Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Magn Reson Med. Author manuscript; available in PMC 2013 September 16.
Published in final edited form as:
PMCID: PMC3774270

B1 and T1 Insensitive Water and Lipid Suppression using Optimized Multiple Frequency-selective Preparation Pulses for Whole-brain 1H Spectroscopic Imaging at 3T


A new method for the simultaneous suppression of water and lipid resonances using a series of dual-band frequency-selective RF pulses with associated dephasing gradients is presented. By optimizing the nutation angles of the individual pulses, the water and lipid suppression is made insensitive to range of both T1-relaxation times and B1 inhomogeneities. The method consists only of preparatory RF pulses and thus can be combined with a wide variety of MRSI schemes including both long and short TE studies. Simulations yield suppression factors, in the presence of ±20% B1 inhomogeneity, on the order of 100 for lipid peaks with three different T1s (300ms, 310ms and 360ms), and water peaks with T1s ranging from 0.8s to 4s. Excellent in-vivo study performance is demonstrated using a 3T volumetric proton spectroscopic imaging (1H-MRSI) sequence for measuring the primary brain metabolites peaks of choline (Cho), creatine (Cr), and N-acetyl Aspartate (NAA).

Keywords: magnetic resonance spectroscopic imaging, water suppression, lipid suppression, metabolite maps


Although proton magnetic resonance spectroscopy (1H-MRS) has become an important clinical and research tool, in vivo studies are often hindered by water and lipid signals that can overwhelm the desired metabolite resonances. Large residual water and lipid resonances distort the spectral baseline and adversely affect accurate quantification.

Water suppression is most commonly achieved with CHEmical Shift Selective (CHESS) pulses (1) in which magnetization from water protons is repeatedly rotated into the transverse plane and spoiled using associated dephasing gradients. The basic method has been improved using the Water suppression Enhanced through T1 effects technique (2) in which an optimization algorithm is used to select sets of CHESS-pulse excitation nutation angles maximizing water suppression over a range of both T1 and B1 variations.

For in vivo brain studies, subcutaneous and bone-marrow lipids are most commonly suppressed using Position RESolved Spectroscopy, PRESS (3), or STimulated Echo Acquisition Mode, STEAM (4), localization techniques that only excite a rectangular volume of interest wholly inscribed within the skull. Although highly effective and robust, the assessment of cortical gray matter near the skull is typically compromised. Subcutaneous lipids can be further suppressed using outer volume suppression (OVS) techniques with very selective saturation (VSS) pulses (5). This technique has the capability to suppress a volume with arbitrary shape, but the suppression depends on B1 inhomogeneity and accurate prescription of the volume to be saturated. To achieve whole brain coverage, both lipid signal extrapolation (6) and Short Tau Inversion Recovery, STIR, techniques have been used (7). However, k-space extrapolation of large lipid signals can be very sensitive to numerical and experimental errors, and suppression of lipids using STIR is limited in the ability to simultaneously null longitudinal lipid peaks with different T1s. Suppression is further compromised in the presence of B1 inhomogeneities. Moreover, the use of a non-spectrally selective STIR pulse leads to metabolite signal losses, reported to be on the order of 30–40% at 1.5T (8). Signal losses are further exacerbated at high fields, ≥3T, due to converging T1 values of the lipid and metabolite resonances (9,10).

Fortunately, at higher magnetic fields, the increase in chemical-shift separation between peaks permits the use of water and lipid-selective pulses to provide suppression of these unwanted signals. Utilizing different chemical shifts, water and lipids have been suppressed with techniques using spectral spatial RF pulses (11) and frequency selective RF refocusing pulses such as BASING (12). Although these techniques do not cause metabolite signal loss, they generally result in prolonged echo time. Without affecting echo times, Smith et, al. (13) proposed a simultaneous frequency-selective water and lipid suppression technique with preparatory hyperbolic secant 90° saturation pulses. Although the lipid suppression obtained was better than using the outer volume suppression techniques, it failed to address issues of either B1 inhomogeneity or multiple T1 relaxation times of lipids.

