GB2337125A - Acoustic noise reduction in MR tomography - Google Patents

Abstract

A multislice NMR tomography method in which gradient echo signals are detected is made quieter by reducing the Lorentz forces, and hence vibrations, in the gradient coils. This is achieved by applying time varying gradients such that the strength of the magnetic field gradients is effected with a sigmoid function, ie the second derivative of the sigmoid function is constant throughout the switching time of the gradients, or at least has no spikes in its changing time dependence.

Description

11 2337125 MAGNETIC RESONANCE IMAGING METHOD FOR NOISE-FREE

INVESTIGATIONS, IN PARTICULAR OF BRAIN ACTIVATION The invention concerns a method of nuclear magnetic resonance tomography (MR) with which an MR-signal is excited from a plurality of slices of an investigational volume through the simultaneous application of a plurality of radio frequency (RF) pulses, each having a narrow excitation profile, in the presence of a slice selection gradient, wherein the phases of the individual RF pulses associated with each individual slice are varied from one excitation step to the next in a unique fashion, and with which signal excitation is effected through gradient reversal in the sense of a gradient echo.

A method of this type is e.g. known in the art from USA-4,843,322.

The most frequently used methods for magnetic resonance imaging investigations of brain activation utilize the principle of BOLD contrast. This technique is based on the changes in tissue susceptibility as a consequence of displacement of the relationship between oxygenated and 1 1 1 '1 2 deoxygenated blood. The method of choice has turned out to be echo - planar - imaging (EPI) which facilitates the taking of a data set covering the entire brain within a short recording time of 1-3 s.

In systems which do not have the capability of generating the high performance time changing magnetic field gradients necessary for EPI, a gradient echo procedure having corresponding longer echo times can also be utilized. This latter technique has the serious disadvantage for localization experiments that only individual slices are thereby investigated, since the recording time for a technically feasible exercise of the procedure in the form of a multislice process assumes values of significantly longer than 20 s and is thereby no longer compatible with the typical time scale for the activation effects to be observed.

The multi-slice capability of gradient echo sequences can be substantially and significantly improved using new methods such as echo shifting (ES-FLASH) (Liu, G., et al. Magn.Reson.Med. 30, 68-75 (1993H or the MUSIC-Sequence (T.L6nneker et al., Magn.Reson.Med. 35, 870-874 (1996)). one must thereby accept the disadvantage that the additional gradient switching steps result in additional gradient-generated noise.

With regard to noise development, a single slice gradient echo recording technique is significantly quieter than MUSIC, ES-FLASH or even EPI. However, when conventionally utilized, same is still so loud that noisesensitive experiments can not be carried out or can only be carried out to an unsatisfactory extent. Examples therefor are investigations of the acoustical cognizance apparatus, speech paradigms as well as investigations during sleep.

The solutions for carrying out more quiet MR experiments proposed up to this point in time are either very difficult c:'I 1 3 from a technical point of view (A New Silent MRI Using A Rotating DC Gradient, Z.H. Cho, et al., Proc.5th Ann.Meeting ISMRM, page 280 (1997)) or are based on techniques having intrinsically low signal to noise ratio. (Ultrafast Silent MRI Using FM DANTE Sequence. Z.H. Cho, et al. Proc. Sth Ann. Meeting ISMRM, page 1822 (1997)).

It is therefore the purpose of the present invention to modify a standard method of the above mentioned kind in as simple a manner as possible such that no noise disturbance or as little noise disturbance as possible occurs for the patient under investigation.

This purpose is achieved in accordance with the invention in a manner which is as surprisingly simple as effective in that the time changing magnetic field gradients used for signal excitation and spatial encoding are applied in such a fashion that the change in the strength of the magnetic field gradient is effected with a sigmoid function whose time dependence is determined in such a fashion that the Lorentz forces caused by gradient switching are minimized.

In this fashion, a method for carrying out magnetic resonant imaging measurements, in particular of brain activation, is facilitated which avoids the magnetic field gradient related noise, which is highly disturbing in conventional methods and which also falsifies the results for a large number of investigations. The method in accordance with the invention is thereby based on the gradient echo sequence, wherein the interfering noise is substantially avoided by using time changing gradient fields having a switching shape which does not generate acoustical vibration or only generates a small amount of acoustical vibration. The excitation of a plurality of slices in one excitation cycle is effected through the application of special radio frequency excitation pulses for slice excitation which excite a plurality of slices simultaneously, however, in such a fashion that the produced signals from each slice

4 have a differing spatial encoding and can thereby be separated for image reconstruction and associated with the respective individual slices.

