeeting of the Society of Magnetic Resonance inMedicine (New York, August 1984), p. 605.

87D. De Klerk, The Construction of High-fieldElectromagnets p.115, Newport Instruments Ltd,,Newport Pagnell, Bucks, 1965.

88M.R. Bendall, A. Connelly and J.M. McKendry,Magn. Reson. Med., 3,157 (1986).

89P. Styles, personal communication.90P. Styles, M.B. Smith, R.W. Briggs and G.K. Radda,

J. Magn. Reson. 62, 397 (1985).9 'A. Haase, J. Magn. Reson. 61,130 (1985).9 2LK. Hedges and D.I. Hoult, Proceedings of the

Fourth Annual Meeting of the Society of MagneticResonance in Medicine (London August 1985),p.1096.

93P. Styles, C.A. Scott and G.K. Radda, Magn. Reson.Med. 2, 402(1985).

9*M. G. Crawley, J. L. Evelhoch and J.J.H.Ackerman, J. Magn. Reson. 64, 20 (1985).

9 5 J . Field, W.M. Brooks, J.M. Bulsing, M.G. Irvingand D.M. Doddrell J. Magn. Reson. 63, 612 (1985).

96R.S. Balaban, D.G. Gadian and G.K. Radda, Kidney

Vol. 8, No. 1/2 41

Int. 20, 575(1981).97G.D. Clarke and R.L. Nunnally, Proceedings of the

Third Annual Meeting of the Society of MagneticResonance in Medicine (New York, August 1984),p.161.

98W.G. Holcomb and J.C. Gore, Proceedings of theThird Annual Meeting of the Society of MagneticResonance in Medicine (New York, August 1984), p.333.

"M.S . Roos, A. Hasenfeld, M.R. Bendall, R.H.Huesman and T.F. Budinger, Proceedings of the

. Third Annual Meeting of the Society of MagneticResonance in Medicine (New York, August 1984), p.632.

100A. Jesmanowicz, T.M. Grist, W. Francisz andJ.S. Hyde, Proceedings of the Fourth Annual Meetingof the Society of Magnetic Resonance in Medicine(London, August 1985), p.489.

'01M.R. Bendall and J.M. McKendry, European patentapplication No. 85305397.3.


103M.R. Bendall and D.T. Pegg, Magn. Reson. Med.2,298(1985).

1(HD.T. Pegg, D.M. Doddrell and M.R. Bendall, J.Chem Phys. 77, 2745 (1982).

105D.T. Pegg and M.R. Bendall, Magn. Reson. Med. 2,453 (1985).

10BM.R. Bendall and D.T. Pegg, J. Magn. Reson. 53,272(1983).

l0?M.R. Bendall, J.A. den Hollander, F.Arias-Mendoza, D.L. Rothman, K.L. Beharand R.G.Shulman, Magn. Reson. Med. 2, 56 (1985).

108D.T. Pegg and M.R. Bendall, J. Magn. Reson., 63,556(1985).

109M.R. Bendall, D.T. Degg, D.M. Doddrell and J.Field, J. Magn.Reson. 51, 520 (1983).

110D.T. Pegg and M.R. Bendall, J. Magn. Reson. 55,114(1983).

1 ' ]\N.P. Aue, S. Miiller and J. Seelig, J. Magn. Reson.61,392(1985).

U 2D.J. Cookson and B.E. Smith, Org. Magn. Reson.16,111 (1981).

113D.W. Brown, T.T. Nakashima and D.L. Rabenstein,J. Magn. Reson. 45, 302 (1981).

1HM.R. Bendall, D.T. Pegg, D.M. Doddrell and D.H.Williams, J. Org. Chem. 47, 3021 (1982).

115D.L Rothman, K.L Behar. H.P. Hetherington, J.A.den Hollander, M.R. Bendall, O.A.C. Petroff and R.G.Shulman, Proc. Nat. Acad. Sci. USA, 82,1633(1985).

116M.R. Bendall. D.T. Pegg, D.M. Doddrell and J.Field, J. Am. Chem. Soc. 103, 934 (1981).

117R. Freeman, T.H. Mareci and G.A. Morris, J. Magn.Reson, 42, 341 (1981).

118S.R. Williams. D.G. Gadian, E. Proctor, D.B.Sprague, D. F. Talbot, I.R. Young and F.F. Brown, J.Magn. Reson. 63, 406 (1985).

119V. Sklenar and Z. Starcuk, J. Magn. Res. 50, 495(1982).

120P.J. Hore, J. Magn. Reson. 55,283 (1983).12'Z. Starcuk and V. Sklenar, J. Magn. Reson. 61,

567 (1985), and refs. therein.122K.M. Brindle. R. Porteous and I.D. Campbell, J.

Magn Reson. 56, 543 (1984).123H.P. Hetherington, M.J. Avison and R.G. Shulman,

Proc. Nat. Acad. Sci., U.S.A. 82, 3115 (1985).124M.R. Bendall and H.P. Hetherington, unpublished

results.125D.L Rothman, F. Arias-Mendoza, G.I. Shulman

and R.G. Shulman, J. Magn. Reson. 60, 430 (1984).126I.D. Campbell and CM. Dobson, J.C.S. Chem.

Comm., 751 (1975).127D.L Rothman, K.L. Behar, H.P. Hetherington and

R.G. Shulman, Proc. Nat. Acad. Sci., U.S.A. 81, 6330(1984).

128H.P. Hetherington, D.L. Rothman and M.R.Bendall, unpublished results.


surface coil techniques for in vivo nmr - weizmann.ac.il · surface coil techniques for in vivo nmr...

Documents