Improved sensitivity to overlapping multiplet signals in in vivo proton spectroscopy using a multiecho volume selective (CPRESS) experiment

Improved Sensitivity to Overlapping Multiplet Signals in in Vivo Proton Spectroscopy Using a Multiecho Volume Selective (CPRESS) Experiment Jiirgen Hennig, Thorsten Thiel, Oliver Speck

A method for volume selective proton spectroscopy is pre- sented based on a multiecho sequence with short refocusing interval tcp. It is demonstrated, that by appropriate choice of tcp on the order of 4-6 ms, signals from overlapping multiplets like the glutamine and glutamate (Glu/Gln) resonances in spectra of the human brain are considerably increased compared with a conventional PRESS volume selection scheme. Thus proton spectra from J-coupled muFiplet signals can be acquired with TE on the order of 20-30 ms avoiding the baseline problems arising at shorter echo times due to broad resonances. This allows to selectively acquire spectra from substances with longer 1, without the confounding effects from J-coupling occurring in conventional volume selection techniques. Key words: proton spectroscopy; CPRESS; PRESS; brain spectroscopy.

INTRODUCTION

The PRESS sequence (1) is a stable and robust technique for localized in vivo proton spectroscopy. The minimum echo time is defined by the timing required for the three slice selection pulses. In addition sufficient time has to be allowed for additional spoiler pulses, which are used to suppress unwanted signals from outside the examined voxel. With conventional actively shielded gradient systems typical values for TE are on the order of 30 ms.

The signal intensity of the observed metabolites is determined by the relaxation time constant T2, which for most relevant metabolites lies on the order of more than 100 ms (2-3). In addition, signals of J-coupled substances suffer additional loss due to the evolution of the signal phase as a function of the coupling constant J. For overlapping multiplets. this dephasing, which-contrary to most other phase effects-is not refocused by spin echo formation, can be appreciable especially at echo times 2 30 ms. The most prominent victims of this coupling effect in in vivo proton spectroscopy are the multiplets corresponding to glutamatelglutarnine (Glu/Gln). The observed signal intensities of these metabolites are much lower than expected from their known concentrations and their long relaxation times T2. Reducing the echo time to less than 10 ms, which has become feasible using the high performance gradient systems becoming rou-

MRM 37816-820 (1997) From Abt. Rbntgendiagnostik, MR Tomography, University Freiburg, Freiburg. Germany. Address correspondence to: Jurgen Hennig, Ph.D., Abt. Rdntgendiagnos- tik, MR-Tomographie, Hugstetterstr.55, 791 06 Freiburg, Germany. Received July 8,1996; revised February 28, 1997; accepted March 3, 1997.

Copyright 0 1997 by Williams & Wilkins All rights of reproduction in any form reserved.

0740-3194/97 $3.00

tinely available within the last few months, leads to a substantial gain in the signal amplitudes of these resonances. It also leads, however, to the observation of broad lines corresponding to rather unspecific resonances with short relaxation times (4). Although the visibility of J-coupled signals is thus enhanced, even crude spectrum quantification can become extremely difficult.

The purpose of this paper is to propose a modified PRESS sequence, which is insensitive to such J-coupling effects. The CPRESS (Carr Purrcell-selected PRESS) sequence makes use of the behavior of coupled spin systems in a multiecho experiment. It allows the acquisition of volume selective spectra with echo times of 30-40 ms, in which broad resonances are effectively suppressed while at the same time, the high signal intensity of signals from J-coupled substances is maintained.

MATERIALS AND METHODS

The time evolution of a ]-coupled spin system undergo- ing multiple refocusing in a Carr-Purcell-Meiboom-Gill multiecho experiment was described in 1966 by Aller- hand (5). The salient result of his analysis is the fact that the signal within an echo train will undergo a modulation with a modulation frequency %J. The amplitude of the modulation will depend on a factor R tcp, where tcp corresponds to the duration of one refocusing interval and R is given by

J corresponds to the coupling constant and 6 represents the chemical shift difference between the coupled multiplets.

From ref. 5 it follows, that for R tcp < 1 the signal modulation will become negligible. The ]-dependent dephasing of the echo amplitude is thus quenched and the signal amplitude of overlapping multiplets will increase. This mechanism has been used to explain the high lipid signal in RARE (FSE, TSE . . .) imaging sequences, which are based on echo trains with short refocusing intervals (6, 7).

