Method of magnetic resonance with excitation by a prewinding pulse

Abstract

A method of magnetic resonance, in which a sample introduced in a measurement volume in an external magnetic field is excited by an excitation pulse and the signal formed by the transverse magnetization thus produced is read out by a receiving coil. The method is characterized in that a prewinding pulse is used as the excitation pulse, which prewinding pulse is characterized in that the formed transverse magnetization M.sub.() of spins of different Larmor frequency after the pulse has a phase .sub.0(), wherein .sub.0() as a function of within a predefined frequency range has an approximately linear course having negative slope, such that the spins refocus after an echo time defined by the pulse without an additional refocusing pulse being necessary.

Claims

1. A method of magnetic resonance for examination of a sample introduced into a measurement volume within an external magnetic field, the method comprising the steps of: a) exciting the sample using an excitation pulse, wherein the excitation pulse is a prewinding pulse generating a transverse magnetization M.sub.() of spins of different Larmor frequency after the excitation pulse, the transverse magnetization having a phase .sub.0(), wherein .sub.0(), as a function of within a predefined frequency range , has an approximately linear course with negative slope, such that spins refocus to form a spin echo after an end of the excitation pulse at an echo time, wherein the echo time is defined by the excitation pulse and is larger than the duration of the excitation pulse, with no additional refocusing pulse or refocusing gradient thereby being necessary; and b) reading out, after a time interval and using a receiving coil, a signal formed by the transverse magnetization produced in step a).

2. The method of claim 1, wherein a time profile of an amplitude and a phase of the prewinding pulse is calculated by solving Bloch equations inversely and the prewinding pulse is produced accordingly.

3. The method of claim 1, wherein the prewinding pulse is calculated using a Shinnar-Le Roux method.

4. The method of claim 1, wherein the prewinding pulse is calculated by an optimal control method.

5. The method of claim 1, wherein the signal is read out spatially encoded in 1, 2, or 3 dimensions by means of time-variable magnetic field gradients, wherein a spatial encoding is performed using at least one of a conventional method, Fourier encoding, radial imaging and spiral imaging.

6. The method of claim 1, wherein the signal is read out after a delay, thereby undergoing a defined, freely selectable T2*-dependent modulation of amplitude and/or phase.

7. The method of claim 5, wherein a produced signal is refocused and read out one or more times at incremental readout times using single or multiple gradient inversion.

8. The method of claim 6, wherein a produced signal is refocused and read out one or more times at incremental readout times using single or multiple gradient inversion.

9. The method of claim 1, wherein the prewinding pulse is applied under a slice selection gradient of strength Amp, wherein the slice selection gradient is switched off after an end of the prewinding pulse and in such a manner that a time integral under the gradient corresponds to an action integral Amp*TE required for self-refocusing.

10. The method of claim 9, wherein a formed signal is read out under a readout gradient GR.

11. The method of claim 1, wherein both the prewinding pulse and the signal occur under a spatially encoded gradient GR.

12. The method of claim 11, wherein repetition of acquisition is carried out in a spatial direction other than 2 or 3-dimensional radial spatial encoding.

13. The method of claim 11, wherein repetition of acquisition is performed with different phase encoding steps between excitation and data readout.

14. The method of claim 9, wherein the prewinding pulse is switched under a time-variable gradient.

15. The method of claim 14, wherein the prewinding pulse is used under a time-variable gradient according to a VERSE method.

16. The method of claim 1, wherein the excitation pulse is chosen to initially take the magnetization to flip angles defined by .sub.z<(t)<n.sub.z.

17. The method of claim 1, wherein the excitation pulse is chosen such that, with a limited maximum pulse amplitude, Rmax()<Rmax(90) is achieved with a negative sign.

18. The method of claim 1, wherein the excitation pulse is chosen to have a constant amplitude and one or more phase jumps.

19. The method of claim 1, wherein an amplitude of the excitation pulse is shaped and changes sign at one or more points.