The first step of the quiet MR method towards solution of the problem in accordance with the invention is based on the fact that the largest fraction of noise from a magnetic resonance imaging investigation is produced by the rapid switching processes of the time changing magnetic field gradients. The local change of the magnetic field by several mT within a time of less than 1 ms leads to the production of very strong Lorentz forces which act on the gradient coils and set them into mechanical vibration. These vibrations are transferred to the air in the magnet via the surface of the gradient system and lead to the generation of sound waves.

The individual under investigation is thereby disturbed by so-called structure-borne noise which acts directly on the head through direct mechanical transfer via the gradient tube, mechanical support structures, and the patient bed as well as by sound transfer through air. The noise threshold which is observed depends, on the one hand, on the acoustical characteristics of the overall system with regard to the transfer of structure-borne noise and also, to a not insignificant extent, on the acoustics of the MR region including the cylindrical magnet opening.

The techniques for noise reduction in accordance with the invention include a reduction in the switching speed of the magnetic field gradients and generation of periodical Lorentz forces which excite vibrations in a frequency range lying outside of the resonance maxima of the investigational volume.

With regard to the excitation of sound waves, the method is assisted by the fact that the vibration characteristics of the gradient tubes normally used are non-linear and do not, below a certain characteristic switching time, lead to sound vibrations which can be sensed. The Lorentz forces causing the emission of sound are proportional to the time change of the local magnetic field. The vibrations of the gradient tube and the sound waves caused thereby primarily depend on the time change of the acting forces. The frequency spectrum of a periodically repeated magnetic resonance imaging sequence can be described by the Fourier transform of the second derivative of the gradient switching scheme. The noise which can be sensed (structure-borne noise or air-transferred noise) then results from the transfer characteristics of the system applied to this sequence.

The designation 11sigmoid function" is understood, within the context of the present invention, as a function whose second derivative does not have any spikes, that is to say, no discrete extrema.

An embodiment of the method in accordance with the invention is particularly preferred with which the change in the phase of the excitation pulse from one excitation step to the next is effected in the form of an increment which is constant for each individual slice and which is different from slice to slice. In this fashion, encoding of the read-out signal, effected in the sense of two-dimensional Fourier transformation, leads to images for individual slices which are displaced with respect to each other after reconstruction.

An improvement is also advantageous with which the phases of the individual RF pulses used to excite the individual slices are displaced with respect to each other in such a fashion that the radio frequency power and thereby the specific absorption rate is reduced. In this fashion, a heating-up of the patient under investigation due to radio frequency effects can be minimized.

This purpose is also served by a variation of the method with which the individual RF pulses for excitation of the 0 \_11 6 individual slices are time displaced.

In a particularly simple improvement of this method variation, the differing dephasing of the spins in the slice selection direction is effected by the presence of generally unknown local field inhomogeneities, such that the signals coming from different regions having differing homogeneities are rephased at least during a part of the data acquisition time.

Alternatively, in another variation of the method, a differing dephasing of the MR signals caused by the time displacement relative to the slice selection gradient is compensated using an appropriate gradient during signal read-out. This method variation can also facilitate bringing the echoes into time alignment within the read-out time window.

A preferred improvement of this method variation provides that the different echo times of the signals caused by the different points of time of excitation of each individual slice are compensated by applying an appropriate magnetic field gradient during signal excitation in the direction of the read gradient. In this fashion, differing dephasing of the signals from the individual slices can be easily compensated for.

An additional advantageous method variation utilizes a gradient in the direction of the slice selection gradient which is incremented in the time period between signal excitation and signal read-out, from one excitation step to the next. In this fashion, the signal sequence of the respective individual slice is given by a linear combination of phase corrected individual signals characteristic for each slice.

A variation in this method also serving this purpose effects the phase changes of the RF pulses from one excitation step 0) 1 1 7 to the next using the principle of Hadamard encoding. This method variation is somewhat simpler from a technical point of view than the others described above. However, it is associated with somewhat reduced imaging quality.