The diagram of the CPRESS experiment is shown in Fig. 1. After slice selective excitation by a 90° pulse and ensuing refocusing by a 180° pulse with an echo interval tcpl under an orthogonal gradient, a multiecho loop with n echoes and a refocusing interval tcpZ is performed before the signal is read out after a final refocusing pulse leading to echo formation after a further time interval tcp3. The total echo time TE is thus given by

TE = tcpl + n tcp2 + tcp3 816

In Vivo MRS with Multiecho Volume Selection (CPRESS) 817

tcpl tcp2 tcp3 I ! I

Rf

7- G Y

I I

FIG. 1. Scheme of the CPRESS volume selection technique. RF denotes radiofrequency pulses and signals, G,, G,, and G, correspond to t h e x. y, and z gradients, respectively. First a slice selective excitation pulse is applied followed by refocusing with a 180" pulse under an orthogonal gradient in a first echo interval tcpl followed by n-fold repetition of a refocusing loop of duration tcp2 and a final echo interval tcp3 to read out the voxel selective signal at a time TE = tcpl + n tcp2 + tcp3. The shaded gradients are crusher gradients used to suppress signal from unwanted coherences. Normally only the second half of the final echo is read out.

The spoiler gradients (shaded) are used for suppres- sion of the unwanted coherences from outside the selected voxel. The distribution of the x, y, and z gradient to the RF pulses shown in Fig. 1 is of course only one of many possibilities. Other arrangements can be chosen according to the performance of the system, the main determining factor being its eddy current behavior. To minimize tcp2, a nonslice-selective refocusing pulse can be used during the refocusing loop, which will, however, increase the intensity of unwanted coherences to be spoiled before the signal readout. It is advisable to use a slice-selective pulse with a profile, which acts the same as the first refocusing pulse before the refocusing loop but with a slightly broader profile to minimize degrada- tion of the voxel shape due to imperfect pulse shapes.

The timing of the sequence is mainly determined by the spoiler gradients necessary to remove unwanted coherences. It has been demonstrated that the area under the spoiler gradient should be on the order of 50-70 mT/m ms for in vivo proton spectroscopy (8). With strong gradient systems using gradients with an amplitude of 25-30 mT/m, effective spoiling can be performed while short refocusing times can still be maintained. It is thus possible to perform the experiment with tcp = tcpl = tcp2 = tcp3 such that the timing of the whole echotrain corresponds to that of a CPMG multiecho sequence. It should be noted that the gradient conditions for coherent superposition of all possible refocusing pathways in such a sequence (9-10) should only be maintained during or until the end of the refocusing loop. If proper gradient symmetry is maintained throughout the whole sequence including spoiler gradients, unwanted coherences from outside the voxel would be refocused as well and might significantly contribute to the observed signal.

For minimizing tcp for a given gradient hardware, it should be noted that the time between the excitation pulse and the first refocusing pulse is the busiest in terms of gradient demand. If this time interval is prolonged

compared with the timing of the rest of the sequence, the echo train will become asymmetrical with respect to refocusing of magnetic field inhomogeneities and chemical shift effects. As has been demonstrated for single shot RARE (U-FLARE) imaging (9), the refocusing pathways during the multiecho loop will then split into two signal groups with a phase difference, which is twice the dephasing during the time increment of the asymmetry. For a chemical shift range of 5 ppm and given the high field homogeneity inside the voxel used in proton spectroscopy, this dephasing will become appreciable, when the asymmetry in the timing significantly exceeds 1-2 ms leading to chemical shift dependent signal attenuation.

EXPERIMENTAL

Experiments were performed using a symmetrical CPRESS sequence with tcp = tcpl = tcp2 = tcp3 on a 2T whole body system (Bruker S200 Avance) equipped with a gradient head coil (30 mT/m, 190 w s rise time). The signal behavior of CPRESS versus PRESS was examined first on a glutamate solution (30 mmol/liter). In addition CPRESS brain spectra (voxel size 2 x 2 x 2 cm3) of normal volunteers (n = 8) were acquired and compared with spectra from a conventional PRESS experiment acquired in the same experimental session and with other- wise identical acquisition parameters. A minimum refocusing time of 4 ms was used leading to a minimal TE of 8 ms for the PRESS experiment.