Description

BRIEF DESCRIPTION OF THE DRAWING

(1) The figures show:

(2) FIG. 1a a schematic view of the pulse sequence of a spin (Hahn) echo sequence, in which the excitation pulse (exc) is a rectangular pulse of negligible length. The accumulated phase ramp is inverted by the refocusing pulse (ref) so that an echo arises at the echo time TE;

(3) FIG. 1b the pulse diagram as in FIG. 1a, but with a Sinc pulse of finite length. Already at the end of the pulse, the spins have accumulated a phase ramp of d/d==tp/2, wherein tp is the length of the pulse;

(4) FIG. 2 the basic principle of the inventive method. A (still to be defined) prewinding pulse Pprew generates transverse magnetization with a linear phase course with a negative slope with reference to the frequency dependence. Due to free induction decay, the magnetization of the excited isochromates develops so that an echo is formed at instant TE, without a further refocusing pulse;

(5) FIG. 3a exemplary inventive pulses that have been calculated in small-angle approximation by regularized matrix inversion. If a larger regularization parameter is chosen, the energy of the B.sub.1 field is reduced;

(6) FIG. 3b-d the frequency response of the inventive pulses. The magnetization has a negative phase ramp (c) in the target range 1000 rad/s1000 rad/s. The linearity thereof declines with larger regularization parameters. The echo time in (d) is calculated from the negative slope TE=d.sub.0/d;

(7) FIG. 4 an exemplary sequence for use of the prewinding pulse in conjunction with an imaging method of 3D Fourier encoding. Rf characterizes the prewinding pulse Pprew, which in this case acts on the entire sample, and the self-refocused signal produced after TE, GR the readout, and GP1 and GP2 the two phase encoding gradients;

(8) FIG. 5a a sequence as in FIG. 4, but with delayed readout for production of a signal with defined T2* weighting, which can be varied in any way;

(9) FIG. 5b a sequence as in FIG. 5a, but with multiple readout of the produced self-refocused signal at incrementally shifted echo times;

(10) FIG. 6a the use of the prewinding pulse Pprew in conjunction with a slice selection gradient GS. With a still constant gradient GS after the end of the pulse, the echo is produced at instant TE. Self-refocusing occurs at instant TE=2TE with linear switch-off of the gradient;

(11) FIG. 6b a sequence as in FIG. 6a, but with spatially encoded readout of the signal as a half-echo under a readout gradient GR. In addition, phase encoding was possible during run-down of the slice gradient;

(12) FIG. 7a a sequence during which the pulse and readout of the signal are performed under a readout gradient GR. The direction of GR can be varied so that the 2 or 3-dimensional k-space is radially sampled. This corresponds to a spherical field of view;

(13) FIG. 7b a sequence such as in FIG. 7a, with the difference that k-space sampling is not performed radially but in the same direction as GR and the other spatial directions are phase-encoded. In this case, a slice is excited in the direction of GR whose thickness results from and the gradient strength of GR;

(14) FIG. 8 an exemplary modulation of the phase course outside the target range . Due to the Fourier shift theorem, echo formation occurs at different instants, depending on the slope d.sub.0/d. In the case of a positive slope, the echo time is, theoretically, before the end of the pulse and therefore not accessible. In the case of the pulses in a) and b), the continuous line represents the real and the dashed line the imaginary part of the pulse. In the associated frequency responses in c) and d), the continuous line represents the absolute value and the dashed line the phase of the frequency response;

(15) FIG. 9 a generic pulse, which consists of two hard pulses with opposite signs;

(16) FIG. 10 the effect of the pulse of FIG. 9. The magnetization is initially tilted into the transverse plane, where it is dephased. A 90 pulse guides the magnetization back into the x-z plane;

(17) FIG. 11A the magnetization from FIG. 10 in the transverse representation for the 90-(90 (M1) and the 90-(100 (M2) pulse combination;

(18) FIG. 11B the magnetization from FIG. 11A)(90-(90 combination) after an evolution time of /(2.Math.). The semicircle represents the precession path of the outermost isochromates ();

(19) FIG. 11C the magnetization from FIG. 11A)(90-(90 combination) after an evolution time of TE1.3.Math./(2). The semicircle represents the precession path of the outermost isochromates ();

(20) FIG. 11D the signal that results from the complex sum of the magnetization M1 or M2 during free precession. The echo (highest point of the signal) is reached at a multiple of the pulse length (1 ms);

(21) FIG. 12A the echo time as a function of the pulse length for the described combination of two hard pulses. It is easy to see that, in the range tp<1, the echo time hardly depends on the pulse length. For longer pulse durations, the behavior slowly transitions to Hahn echoes with TE=tp;

(22) FIG. 12B the part relevant for the inventive pulse from FIG. 12A enlarged. Both the 90-(90 (TE1), and the 90-(100 combination (TE2) are shown. It can be seen that their behavior for tp<<1 greatly differ from each other. However, for 0<tp<0.7, both curves are above the original straight lines of slope 1;

(23) FIG. 12C the signal intensity at the echo time for the two hard pulse combinations. It can be seen that I2 (90-(100 achieves higher values in total. This can be understood as arising because the total flip angle for the on-resonant spins is not equal to zero;

