Method for the operation of a nuclear magnetic resonance tomography apparatus for separating water and fat signals

[0002] 1. Field of the Invention

[0003] The present invention is directed to a method for operating a nuclear magnetic resonance tomography apparatus, and in particular to such a method which allows separation of water signals and fat signals from each other.

[0004] 2. Description of the Prior Art

[0005] U.S. Pat. No. 4,769,603 discloses a pulse sequence that is usually referred to with the acronym “FISP”. A specific version of this pulse sequence, wherein the gradients in all spatial directions are reset before each excitation, is referred to as “True FISP Sequence”. The relatively long transient time given this pulse sequence is reduced, as taught in U.S. Pat. No. 5,541,514, emitting a radio-frequency pulse having the flip angle α/2 preceding a FISP sequence, with successive excitations having a flip angle α.

[0006] The True FISP sequence exhibits excellent properties for MR imaging, namely high spatial resolution, short measuring time in the range of seconds, and high signal-to-noise ratio, particularly for tissue having liquid consistency.

[0007] The True FISP Sequence exhibits a strong T1/T2 weighting. Pure water having a ratio T1/T2=1 yields the maximum SSFP (Steady State Free Procession) signal MO/2 given FISP. MO represents the magnetization in the relaxed condition dependent on the proton density. Fat exhibits a T1/T2 ratio of 270:70 ms. Even though this ratio is clearly higher than that of pure water, fat still exhibits an extremely high signal intensity in the FISP sequence. For comparison, muscle tissue has a T1/T2 ratio of 12:1. In conventional sequences wherein a high signal intensity of fat is likewise disturbing, this problem is usually solved by fat saturation in a preparation phase. A spectrally selective radio-frequency pulse is emitted into the examination subject. In order to destroy the magnetization produced by this pulse, the radio-frequency pulse is enclosed by pulses known as spoiler gradient pulses. Such a preparation phase, however, lasts for a relatively long time (for example, 10 ms), and thus it is unsuitable for FISP, which has a typical repetition time TR=5 ms. Given the necessary lengthening to, for example, 15 milliseconds, pronounced artifacts would be obtained due to field inhomogeneities.

[0008] It is an object of the present invention to enable a separation of the fat and water signals in a FISP sequence.

[0009] The above object is achieved in accordance with the principles of the present invention in a method for operating a nuclear magnetic resonance tomography apparatus wherein radio-frequency pulses are emitted into an examination subject with a repetition time, in order to excite nuclear spins in the examination subject, nuclear resonance signals resulting from the nuclear spins are read out between the radio-frequency pulses under a readout gradient in a readout window, the nuclear resonance signals are phase-encoded before the readout window and the phase-encoding is reset before the next radio-frequency pulse, and wherein the repetition time is selected so that fat-bonded protons experience a phase rotation of approximately n·180° due to the effect of the chemical shift compared to water-bonded protons, wherein n is an odd number.

[0010] FIG. 1 shows an exemplary pulse sequence in accordance with the invention.

[0011] FIG. 2 is a spatial illustration of the magnetization vector.

[0012] FIG. 3 shows the magnitude signal of the transverse magnetization as function of the spin dephasing.

[0013] FIG. 4 shows the development over time of the y-component My of the magnetization given a repetition time TR=4.8 ms.

[0014] FIG. 5 shows the corresponding development over time of the transversal magnetization Mt.

[0015] FIG. 6 shows the development over time of the y-component My of the magnetization given a repetition time TR=2.4 ms.

[0016] FIG. 7 shows the corresponding development over time of the transversal magnetization Mt.

[0017] FIG. 8 is an intensity profile of an oil/water phantom without oil suppression.

[0018] FIG. 9 is an intensity profile of an oil/water phantom with oil suppression.

