Nuclear Spin Tomography Method

The invention relates to a nuclear spin tomography method according to the preamble of patent claim 1. Such a method is described e.g. in U.S. Pat. No. 4,070,611.

Methods which permit the image-representation of the interior of bodies from the exterior pursuant to application of nuclear magnetic resonance effects have also become known under designations such as Zeugmatography, Spin-Imaging, Spin-Mapping or FONAR. With the aid of such methods it is possible to image the density of specific isotopes as well as the distribution of the nuclear magnetic relaxation times over a body cross-section. For medical diagnostics, the hydrogen bound to water is of particular importance. One can thus obtain, in a non-invasive manner, data regarding the physical and chemical state of living tissue. As the measuring basis for this purpose the known phenomena of nuclear magnetic resonance are employed.

In carrying out diagnostic methods of the present type, the body to be examined is exposed to a DC magnetic field B.sub.o and the nuclear isotopes to be be detected are excited to precession by means of a high frequency field B.sub.1. B.sub.1 must here have the frequency .omega.=.gamma..multidot.B.sub.o given by the gyromagnetic ratio. The allocation of the nuclear resonance signal--detectable in a receiver coil as a consequence of the precession movement--to the location of its origin proceeds through the application of linear magnetic field gradients, such as are described e.g. by P. C. Lauterbur in "Nature" 242 (1973), pages 190/191. If G.sub.x is the intensity of the field gradient in the x-direction, the local nuclear resonance frequency .omega. (x) results at:

It is thus possible to allocate the nuclear resonance signal, on the basis of its frequency, to a specific location along the x-direction. Though successive application of linear field gradients G.sub.x, G.sub.y, and G.sub.z, in all three spatial directions x, y, z, also a three-dimensional localizing in an extended subject is possible.

A simple implementation of the method disclosed in the above-captioned U.S. Pat. No. 4,070,611 presents difficulties, since a very brief rise time of the gradient current pulses is required, on the one hand, and each virtually unavoidable basic field inhomogeneity leads to a reduction of the signal, on the other hand. The method described in the present invention is not, or only insignificantly, influenced by these two difficulties.

Accordingly, the invention has the objective, in the case of a method according to the preamble of claim 1, with a given technically unavoidable inhomogeneity of the DC magnetic field, of producing a sectional image of a body which is optimum with regard to signal-to-noise ratio and freedom from artifacts.

This objective is achieved in accordance with the invention by virtue of the fact that an optimally brief pulse train is employed, in that first, with a switched-on Z-gradient, by means of a shaped 90.degree.-Hf-pulse, a layer at a specific position along the Z axis (z=const) is excited, that following termination of the Hf-irradiation of the period (or duration) T.sub.z, an x-gradient of the period T.sub.x /2 (T.sub.x e.g. 10 ms) and simultaneously a y-gradient (G.sub.y) with variable length are switched-on, whereupon, at the constant time T.sub..pi.1, a 180.degree.-Hf-pulse follows, which is followed simultaneously by a z-gradient of the length T.sub.z /2 (T.sub.z e.g. 10 ms) and a y-gradient of variable length, so that the difference between the y-gradient effect .intg.G.sub.y (t)dt before the 180.degree.-Hf-pulse and the y-gradient effect subsequent thereto varies from (-N/2)-times to (N/2-1)-times an increment T.sub.x G.sub.x .multidot.1/N, whereupon that particular x-gradient of the period T.sub.x follows, during which time the entire nuclear spin echo is being measured, which arises at the fixed time t.sub.E1 as symmetry center, and that, for an N.times.N-image (N=number of image points), N of such operating sequences (or cycles) are measured, whereby the y-gradient effect varies in the above-cited fashion.

In the following further explanations of details and advantages of the invention, reference is made to known terms from nuclear magnetic resonance pulse spectroscopy; for example, 90.degree.- and 180.degree.-pulse, and to the exemplary embodiments illustrated in the Figures on the accompanying drawing sheets; and other objects, features and advantages will be apparent from this detailed disclosure and from the appended claims.