(24) FIG. 12D the dimensionless factor R, for comparison with the ideal Hahn echo (R=1). For tp<0.7, echo times can be produced that are greater than the length of the pulse;

(25) FIG. 13 a generic pulse, which is composed of a /2 and a /2 pulse (a) with .sub.1=B.sub.1. The result is a flip angle of 0 for the on-resonant spin isochromate and a linear rise of the flip angle with the off-resonance frequency (b). The phase makes a jump of , so that the isochromate pairs comprising and gradually converge as a function of their absolute value (c). The instant of the maximum signal (echo) primarily depends on the available frequencies w;

(26) FIG. 14 a generic pulse, which is composed of a /2 and a /20.2 pulse (a). Only the real part of the pulse is shown, the imaginary part is negligible in the case of this special pulse, which is not necessarily the case for all inventive pulses. The result is a flip angle of 3 for the on-resonant spin isochromate and a rise of the flip angle with the absolute value of the off-resonance frequency (b). The phase now no longer makes a jump, as compared with FIG. 4, but describes a steep edge (c). The normalized slope R achieves values |R|>10, so that the spins refocus near to the resonance frequency only a good 5 ms thereafter (d);

(27) FIG. 15 the principle of inventive pulses: The phase distance of the spins, which were accumulated near to the equator (A), is enlarged by the transition to smaller flip angles (B). In (B), the projection of the magnetization onto the equator is additionally shown to illustrate the enlargement of the split (dashed). If the magnetization is guided through the pole, the phase course is inverted, which results in the formation of an echo (dotted); and

(28) FIG. 16 a pulse, which was calculated by means of Optimal Control (a) in a limited range of off-resonance frequencies, it shows an approximately linear phase course at the end of the pulse for an approximately constant flip angle (b, c). In the optimized frequency range, the |R| values are slightly above 1 (d).

DESCRIPTION OF THE PREFERRED EMBODIMENT

(29) As already mentioned above, the inventive method is based on the concept of a spin echo formation without a refocusing pulse, which is achieved by using a special excitation pulse, which produces dephasing of negative slope with respect to the frequency (d.sub.0/d<0) over a desired frequency range . After excitation with such a pulse, rephasing of the spins is performed with free precession and, without a refocusing pulse, a spin echo is formed at the echo time TE=d.sub.0/d, which is thus defined by the pulse.

(30) This principle of the invention is outlined in FIG. 2.

(31) Such a pulse designated phase pre-winding pulse can be generated by solving the Bloch equation for the stated conditions, wherein in the following consideration, T1 and T2 relaxation are ignored and only the free precession of the magnetization in the presence of the static magnetic field inhomogeneities and the effects caused by the time-variable B.sub.1 field of the excitation pulse are considered:

(32) In the so-called small-angle approximation, the non-linear Bloch equation can be linearized. In this approximation, a Fourier relation exists between the pulse and its frequency response. If the relaxed magnetization {right arrow over (M)}.sub.0=(0,0,1).sup.T is assumed before the pulse, the following results for complex transverse magnetization M.sub.=M.sub.x+iM.sub.y after the pulse
{right arrow over (M)}.sub.=E.Math.{right arrow over (x)}.(4)

(33) The vector {right arrow over (M)}.sub. has the elements M.sub.,k=|M.sub.|.Math.exp(i.sub.kTE), E is a matrix with the elements E.sub.k,n=exp(i.sub.knt) and {right arrow over (x)} describes the sought pulse. .sub.k extends in sufficiently small steps to satisfy the Nyquist criterion over the frequency range , in which the pulse is to achieve the desired effect. The absolute value of the transverse magnetization results from the desired flip angle |M.sub.|=sin(). The so-called dwell time t describes the digital sampling rate of the variable B.sub.1 field so that the pulse {right arrow over (x)} has length n.Math.t. Because E is generally not quadratic and is ill-conditioned, a regularized Moore-Penrose pseudoinverse is used to provide the solution.
{right arrow over (x)}=E.sup.H(EE.sup.H+.sup.21).sup.1{right arrow over (M)}.sub..(5)

(34) The Tikhonov regularization used here minimizes the irradiated energy of the B.sub.1 field. The regularization parameter trades the accuracy of the inversion off against minimization of the energy of the B.sub.1 field.

(35) The solution obtained in this way can be used either directly as a pulse in linear approximation, or be applied to the non-linear Block equation by means of the Shinnar-Le Roux method to solve the equation exactly (see reference [9]).

(36) Alternately, it is possible to calculate inventive pulses by means of the Optimal Control method (see reference [3]). The pulse is represented as a control that is optimized to minimize a cost function.

