MAGNETICRESONANCE IN MEDICINEzS,65-73 ( 1992)
Glycogen Detection by in Vivo 13CNMR A Comparison of Proton Decoupling and Polarization Transfer M.SANER, G. MCKINNON,AND P. BOESIGER Institute of Biomedical Engineering and Medical Informatics, Universityand ETH Zurich, Moussonstrasse 18, CH-8044 Zurich, Switzerland Received July 12, 1991; revised February 5, 1992; accepted February 5, 1992 The performance of gated proton decoupling and polarization transfer with respect to glycogen detection by 'F NMR was investigated. Experiments were performed on a 1.5T whole-body scanner using a 13Csurface coil in combination with a proton head coil. Spectra were acquired from a glycogen phantom and from the lower leg of a healthy volunteer using proton decoupling and the polarization transfer method SINEFT. The signal strength of the C , resonance of glycogen was determined and compared to a reference spectrum acquired without any form of sensitivityenhancement.In the phantom experiment both decouplingand SINEPT produced a signal gain of 3.5. Under in vivo conditions, the signal gain was approximately 2.5 for both techniques. We conclude that decoupling and polarization transfer are equivalently useful techniques for glycogen detection. o 1992 Academic Pms, Inc.
INTRODUCTION
Glycogen, the storage form of glucose, plays a central role in human energy metabolism. Normal concentrations of glycogen, expressed by the concentration of its glucosyl units, range from 60 to 120 mmol/liter (mean 84 mmol/liter) in human skeletal muscle (1,2) and from 230 to 300 mmol/liter in the human liver ( 3 , 4 ) . Studieson rats (5-7) and rabbits (8)have demonstrated that glycogen can be detected under in vivo conditions by I3C NMR. It has also been shown that glycogen is 100% NMR visible despite its large molecular weight (9, 10). Detection of glycogen in human skeletal muscle and liver by 13C NMR spectroscopy has been reported by several groups ( 11-22). The main problem of detecting glycogen is the low sensitivity which is a consequence of the low natural abundance ( 1.1%) of I3C and its low gyromagnetic constant y (a factor of 4 smaller than protons). While a few glycogen studies on humans have been performed without sensitivity enhancement techniques ( 11-15), mainly proton decoupling (16-22), but also polarization transfer (23-27), has been applied to increase the low sensitivity of 13C. Decoupling is used routinely in high-resolution I3CNMR. By applying RF irradiation at the proton resonance frequency the heteronuclear spin-spin coupling is suppressed and multiplets collapse into single resonance lines. Thus spectra are simplified considerably, and simultaneously the signal-to-noise ratio is improved. For a CH doublet (such as the C1 resonance of glycogen) a signal gain by a factor of 2 can be obtained by decoupling during the signal acquisition period. 65
0740-3 194192 $5.00 Copyright Q 1992 by Academic Rss,Inc. All rights of reproduction in any form reserved.
66
SANER, MCKINNON, AND BOESIGER
Polarization transfer by RF pulses is a relatively new technique which can be used for sensitivity enhancement and spectral editing. Here the spin system is driven into a state of nonequilibrium populations by several RF pulses, which leads to increased population differences for the 13C transitions and therefore to an increased signal intensity in the 13C spectrum. As opposed to decoupling there is no RF irradiation during signal acquisition. For a CH doublet a gain factor of yIH/y"' x 4 can be obtained. Commonly used pulse sequences are, INEPT (insensitive nuclei enhanced by polarization transfer) (28, 2 9 ) , SINEPT (sine-dependent polarization transfer) (30), and DEPT (distortionless enhancement by polarization transfer) (31). Here we report on a comparison of proton decoupling and polarization transfer, by the SINEPT pulse sequence, with respect to their sensitivity enhancement for the glycogen C, resonance. SINEPT was chosen for several reasons. The sequence is very simple, since it consists of only three RF pulses. It requires no 180" I3C pulse (as opposed to INEPT or DEPT) and can therefore easily be performed with a 13Csurface coil. Moreover, by a sequential arrangement of the RF pulses the sequence can be implemented on a single-channel MR system. METHODS