[0019] As an exemplary embodiment of the invention, FIG. 1 shows a pulse sequence that—with the exception of the time sequence explained below—is disclosed by U.S. Pat. No. 5,541,514. The pulse sequence begins with a radio-frequency pulse RF1 that is emitted under the influence of a slice selection gradient GS1, so that only one slice of the examination subject is excited. The dephasing caused with the slice selection gradient GS1 is in turn canceled by the oppositely directed gradient GS2. A second radio-frequency pulse RF2 that is emitted under a slice selection gradient GS3 follows. In the next step, the nuclear spins are phase-encoded by a phase-encoding gradient GP1 and are dephased in readout direction by a gradient GR1. Further, a gradient pulse GS4 directed opposite to the slice selection gradient GS3 is emitted in the slice selection direction.

[0020] In the next step, a nuclear magnetic resonance signal S is read out under a readout gradient GR2. Finally, a gradient pulse GS5 is applied in the negative slice selection direction, and a gradient pulse GR3 is emitted in the negative read out direction. The phase-encoding due to the gradient pulse GP1 is in turn reset by a gradient pulse GP2 in the opposite direction. Subsequently, a further radio-frequency pulse RF3 is emitted under a slice selection gradient GS6. As is typical of the True FISP sequence, the amplitude-time integral in all gradient directions is equal to zero between two excitations and the gradients exhibit a symmetry in every time window between two excitations.

[0021] The repetition time, i.e. the time between two excitations with radio-frequency pulses, is referenced TR in FIG. 1. The radio-frequency pulse RF1, serving as a preparation pulse, has a spacing of TR/2 from the radio-frequency pulse RF2 for reasons described in detail in the aforementioned U.S. Pat. No. 5,541,514. Further, the phase alternation of the radio-frequency pulses is critical, so that an excitation is obtained according to the relation α/2₋TR/2−[α₊−TR−α₋−TR . . . ]n.

[0022] The manner by which the above-described excitation influences the magnetization in the rotating coordinate system x, y, z is shown in FIG. 2. The magnetization vector initially is in the direction of the fundamental field, i.e. in the z-direction, and is deflected in the negative direction to the position I by the first radio-frequency pulse RF1 having the flip angle α/2 (referred to in brief below as α/2 pulse). A deflection by the angle α to the position II ensues due to the first a pulse RF2. Subsequently, the magnetization is swept back and forth between the positions I and II due to the radio-frequency pulses RF of alternating polarity. A steady state condition of the magnetization thus is obtained.

[0023] This consideration, however, only applies as long as no dephasing of the spins occurs. The behavior of the steady state magnetization can be described as a function of the dephasing angle with the equations of Freeman and Hill that are recited in “Phase and Intensity Anomalies in Fourier Transformation NMR,” Freeman et al., Journal of Magnetic Resonance4, pages 366-383, 1971. FIG. 3 shows the magnitude of the transverse magnetization ({square root over (Mx²+L +My²+L )}) as function of the spin dephasing for water (broken line) and fat (solid line). Independently of the material, the signal amplitude disappears given phase-alternating radio-frequency-excitation at those locations at which the spin dephasing amounts to φ=n·π with an uneven n. In the ideal, uniform fundamental magnetic field B₀, the dephasing within the repetition time TR is equal to zero, and the signal is high given phase alternation of the α pulses. In a real basic magnetic field with an inhomogeneity ΔB₎, the spin dephasing over time amounts to θ (t)=2π·γ·ΔB₀T.

[0024] The spin dephasing θ (t) produced by the field inhomogeneity viewed over the entire TR interval must be far smaller than π so that the signal amplitude does not drop off. The demands made of the field homogeneity given True FISP sequences are thus extremely high.