In FIG. 1 the block circuit diagram of an apparatus is schematically illustrated which is suitable for carrying out the inventive nuclear spin tomography method;

In FIG. 2 there is illustrated, in a diagram plotting the z-gradient signals G.sub.z (t) as a function of time, the sequence of the Hf- and z-gradient pulses, of a favorable pulse train;

In FIG. 3 there is illustrated the corresponding sequence of the x-gradient pulses, G.sub.x (t);

In FIG. 4 there is illustrated the corresponding sequence of y-gradient pulses for positive gradient effects k.sub.y ;

In FIG. 5 there is illustrated the corresponding sequence of the y-gradient pulses for negative gradient effects k.sub.y ;

In FIGS. 6 and 7, in a diagram, alternative switching sequences for the y-gradient are illustrated, whereby the latter is switched-on as long as possible;

In FIGS. 8 and 9, there are illustrated, in diagrams, the pulse trains for projection-reconstruction-methods,

In FIG. 10 a variant for the y-gradient coil is illustrated,

In FIGS. 11 and 12, an alternative pulse train is illustrated in a diagram for which the current supply of the y-gradient must be designed in a bipolar fashion, and

In FIGS. 13a-13d an example is illustrated of the excitation diagram for the purpose of simultaneous measurement of several layers.

In FIG. 1, 1 and 2, as well as 3 and 4, characterize coils with which a DC magnetic field B.sub.o is generated in which, in the case of application for medical diagnostics, the body 5, to be examined of a patient is disposed. Moreover, gradient coils are associated with said patient which are provided for the purpose of generating independent, mutually perpendicular magnetic field gradients of the directions x, y, and z, according to the directions indicated at 6 in FIG. 1. In the Figure, for the purpose of clarity, only coils 7 and 8 are illustrated which, together with a pair of opposite similar coils, serve the purpose of generating an x-gradient. The similar, not-illustrated y-gradient coils are disposed parallel to the body 5 and above as well as below said body and those for the z-gradient field are disposed transversely to the longitudinal axis at the head end and at the foot end. The arrangement, moreover, additionally contains a high frequency coil 9 serving the purpose of excitation of the nuclear resonance and reception of the nuclear resonance signals.

The coils 1 through 4, 7, 8 and 9, bordered by a dash-dot line 10, represent the actual examination instrument. It is operated from an electric arrangement which comprises a power supply unit 11 for the operation of the magnetic coils 1 through 4, as well as a gradient current supply 12 to which the coils 7 and 8, as well as the additional gradient coils, are connected. The coil 9, serving the purpose of excitation as well as reception of the signal, is connected, via a modulator 13 and a transmitter amplifier 14, to a radio-frequency oscillator 15, on the one hand, and, via a signal amplifier 16 and a phase-sensitive rectifier 17, to a process computer 18, on the other hand, from which, for the purpose of output of the image information, a viewing screen apparatus 19 is supplied. For controlling the arrangement, moreover, additionally connections 20, 21, and 22, are provided between the computer 18 and the modulator 13, between the oscillator 15 and the rectifier 17, as well as between the computer 18 and the gradient current supply 12.