Measurements were performed on a Philips Gyroscan S15 ( 1.5 T ) imaging/spectroscopy system. An additional transmitter channel, consisting of a frequency synthesizer (SMG, Rhode & Schwarz), a homebuilt phase modulator, and a power amplifier ( 150 LA, Amplifier Research), was used for proton decoupling. Polarization transfer experiments were performed without hardware modifications. The phantom consisted of a rectangular bottle filled with 90 mmol/liter glycogen (Fluka AG, Buchs, Switzerland) and 66 mmol/liter sodium chloride. This NaCl concentration produced approximately the same loading of the 13Ccoil as that of the leg of a volunteer. The in vivo spectra were acquired from the lower leg of a healthy volunteer (age 33, weight 57 kg, male) who did not follow any specific diet. The examination was performed in the afternoon, about 2 h after a meal. A proton head coil was used both for decoupling and polarization transfer experiments. The quality factor Q of the empty coil was 400, while the Q of the coil loaded with the leg of a volunteer was typically in the range of 70-90. A two-turn, 7-cmdiameter surface coil was used for 13Cexcitation and detection. The coil was constructed from 3-mm copper wire and mounted inside a Plexiglas box. A balanced tuning configuration was used in order to minimize dielectrical losses. The quality factor of the empty 13Ccoil was 300, while the loaded Q was typically between 100 and 120. The three pulse sequences which were investigated are shown in Fig. 1. A simple pulse-acquire experiment (Fig. 1A)without any sensitivity enhancement was performed to obtain a reference spectrum. A fast half-passage adiabatic excitation pulse (hyperbolic tangent type; length, 3 ms; frequency sweep range, 1500 Hz) was used for I3Cexcitation. The same pulse was also used as the readout pulse for decoupling and polarization transfer experiments. Gated CW proton decoupling (Fig. 1B) was performed during the signal acquisition time of 64 ms. The proton frequency was set on-resonance for the H1 proton of glycogen at 5.4 ppm (32). The decoupler power delivered to the coil was 32 W, which resulted in a decoupler field yB2/27r of approximately 400 Hz for the phantom and
PROTON DECOUPLING VS POLARIZATION TRANSFER
67
A
Decoupling
B
'H
FIG. 1. Pulse sequences used for glycogen detection: (A) simple pulse-acquire method; ( B ) gated CW proton decoupling; (C) polarization transfer by SINEPT-SP.
230 Hz for the in vivo measurement. The power absorbed by the tissue is given by Ptissue = ptotal- ( 1 - Qloaded/Qunloaded). For Qloaded/Qunloaded = 90/400 this amounts to 24.8 W. Taking into account the decoupler duty cycle of 47.4%, a mean power absorption of 1 1.8 W results. This power is absorbed by the tissue inside the decoupler coil which had a mass of approximately 3 kg. The average power deposition is therefore 3.9 W/kg. The FDA guidelines (see, e.g., ( 1 9 ) and references therein) require the specific absorption rate SAR in the extremities to be below 8 W/kg for any gram of tissue. We estimated the ratio of maximum SAR to average SAR to be in the order of 2 for our experimental setup. Thus the maximum power absorption was just below FDA guidelines in this experiment. We would like to point out that a volume coil with a homogeneous BI field has the lowest ratio between average and maximum SAR. This ratio will in general be much greater than 2 if a surface coil is used for decoupling (see Fig. 6 in ( 1 9 ) ) . The SINEPT-SP experiment (SINEPT with sequential pulses) shown in Fig. 1C is a modified version of the original SINEPT pulse sequence (30). Because of the sequential arrangement of RF pulses the experiment can easily be implemented on single transmitter channel MR systems. The experiment was performed with tl = 2.5 ms and t2 = 8 ms. The first delay t l was chosen accordingto tPpt = (l/?rJ)arctan(a J T 2 ) ( 3 4 ) .The coupling constant f is 170 Hz for C, of glycogen, and a value of 8 ms for the transverse relaxation time TiHfor the HI proton of glycogen was assumed ( 3 3 ) . The second delay t 2 , which can be chosen arbitrarily, was set to the minimal settling time required for the GYROSCAN demodulator after a change of system frequency from 64 to 16 MHz. In order to maximize the sensitivity for one line of the glycogen doublet, no attempt was made to separate signal contributions from polarization transfer and initial 13C