(37) The desired frequency response can optionally be exactly satisfied, or the distance from the desired frequency response can be minimized. Satisfying the Bloch equation is ensured by a Lagrange multiplier. The Bloch equation is therefore also exactly satisfied in the optimal control method.

(38) FIG. 3(a) shows a solution in small-angle approximation by way of example. In this case, 1000 rad/s.sub.k1000 rad/s, |M.sub.|/M.sub.0=0.1, and TE=1.5 ms (measured from the end of the pulse) were selected. The dwell time is 1 s and the pulse length 4 ms. FIGS. 3(b-d) show the frequency response of the pulse that was calculated by means of Bloch simulations (see reference [9]).

(39) In the frequency response in FIG. 3(b), the target range is a local minimum with respect to the flip angle. Outside this frequency range, the flip angle is comparatively large. This means the longitudinal magnetization is largely destroyed and can scarcely recover with short repetition times (TR<<T1). This can be used in a targeted way to suppress signals of unwanted Larmor frequencies (e.g. fat using the chemical shift). Analogously, it is possible to choose a very small flip angle for unwanted frequency ranges, which also causes signal suppression in these ranges. Optimization of the inventive pulses for suppression of certain Larmor frequencies can be achieved, for example, by additional parameters in the cost function (in all methods presented), or by Lagrange multipliers.

(40) A preferred group of applications of the inventive method is shown in FIG. 4. A global prewinding pulse is used in conjunction with three-dimensional Fourier spatial encoding. Here, Rf shows the prewinding pulse Pprew and the echo formed after time TE. GR corresponds to the readout gradient for spatial encoding along the freely selectable first spatial coordinate, GP1 and GP2 the phase encoding gradients in the two other orthogonal spatial directions. If the Pprew pulse were replaced with a conventional excitation pulse, this would be the same as a 3D gradient echo method.

(41) In a preferred implementation of this method, &is chosen such that it covers the entire range of the variation of the Larmor frequency of the nuclear spins caused by field inhomogeneities in the volume to be imaged. Unlike conventional gradient echo methods in which the formed signal is phase- and amplitude-modulated by T2* decay caused by inhomogeneities, with use of Pprew, the formed signal is refocused with respect to the field inhomogeneities. Such an implementation is especially suitable for examination of inhomogeneous samples without T2* signal loss.

(42) By analogy with the methods shown, many of the other methods known from the literature for spatial encoding (e.g. radial imaging, spiral imaging, and others) can be combined with the inventive prewinding pulse.

(43) The instant of spontaneous refocusing TE does not necessarily have to match the gradient echo instant caused by inversion of the readout gradient (FIG. 5a). In the case of delayed formation of the gradient echo at an instant TE, a signal is formed that is influenced in its amplitude and/or phase by T2* and/or chemical shift within the time difference TE-TE. This can be used, in particular, to implement fast data readout with a small time difference TE-TE, which is not accessible to direct readout. Due to variation of TE or also due to multiple inversion of the readout gradient GR (FIG. 5b), multiple signals with different signal modulation caused by T2* can be produced, according to the principle of the multipoint Dixon (see references [4] and [12]) or also PEPSI method (see reference [10]), and parameter maps of the local field homogeneity and/or the local T2* times can thereby be determined. Unlike conventional methods, in this case the zero point of the T2* development is directly accessible to measurement by selection of a TE=TE.

(44) In a further preferred implementation of the methods shown in FIGS. 4 and 5, is chosen such that it covers the range of the resonance frequency of water or fat. The formed signal is then rephased at the echo time instant TE only for the desired range (water or fat): signals of the unwanted range are dephased and, depending on the amplitude profile in the relevant frequency range, suppressed.

(45) A further group of preferred applications is shown in FIG. 6. Here, the prewinding pulse Pprew is used in the presence of a slice selection gradient with amplitude Amp. If the gradient is left unchanged and constant after the end of the pulse (dotted line in FIG. 6a), the self-refocusing is performed at instant TE. If the gradient is switched off after the pulse (t=0), self-refocusing is performed at an instant TE, which is then reached when the area under the gradient between 0 (end of the pulse) and TE corresponds to the area under the constant gradient until instant TE:

(46) $\begin{array}{cc}{}_0^{{\mathrm{TE}}^{}}\mathrm{GS}\left(t\right)\mathrm{dt}=\mathrm{Amp}.Math.\mathrm{TE}.& \left(6\right)\end{array}$

(47) With linear decline of GSshown in FIG. 6aTE=2 TE. However, in this case, only the variations of the Larmor frequency are refocused due to the slice gradient at instant TE. Signals with off-resonances undergo T2*-dependent signal modulation in accordance with the time difference TE-TE.