In this work, we extend the WET technique (2) with the use of multiple dual-band spectrally-selective RF pulses to simultaneously obtain robust suppression of water and lipids with different T1s in the presence of B1 inhomogeneity. The suppression scheme consists of only preparatory RF pulses, therefore it doesn’t pose any limitation on the echo time. This suppression technique is combined with a volumetric 1H-MRSI readout to achieve whole brain spectroscopic imaging. High-quality metabolite spectra with whole brain coverage and robust lipid and water suppression are demonstrated in an in-vivo study.


Choline (Cho), creatine (Cr), and N-acetyl Aspartate (NAA), with chemical shifts of 3.2, 3.0 and 2.0ppm respectively, were the primary proton brain metabolites targeted for MRSI in this study. In contrast, chemical shifts of the largest in-vivo lipid resonances range from 0.9 to 1.5ppm. In addition, lipids are composed of complex chemical compounds with a range of T1s. Duewell et, al. (14) studied relaxation times of lipids in musculoskeletal tissues using inversion recovery techniques and reported T1s of 310ms and 390ms at 1.5T and 4T respectively. Water spins have a chemical shift of 4.7ppm but multiple T1 relaxation times of 1099ms and 1348ms at 1.5T and 4.0T respectively in gray matter and 741ms and 904ms in white matter. At the same field strengths, water T1s range from 3s to 4s in cerebral spinal fluid (1518).

Typical B1 inhomogeneity on a clinical 3T scanner is on the order of ±20% (19). Under these conditions, the true nutation angle for each presaturation pulse will be space-variant, that, if not compensated, results in compromised water- and lipid-suppression. In order to have robust suppression of water and lipid magnetization with different T1s under the condition of B1 inhomogeneity, we propose a four presaturation RF pulse scheme as shown in Figure 1. The nutation angles of the four presaturation pulses are numerically optimized to maximally suppress magnetization with representative T1s of lipid and water spins in the presence of ±20% B1 inhomogeneity at the time of 90 degree excitation. Although exact T1 values of lipid and water spins in-vivo are hard to determine, representative values can be obtained from literature (1418). The four different nutation angles were found by minimizing the maximum absolute value of residue longitudinal magnetization for three lipid T1s, 300/310/360 ms and two water T1s, 1/3s.

Figure 1
Scheme of multiple nutation angle frequency selective pulses for both water and lipid suppression. Dephasing gradients were applied after each preparatory RF pulses.

To rotate water and lipid magnetization simultaneously while not disturbing the metabolites of interest, each presaturation RF pulses was designed with the same dual-band spectrally selective excitation profile as shown in Fig. 2. In particular, at 3T, NAA and the closest lipid peak are separated by 89 Hz. In this design, we chose a minimum-phase RF pulse design to achieve a sharp transition with two 250 Hz rotation bands separated by 500 Hz such that one rotation band covers the lipid peaks while the other excites the water signal, leaving all three metabolites of interest undisturbed. To reduce TR and relaxation effects, a minimum RF pulse duration was chosen under the condition that the transition band is narrow enough to rotate lipid spins but not NAA. A 20 ms RF pulse length was selected. The Shinnar-Le Roux algorithm (20) was used for the RF pulse design, in which the transition bandwidth is traded off against ripple amplitudes in the pass and stop bands. The designed RF pulse has a transition band of 50 Hz and pass-band ripple amplitude of 2% The four pulses are separated by 30ms intervals, during which crusher gradients are applied to dephase the transverse magnetization.

Figure 2
Dual band minimum phase RF pulse and its spectral profile for a nutation angle of 180°.