68
SANER, MCKINNON, AND BOESIGER
magnetization. Therefore, neglecting relaxation effects, a -35 antiphase doublet can be expected for a CH group, which results from the superposition of a -4:4 antiphase doublet (from polarization transfer) and a 1: 1 in-phase doublet (from initial I3Cmagnetization). Localized shimming was performed on a volume of interest of 8 X 8 X 5 cm using the Autoshim function of the GYROSCAN spectroscopy software. Proton linewidths (FWHM) were 10 Hz in the phantom and 14 Hz in the leg. For each spectrum 5 12 complex datapoints were acquired with a sampling rate of 8 kHz. For the phantom, 2048 measurements were averaged and for the in vivo spectra, 4096 measurements. The repetition time was 135 ms in both cases, resulting in a total measurement time of about 10 min for the in vivo spectra. Time domain data was zero-filled to 1024 points and multiplied by a Gaussian filter with 20-Hz line broadening. Manual phase correction and a polynomial baseline correction was performed on all spectra. RESULTS
Spectra acquired from a glycogen phantom are shown in Fig. 2. On the left side the theoretical doublet intensity ratios for the three pulse sequences are shown: the 1: 1 doublet of the one-pulse experiment is expected to be converted into a singlet of intensity 2 by gated decoupling and into a - 3 5 antiphase doublet by SINEPT. Spectrum A is a reference spectrum which was acquired without any form of proton irradiation. The two lines of the glycogen C , doublet can be observed at 95 and at
1
1
1
1
1
1
1
1
1
1
260 240 220 200 180 160 140 120 100 80
1
1
1
1
60
40
20
0
[ppm]
FIG.2. Spectra acquired from a glycogen phantom using different methods:(A) one-pulse spectrum (only I3C excitation pulse); (B) spectrum acquired using CW proton decoupling, the 'H frequency was set onresonance for the H I proton of glycogen; (C) SINEPT-SP spectrum, the pulse sequence was optimized for the CI resonance of glycogen. Measurements (2048) were performed with a repetition time of 135 ms. A Gaussian filter with 20 Hz line broadening was applied.
69
PROTON DECOUPLING VS POLARIZATION TRANSFER
106 ppm. The signal-to-noise ratio (signal height divided by twice the rms noise) is approximately 4. The other glycogen resonances ( C2 at 72 ppm, C3 at 74 ppm, C4 at 78 ppm, C5at 72 ppm, and C6 at 6 1 ppm, according to ( 9 ) )overlap in part and cannot be identified unambiguously. Spectrum B was obtained using CW proton decoupling. As expected, the C , resonance appears as a singlet, centered at 101 ppm. The signal gain with respect to the reference spectrum is 3.5. A gain by more than a factor of 2 (which could be expected for gated decoupling) was also observed in (20) and explained by the suppression of C-H long-range coupling. We assume that due to the relatively high decoupler duty cycle in our experiment there is also a partial nuclear Overhauser enhancement (NOE). In the SINEPT-SP spectrum (C) the C, resonance of glycogen can be observed as an antiphase doublet centered at 101 ppm. There is positive interference for the right line of the doublet and partial cancellation for the left line. The signal gain with respect to the reference spectrum is 3.5 for the right line. Spectra acquired from the lower leg of a healthy volunteer are shown in Fig. 3. Full-scale spectra are shown on the left, an expanded view on the right. All three spectra are dominated by lipid resonances: carboxyl groups (COOR) at 172 ppm, mono- and polyunsaturated fatty acids (C = C ) at 130 ppm, C2 of glycerol at 69 ppm, C1,3of glycerol at 62 ppm, and (CH,), carbons of fatty acids at 30 ppm (35).The resonance of (Phospho)Creatin at 157 ppm is also visible in all three spectra.
A
1
I
I
I
200
150
100
50
I
0 bpml
I
180
l
l
160 140
I
I
I
120
100
80
bpml
FIG.3. In vivo I3C spectra acquired from the lower leg of a volunteer: (A) one-pulse spectrum (only "C excitation pulse); ( B ) proton decoupled spectrum; ( C ) SINEPT-SP spectrum. Measurements (4096) were performed with a repetition time of 135 ms. A Gaussian filter with 20 Hz line broadening was applied.