(48) Such an implementation is especially advantageous for achieving very short echo times TE. In a magnetic resonance tomograph, a certain dead time between the end of the RF pulse and the beginning of data readout is generally necessary due to the technical characteristics of the coils and receiving electronics. The dead time is typically in the range of a few microseconds to approx. 100 s. Even if extremely short, hard pulses are used, with subsequent readout of the formed signal as free induction decay (FID), the instant of signal coherence (=start of the FID) is not read out, which results in T2*-dependent signal attenuation. Moreover, the spectrum formed by Fourier transformation of an incomplete FID exhibits a complex, frequency-dependent phase that is difficult to correct.

(49) Due to use of a prewinding pulse, the FID can be read out completely. This permits spectroscopic measurements with very short echo times TE.

(50) FIG. 6b shows the method shown in FIG. 6a in conjunction with a spatial encoding gradient GR. This is switched on at instant TE; the signal is therefore read out as a so-called half echo. For 2 (or 3) dimensional spatial encoding, the principle of radial spatial encoding is advantageously used, wherein the direction of the readout gradient is continuously changed in successive acquisition steps.

(51) FIG. 7 finally shows an implementation of the method under a constant readout gradient GR. The spectrum generated by Fourier transformation of the formed echo exhibits a constant phase over the frequency range , wherein, over the gradient GR, corresponds to a range r in real space. is thereby chosen such that r is located symmetrically around the zero point of the gradient. If the acquisition is repeated while continuously changing the direction of the gradient GR according to 3-dimensional radial spatial encoding, a three-dimensional data set can be produced (FIG. 7a). For localized imaging in a spherical volume with diameter r, the acquisition can be performed with undersampling (see reference [13]). Signals outside this volume appear dephased and weakened by dephasing and can be ignored. In particular, spin echo type acquisitions with very short echo times can thereby be produced. In addition to spatial encoding with isotropic angle distribution of the acquired projections, in an analogous way to known methods, this method is suitable for fast acquisition, possibly with anisotropic distribution of the spatial angles (golden angle) (see reference [11]). With use of fast repetition and a pulse profile of the prewinding pulse with large flip angles outside (as in FIG. 3), additional signal saturation can be achieved in the unwanted marginal range.

(52) Moreover, GR can be left constant in one direction. In this case, a slice is excited whose thickness results from and the gradient strength. The two remaining spatial directions can be encoded with phase encoding between excitation and data readout (FIG. 7b). During phase encoding, the readout gradient GR can optionally remain switched on or be switched off for a short time. The echo time results depending on the area under GR.

(53) The previous description was limited to discussion of the possibilities of prewinding pulses with reference to the characteristic feature of the negative slope of the phase course in the range . Methods are described below that are concerned with controlled modulation of the frequency ranges outside .

(54) As shown in FIG. 8, with relevant specifications of the achieved phase and amplitude course outside , numerous pulses can be generated that exhibit an identical phase and amplitude course within . To suppress the generally unwanted signals outside , it is generally advantageous to achieve the strongest possible and positive slope of () for these ranges while at the same time keeping the amplitude low. As already described, use of a profile with high amplitude in the unwanted range in conjunction with short repetition times TR during data acquisition can be used for signal saturation outside .

(55) With use of small-angle approximation (linear approximation of the Bloch equation)as in the examples shownthe duration of the pulse is directly correlated with the slope d.sub.0/d in the range due to the properties of the Fourier transformation. Pulses with short duration and large negative slope d.sub.0/d can be generated with the non-linear methods (optimal control, etc.) already mentioned.

(56) It should also be pointed out that, in those methods in which the prewinding pulse is used below a spatial encoding gradient (FIGS. 6 and 7), a change of the pulse and gradient shape according to the principle of the VERSE method (see reference [3]) can be used to minimize either the pulse duration or possibly also the maximum pulse power.

(57) A further preferred implementation of an inventive pulse will first be explained by means of a generic pulse that is composed of two hard pulses. The transition to more practical pulses will then be illustrated, including an estimation of the possibilities and limits of inventive pulses.

(58) A generic pulse consists of the sequence of 2 radio-frequency pulses RF.sub.1 and RF.sub.2 with flip angle .sub.1 and .sub.2 and negligible length, these having a distance tp (FIG. 9) between them. This is to act upon a set of spins, which are distributed over a frequency interval =.sub.0+ with constant amplitude |M()|, wherein is small compared to /tp.