Spectra were processed as follows: An FID of the water signal from the voxel of interest was acquired (eight averages) and used as reference for a point-by-point correction of the metabolite signals before Fourier transfor- mation (FT) to correct for residual eddy current effects. FT was then performed with 1.5 Hz Lorentzian line- broadening. No additional postprocessing was performed except for a slight zero order phase correction.

For quantitation of the in vivo spectra the LC algorithm (Version 5.1-6) by Provencher (11) was used. Reference spectra for the evaluation were acquired on solutions of the corresponding metabolites using identical measurement parameters as in the in vivo spectra. They consist of the following 11 major metabolites: alanine, aspartate, choline (Cho), creatine (Cre), yaminobutyric acid, glutamine (Gln), glutamate (Glu), myo-hositol (Myl), lactate, N-acetylaspartate (NAA), and taurine. The analysis window in the ke- quency domain was kom 4.2 to 1.0 ppm and no prior knowledge about phase correction values was used. Cali- bration of the results for absolute quantitation was not at- tempted for these initial comparative measurements.

RESULTS

Figure 2 shows a comparison of in vitro PRESS and CPRESS spectra of glutamate with identical TE. The CPRESS spectra were acquired with tcp = 5 ms. The variation of TE was achieved by increasing the number of refocusing loops n. The spectra illustrate the temporal evolution of the phase of the coupled multiplets in the PRESS spectra leading to inversion of some of the overlapping lines for echo times greater than 40 ms. The CPRESS spectra in comparison show identical shapes for

818 Hennig et al.

PRESS CPRESS

- - -~ ~. -- 5 0 40 30 2 0 10 50 4 0 3 0 2 0 1 0

PPm PPm

FIG. 2. Comparison of conventional PRESS spectra (left) with spectra acquired with CPRESS (right) as a function of the echo time TE. Spectra were acquired from a 2 x 2 x 2 ml voxel in a vessel filled with glutamate solution (30 rnrnoVliter). For CPRESS an echo interval tcp of 5 m s was used throughout the experiments. Longer echo times were reached by increasing the number n of CPMG refocusing intervals.

the multiplet signals even for echo times as long as 270 rns.

For quantitation of the spectra, integration was performed over the range of 1.7-2.7 ppm. The results are shown in Fig. 3. It is clearly demonstrated, that CPRESS

I(a.u.)

500

te [ms] 0 0 50 100 150 200 250

FIG. 3. Comparison of signal intensities from the spectra shown in Fig. 2. The intensity values were derived as the integrals over the spectra in the range between 1.7-2.7 pprn. The circles correspond to the values of the CPRESS measurements, squares are intensities measured by PRESS.

gives much higher signals for echo times TE 2 30 ms. The very fast signal decay of the in vitro PRESS spectra could be ameliorated at least in principle by taking the integral over the magnitude rather than the real part of the spectra. Magnitude calculation leads, however, to very broad “feet” of the spectra, which severely affects the definition of the spectrum baseline and which would be unfeasible for in vivo applications anyway.

Figure 4 shows the results of in vivo experiments performed on a normal volunteer. The in vivo PRESS spectra again show the fast decay of the multiplets in the range between 2.0-2.4 pprn and 3.6-4.0 pprn mainly attributed to signals from glutamate and glutamine. The CPRESS spectra acquired with tcp = 5 ms demonstrate high intensities for these multiplets even at echo times TE = 135 ms.

Figure 5 shows the result of fitting glutamate reference spectra to the in vivo spectra using the LC model by Provencher. For comparison, in vitro glutamate spectra with TE = 10 ms and TE = 60 rns were taken as reference spectra for the fitting algorithm. It is demonstrated that both reference spectra lead to nearly identical results over the whole range of echo times from 10-270 ms. This makes data evaluation for measurements at different echo times (i-e., for T, measurements) considerably eas- ier and is in marked contrast to PRESS spectra, were a different set of model functions is required for every echo time (Fig. 2).