Subsequently, following switching-on of the basic field with the aid of the gradient current supply and the Hf-transmitting- and receiving-system, for example the sequence sketched in FIG. 2, is executed. First, the excitation of a layer with z=const takes place by means of a selective 90.degree.-Hf-pulse. To this end, the z-gradient of a corresponding time function 30 is switched-on. Simultaneously, a Hf-pulse with an envelope as indicated at 31 is radiated-in. The frequency of the Hf-pulse conforms, according to .omega.(z)=.gamma.(B.sub.o +G.sub.z .multidot.z), to the position z of the layer to be excited. The duration of the pulse amounts to N times a basic time raster .DELTA.t which is adjusted depending upon the intensity of the gradients and the thickness of the desired layer and which lies on the order of magnitude of 100 .mu.s. N is, for example, 128. Gradient current pulses G.sub.x follow with a length of (N.sub.G.sup.- +N/2).multidot..DELTA.t, designated by 35 and 36, whereby the number N.sub.G.sup.-, must be so selected that, after the time N.sub.G.sup.- .multidot..DELTA.t, the gradient current has reached its constant value after the rising edge. Simultaneously the y-gradient for a time N.sub.G.sup.- .multidot..DELTA.t, referenced here with 40, is switched on. After a 180.degree.-Hf-pulse 32, the z-gradient for N/2.multidot..DELTA.t, referenced with 33, FIG. 2, may be switched on. Simultaneously the y-gradient for a time of (N.sub.G .sup.+ +N.sub.y.sup.+).multidot..DELTA.t is switched on (41,42, FIG. 4), whereby, for N.sub.y.sup.+ =N/2-1, the difference of the effects of the y-gradient prior to and after the 180.degree. pulse, as desired, in the first scan is equal to G.sub.y .multidot..DELTA.t.multidot.(N/2-1). It is assumed that the gradient effect during the time N.sub.G.sup.- .multidot..DELTA.t (40, FIG. 4) prior to the 180.degree.-pulse (32, FIG. 2) is equal to the gradient effect during the time N.sub.G.sup.+ .multidot..DELTA.t (41, FIG. 4) after the 180.degree.-pulse. During the following x-gradient pulse of the period (N.sub.G.sup.+ +N+N.sub.G.sup.-).multidot..DELTA.t with the references 37, 38, 39, FIG. 3, an echo of the nuclear resonance signal results which is scanned via the high frequency electronics and output. It is additionally possible, with the aid of an additional 180.degree.-pulse 34, FIG. 2, and an x-gradient pulse 37', 38', 39', FIG. 3, to generate and evaluate an additional echo which corresponds to the Fourier spectrum of the examined layer for an inverted wave count vector in y-direction (-k.sub.y). In further scans the experiments (examination sequences) are then analogously repeated, whereby the period of the y-gradient pulse 42, FIG. 4, is reduced to zero step-by-step and, in the then following scans, the y-gradient pulse 40', FIG. 5, after the 180.degree.-pulse 32, FIG. 2, constantly maintains the length N.sub.G.sup.+ .multidot..DELTA.t, whereas the y-gradient pulse 42, FIG. 5, prior to the 180.degree.-pulse is extended from zero in step-by-step fashion to N/2.multidot..DELTA.t. When a complete data set for all desired lengths of the y-gradient pulses is present, then the data is subjected in the process computer to a two-dimensional Fourier-transformation, whereupon the image which has thus resulted is transmitted to the viewing screen apparatus 19.

The limitation to only one layer keeps the overall measuring time for an image brief, which is important for medical application. Moreover, the demands made of the dynamic (or contrast) range of the pre-amplifier 16 and the analog-to-digital converter, which are considered in FIG. 1 as part of the process computer, are smaller than in the case of methods which simultaneously utilize the signal from the entire volume range.

As proposed by A. Kumar, D. Welti, and R. R. Ernst (J. Magn. Res. 18 (1975) 69), the distribution of a nuclear resonance parameter can be calculated as a function of the location (x,y) by means of two-dimensional Fourier transformation from measured signal values S(t,N.sub.y), if said signal values represent the Fourier coefficients S(k.sub.x,k.sub.y) of this distribution. The variables k.sub.x and k.sub.y describe the wave numbers of the respective direction and are given by ##EQU1##

If one describes the phase position of the transverse magnetization-generated by the 90.degree.-Hf-pulse-of a volume element at the location (x,y) by a complex number in the reference system rotating with .gamma..multidot.B.sub.o, then measurement is to be conducted as follows:

The measurement of the complex signal S(k.sub.x, k.sub.y) proceeds in terms of apparatus through phase-sensitive rectification with the transmitter oscillator as reference. In order to obtain a real N.times.N-image with good sharpness and good contrast, a value supply as follows suffices:

One would have to take the real part of the complex two-dimensional-Fourier-transformed signal, the "absorption", as image value, and one would have to reject the imaginary component, the "dispersion", since the latter, on account of its broad wings, is not suitable for the imaging.

However, in practice, there is added to the above ideal signal an additional unknown phase factor which originates from basic field-inhomogeneities and -drifts and from non-perfect electronics adjustment. In the computer processing this effects a dispersive admixture to the real image value, and hence a quality deterioration.

It is therefore better to measure, also for k.sub.y, a complete value supply; i.e., double as many pulse trains

which is often in practice necessary anyway for the obtainment of a usable signal-to-noise ratio. The double Fourier transformation thus yields a complex image value with imaginary part equal to zero in the ideal instance; i.e., an absolute sum-resolved locally in a sharp and high-contrast manner-of the complex image value, in case the above-cited interferences cause a phase rotation or a phase shift as a function of frequency over the image plane.

The pulse train employed according to FIG. 2 permits the generation of the entire value range of the wave numbers k.sub.x, k.sub.y (thus also the negative ones) with only one polarity of the gradient fields. The electronics for the rapid switching of the gradient fields is very costly and would have to be separately installed per se for each polarity in case the polarity would have to be changed within a pulse train for the purpose of mensuration of positive and negative k.sub.x - and k.sub.y -components.