Further advantages of the invention can be extracted from the description and the drawing. The above mentioned features and those to be described further below can be utilized in accordance with the invention individually or collectively in arbitrary combination. The embodiments shown and described are not to be considered as exhaustive enumeration, rather have exemplary character for illustration of the invention.

The invention is represented in the drawing and is described more closely below with reference to an embodiment.

Fig. 1 shows the time dependence of gradient switching and the associated first and second derivatives, wherein, in column A, a typical switching scheme for conventional gradient switching is shown, in column B a switching scheme for a parabolic switching shape, and in column C a switching scheme for a sinusoidal switching shape; Fig. 2 shows the time dependence of an excitation and measurement sequence in accordance with the invention with three simultaneous slice selection pulses having frequencies fl, f2, f3, a slice selection gradient Gs, a read gradient Gr, a phase encoding gradient Gp as well as a gradient echo signal Rf, wherein the change of the gradient amplitudes is effected using a parabolic switching shape; Fig. 3 shows a switching scheme such as in Fig. 2 utilizing a saturation pulse S in the presence of a slice selection gradient in the direction of Gp; 11 1 _. ' I; i 8 Fig. 4 shows a sequence as in Fig. 2, however, with slice selection pulses fl, f2, f3 time-displaced with respect to each other; Fig. 5 shows the sequence of Fig. 4 using, however, small compensation gradients in the slice selection direction Gs; Fig. 6a shows a schematic sequence of voxels for a slice plane perpendicular to the slice selection gradient Gs with read-out using a read gradient Gr in a direction perpendicular to the slice selection gradient Gs; Fig. 6b as in Fig. 6a, however, with a read gradient Gr which is not perpendicular to the slice selection gradient Gs; Fig. 7 shows the sequence such as in Fig. 5, however, with an additional compensation gradient in the direction of the read gradient Gr; Fig. 8 shows the sequence of Fig. 4, however, with incremental compensation gradient in the direction of the slice selection gradient Gs; Fig. 9 shows a sequence for "spiral imagingll; and Fig.10 shows a sampling scheme in k space associated with "spiral imagingll.

Fig. 1 shows, in (A), a typical switching scheme used in MR sequences with which a gradient is initially applied with a linear increase up to a certain amplitude and, after a certain time, is changed to a different, in this case, negative amplitude value and finally returned to zero.

9 Typical times for such processes are in the millisecond range or less. The first derivative of the gradient amplitude as a function of time reflects the Lorentz forces leading, in turn, to mechanical deflection and deformation of the gradient tube. Acoustical vibrations are caused by the changes in the deformations and deflections reflected by the second derivative. The sequence of observed spikes in the second derivative leads to acoustical vibrations having frequencies around 1 kHz, in consequence of the time sequence in the millisecond region.

As shown in (B), a gradient change having a parabolic switching characteristic leads to a linear change and thereby to a second derivative having constant values within the switching time. The distribution of the active amplitudes over the entire switching region leads to a substantial reduction in acoustical amplitudes. A sinusoidal switching (C) likewise avoids the spikes which cause acoustical noise.

The passive acoustical damping of a magnetic resonance imaging system typically has a non-linear damping characteristic, wherein low frequency components are more strongly damped than high frequencies. This damping characteristic decreases monotonically with frequency and can be substantially influenced by the properties of the patient tunnel and of the investigational volume so that the overall transfer properties are a complicated function of the excitation frequency.

A measurement of the overall transfer function can be advantageously carried out through a frequency dependent determination of the acoustical pressure, wherein the gradients are switched sinusoidally and the frequency of the sinusoidal modulation is varied in the relevant region. In particular, in order to determine the high frequency components, the acoustical noise spectrum can be 1 1- -1 alternatively determined through the application of a short acoustical pulse. This, however, fails to consider those acoustical components which generate standing waves via resonance effects. In particular, for volumes not having acoustical damping elements, the measurement can be more advantageously carried out through periodic repetition of gradient pulses. This also allows measurement of standing acoustical waves. With highly non-linear acoustical characteristics, the acoustical spectrum is preferentially determined directly from the sequence utilized for the magnetic resonance imaging measurement.