70
SANER, MCKINNON, AND BOESIGER
In the reference spectrum ( A ) both lines of the glycogen C I doublet can be observed at 95 and 106 ppm. The C2 - C6 resonances of glycogen are overlapped by the intense glycerol lines. In the proton-decoupled spectrum ( B ) the C I resonance of glycogen shows up as a singlet at 101 ppm. The signal gain with respect to one line of the doublet in Fig. 3A is approximately a factor of 2.3. At 118 ppm the C4 resonance of histidine is visible (22). Interestingly there is also a slight signal enhancement for the quaternary ( P ) C r peak at 157 ppm. Additionally there is an unassigned resonance at 142 ppm which can also be observed in the muscle spectra of a marathon runner in (20) but not in the muscle spectra shown in (22). In the SINEPT-SP spectrum ( C ) resonances appear as antiphase multiplets. They are not point-symmetrical since the signal contributions from initial 13Cmagnetization and polarization transfer are superimposed. For the same reason quaternary resonances (e.g., at 172 ppm) do not vanish. There is positive interference for the right line of the glycogen C I doublet at 95 ppm which can be identified unambiguously. The signal intensity is comparable to the singlet in the decoupled spectrum, the gain with respect to the reference spectrum is approximately 2.5. The left line of the doublet at 106 ppm nearly vanishes, but is positive, as opposed to the phantom spectrum, where it is clearly negative. This indicates that the contribution from polarization transfer is smaller than in the phantom measurement. The massive signal gain for the lipid doublet at 130 ppm indicates the correct setting of the proton transmitter frequency. The corresponding lipid protons have the same chemical shift (5.4 ppm) and approximately the same coupling constant as the H I proton of glycogen ( 3 2 ) . The efficiency of polarization transfer for C , of glycogen is significantly smaller than the maximum value predicted by theory, which has been observed before (2 5 , 3 3 ) . There are two loss mechanisms which contribute to this effect: there is no refocusing of IH magnetization in the SINEPT experiment (as opposed to INEPT (28, 2 9 ) ) . Thus main field inhomogeneities will cause a signal loss by dephasing. However, t l is only 2.5 ms and a water linewidth of 14 Hz corresponds to a decay time constant of 1/xAu = 23 ms (assuming a Lorentzian lineshape). Thus, dephasing during t l only accounts for 10% of the signal loss. We therefore conclude that the principal loss is caused by TT relaxation of the proton magnetization between the two 90" pulses. As an estimate of the relaxation time constant T ; , a value of 8 ms was found, based on a series of SINEPT-SP experiments with variable t l ( 3 3 ) .Using this value, the efficiency of the t l period, E l = sin( x J t l ) exp( -rl / T ; ) , can be calculated. A value of 71% is obtained. For the signal decay during the f 2 interval, where the spin system is in a state of longitudinal two-spin order. a time constant of 50 ms was found. With t2 = 8 ms the efficiency in this interval is 85%. Thus an overall efficiency of polarization transfer of 60% can be expected. corresponding to a signal gain by a factor of 2.4. If the contributions from initial 13C magnetization and polarization transfer are superimposed, this results in a doublet intensity ratio of - 1.4:3.4. While the phantom measurement is in good agreement with these values, the in vivo experiment shows an even smaller con-
-
PROTON DECOUPLING VS POLARIZATION TRANSFER
71
tribution from polarization transfer, resulting in a doublet intensity ratio of approximately 0:2. We assume that the T ; of the glycogen H I proton may be even shorter than 8 ms under in vivo conditions. DISCUSSION