The longitudinal magnetization Mz can be calculated analytically in an inversion recovery experiment. Assuming Mz before the first rotation RF pulse is unity, Mz at time t after the rotation RF pulse is:

equation M1

where θ1 is the nutation angle of the RF pulse and T1,i is the T1 relaxation time of the ith spin of interest.

Defining Mz at the time before the second rotation pulse as Mz,1, the Mz at time t after the second rotation pulse is

equation M2

By induction, Mz at time t after the nth rotation pulse is:

equation M3

Setting the time interval between adjacent rotation pulses and the time interval between the last rotation pulse and the excitation pulse, e.g the 90° slice-selective pulse in a PRESS or STEAM sequence, as a constant, Mz at the time of excitation can be written as:

equation M4


equation M5


equation M6

representing a ±20% range of B1 inhomogeneity.

The optimization problem can be stated as finding a set of angles θn (n = 1…4) such that the maximum absolute value of Mz,4 for a set of spins with different T1s is minimized for b[set membership][−0.2 0.2]. A nonlinear minimization algorithm (Nelder-Mead) with multiple starting searching points is used to yield best results. For each searching step of θn (n = 1…4) in the Nelder-Mead algorithm, it is required to find the maximum absolute value of Mz,4 for a target T1 subject to b[set membership][−0.2 0.2]. Since Mz,4 is a non convex function of θn, it is impractical to find the global maximum. However, the problem can be solved when Mz,4 is approximated with a polynomial expansion.

To the first order of approximation,

equation M7


equation M8


equation M9

Mz,4 is thus a polynomial function of b and the problem reduces to one of finding the maximum absolute value of a multivariable polynomial function subject to a polynomial inequality. The global optimal is found using GloptiPoly optimization program with SeDuMi in Matlab, (Mathworks, MA, USA) (21,22).

In this study, three T1s, 300/310/360ms were used for the lipid peaks and two T1s, 1/3s were assumed for water resonance. The optimal nutation angles for the four presaturation pulses were found to be 75, 97, 74 and 148 degrees.

In order to test the performance of the proposed suppression scheme, the four dual-band presaturation pulses with optimized nutation angles were synthesized separately using Shinnar-Le Roux algorithm and incorporated into a PRESS MRSI sequence. All experiments were conducted on a 3T echo-speed whole-body magnet (GE Healthcare; Waukesha, WI, USA). Since the presaturation RF pulses are frequency selective and thus sensitive to B0 inhomogeneities, high order shimming was performed over the region of interest before scanning (23). At 3T, the NAA peak and the closest lipid peak are separated by 89 Hz while the transition band of the pulse is 50 Hz. Across the region of interest, B0 variations could cause frequency shifts larger than the tolerance of the dual-band presaturation pulses even after high order shimming. In order to prevent NAA peaks being saturated, during spectral-prescan, the modulating frequency of the four dual-band pulses was initially set on lipids. This modulating frequency was gradually increased to a point where NAA spins were undisturbed while most lipid spins were suppressed.

The new suppression scheme was incorporated into an MRSI sequence and tested in-vivo on a healthy 31 year old male subject. Informed consent, approved by Institutional Review Board was obtained from the subject. Data were collected with standard CHESS water suppression, and the four dual-band presaturation pulses with optimal flip angles. The PRESS box was prescribed to encompass the entire slice of the brain with TR=1.5s, TE=144ms, 24cm FOV, 16×16 matrix size, 4.5 cc voxel and 13 minute acquisition time.

Results and Discussion

The optimization algorithm using different starting rotation angles converged to similar optimized values. The time of the optimization procedure is on the order of one minute using a Pentium PC. Suppression factors for spins with T1s ranging from 100ms to 4s in the presence of ±20% B1 inhomogeneity are shown in Figure 3. As shown in the figure, residual Mzs are on the order of 1% for all spins with T1s longer than 1s.

Figure 3
Simulated absolute values of residual lipid magnetization at targeted T1s. Values are in linear scale.