Since the acoustical spectrum results, in first approximation, from the second derivative of the gradient switching sequence, rapid changes in the gradient switching procedure should be avoided. This can be achieved by replacing the conventional t rape zoidal -shaped gradient switching with a sigmoid switching shape. In dependence on the acoustical transfer function, it can thereby be advantageous to effect the switching using a constant second derivative (that is to say, quadratically increasing and decreasing functions) or having a time changing second derivative (e.g. sinusoidal). The switching time for change of the gradient amplitudes can then be adjusted to an acceptable noise threshold.

However, the switching times which can thereby be tolerated no longer permit measurement sequences which require a plurality of rapidly changing switching processes. Towards this end, this principle is not compatible with EPI, MUSIC, and only to a limitable extent, with ES-FLASH.

Simple single slice recording can be effected with gradient echo sequences using slower switching processes for the magnetic field gradients. This, however, does not solve the purpose in accordance with the invention for possible investigation of multi-slices with adequate spatial and time resolution.

11 With the method in accordance with the invention, the gradient echo sequence must be modified in such a fashion that investigation of a plurality of slices is possible, however, with no or only small increase in the number of magnetic field gradient switching steps.

one possibility for multi-slice recording is thereby given by the principle of multi-slice selective pulses as is known in the art e.g. (J. Henning Magn.Reson.Med. 25: 289-298 (1992H and (G.H.Glover, US Patent 4, 843,322 (1989)). Such pulses are based on the overlap of pulses each generating a single slice.

The overlap of the individual pulses does not thereby require a plurality of radio frequency transmitters and radio frequency coils. The overlap can be calculated as the simple sum of the individual signals so that, from a technical point of view, only one single radio frequency pulse must be utilized whose shape corresponds to the required overlap.

Within the context of a two-dimensional Fourier transformation, the socalled k-space matrix is filled up row by row through repetition of the recording while varying the phase gradient. The phase position of the pulses of each individual slice is stepped with each excitation pulse relative to the reference phase by an amount which is constant from one excitation step to the next but which, however, is different for each excited individual slice. For m differing excited slices one advantageously chooses the phase increment Pi of the ith slice as Pi = 2 7r ( i - 1) /m (1) 12 In the kth recording step, the ith slice therefore has a phase shift of PI, k = 27r k (i-l) /m (2) In this case, following Fourier transformation, the images of the individual slices are displaced with respect to each other, in each case, by 1/m of the image size in the phase encoding direction. An image free of overlaps can then be obtained by correspondingly increasing the imaging field for data recording with the number of imaging pixels remaining unchanged. The imaging resolution is thereby reduced, wherein this magnification, in a preferred implementation, occurs only in the phase encoding direction of the image.

The switching scheme of a sequence of this type is shown in Fig. 2. After switching-on a slice selection gradient Gs, three slice selection pulses having frequencies fl, f2 and fS are simultaneously applied the band width of which is less than or equal to the difference between the corresponding frequencies and thereby, in combination with Gs, simultaneously excites spins in a plurality of separate slices.

After termination of the slice selection pulse, a signal in the sense of a gradient echo signal Rf is produced through application of a read gradient Gr as well as a phase encoding gradient Gp, wherein the change of the gradient amplitudes is, in correspondence with the method in accordance with the invention, effected with a switching shape which is gradual (in this case parabolic). The amplitude of the phase encoding gradient is changed from one excitation to the next in order to thereby obtain a data set for two-dimensional Fourier transformation, following a plurality of repetitions.

In investigations which are intended to only cover a portion 13 of the head, those portions of the body which are not to be observed can be subjected to signal saturation in a manner which is known in the literature through saturation procedures. A conventional method therefor comprises a saturation pulse S in the presence of a slice selection gradient in the direction of Gp, wherein the amplitude of S is chosen in such a fashion that the z-magnetization of the spins selected by the bandwidth and frequency of S as well as the amplitude of Gp is precisely zero at the point of time of application of the excitation pulse so that no contribution to the signal is provided (Fig. 3).

An alternative for avoiding undesired overlap of the images from differing slices is also effected by an m-fold expansion of the imaging field in the phase encoding direction with correspondingly reduced resolution in this direction. An m-fold increase in the imaging field while maintaining the imaging resolution in the phase direction is possible, however, leads to an increase in the measuring time due to the necessarily larger number of measuring steps with differing phase encoding.