(59) Initially, case 1 is considered with .sub.1=90 and .sub.2=90 with negligible length of the single pulses. As shown in FIG. 10, the z-magnetization (FIG. 10A) is initially rotated through RF.sub.1 to the y-axis (FIG. 10B). The isochromates fan out in the following time tp by an angle .sub.tp=+*tp (FIG. 10C). Due to the pulse RF.sub.2, the magnetization is rotated back in the z-direction again (FIG. 10D). As can be seen from FIG. 10D, the total magnetization is now in the x-z plane.

(60) The further development of the magnetization is considered in the transverse plane. FIG. 11A shows the magnetization M1(0) at the end of the total pulse (t=0) according to FIG. 10D. The two end points according to =.sub.0+ are drawn as thick dots for better illustration. In time tc=/(2), the two isochromates accumulate, with maximum frequency (+), a phase of /2 and meet on the x-axis (FIG. 11B). The semicircle corresponds to the precession path of the isochromates with +. Isochromates with lower off-resonance frequency precess correspondingly more slowly. The total signal according to the sum of the transverse component of all isochromates is I1(tc). By numerical or analytical solution of the signal equation it can be shown that the maximum of the total signal and therefore the intensity I1(TE) of the spontaneous echo is achieved at an instant t=TEN 1.3*tc (FIG. 11C).

(61) From this illustration, it is immediately clear that TE depends exclusively on and not on the pulse duration.

(62) FIG. 11D shows the development over time of the signal intensity I1 for =2=628 rad/s for tp=1 ms. The signal maximum is reached at instant TE=6.64 ms; the signal intensity is approx. 13% of the available magnetization.

(63) FIG. 12A shows the dependency of the echo time TE on the pulse duration tp. The pulse duration tp is expressed in units of 2/. It becomes clear that the derivation shown in FIGS. 11 and 12 for pulse durations in the range of weak dephasing (tp0.7.Math.2/) is valid. With stronger dephasing during tp, the echo time approximates to the Hahn echo regime and TE=tp applies.

(64) The range of weak dephasing that is relevant for the inventive method is shown magnified in FIG. 12B.

(65) FIG. 12C shows the echo amplitude I1(TE) as a function of . For strong dephasing, this approximates to the value of 50% expected for a refocusing pulse of 90.

(66) FIG. 12D shows the dimensionless factor R=tp/TE as compared with the formation of a conventional spin echo (TE=tp).

(67) The illustration shown so far permits an intuitive understanding but is not necessarily an optimum implementation. As shown in FIG. 12C, the signal intensity is very small. An increase in the echo amplitude can be achieved easily by selecting a magnitude of .sub.2>90 (.sub.2 smaller than 90 therefore). As shown in FIG. 10A for .sub.2=100, the magnetization M2(0) formed at instant t=0 already exhibits an x-magnetization differing from 0, which results in an increase of the signal I2 subsequently formed (FIG. 10C). Because the angle of aperture of the isochromates at .sub.2=90 is smaller than case 1 and therefore smaller than 180, the echo time is reduced. In FIGS. 12B and D, this is shown based on the diagrams for TE2 vs. tp or R2 vs. tp (dashed lines).

(68) For a given =2 and a desired echo time TE, the flip angle .sub.2 that is optimal for the specific application can be calculated in this way.

(69) In the previous consideration, the pulses were assumed to be delta pulses. In practical implementation, this is impossible. An efficient implementation is achieved with a pulse with constant amplitude and a total length 2*tp, at which .sub.1=.sub.2=90 applies, so that the phase pulse is shifted by 180 in the middle of the pulse. Such a pulse is depicted in FIG. 13. .sub.1 corresponds therein to the nutation frequency and, for the pulse shown, .sub.1.Math.tp=/2 applies.

(70) Of course, other flip angles can be chosen for both of the pulse parts. FIG. 13b, c shows the resulting flip angle and the phase as a function of the off-resonance frequency. In accordance with FIG. 11A, the latter exhibits a jump by in the on-resonance frequency.

(71) Because, for the inventive implementation, tp<<2/ is chosen, the bandwidth of the selected pulses is sufficiently large so that the calculations derived from the hard pulse approximation and shown in FIG. 10-12 still apply as a good approximation. However, the end of the pulse is shifted by tp/2 and TE is therefore shortened by tp/2.

(72) The previous calculations refer to the case of magnetization that is constant in the range =.sub.0+; such a case can, for example, be achieved by applying a constant gradient over a homogeneously filled vessel.

(73) If the distribution of the isochromates arises due to local field inhomogeneities, this is generally not satisfied; for stochastic local field changes, a Cauchy distribution of the isochromates is frequently assumed. This changes the numeric values of the calculations shown, but the line of argument is still valid in qualitative terms. By determination of TE for given parameters tp, .sub.1, and .sub.2, a measure of the type of microscopic field distribution can be found.