CPRESS PRESS

te [ms]

10

m/c-h I 40 n kf3

60 fiU,L 10

4.0 3.0 2.0 1.0 PPm

4.0 3.0 2.0 1.0 ppm

FIG. 4. PRESS and CPRESS spectra from the brain of a healthy volunteer (2 x 2 x 2 ml). All spectra were acquired in a single session at identical voxel position. A repetition time of TR = 1500 rns was used, 64 averages were acquired. Note the high intensity of the Glu/Gln resonances in the CPRESS spectra even at TE =

135 ms. n is the number of CPMG refocusing intervals.

In Vivo MRS with Multiecho Volume Selection (CPRESS)

400 ~

300

200 ~

100 -

819

I(a.u.)

800 1

6oo 500 i a -.

I 1 ' 1

0 50 100 150 200 250 te[ms]

FIG. 5. Intensity values in arbitrary units for glutamate calculated with the LC model algorithm for the CPRESS spectra shown in Fig. 4. The evaluation was performed with model spectra acquired at TE = 10 ms (circles) and TE = 60 ms (squares). The curve represents a biexponential fit as described in the text.

The signal intensities can be fitted with a biexponential decay curve, where the faster decaying component (T , = 16 ms] can be attributed to macromolecules and the slow component (T , = 255 ms) corresponds to the effective T, of glutamate. No significant difference was observed between the T, values measured for singlet resonances like NAA with CPRESS versus PRESS, which were measured to be in the range of 350-400 ms (N- acetylaspartate) and 220-270 ms (creatine/phospocreat- ine) in accordance with literature values. An exact deter- mination of absolute concentrations and relaxation rates was beyond the purpose of this study.

The results of the ratios of relevant metabolites to creatine are listed in Table 1. It is demonstrated, that the measured (Glu + Gln)/Cre is significantly higher with CPRESS versus PRESS with increasing TE. The standard deviations given reflect averaged values of the intraindi- vidual fits [n = 8) and not inter-individual variations. A separate set of model spectra acquired with identical parameters as the in vivo experiments was used for each experimental condition.

Figure 6 shows CPRESS spectra acquired with identical echo time TE = 60 ms but with different tcp. The

spectrum at the bottom with tcp = 30 ms corresponds to a conventional PRESS experiment. It is demonstrated that the multiplet signals start to be appreciably reduced at tcp = 6 ms and more or less vanish for longer refocusing intervals.

DISCUSSION

The CPRESS sequence significantly increases the visibility of overlapping multiplet signals like those of glutamate/glutamine. The TE dependence of the observed multiplet signals corresponds to an exponential decay with a T2 relaxation time as expected for small molecules and is well within the range of other observed metabolite signals. The fast signal drop with TE as observed in conventional PRESS experiments is thus avoided.

A very convenient feature of CPRESS is the fact that the spectral shape of multiplets does not change with TE, when tcp is sufficiently short. This allows spectrum quantitation at any echo time using the LC model with a single reference spectrum. It is very encouraging to note that glutamate could be reliably fitted to the in vivo spectra in spite of the heavy overlap with signals from glutamine. In this study no attempt has been made to convert the arbitrary intensity units delivered by the fitting algorithm into absolute concentration values. With appropriate calibration techniques described in the literature using an internal or external reference, absolute quantitation of molecules with coupled spin systems appears to be feasible, although further studies are required to establish the usefulness of CPRESS for absolute quantitation.

CPRESS can thus be used to acquire proton spectra with high Glu/Gln with a TE of 20-40 ms. At these echo times, the broad resonances from macromolecules that prohibit quantification of signals in spectra acquired with very short TE have already vanished. An additional benefit to the increase in signal-to-noise is the fact that the phase of multiplet signals acquired with short tcp is nearly identical to that of singlets. Phase correction for these signals is therefore much more easy and reliable compared with PRESS spectra.