In the nutation angle optimization process, we assumed 3T lipid spins with three T1s values (300, 310, 360ms). Although this is an approximation of the true values, the robustness of the technique for a large range of T1s, as demonstrated from simulation in Figure 3 and experimental results in Fig. 5, is clearly demonstrated.

Figure 5
Grid plot of spectra from an in-vivo scan with 1500ms TR, 24cm FOV, 16×16 matrix size, 4.5 cc voxel and 13 minute acquisition time. The spectra were obtained using multiple dual-band RF pulses for both water and lipid suppression.

The proposed method was performed on 4 healthy subjects after obtaining informed consent approved by IRB. Effective water and lipid suppression has been achieved in all in-vivo studies without disturbing metabolites of interest. Figure 4 and Figure 5 show the grid plots of spectra from one representative study, acquired with standard CHESS pulses and with four dual-band presaturation pulses, from voxels throughout the FOV on the same scale. With conventional CHESS, NAA peaks are greatly contaminated by the much larger lipid peaks for the regions close to the subcutaneous fat, making it difficult for interpretation and quantification. With the four dual-band presaturation pulse approach, it can be shown from the full grid plots that the proposed method achieved significantly better water suppression than conventional CHESS pulses in almost all voxels. Comparing the NAA peaks in Figure 4 without lipid suppression and in Figure 5 with the four dual-band frequency-selective presaturation pulses, we found that NAA peaks were not affected and at the same time, residual lipid peaks in the suppression band are much smaller than the NAA peaks in all voxels throughout the FOV. Therefore metabolite spectra throughout the brain can be obtained without losing cortical gray matter regions that are normally lost with a PRESS box prescribed inside the skull. As water and lipids have a range of T1s for in-vivo studies and experience B1s with different amplitudes, sufficient water and lipid suppression across the whole FOV demonstrates the robustness of this method.

Figure 4
Grid plot of spectra from an in-vivo scan with 1500ms TR, 24cm FOV, 16×16 matrix size, 4.5 cc voxel and 13 minute acquisition time. The spectra were obtained using CHESS pulses for water suppression with no lipid suppression.

The proposed suppression scheme suppresses lipids using frequency selective RF pulses and dephasing gradients. A saturation scheme with four RF pulses was chosen as a result of the balance between preparation time and suppression effects. From simulation, using 5 or more preparation RF pulses doesn’t provide significantly improved suppression while using 3 or less preparation RF pulses results in compromised suppression. Since chemical shifts of lipids overlap with lactate, the suppression technique prevents lactate detection. The four dual-band presaturation RF pulses don’t pose significant SAR restrictions on the MRSI sequence. Because all of the dual-band pulses are frequency selective, they are relatively long (20 ms) with low peak RF amplitude. As a result, the SAR of the proposed pulse sequence is significantly lower than a conventional fast spin echo approach, commonly used in both head and body imaging. As the SAR is proportional to the sum of squares of the nutation angles, the SAR of the proposed 4 dualband RF pulses are much lower than water suppression techniques such as WET (2) and VAPOR (24) which utilize multiple RF pulses. Moreover, the 4 dualband RF pulses proposed achieve both water and lipid suppression and the SAR of the 4 dualband RF pulses is lower than conventional CHESS pulses combined with OVS RF pulses for lipid suppression.