One problem associated with the application of a multi-slice pulse, is that, in the event of a simple coherent overlap, it has m-times the amplitude of an equivalent individual slice pulse. The overall radio frequency power thereby utilized increases with m2 which, in particular with apparatusses having high field strengths of 2T and more, as are preferred for the applications of interest, can lead to unacceptably high SAR (specific absorption rate).

J.Hennig Mag.Reson.Med. 25; 289-298 (1992) has already shown that this problem can be reduced using non-coherent overlap, wherein the radio frequency power increases only linearly with the number of slices.

Another possibility for reduction in power is use of the mutually time displaced pulse shapes for excitation of the 1 1 1 1 __--- 14 individual slices, as shown in Fig. 4. In this type of implementation, the spins from the individual slices are exposed to the dephasing effect of the slice selection gradient for differing lengths of time due to the time displacement of the individual excitation pulses. An exact rephasing thereby occurs only for spins from a particular slice, whereas the spins from each of the other slices have a dephasing transverse to the slice plane.

The magnitude of this dephasing depends an the time displacement of the pulses with respect to-each other as well as on the strength of the slice selection gradient utilized. A dephasing of approximately 900 can, however, thereby be viewed as still tolerable, since the associated signal loss for a dephasing of this type could, in any event, already occur for the sequences utilized due to magnetic field inhomogeneities.

This dephasing can be avoided if, during the time of read-out of the data, a small compensation gradient is applied in the direction of the slice selection gradient such that each signal to be observed is refocussed within the data acquisition window at a particular time which is slightly different from slice to slice. (Fig. 5).

one should, however, note that the additional gradients cause the direction of the slice selection gradient and the read gradient to no longer be perpendicular to each other.

As schematically shown in Fig. 6a, the perpendicular to the slice selection gradient Gs normally determines the direction of the slice plane (horizontal voxel edge). The perpendicular to the applied read gradient Gr which, for its part, is orthogonal to the slice selection gradient Gs determines the view direction on the voxel (vertical voxel edge). Therefore, when the slice selection gradient Gs and the read gradient Gr are orthogonal to each other, a rectangular voxel results.

Fig. 6b schematically shows, when the slice selection gradient Gs and the read gradient Gr are no longer orthogonal, that the signals of a voxel consolidated to one imaging point no longer originate from a parallelpiped perpendicular to the imaging plane. The edges of the voxel are tilted towards the imaging plane defined by the slice selection and read gradients.

Appendix A shows the conditions satisfied for each additional gradient and the resulting imaging geometry. The point of time of the nominal refocussing relative to the read and slice selection gradients differs in this implementation. As long as the difference among the echo read-out times remains within a range of +/- 20-., the resulting contrast difference among the individual images of the slice is, however, not of consequence for the application at hand.

The differing read-out times can be corrected if an appropriate gradient is switched in the direction of the read gradient during application of the pulse. The completely compensated sequence is shown in Fig. 7.

In analogy thereto, the differing dephasing of the spins in the individual slices can also be compensated by using a smaller gradient in a slice selection plane which is incremented from recording step to recording step in analogy to the phase encoding gradient (Fig. 8). This leads to a tilting of the edges of the voxel in the direction of the phase encoding gradient.

One should note that, for both variations, the nonorthogonality of the imaging plane relative to the slice selection gradient can only be tolerated for small tilting angles Z (Z << 300, preferentially of 100 or less), since a strong image blurring otherwise occurs due to the loss in edge information. In general, the acceptable value of Z is 16 smaller the thicker the slice under investigation relative to the imaging resolution.

Finally, one should mention that the basic concept of the method can also be applied to processes which are not based on image encoding in accordance with the two-dimensional Fourier transformation method. Such techniques are e.g. known in the art through the methods of signal encoding in the sense of filtered back projection or in the sense of socalled "spiral imagingll. The latter is particularly well suited as a "quiet sequence" due to the type of signal encoding. Both gradients are thereby changed in the image plane in such a manner that the signal encoding assumes a spiral in k space (Fig. 9).

Sequences designated up to this point as Gr and Gp are now designated as Gx and Gy, since, in this case, there is no obvious differentiation between a read and a phase encoding gradient. The three gradients Gs, Gx and Gy are, however, still mutually perpendicular magnetic field gradients.