An additional potential application of CPRESS is the possibility to determine diffusion constants of metabolites by acquiring spectra with variable tcp at a constant

Table 1 Ratios of Metabolite Concentrations to Creatine for Relevant Metabolites as a Function of the Echo Time TE for PRESS versus CPRESS

TE in m s NadCre Cho/Cre My l/Cre Glu/Cre Gln/Cre (Glu + Gln)/Cre PRESS 10 1.33 f 0.08 0.13 2 0.01 0.59 2 0.05 0.61 f 0.08 0.53 ? 0.14 1.14 f 0.22 30 1.39 f 0.07 0.13 % 0.01 0.52 2 0.05 0.38 5 0.06 0.38 5 0.09 0.76 f 0.15 60 1.34 f 0.07 0.12 ? 0.01 0.25 -t 0.1 1 0.05 f 0.08 0.10 c 0.10 0.16f 0.11

10 1.33 f 0.08 0.13 2 0.01 0.59 ? 0.05 0.61 f 0.08 0.53 c 0.14 1.14 f 0.22 20 1.56 2 0.10 0.14 2 0.02 0.68 f 0.07 0.56 5 0.08 0.64 ? 0.16 1.21 ? 0.24 30 1.51 f 0.10 0.13 2 0.02 0.56 -C 0.08 0.44 2 0.08 0.73 c 0.18 1.16 2 0.26 40 1.57 f 0.12 0.14 2 0.02 0.59 f 0.08 0.52 5 0.08 0.44 2 0.17 0.96 -t 0.26 60 1.43 f 0.1 1 0.15 2 0.02 0.59 f 0.07 0.52 5 0.08 0.47 z 0.21 0.98 -t 0.30 80 1.39 ? 0.1 0 0.12 2 0.01 0.61 z 0.08 0.47 f 0.08 0.34 ? 0.15 0.81 2 0.25

CPRESS

The data represent average values of measurements in eight volunteers. Errors also represent averages of intra-individual standard deviations (and not the inter-individual spread of data). Echo time variations for CPRESS were performed by increasing the number of CPMG refocusing intervals at a constant echo spacing of tcp = 5 ms.

820 Hennig et al.

n

7.5

10

15

30

FIG. 6. CPRESS spectra with TE = 60 m s as a function of tcp. The signal from Glu/Gln drops off appreciably for tcp > 5 ms. Acquisition parameters were identical to those in Fig. 4.

overall echo time TE (12). A fast refocusing loop can of course also be implemented into a STEAM localization experiment or a CSI experiment.

I t should be noted, that the coupling pattern of strongly coupled multiplets depends on the difference in chemical shift [see Eq. 11 and thus on the field strength. For glutamatelglutarnine i t has been reported, that !-coupling dependent signal loss on high-field systems (7T) occurs already at echo times as short as 10 ms (13). For a larger chemical shift separation. shorter echo intervals tcp have to be chosen to avoid modulation-dependent signal loss. On high field systems (Bl, >> 2T), the application of CPRESS therefore might be limited by SAR problems. In the field range of conventional magnets. the realization of tcp values between 4-6 nis according to Fig. 6 consti- tutes no real problem at least as long as the number of refocusing loops and thus the overall echo time stays within reasonable limits.

A number of studies have demonstrated the impor- tance of changes in the concentration of Glu/Gln (14-21). CPRESS promises to significantly increase the signal- to-noise ratio and the reproducibility of such spectra. The ability to distinguish fast initial signal decay due to !-coupling effects from true T , decay also offers consid- erable improvements in the analysis of broad underlying signals commonly attributed to macromolecules.

REFERENCES

1 . P. A . Bottomlev. T. ti. Fiistor. W. M. Leue. In vivci nuclear magnetic rosonance chemical shift imaging hy selective irradiatinn. Proc. N. Y. A m d . Sci. 81. tiB5ti-GHRIl (l9R4).

2. 1. Frahm. H. Druhn. M. L. Gyngell. K . D. Merboldt. W. Hanicke. R.

Sauter, Localized proton NMR spectroscopy in different regions of the hunian brain in vivn. Relax- ation times anti concentrations of r:nrc!hral metabolites. Mogn. Reson.

R. h i s . T. Ernst, B. D. Ross, Devel- opment of the human brain: in vivo quantification of metabolite and water content with proton magnetic resonance spectroscopy. ibfugn. ReSOJl.

K. L. Bohar. D. L. Rothman. D. D. Spencer. 0. A. C. Petroff. Analysis of macromolecule resonances in 'H NMR spectra of human brain. Mqyn. Reson. Aged. 32, 294-302 (1994). A. Allerhand. Analysis of Carr-Pur- cell spin-r!c:ho NMR expcrimrnts on multiple-spin systems. 1. The effec:t of homonuclear coupling. 1. Chem. Phys. 44, 1-9 (1966). P. S. Melki. R. V. Mulkern. L. P. Panych. F. A. Jolesz. Comparing the PAISE method with conventional dual-echo sequences. 1. Mogn. Rf!-

R. M. Henkclman, P. A. Hardy. J. E. Bishop. C. S. Poun, D. B. Plewes. Why fat is bright in RARE and fast spin-echo imaging. I . Mogn. Reson. ImoEing 2, 533-540 (1992).