The proposed suppression scheme can be used in many spectroscopic imaging applications. Since the method proposed uses frequency-selective RF pulses for water and lipid suppression, effective lipid suppression can’t be achieved without affecting NAA in the case of extreme B0 inhomogeneity. As the NAA singlet, 2ppm, is separated from the closest lipid peak, 1.3ppm, by 89 Hz at 3T and the transition bandwidth of the frequency-selective pulses is 50 Hz, the tolerance of the B0 inhomogeneity using this technique is 39 Hz. For in-vivo studies, B0 inhomogeneity is determined by many factors such as shimming, local susceptibilities, etc. In a spectroscopic quantification study at 1.5T, Pfefferbaum et, al. (25) successfully obtained complete metabolite maps of Cho, Cre and NAA using spectra with B0 shifts less than 5 Hz for brain regions above AC-PC line. Smith et, al. (13) also addressed this issue using a frequency-selective technique at 3T with a similar transition bandwidth of 50 Hz as used here. R3.3 In our center, for brain regions superior to thalamus, B0 shifts are within 20Hz for most voxels with high order shimming at 3T. Therefore, the frequency-selective RF pulses proposed provided sufficient margin for effective suppression of lipids without disturbing NAA. However, particular care is needed in setting the modulating frequency for these RF pulses for in-vivo studies. For applications with less demanding transition band requirements, such as 1H MRSI of breast or prostate, the modulating frequency of the RF pulses can be set with less precision.

Although the effectiveness of the achievable water and lipid suppression was demonstrated in an in-vivo brain 1H MRSI study of a normal volunteer using an echo time of 144ms and targeting three primary NMR-detectable proton metabolites of Cho, Cr, and NAA, this suppression scheme constitutes only preparatory RF pulses and dephasing gradients and therefore has the flexibility to be combined with both short and long TE MRSI studies.


We have designed and implemented an extension of the WET technique (2) to provide robust suppression of water and lipids using multiple dual-band frequency-selective presaturation RF pulses for in vivo 3T 1H MRSI studies. The primary advantages of the method are: 1) robust simultaneous suppression of water and lipid resonances with different T1 relaxation times, 2) no loss of metabolite signals or compromised spatial coverage, 3) robust suppression in the presence of ±20% B1 inhomogeneities, 4) use of only presaturation RF pulses and associated dephasing gradients such that the method can be readily combined with a variety of MRS localization schemes.


The study is supported by grants NIH RR09784, CA098523, and the Lucas Foundation.