We shall now consider the generation of the desired k.sub.x and K.sub.y -values by means of the proposed pulse train in FIG. 3. The x-gradient pulse 35, 36, before the 180.degree.-Hf-pulse, with the period (N.sub.G.sup.- +N/2).multidot..DELTA.t, whereby .DELTA.t is a fixed time raster, produces a magnetization M.sub.xy.sup.- .about.exp (i k.sub.x.sup.- .multidot.x), with k.sub.x.sup.- =(N.sub.G.sup.- +N/2). .DELTA.t.multidot..gamma.G.sub.x where G.sub.x is the intensity of the x-gradient field. The first y-gradient pulse 40, or 42' (FIG. 5) respectively, additionally generates a phase/frequency characteristic in the y-direction:

in the first variant (FIGS. 4, 5). The subsequent 180.degree.-Hf-pulse 32 reverses, no matter what phase it has, the signs of the phases in the M.sub.xy -magnetization.

The y-gradient is switched on for the second time and delivers, at the end of the pulse 42, or 40', respectively, the desired wave number k.sub.y for N.sub.G.sup.- =N.sub.G.sup.+ :

The additionally introduced intervals N.sub.G.sup.+ and N.sub.G.sup.- 41 and 41' avoid the necessity of extremely brief pulses of the length 1.multidot..DELTA.t, 2.multidot..DELTA.t, etc., being switched; it is then, namely, technically difficult to exactly obtain the desired pulse effect .intg.G.sub.y (t)dt=G.sub.y .multidot.n.multidot..DELTA.t, since, in the case of brief time intervals, the gradient current is again already switched off before current and field have attained their stationary value. The additional intervals in the y-gradient pulse thus ensure the linearity of the progression (or course) of the wave number k.sub.y as a function of the variable clock pulse number N.sub.y.

The second x-gradient commences with a time interval N.sub.G.sup.+ .multidot..DELTA.t, referenced with 37 in FIG. 3, while the gradient field stabilizes itself and the preamplifier at the receiver coil recovers from the 180.degree.-Hf-pulse and the switching edge of the x-gradient. In the case of brief drop-off edges of the gradient pulses, N.sub.G.sup.+ =N.sub.G.sup.- ; otherwise, N.sub.G.sup.+ can be selected a few clock pulses .DELTA.t longer than N.sub.G.sup.- in order to compensate for the contribution of the drop-off edge of the prior x-gradient-forward pulse; instead, the scanning of the measured values can also be displaced (or shifted) by a corresponding time delay.

If the first measured value at commencement of the interval 38 is received, the magnetization has the form: M.sub.xy .about.exp [i .gamma.G.sub.y N.sub.y .multidot..DELTA.t.multidot.y].

and hence also the signal has the previously demanded form (or shape)

with the desired value range for k.sub.x, k.sub.y. For k.sub.x =0; i.e., at the time t=t.sub.E1, all volume elements relating to x are without phase shift, so that a signal maximum, a "nuclear spin echo" results.

The z-gradient 33 after the 180.degree.-Hf-pulse 32 serves the purpose of minimizing the phase shift over the layer thickness .DELTA.z in order to thus generate an echo which is as strong as possible. In the case of symmetric 90.degree.-Hf-pulses, this is accomplished by a negative z-gradient, or, as in our case, by a positive z-gradient after a 180.degree.-Hf-pulse, which is approximately half as long as the 90.degree.-Hf-pulse 31 (equal z-gradient field intensity assumed).

The utilization of a central 180.degree.-Hf-pulse for the generation of the signal has, in addition to the monopolar gradients, yet a further, very important advantage. At the time t.sub.E1 of the echo the phase error, through a basic field inhomogeneity .DELTA.B.sub.o, which added up to .gamma..multidot..DELTA.B.sub.o .multidot.t.sub.E1 /2 before the 180.degree.-Hf-pulse, has again become zero; likewise, the phase shift within a volume element, due to .differential..DELTA.B.sub.o /.differential.x, .differential..DELTA.B.sub.o /.differential..sub.y, .differential..DELTA.B.sub.o /.differential..sub.z, at this time is zero. The signal is thus measured with the lowest-possible phase errors, and therefore also with an intensity hardly impaired by the basic field inhomogeneity. The signal is attenuated only by the unavoidable phase shift which originates from the intended (or desired) gradient fields G.sub.x and G.sub.y.