M d . 11, 47-ti3 (1989).

Msd. 30, 424-437 [1993).

son. rmlnging 1, 319-326 ( 1 ~ 9 1 ) .

_ _ 8. T. Ernst. L. Chang. Elimination of artifacts in shart echo time 'H MR

spectroscopy of the fmntal lobe. M a p . Reson. Med. 36,462-468 (1996). 9. D. G. Norris. 1'. Bornert, T. Reese. D. Leibfritz. On the application of

ultrafast RARE experiments. Mogn. Reson. Med. 27, 142-164 (1992). 10. J. Hennig. Echoes-how to generate. recognize. use or avoid them in

MR-imaging sequences. Concepts M o p . Reson. 3, 179-192 (1991). 11. S. W. Provoncher. Estimation of metabolite concentrations from lo-

calized in vivo proton NMR spectra. Mogn. Reson. Med. 30,672-679

12. 13.

14.

15.

16.

17.

18.

19.

20.

21.

(1993). E. L. Hahn, Spin echoes. Phys. Rev. 80, 580-594 (1950). W. Q. Both. R. Lamerichs. P. R. Luyten, W. M. M. J. Bovne. Optimal sigiial strength for glutamate and glutamine in the PRESS and STEAM sequence at 7 and 1.5 T. in "Proc.. SMRM. 12th Annual Meeting, 1993," p. 517. R. Kreis. B. D. Ross, N. A. Farrow, 2. Ackerman. Metabolic disorders of the brain in chronic hepatic encephalopathy detected with H-1 MR spectroscopy. Radiology 182, 19-27 (1992). R. A. Moats, T. Ernst, T. K. Shonk. B. D. Ross. Abnormal cerebral metabolite concentrations in patients with probable Alzheimer dis- [vase. M o p . Reson. Med. 32, 110-115 (1994). B. 0. Ross. S. Jacobson. F. Villamil. J . Kornla. R. Kreis. T. Ernst. 1'. Shonk. R. A. Moats. Subclinical hepatic encephalopathy: proton MR spectroscopic ahnormalities. Rodiolog.~ 193, 457-463 (1994). S. D. Taylor-Robinson, 1. Sargentoni. C. D. Marcus. M. Y, Morgan, D. J. Bryant. Regional variations in cerebral proton spectroscopy in patients with chronic hepatic encephalopathy. Mefob. Brain Dis. 9. 347-359 (1994). D. Hailssinger. J. Laubcnberger, S. vom Dahl, 1'. Ernst, S. Baver. M. Langer. W. Gcrok. J. Hennig. Proton magnetic resonance spectroscopy studies on human brain myo-inositol in hypo-osniolarity and hepatic encephalopathy. Gastroenterology 107, 1475-1480 (1994). S. D. Taylor-Robinson. R. A. Weeks. J . Sargentoni. C. D. Marcus, D. J . Bryant. A. E. Harding, D. J , Brooks, Evidence for glutamate excit~~trix- icity in Huntington's disease with proton magnctir resonance spectroscopy. Loncet 343. 11 70 (1994). tk:2R. A. Moats. L. Watson, T. Shonk. S. Tokuyama. D. Braslati. R. Eto. J. C. Mandigo. B. D. Ross, Added value of automated clinical proton MR spectroscopy of the brain. \. Conrp. Assist. 7'oniogr. 19, 480-491 (1995). P. Mnhanakrishnan, A. H. Fowler. J. P. Vonsattel. M. M. Husain, P. R. lolles. P. Liem. R. A. Komoroski. An in vitro 'H nuclear magnetic resnnance study of the temporoparietal cortex of Alzheimcr brains. Esp. Brain RPS. 102, 503-510 (1995).

Documents

Improved sensitivity to overlapping multiplet signals in in vivo proton spectroscopy using a multiecho volume selective (CPRESS) experiment