The two methods investigated in this study, proton decoupling and polarization transfer using SINEPT, both produce a signal gain by more than a factor of 2 for the C, resonance of glycogen under in vivo conditions. This corresponds to a reduction in measurement time by more than a factor of 4. Therefore we consider decoupling and SINEPT to be alternatives for this application. The actual acquisition time for a spectrum from the human calf can be made as small as 10 min for either one of the two methods. Thus dynamic studies with a reasonably fine temporal resolution become feasible. The major advantages of proton decoupling are the simplification of the spectra by the conversion of multiplets into single lines and the robustness of the method. However. decoupling has also two major drawbacks: there is an inherent danger of tissue damage by excessive absorption of R F power and the technique requires a double channel transmitter system which is not available on the vast majority of installed whole-body MR systems. In contrast, due to the sequential arrangement of excitation pulses, SINEPT-SP can be performed on any single-channel MR scanner which allows a fast change of the system frequency. The hardware requirements are minimal: either a double-tuned coil or two coils ( ‘H and 13C)and a frequency-selective splitter/combiner (36) are required but no additional transmitter hardware. As opposed to decoupling there is no extended period of ‘H irradiation in the SINEPT-SP pulse sequence. The two discrete 90” proton pulses do not produce any significant tissue heating. Therefore short repetition times in the order of 100 to 150 ms can be used. The minimum is not dictated by the need to reduce the duty cycle of the decoupler (20, 22). The major drawbacks of SINEPT-SP are: ( i ) the method can be optimized for one resonance only; (ii ) the splitting of resonance lines into multiplets is not removed, and therefore crowded spectra are obtained; (iii) the method produces so-called antiphase multiplets, i.e., there are resonance lines with negative intensities in the spectrum and there is a possibility for signal cancellation; (iv) SINEPT is not a robust technique since polarization transfer is highly dependent on the ‘H transmitter frequency. However. for the specific application of glycogen detection, the consequences of these properties are not too severe: ( i ) there is only one resonance of interest, the C , line at 101 ppm; (ii) the CI doublet of glycogen is located in a clear “spectral window,” relatively far away from intense lipid resonances; (iii) the only resonance which could possibly interfere with the C, resonance of glycogen centered at 101 ppm is the histidine resonance at 1 18 ppm (20,22).However, if the right line (at 95 ppm) of the glycogen doublet is selected for enhancement, any cancellation of the left line becomes irrelevant; (iv) by localized shimming over a relatively small volume and monitoring the lipid resonance at 130 ppm the correct setting of the ‘H transmitter frequency can easily be verified. In conclusion, we have shown that both proton decoupling and polarization transfer by SINEPT-SP can be used for the in vivo detection of glycogen. For the C1resonance
72
SANER, MCKINNON, AND BOESIGER
both methods produce an increase in signal strength by more than a factor of 2. We believe that proton decoupling is the method of choice, provided that the necessary hardware is available and that excessive RF power deposition can be excluded. Polarization transfer by SINEPT-SP is a powerful alternative. It is an inherently safe technique which is equivalent with respect to sensitivity, but less demanding in terms of hardware requirements. ACKNOWLEDGMENTS We thank Dr. Nicolau Beckmann from the Biozentrum, Basel, and Beat Werner from the University Children’s Hospital, Zurich, for helpful discussions. Support through the Swiss National Science Foundation (NFP-18) is gratefully acknowledged. REFERENCES
E. HULTMAN AND J. BERGSTROM, Acra Medica Scand. 182, 109 ( 1967). E. HULTMAN, Scand. J. Clin. Lab. Invest. 19, 209 ( 1967). L. NILSSONAND E. HULTMAN, Scand. J. C h . Lab. Invest. 32, 325 ( 1973). L. NILSSONAND E. HULTMAN, Scand. J. Clin. Lab. Invest. 33, 5 ( 1974). J. R. ALGER,L. 0. SILLERUD, K. L. BEHAR,R. J. GILLIES,AND R. G. SHULMAN, Science 214, 660 (1981). 6. P. CANIONI, J. R. ALGER, AND R. G. SHULMAN, Biochemistry 22,4974 ( 1983). 7. N. V. REO, B. A. SIEGFRIED, AND J. J. H. ACKERMAN, J. Biol. Chem. 259, 13664 ( 1984). 8. J. R. ALGER, K. L. BEHAR,D. L. ROTHMAN, AND R. G. SHULMAN, J. Magn. Reson. 56, 334 (1984). 