With a recording time of 30-50 ms, which is acceptable for the application at hand, while maintaining the boundary conditions of a sufficiently slow time change of the magnetic field gradients for the suppression of noise, the so-called k space can no longer be sampled with spirals which are sufficiently dense for image reconstruction. only typically 1-4 windings per recording step can be sampled. Using a segmented recording process, a sufficiently complete sampling can be achieved through repetition and corresponding changes of the spiral (Fig. 10).

The recording of a plurality of simultaneous slices in correspondence with the above description is then, in the simplest case, carried out in such a fashion that, for m slices in correspondence with the above described method, the recording is repeated m-times for each partial spiral, wherein the phase of the pulses of each individual slice is,

17 as described, stepped forward from one recording to the next. The signal for the individual slices can be calculated from a linear combination of the recorded signals. For determination of the signal Si of the ith slice, the m signals recorded in each case are thereby phase-corrected in such a fashion that all signal portions of the ith slice are coherent. For the kth partial recording Sk (with k = 1 to m) this correction corresponds consequently to (2); -PI,k = -27r k(i-l)/m Si is thereby given by m Si = T S k exp (-27r k(i-l)/m) k=l Since, with this procedure, the m individual signals of the other slices are equallydephased, they are averaged out by the summing-up. The m-fold repetition of each partial spiral thereby leads to an increase in the recording time by a factor m. This is, however, compatible with the boundary condition of a recording in less than 20 s due to the intrinsically more rapid data taking in consequence of the lower number of partial spirals for achieving sufficiently dense recording of the k space matrix.

In contrast to this sequential recording of m individual slices which can be carried out in the same measuring time, this type of recording offers an improvement in signal to noise ratio by a factor m 1/2. For this implementation as well, it can be advantageous to successively compensate, from slice to slice, for the differing dephasing of the signals from the individual slices caused by a time displacement of the excitation pulses using a weak gradient in the slice selection plane during data read-out. Due to 1 1 18 the special imaging properties of non-rectilinear k space sampling procedures, possible residual portions of each signal not fully compensated by the linear combination are spread diffusely across the imaging plane and therefore lead to relatively non-disturbing artefacts. Implementation of the method can also be carried out with simultaneous individual pulses in analogy to Fig. 2.

The same strategy can also be utilized for other spatial encoding procedures such as filtered back projection and even for non-Fourier encoding procedures such as wavelet encoding, singular value decompensation (SVD) and the like.

Finally, one should mention that a unique encoding of the signals from the individual slices can also be achieved by programming the phase of the excitation pulse in accordance with the principle of Hadamard encoding.

In conclusion, one mentions that the measuring principle described herein can also be utilized for carrying out MR investigations which do not concern obser-vation of brain activation. The described sequences can e.g. be thereby utilized in all cases where the gradient noise associated with conventional MR investigations is unpleasant to the patient or should be avoided for other reasons. Examples are investigations of sleeping patients and experimentees or special cases such as small children or particularly sensitive people.

Finally, one should mention that the principle of successive refocussing of the individual signals through application of a gradient in the slice plane and thereby the observation of an image with voxels tilted relative to the imaging plane as shown in Fig. 5, is also suitable for compensating differing dephasing of signals which is not caused intentionally. It can also be utilized to compensate for non-intentional and unknown dephasings as for example caused by local differing field homogeneities, at least over a portion of the - 1 1 19 recording window, for improved imaging quality even in conventional recordings such as conventional gradient echo sequences and also in echo planar imaging techniques.

i 1 Append-ix A The band width of the individual pulses of each slice is BW, the number of slices n, with the slices being directly adjacent to each other so that the separation from slice middle to slice middle is BW, in each case. with a time displacement dt of dt = 1 / (n BW) (AI) for the sequence of the individual pulses, a complete loss in coherence for the shape of the individual pulses occurs, since, with this time separation, the relative phase of the pulses associated with neighbouring slices is precisely 3600/n, so that the vectors of the radio frequency field are evenly distributed in the plane.

With a slice thickness SL, the calculated gradient field Gs of the slice selection gradient utilized is

Gs = BW/SL l/n whereby n is the gyromagnetic ratio.