Through the--if possible--simultaneous influence of the gradient fields G.sub.x and G.sub.y it has been successful in accordance with the invention to keep the pulse train until the echo, very brief. The signal is altogether essetially determined by the genuine transverse relaxation time T.sub.2 of the observed nuclei in their material surrounding:

Since, with each pulse train, the measured values are detected at the same time following excitation, all k.sub.y -Fourier components in the signal are also equally weighted.

Through an additional 180.degree.-Hf-pulse 34, FIG. 2, according to the above-described measuring operation, a second echo can be generated. Directly following this second 180.degree.-pulse, the magnetization has the form:

Symmetrically to the second 180.degree.-Hf-pulse the x-gradient is again switched on corresponding to the first readout interval (37', 38', 39'). Whereas the wave numbers k.sub.x again run from .gamma.G.sub.x .multidot.(-N/2).multidot..DELTA.t to .gamma.G.sub.x .multidot.(N/2-1).multidot..DELTA.t, as in the case of the first echo, the wave vector k.sub.y is now inverted, which must be taken into consideration in the computational processing. Thus, in the second echo the following results:

In the same manner, also additional echoes can then be generated, whereby the odd-numbered and the even-numbered echoes respectively comprise the same information.

In the case of Hf-pulses it is advisable, as is known, to change the sign from pulse to pulse in order that effects of small Hf field intensity deviations from the nominal value cannot be added up.

Since, in the generation of the wave number k.sub.y, only the difference between the y-gradient-pulse lengths before and after the first 180.degree.-Hf-pulse is decisive, further variants for this are conceivable. A second variant is illustrated in FIGS. 6, 7. Whereas, in the first variant, the y-gradients, at least N.sub.G, but otherwise as briefly as possible, are connected, in the case of the second variant, they are connected as long as possible; this is advantageous if the building-up time of the y-gradient field is very long. For positive k.sub.y -values the switching-on interval before the 180.degree.-pulse varies from N.sub.G.sup.- .multidot..DELTA.t to (N.sub.G.sup.- +N/2-1).multidot..DELTA.t, given a constant switching-on time of (N.sub.G.sup.+ +N/2).multidot..DELTA.t after the 180.degree.-pulse. Conversely, for negative k.sub.y -values, the switching-on period after the 180.degree.-pulse is altered from (N.sub.G.sup.+ +N/2-1).multidot..DELTA.t to N.sub.G.sup.+ .multidot..DELTA.t, whereas previously the switching-on time amounts constantly to (N.sub.G.sup.- +N/2).multidot..DELTA.t.

In FIG. 10 the third variant of a possible circuit diagram for the y-gradient coil is illustrated which can be employed instead of the variant 1 of FIGS. 4 and 5. Positive as well as negative wave numbers k.sub.y are attained through sole variation of the second y-gradient, whereas the first y-gradient before the 180.degree.-Hf-pulse exhibits constant length. In order to realize a total of N length gradiations, e.g. the time steps .DELTA.t=.DELTA.t/2 are selected half as long as in the case of variant 1; as compensation the amplitude G.sub.y of the y-gradient field must then be doubled: G.sub.y =2.multidot.G.sub.y, and in general, G.sub.y =2.multidot.G.sub.x. However, it is also possible to operate with an unaltered time raster and unaltered gradient intensity. Only then, the 180.degree.-Hf-pulse must be delayed by an interval N/2.multidot..DELTA.t, and the readout interval delayed by N.multidot..DELTA.t relative to the previous timing in order to still fully utilize all advantages of the spin echo method. The stepwise changing of the duration of the y-gradient solely after the 180.degree.-Hf-pulse is more favorable if the Hf field is inhomogeneous.

The doubling of the field amplitude of the y-gradient can also be avoided if its current supply is designed in bipolar fashion, such as is illustrated in variant 4 in FIGS. 11 and 12. Through the additional intervals N.sub.G.sup.- and N.sub.G.sup.+ the precise preparation of the small k.sub.y values is unproblematical. Naturally precaution must be exercised to ensure that the amount of the negative gradient current is equal to that of the positive.

A pulse train for the production of a tomographic (or layer) image according to the projection-reconstruction method is illustrated in FIGS. 8, 9. It results from the pulse train according to FIG. 2 for the 2D-Fourier method essentially through the omission of the y-gradient pulses. Instead of the x-gradient field, a gradient field of equal intensity, but variable direction .phi. is generated, whereby the angle .phi. covers the x-y-plane and, with each operating cycle, is increased by an amount .DELTA..phi.. In case the rotation of the .phi.-gradient field occurs through mechanical rotation of the coil arrangement, a monopolar current supply suffices. In the case of two coil arrangements orthogonal to one another, the one amplitude G.sub.x must vary from cycle to cycle with cos .phi., the other G.sub.y with sin .phi.. If .phi. covers the angular range of 0 through 180.degree., then the current supply of the x-gradient coil must be designed in bipolar fashion.