9. L. 0. SILLERUD AND R. G. SHULMAN, Biochemistry 22, 1087 ( 1983). 10. R. A. SHALWITZ, N. V. REO, N. N. BECKER,AND J. J. H. ACKERMAN,Magn. Reson. Med. 5, 462 (1987). 11. T. JUE,A. B. LOHMAN,R. J. ORDIDGE,AND R. G. SHULMAN, Magn. Reson. Med. 5, 377 (1987). 12. M. J. AVISON,D. L. ROTHMAN, E. NADEL,AND R. G. SHULMAN, Proc. Natl. Acad. Sci. USA 85, 1634 (1988). 13. T. JUE, M. AVISON,D. L. ROTHMAN, E. NADEL,J. HAMM,AND R. G. SHULMAN, “Abstracts of the Sixth Meeting of the SMRM, New York, 1987,” p. 584. 14. H. BOMSDORF, T. HELZEL,D. KUNZ,P. ROSCHMANN, 0. TSCHENDEL, AND J. WIELAND, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 22. 15. T. JUE,D. L. ROTHMAN, B. A. TAVITIAN, AND R. G. SHULMAN, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 248. 16. A. HEERSCHAP, P. R. LUYTEN,J. I. VAN DER HEIJDEN, AND J. A. DEN HOLLANDER, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 326. 17. T. JUE,B. A. TAVITIAN, D. L. ROTHMAN, AND R. G. SHULMAN, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 356. 18. T. JUE,D. L. ROTHMAN,G. I. SHULMAN, B. A. TAVITIAN, R. A. DEFRONZO,AND R. G. SHULMAN, Proc. Nut1 Acad. Sci. USA 86,4489 ( 1989). 19. P. A. BOTTOMLEY, C. J. HARDY,P. B. ROEMER,AND 0. M. MUELLER, Magn. Reson. Med. 12, 348 (1989). 20. A. HEERSCHAP, P. R. LUYTEN,J. I. VAN DER HEYDEN.L. J. M. P. OOSTERWAAL, AND J. A. DEN HOLLANDER, NMR Biomed. 2, 124 ( 1990). 21. G. I. SHULMAN, D. L. ROTHMAN, T. JUE,P. STEIN,R. A. DEFRONZO,AND R. G. SHULMAN, N Engl. J. Med. 322,223 (1990). 22. N. BECKMANN, J. SEELIG,AND H. WICK,Magn. Reson. Med. 16, 150 ( 1990). 23. S. D. SWANSON AND H. N. YEUNG,“Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 485. 24. M. SANER,G. MCKINNON, AND P. BOESIGER, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 65, (Works in Progress) ( 1988). 25. H. BOMSDORF AND J. WIELAND, “Abstracts of the Eighth Meeting of the SMRM, Amsterdam, 1989,” p. 184. I. 2. 3. 4. 5.
PROTON DECOUPLING VS POLARIZATION TRANSFER
73
26. A. HEERSCHAP,P. LUYTEN,R. VAN STAPELE, AND J. DEN HOLLANDER,“Abstracts of the Eighth Meeting of the SMRM, Amsterdam, 1989,” p. 1016. 27. N. BECKMANN, S . MOLLER,AND J. SEELIG,“Abstracts of the Eighth Meeting of the SMRM, Amsterdam, 1989,” p. 250. 28. G. A. MORRISAND R. FREEMAN, J. Amer. Chem. Soc. 101,760 ( 1979). 29. D. P. BURUM A N D R. R. ERNST,J. Mugn. Reson. 39, 163 (1980). 30. H. J. JAKOBSEN, 0. W. WRENSEN, A N D H. BILDS~E, J. Mugn Reson. 51, 157 (1983). 31 D. M. DODDRELL,D. T. P E W , A N D M. R. BENDALL,J. Mugn. Reson. 48,323 (1982). 32. M. J. AVISON, D. L. ROTHMAN,E. NADEL,G. I. SHULMAN, AND R. G. SHULMAN, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 503. 33. M. SANER,G. MCKINNON,AND P. BOESIGER,“Abstracts of the Eighth Meeting of the SMRM, Amsterdam, 1989,” p. 602. 34. N. MOLLERAND A. BAUER,J. Mugn. Reson. 82,400 (1989). 35. CH. T. MOONEN,R. J. DIMAND,AND K. L. Cox, Mugn. Reson. Med. 6, 140 ( 1988). 36. M. SANER,G. MCKINNON,AND P. BOESIGER,“Abstracts of the Eighth Meeting of the SMRM, Amsterdam, 1989,” p. 624.
Glycogen Detection by in Vivo 13CNMR A Comparison of Proton Decoupling and Polarization Transfer M.SANER, G. MCKINNON,AND P. BOESIGER Institute of Biomedical Engineering and Medical Informatics, Universityand ETH Zurich, Moussonstrasse 18, CH-8044 Zurich, Switzerland Received July 12, 1991; revised February 5, 1992; accepted February 5, 1992 The performance of gated proton decoupling and polarization transfer with respect to glycogen detection by 'F NMR was investigated. Experiments were performed on a 1.5T whole-body scanner using a 13Csurface coil in combination with a proton head coil. Spectra were acquired from a glycogen phantom and from the lower leg of a healthy volunteer using proton decoupling and the polarization transfer method SINEFT. The signal strength of the C , resonance of glycogen was determined and compared to a reference spectrum acquired without any form of sensitivityenhancement.In the phantom experiment both decouplingand SINEPT produced a signal gain of 3.5. Under in vivo conditions, the signal gain was approximately 2.5 for both techniques. We conclude that decoupling and polarization transfer are equivalently useful techniques for glycogen detection. o 1992 Academic Pms, Inc.