[0025] However, it is not only a magnetic field inhomogeneity but also the chemical shift δ that leads to a dephasing, i.e. the shift of the resonant frequency ω due to the chemical bond. The chemical shift δ for protons that are bonded in lipid molecules has a value of approximately 3.3 ppm compared to free water, i.e. the resonant frequency ω of protons bonded in lipid molecules is 2π·γ·δB₀smaller than that of free water. Given a field strength B₀=1.5 T, this difference amounts, for example, to 210 Hz. The invention is based on selecting the repetition time TR such that the signal from the lipid molecules (i.e. fat) exhibits a phase rotation of 180° or an odd-numbered multiple of 180° compared to the signal from free water. In the illustration of FIG. 2, this corresponds to an angular rotation by 180° around the z-axis. When, as described above, phase-alternating radio-frequency pulses are emitted, then the fat molecules no longer experience any alternation of the radio-frequency due to the phase rotation by 180°. Whereas the magnetization for free water flips back and forth between the conditions referenced I and II in FIG. 2 and adequate magnetization is thus always available the magnetization of fat becomes very low after a few excitation pulses due to the 180° phase rotation.

[0026] When, on the other hand, the repetition time is selected according to the above-presented rule and no phase alternation of the excitation is implemented, then the magnetization of water, and thus the water signal, will already disappear after a few excitations. By contrast, the following situation arises for the magnetization of fat: Assume the fat magnetization initially is in condition I according to FIG. 2. At the time of the next α pulse (i.e., after TR), the magnetization is rotated by 180° around the z-axis due to the dephasing caused by chemical shift, i.e. into position II. Due to the α pulse that now follows, the magnetization is again rotated into position I, etc. Ultimately, thus, the alternation between the positions I and II occurs here for fat signals, so that these supply a high signal contribution in the steady state.

[0027] A fat image or water image thus can be obtained dependent on whether the aforementioned pulse sequence is operated with or without phase alternation of the radio-frequency excitation.

[0028] On the basis of the above particulars, a repetition time of TR=1/(2γ·δ·B₀)=2.4 ms or an odd-numbered multiple thereof arises for a basic magnetic field of 1.5 T. Such repetition times can still be realized with high-performance nuclear magnetic resonance tomography systems.

[0029] FIG. 6 shows how the magnetization My develops in the y-direction for fat and water when the aforementioned repetition time TR=2.4 ms is adhered to. The following were assumed as parameters:

[0030] B₀=1.5 T, TR=2.4 ms, α=90°, 67 =3.3 ppm, T1=200 ms, T2=60 ms.

[0031] One can see that the water magnetization referenced Myw reaches a steady state condition at a relatively high level. The fat magnetization referenced Myf, however, oscillates greatly after every repetition due to the 180° phase rotations and decays to zero. The same is true of the magnetization Mxf registered with the same parameters. The magnitude of the water magnetization

[0032] Mtw={square root over (Myw²+L +Mxw²+L )}

[0033] and the magnitude of the fat magnetization

[0034] Mtf={square root over (Myf²+L +Mxf²+L )}

[0035] are shown in FIG. 7, which illustrates transversal magnetization Mt.

[0036] For comparison, the corresponding magnetization curve when the pulse sequence is operated with a repetition time TR=4.8 ms is shown in FIG. 4 for My and in FIG. 5 for Mt. In this case, a phase rotation of the fat signals respectively ensued by 360°, so that the magnetization of the fat protons, just like that of the water protons, flips back and forth between two conditions with every repetition. Here, the magnetization Myf and Mxf of fat oscillates greatly at the start; however, it proceeds to the same steady state condition as the magnetization Myw or Mxw of water.

[0037] For demonstrating the affect of the above-presented pulse sequence, FIGS. 8 and 9 show the intensity profiles of an oil/water phantom, without the inventive fat suppression (in this case, oil suppression) in FIG. 8 and with the inventive fat suppression in FIG. 9. The oil/water phantom has an oil-filled core in a water-filled cylinder as well as an oil-filled outside cladding. A comparison of the two diagrams shows that the signal S from the oil-filled core as well as the signal S from the oil-filled cladding have become significantly lower. MR images with clear suppression of the fat part thus are obtained in a simple way without major interventions to the system.