All advantages otherwise illustrated in the case of the 2D-Fourier method according to FIG. 2, which do not relate to the y-gradient, apply also to the application in the case of the projection-reconstruction-method. They are here listed only in the key-word fashion, since they have already been described therein: selective excitation of a z.sub.o -const layer including refocusing by a second, homopolarized z-gradient pulse.

Through the phase-sensitive mensuration of entire, symmetric echoes, the spectrum of the signal of an operating sequence (or cycle) is real under ideal conditions; in the case of interferences due to electronic phase errors and basic field-inhomogeneities, the sum of the complex spectrum can again be employed as real projection without deterioration of the imaging properties.

The additional intervals N.sub.G ensure the constancy of the gradient field intensity during the measured value detection. The signal intensity is, on account of the central 180.degree.-Hf-pulse, hardly attentuated by basic field inhomogeneities; several echoes can be measured in one operating sequence (or cycle). All described imaging methods can be modified to the extent that, through the selective 90.degree. pulse, several parallel layers are simultaneously excited. The sign, or differently expressed the phase position, of the excitation is then successively varied for the individual, excited layers within an otherwise equal measuring sequence in order, with the aid of suitable decoding algorithms, to be able to obtain the signal conributions of the individual layers from the overall signal. Since, in the case of N parallel, simultaneously excited layers, at least N measuring cycles with a different excitation sign for a decoding are necessary, one obtains no abbreviation of the overall measuring time in comparison with the successive measurement of the individual layers. However, one obtains a signal-to-noise ratio which is higher by the factor of .sqroot.N.

FIG. 13 illustrates as an example a suitable excitation sequence for the simultaneous mensuration of four layers. If S.sub.a -S.sub.d designate the aggregate signals obtained with these experiments, and S.sub.I to S.sub.IV designate the signals expected with excitation of each individual layer, then the following is valid:

Through application of the inventive excitation sequence one thus obtains, in four measurements, the signals of the four planes with a signal-to-noise ratio corresponding to an averaging over four measuring cycles. A generalization N simultaneously excited layers is readily possible corresponding to the method disclosed in the German OS 29 36 465 for N parallel lines in one plane.

It will be apparent that many modifications and variations may be made without departing from the scope of the teachings and concepts of the present invention.

With respect to the timing relationships in the various illustrative embodiments, the following may be noted:

(a) Scanning along successive lines of a selected layer is effected by the relative duration of an orthogonal field gradient (e.g. G.sub.y) before and after the brief 180.degree. high frequency pulse (e.g. 32, FIG. 2) in successive measurement cycles. The duration of the 90.degree. high frequency pulse in the modifications of FIGS. 2-5, 6-7, 10 and 11-12 may be about ten milliseconds, e.g. N.multidot..DELTA.t where N=128 and .DELTA.t=100 microseconds. The intervals between the center of the 90.degree. pulse at t-0 and the brief 180.degree. pulse at t.sub..pi.1 substantially corresponds to the interval between times t.sub..pi.1 and t.sub.E1.

(b) The intervals between t-0, t.sub..pi.1 and t.sub.E1 are relatively brief. For example, each interval may be of of the order 1.5N.multidot..DELTA.t (e.g. fifteen milliseconds).

(c) The orthogonal field gradients (e.g. G.sub.x and G.sub.y for the variants of FIGS. 3-5, 6-7, 10 and 11-12) may be switched on simultaneously following termination of the high frequency irradiation, and occupy a maximum time interval which may be less than the duration of the 90.degree. pulse. Thus the G.sub.x pulse 35,36 may be on continuously for a time interval of slightly more than N/2.multidot..DELTA.t, e.g. slightly more than five milliseconds.

(d) Multiple 180.degree. pulses may follow each other, e.g. at intervals of the order of 3N.multidot..DELTA.t or about thirty milliseconds.

The following documents are incorporated herein by reference to way of background:

U.S. Pat. No. 4,070,611 dated Jan. 24, 1978 and U.S. Pat. No. 4,390,840 dated June 28, 1983.