(74) In an analogous way to the hard pulse combination 90-(100, the pulse from FIG. 13 can be modified to become a /2(/20.2) (FIG. 14). In this case, the resulting flip angle for all isochromates is greater than zero and the phase does not exhibit a jump but a continuous ramp. For dimensionless description, the factor R stated above can also be described by the slope of the phase as a function of the frequency:

(75) $R=\frac{1}{\mathrm{tp}}\frac{d{}_0}{d}.$

(76) Therein, .sub.0, denotes the phase at the end of the pulse. If R>0, this means that all isochromates are in phase after the excitation. If R<0, a spin echo is formed at instant TE=R.Math.tp. In the regime of the Hahn echo, R=1 always applies for delta pulses and R>1 always applies for real pulses.

(77) The principle of the inventive pulses can generally be described by the fact that the magnetization is not directly taken to the desired target flip angle .sub.z, but to flip angles (t), which are defined by .sub.z<(t)<n.sub.z. At these flip angles, the spins dephase. As shown below, the dephasing can be strengthened by the transition to smaller flip angles .sub.z. If the magnetization ramp d/d is inverted, a spin echo arises at the end of the pulse. The echo time is not limited by a lower limit because complete dephasing is not a precondition for it (cf. Hahn echoes). As is shown below, the echo time (measured as of the end of the pulse) is also not limited to TE<T.sub.P if the target flip angle .sub.z/2 is selected. T.sub.P describes the length of the pulse.

(78) To estimate the possible echo times, the on-resonant spin isochromate in the rotating reference system at .sub.0=0 and a further isochromate with frequency on the Bloch sphere with radius 1 are considered. For geometrical considerations, B.sub.1 fields of any strength must be able to take effect. The flip angle can therefore be changed at any rate. However, all B.sub.1 fields act on all isochromates. A distance between isochromates as defined by the Euclidean norm can therefore only be achieved by precession of the spins as a consequence of their different Larmor frequencies and can therefore only be indirectly influenced.

(79) In spherical coordinates, the position of each isochromate on the Bloch sphere can be described by the flip angle and the phase .

(80) Irrespective of the flip angle, the two isochromates are separated after an infinitesimal time step dt by the phase
d=.Math.dt.

(81) Through a change to the Cartesian coordinate system, the distance between the two isochromates in the Euclidean norm is described by
d=.Math.dt sin
where describes the flip angle at which d was produced. By derivation, it can be shown that the greatest change in the distance d is achieved at .sub.max=/2. In this case, the following applies
d.sub.max=.Math.dt.

(82) If the magnetization is now folded toward the target flip angle .sub.z/2 and the Euclidean distance between the two isochromates is again described in spherical coordinates, the maximum achievable phase difference yields

(83) $d{}_{\max }\left({}_z\right)=\frac{.Math.\mathrm{dt}}{\sin {}_z}.$

(84) This calculation has been set up for infinitesimally small times of free precession. It retains its validity as long as <<sin , is satisfied. It must be understood as the upper limit for finite times and larger produced phase differences. The sign of the phase is irrelevant because, in this theoretical estimation, any speed of rotation of the magnetization is permitted and the sign can be changed at any rate. In the description above, delta-shaped pulses, separate from a time of free precession are assumed. This does not entail any loss in generality, because any pulse in a limit value consideration can be seen as a concatenation of infinitesimally small delta-shaped pulses followed by an infinitesimally small time of free precession.

(85) Because of this geometrical estimation, it can be seen that, for .sub.z=/2, no more than |R|=1 can be achieved. However, for smaller flip angles larger values are also possible, i.e. a pulse of length T.sub.P can generate a phase ramp that generates a spin echo after a time TE>T.sub.P after the end of the pulse. However, the estimation only applies to nn2 , and not therefore to the pulses in FIGS. 9 and 13, which are a limiting case.

(86) Inventive pulses make use of the described effect by initially taking the magnetization to flip angles close to /2, where a large Euclidean distance quickly arises between the isochromates. The magnetization is then taken to the target flip angle .sub.z, which is further from /2. This results in phase ramp d/d with a large absolute value. This concept is illustrated in FIG. 16. A sign inversion can be achieved, for example, by taking the magnetization to the negative flip angle .sub.z and therefore inverting the phase ramp. Due to free precession, the negative phase ramp then automatically results in a spin echo whose instant is determined by the slope of the phase ramp.