INTRODUCTION
Glycogen, the storage form of glucose, plays a central role in human energy metabolism. Normal concentrations of glycogen, expressed by the concentration of its glucosyl units, range from 60 to 120 mmol/liter (mean 84 mmol/liter) in human skeletal muscle (1,2) and from 230 to 300 mmol/liter in the human liver ( 3 , 4 ) . Studieson rats (5-7) and rabbits (8)have demonstrated that glycogen can be detected under in vivo conditions by I3C NMR. It has also been shown that glycogen is 100% NMR visible despite its large molecular weight (9, 10). Detection of glycogen in human skeletal muscle and liver by 13C NMR spectroscopy has been reported by several groups ( 11-22). The main problem of detecting glycogen is the low sensitivity which is a consequence of the low natural abundance ( 1.1%) of I3C and its low gyromagnetic constant y (a factor of 4 smaller than protons). While a few glycogen studies on humans have been performed without sensitivity enhancement techniques ( 11-15), mainly proton decoupling (16-22), but also polarization transfer (23-27), has been applied to increase the low sensitivity of 13C. Decoupling is used routinely in high-resolution I3CNMR. By applying RF irradiation at the proton resonance frequency the heteronuclear spin-spin coupling is suppressed and multiplets collapse into single resonance lines. Thus spectra are simplified considerably, and simultaneously the signal-to-noise ratio is improved. For a CH doublet (such as the C1 resonance of glycogen) a signal gain by a factor of 2 can be obtained by decoupling during the signal acquisition period. 65
0740-3 194192 $5.00 Copyright Q 1992 by Academic Rss,Inc. All rights of reproduction in any form reserved.
66
SANER, MCKINNON, AND BOESIGER
Polarization transfer by RF pulses is a relatively new technique which can be used for sensitivity enhancement and spectral editing. Here the spin system is driven into a state of nonequilibrium populations by several RF pulses, which leads to increased population differences for the 13C transitions and therefore to an increased signal intensity in the 13C spectrum. As opposed to decoupling there is no RF irradiation during signal acquisition. For a CH doublet a gain factor of yIH/y"' x 4 can be obtained. Commonly used pulse sequences are, INEPT (insensitive nuclei enhanced by polarization transfer) (28, 2 9 ) , SINEPT (sine-dependent polarization transfer) (30), and DEPT (distortionless enhancement by polarization transfer) (31). Here we report on a comparison of proton decoupling and polarization transfer, by the SINEPT pulse sequence, with respect to their sensitivity enhancement for the glycogen C, resonance. SINEPT was chosen for several reasons. The sequence is very simple, since it consists of only three RF pulses. It requires no 180" I3C pulse (as opposed to INEPT or DEPT) and can therefore easily be performed with a 13Csurface coil. Moreover, by a sequential arrangement of the RF pulses the sequence can be implemented on a single-channel MR system. METHODS
Measurements were performed on a Philips Gyroscan S15 ( 1.5 T ) imaging/spectroscopy system. An additional transmitter channel, consisting of a frequency synthesizer (SMG, Rhode & Schwarz), a homebuilt phase modulator, and a power amplifier ( 150 LA, Amplifier Research), was used for proton decoupling. Polarization transfer experiments were performed without hardware modifications. The phantom consisted of a rectangular bottle filled with 90 mmol/liter glycogen (Fluka AG, Buchs, Switzerland) and 66 mmol/liter sodium chloride. This NaCl concentration produced approximately the same loading of the 13Ccoil as that of the leg of a volunteer. The in vivo spectra were acquired from the lower leg of a healthy volunteer (age 33, weight 57 kg, male) who did not follow any specific diet. The examination was performed in the afternoon, about 2 h after a meal. A proton head coil was used both for decoupling and polarization transfer experiments. The quality factor Q of the empty coil was 400, while the Q of the coil loaded with the leg of a volunteer was typically in the range of 70-90. A two-turn, 7-cmdiameter surface coil was used for 13Cexcitation and detection. The coil was constructed from 3-mm copper wire and mounted inside a Plexiglas box. A balanced tuning configuration was used in order to minimize dielectrical losses. The quality factor of the empty 13Ccoil was 300, while the loaded Q was typically between 100 and 120. The three pulse sequences which were investigated are shown in Fig. 1. A simple pulse-acquire experiment (Fig. 1A)without any sensitivity enhancement was performed to obtain a reference spectrum. A fast half-passage adiabatic excitation pulse (hyperbolic tangent type; length, 3 ms; frequency sweep range, 1500 Hz) was used for I3Cexcitation. The same pulse was also used as the readout pulse for decoupling and polarization transfer experiments. Gated CW proton decoupling (Fig. 1B) was performed during the signal acquisition time of 64 ms. The proton frequency was set on-resonance for the H1 proton of glycogen at 5.4 ppm (32). The decoupler power delivered to the coil was 32 W, which resulted in a decoupler field yB2/27r of approximately 400 Hz for the phantom and
PROTON DECOUPLING VS POLARIZATION TRANSFER
67
A
Decoupling
B
'H
FIG. 1. Pulse sequences used for glycogen detection: (A) simple pulse-acquire method; ( B ) gated CW proton decoupling; (C) polarization transfer by SINEPT-SP.