(A2) The gradient integral between the middle of the first and last pulses in time is given by Gs dt = 1/(n SL) l/n (A3) and is thereby only dependent on the distance to the outermost slice, in each case, and not, however, on the bandwidth of the pulse utilized.

For a signal read-out time AQ, the strength of the gradients Gsr which are to be applied in the slice selection direction during recording is given by Gsr = Gs dt/aq 2 For a recording time aq for MX data points, a spectral width SW of the recording frequency is SW = MS/(2 aq).

The strength of the read gradient Gr to achieve an observation window of size FOV is thereby given by Gr = MX/ (2 aq FOV) l/n For typical values of BW = 1 kHz, SL = 5 mm, n = 3, aq = 25.6 ms, FOV = 25 cm and MX = 256 Gsr = 0.061 mT/m.

The read gradient assumes values of Gr = 0.469 mT/m.

(A4) The tilt angle a of the voxel relative to the image plane thereby assumes a value of arctan a = Gsr/Gr and thereby a = 7.410.

A tilt angle of this kind is completely acceptable and leads to no significant reduction in image quality. In the event of a larger Gsr, the refocussing of the signals of the individual slices occurs during a time interval which is smaller than aq. One must therefore take care that the negative gradient in the slice selection direction between the time of excitation by the radio frequency pulse and data read-out has an amplitude which is sufficient to rephase the spins of all slices at a defined point in time lying within the acquisition window.

2.

22

Claims (1)

1. Claims

   Method of nuclear magnetic resonance (MR) imaging with which an MR signal from a plurality of slices of an investigational volume is excited through the simultaneous application of a plurality of radio frequency (RF) pulses, each having a narrow excitation profile, in the presence of a slice selection gradient, wherein the phases of the individual RF pulses associated with each individual slice are varied from one excitation step to the next in an unique fashion for each slice and with which signal production is effected by gradient reversal in the sense of a gradient echo characterized in that the time changing magnetic field gradients utilized for signal excitation and spatial encoding are applied in such a fashion that the time change in the strength of the magnetic field gradients is effected with a sigmoid function the time dependence of which is determined in such a fashion that the time change in the Lorentz forces, associated with the currents in the gradient coils, caused by gradient switching are minimized in that the second derivative of the sigmoid function is constant throughout the switching time of the gradients or at least has no spikes in its changing time dependence.

   Method according to claim 1, characterized in that the change in the phase of the excitation pulses from one excitation step to the next is effected in the form of a constant increment for each individual slice, which is different from slice to slice.

   h.

   23 3. Method according to any one of the preceding claims, characterized in that the phases of the individual RF pulses for excitation of the individual slices are displaced with respect to each other in such a fashion that the radio frequency power and thereby the specific absorption rate is reduced.

   4. Method according to any one of the preceding claims, characterized in that the individual RF pulses for excitation of the individual slices are applied in a time-displaced fashion. ' Method of claim 4, characterized in that the individual dephasing of spins in the slice selection direction is effected by the presence of generally unknown local field inhomogeneities such that the signals coming from regions having different homogeneity are rephased during at least a part of the data acquisition time.

   Method according to claim 4, characterized in that the differing dephasing of the MR signals relative to the slice selection gradient caused by the time displacement are compensated for by applying an appropriate gradient during signal read-out.

   7.

   Method according to any one of the claims 4 through 6, characterized in that the differing echo times of the signals caused by the differing points in time of excitation of each individual slice are compensated for by applying an appropriate magnetic field gradient during signal excitation in the direction of the read gradient.

   Method according to any one of the claims 4 through 7, characterized in that a gradient, applied in the direction of the slice selection.gradient, is incremented from one excitation step to the next in the time interval between signal excitation and signal read-out.

   24 Method according to any one of the claims 1 or 3 through 8, characterized in that the spatial encoding of the MR signals from the image plane is effected in the sense of the "spiral imagingI, method and the underlying sequence for measurement of m individual slices is repeated m-fold times, wherein the phase of the RF pulse utilized for excitation of each individual slice is effected from one excitation step to the next in the form of an increment, constant for each individual slice, which differs from slice to slice.

   10. Method according to any one of the claims 1 or 3 through 8, characterized in that the phase change of the RF pulses is effected from one excitation step to the next in accordance with the principle of Hadamard encoding.