(87) To achieve complete rephasing of all spins with R=1 with a flip angle of .sub.z=/2, magnetization must first be excited around /2 and then refocused with a pulse. Overall, therefore, a total rotation of 3/2 is required. With very short total pulse lengths and limited B.sub.1 amplitude, this total rotation can no longer be achieved. In limiting cases, in which the required rotation can just about be performed, the individual excitation and refocusing components are so long that R>>1 results. By choosing flip angle .sub.z/2, inventive pulses make it possible to break through these limits with finite B.sub.1 fields and to achieve longer echo times than were previously possible in prior art.

(88) A flip angle that is as constant as possible over the entire relevant frequency spectrum and simultaneous refocusing of all spins is usually desired. These conditions are only very poorly met by the generic pulses described so far. However, the principle can easily be applied to shaped pulses, which are optimized with respect to constant flip angle and linear phase course. By way of example, FIG. 16 shows a pulse that has been optimized by means of the optimal control method in a range of =1250 rad/s onto a flip angle of 3 and R=1.3. The pulse duration is 0.5 ms.

LIST OF REFERENCES

(89) [1] Bloom, A. L., 1955. Nuclear Induction in Inhomogeneous Fields. Physical Review 98, 1105-1111. [2] Conolly, S., Nishimura, D., Macovski, A., 1986. Optimal control solutions to the magnetic resonance selective excitation problem. IEEE Trans Med Imaging 5, 106-115. [3] Conolly, S., Nishimura, D., Macovski, A., Glover, G., 1988. Variable-rate selective excitation. Journal of Magnetic Resonance (1969) 78, 440-458. [4] Glover, G. H., 1991. Multipoint dixon technique for water and fat proton and susceptibility imaging. Journal of Magnetic Resonance Imaging 1, 521-530. [5] Hahn, E. L., Maxwell, D. E., 1952. Spin echo measurements of nuclear spin coupling in molecules. Physical Review 88, 1070. [6] Hennig, J., Nauerth, A., Friedburg, H., 1986. RARE imaging: a fast imaging method for clinical MR. Magn Reson Med 3, 823-833. [7] Kiliptari, I. G., Tsifrinovich, V. I., 1998. Single-pulse nuclear spin echo in magnets. Physical Review B 57, 11554-11564. [7a] K. O. Lim, J. Pauly, P. Webb, R. Hurd, A. Macowski, Short TE Phosphorus spectroscopy using a spin-echo pulse, MRM 32, 98-103 (1994). [7b] P. Balchandani, M. Yamada, J. Pauly, P. Yang, D. Spielman, MRM 62, 183 (2009). [7c] X. Wu, P. Xu, R. Freeman, Delayed Focus Pulses for MRI: An evolutionary Approach Mag. Res. in Med. vol. 20 pp. 165-170 (1991). [7d] J. Shen, Delayed-focus pulses optimized using simulated annealing, J. Magn. Reson. 149, 234-238 (2001). [8] Meyer, C. H., Pauly, J. M., Macovskiand, A., Nishimura, D. G., 1990. Simultaneous spatial and spectral selective excitation. Magnetic Resonance in Medicine 15, 287-304. [9] Pauly, J., Le Roux, P., Nishimura, D., Macovski, A., 1991. Parameter relations for the Shinnar-Le Roux selective excitation pulse design algorithm [NMR imaging]. IEEE Trans Med Imaging 10, 53-65. [10] Posse, S., Dager, S. R., Richards, T. L., Yuan, C., Ogg, R., Artru, A. A., Mller-Gartner, H.-W., Hayes, C., 1997. In vivo measurement of regional brain metabolic response to hyperventilation using magnetic resonance: Proton echo planar spectroscopic imaging (PEPSI). Magnetic Resonance in Medicine 37, 858-865. [11] Prieto, C., Uribe, S., Razavi, R., Atkinson, D., Schaeffter, T., 2010. 3D undersampled golden-radial phase encoding for DCE-MRA using inherently regularized iterative SENSE. Magnetic Resonance in Medicine 64, 514-526. [12] Reeder, S. B., McKenzie, C. A., Pineda, A. R., Yu, H., Shimakawa, A., Brau, A. C., Hargreaves, B. A., Gold, G. E., Brittain, J. H., 2007. Water-fat separation with IDEAL gradient-echo imaging. Journal of Magnetic Resonance Imaging 25, 644-652. [13] Scheffler, K., Hennig, J., 1998. Reduced circular field-of-view imaging. Magnetic Resonance in Medicine 40, 474-480. [14] U.S. Pat. No. 7,425,828 B2 [15] US 2005/0127911 A1