1. Haase A, Frahm J, Hanicke W, Matthaei D. 1H NMR chemical shift selective (CHESS) imaging. Phys Med Biol. 1985;30(4):341–344. [PubMed]
2. Ogg RJ, Kingsley PB, Taylor JS. WET, a T1- and B1-insensitive water-suppression method for in vivo localized 1H NMR spectroscopy. J Magn Reson B. 1994;104(1):1–10. [PubMed]
3. Bottomley PA. Spatial localization in NMR spectroscopy in vivo. Ann N Y Acad Sci. 1987;508:333–348. [PubMed]
4. Frahm J, Bruhn H, Gyngell ML, Merboldt KD, Hanicke W, Sauter R. Localized high-resolution proton NMR spectroscopy using stimulated echoes: initial applications to human brain in vivo. Magn Reson Med. 1989;9(1):79–93. [PubMed]
5. Luo Y, de Graaf RA, DelaBarre L, Tannus A, Garwood M. BISTRO: an outer-volume suppression method that tolerates RF field inhomogeneity. Magn Reson Med. 2001;45(6):1095–1102. [PubMed]
6. Haupt CI, Schuff N, Weiner MW, Maudsley AA. Removal of lipid artifacts in 1H spectroscopic imaging by data extrapolation. Magn Reson Med. 1996;35(5):678–687. [PMC free article] [PubMed]
7. Bydder GM, Young IR. MR imaging: clinical use of the inversion recovery sequence. J Comput Assist Tomogr. 1985;9(4):659–675. [PubMed]
8. Ebel A, Govindaraju V, Maudsley AA. Comparison of inversion recovery preparation schemes for lipid suppression in 1H MRSI of human brain. Magn Reson Med. 2003;49(5):903–908. [PubMed]
9. de Bazelaire CM, Duhamel GD, Rofsky NM, Alsop DC. MR imaging relaxation times of abdominal and pelvic tissues measured in vivo at 3.0 T: preliminary results. Radiology. 2004;230(3):652–659. [PubMed]
10. Traber F, Block W, Lamerichs R, Gieseke J, Schild HH. 1H metabolite relaxation times at 3.0 tesla: Measurements of T1 and T2 values in normal brain and determination of regional differences in transverse relaxation. J Magn Reson Imaging. 2004;19(5):537–545. [PubMed]
11. Spielman D, Meyer C, Macovski A, Enzmann D. 1H spectroscopic imaging using a spectral-spatial excitation pulse. Magn Reson Med. 1991;18(2):269–279. [PubMed]
12. Star-Lack J, Nelson SJ, Kurhanewicz J, Huang LR, Vigneron DB. Improved water and lipid suppression for 3D PRESS CSI using RF band selective inversion with gradient dephasing (BASING) Magn Reson Med. 1997;38(2):311–321. [PubMed]
13. Smith MA, Gillen J, McMahon MT, Barker PB, Golay X. Simultaneous water and lipid suppression for in vivo brain spectroscopy in humans. Magn Reson Med. 2005;54(3):691–696. [PubMed]
14. Duewell SH, Ceckler TL, Ong K, Wen H, Jaffer FA, Chesnick SA, Balaban RS. Musculoskeletal MR imaging at 4 T and at 1.5 T: comparison of relaxation times and image contrast. Radiology. 1995;196(2):551–555. [PubMed]
15. Bottomley PA, Foster TH, Argersinger RE, Pfeifer LM. A review of normal tissue hydrogen NMR relaxation times and relaxation mechanisms from 1–100 MHz: dependence on tissue type, NMR frequency, temperature, species, excision, and age. Med Phys. 1984;11(4):425–448. [PubMed]
16. Breger RK, Rimm AA, Fischer ME, Papke RA, Haughton VM. T1 and T2 measurements on a 1.5-T commercial MR imager. Radiology. 1989;171(1):273–276. [PubMed]
17. Bryant RG, Mendelson DA, Lester CC. The magnetic field dependence of proton spin relaxation in tissues. Magn Reson Med. 1991;21(1):117–126. [PubMed]
18. Fischer HW, Rinck PA, Van Haverbeke Y, Muller RN. Nuclear relaxation of human brain gray and white matter: analysis of field dependence and implications for MRI. Magn Reson Med. 1990;16(2):317–334. [PubMed]
19. Vaughan JT, Garwood M, Collins CM, Liu W, DelaBarre L, Adriany G, Andersen P, Merkle H, Goebel R, Smith MB, Ugurbil K. 7T vs. 4T: RF power, homogeneity, and signal-to-noise comparison in head images. Magn Reson Med. 2001;46(1):24–30. [PubMed]
20. Pauly JLRP, Nishimura D, Macovski A. Parameter relations for the Shinnar-Le Roux selective excitation pulse design algorithm. Medical Imaging, IEEE Transactions on. 1991;10(1):53–65. [PubMed]
21. Lasserre JB. Global Optimization with Polynomials and the Problem of Moments. SIAM Journal on Optimization. 2000;11(3):796–817.
22. Sturm JF. Using SeDuMi 1.02, a MATLAB toolbox for optimization over symmetric cones. 1999;11–12:625–653.
23. Kim DH, Adalsteinsson E, Glover GH, Spielman DM. Regularized higher-order in vivo shimming. Magn Reson Med. 2002;48(4):715–722. [PubMed]
24. Tkac I, Starcuk Z, Choi IY, Gruetter R. In vivo 1H NMR spectroscopy of rat brain at 1 ms echo time. Magn Reson Med. 1999;41(4):649–656. [PubMed]
25. Pfefferbaum A, Adalsteinsson E, Spielman D, Sullivan EV, Lim KO. In vivo spectroscopic quantification of the N-acetyl moiety, creatine, and choline from large volumes of brain gray and white matter: effects of normal aging. Magn Reson Med. 1999;41(2):276–284. [PubMed]