230 Hz for the in vivo measurement. The power absorbed by the tissue is given by Ptissue = ptotal- ( 1 - Qloaded/Qunloaded). For Qloaded/Qunloaded = 90/400 this amounts to 24.8 W. Taking into account the decoupler duty cycle of 47.4%, a mean power absorption of 1 1.8 W results. This power is absorbed by the tissue inside the decoupler coil which had a mass of approximately 3 kg. The average power deposition is therefore 3.9 W/kg. The FDA guidelines (see, e.g., ( 1 9 ) and references therein) require the specific absorption rate SAR in the extremities to be below 8 W/kg for any gram of tissue. We estimated the ratio of maximum SAR to average SAR to be in the order of 2 for our experimental setup. Thus the maximum power absorption was just below FDA guidelines in this experiment. We would like to point out that a volume coil with a homogeneous BI field has the lowest ratio between average and maximum SAR. This ratio will in general be much greater than 2 if a surface coil is used for decoupling (see Fig. 6 in ( 1 9 ) ) . The SINEPT-SP experiment (SINEPT with sequential pulses) shown in Fig. 1C is a modified version of the original SINEPT pulse sequence (30). Because of the sequential arrangement of RF pulses the experiment can easily be implemented on single transmitter channel MR systems. The experiment was performed with tl = 2.5 ms and t2 = 8 ms. The first delay t l was chosen accordingto tPpt = (l/?rJ)arctan(a J T 2 ) ( 3 4 ) .The coupling constant f is 170 Hz for C, of glycogen, and a value of 8 ms for the transverse relaxation time TiHfor the HI proton of glycogen was assumed ( 3 3 ) . The second delay t 2 , which can be chosen arbitrarily, was set to the minimal settling time required for the GYROSCAN demodulator after a change of system frequency from 64 to 16 MHz. In order to maximize the sensitivity for one line of the glycogen doublet, no attempt was made to separate signal contributions from polarization transfer and initial 13C
68
SANER, MCKINNON, AND BOESIGER
magnetization. Therefore, neglecting relaxation effects, a -35 antiphase doublet can be expected for a CH group, which results from the superposition of a -4:4 antiphase doublet (from polarization transfer) and a 1: 1 in-phase doublet (from initial I3Cmagnetization). Localized shimming was performed on a volume of interest of 8 X 8 X 5 cm using the Autoshim function of the GYROSCAN spectroscopy software. Proton linewidths (FWHM) were 10 Hz in the phantom and 14 Hz in the leg. For each spectrum 5 12 complex datapoints were acquired with a sampling rate of 8 kHz. For the phantom, 2048 measurements were averaged and for the in vivo spectra, 4096 measurements. The repetition time was 135 ms in both cases, resulting in a total measurement time of about 10 min for the in vivo spectra. Time domain data was zero-filled to 1024 points and multiplied by a Gaussian filter with 20-Hz line broadening. Manual phase correction and a polynomial baseline correction was performed on all spectra. RESULTS
Spectra acquired from a glycogen phantom are shown in Fig. 2. On the left side the theoretical doublet intensity ratios for the three pulse sequences are shown: the 1: 1 doublet of the one-pulse experiment is expected to be converted into a singlet of intensity 2 by gated decoupling and into a - 3 5 antiphase doublet by SINEPT. Spectrum A is a reference spectrum which was acquired without any form of proton irradiation. The two lines of the glycogen C , doublet can be observed at 95 and at
1
1
1
1
1
1
1
1
1
1
260 240 220 200 180 160 140 120 100 80
1
1
1
1
60
40
20
0
[ppm]
FIG.2. Spectra acquired from a glycogen phantom using different methods:(A) one-pulse spectrum (only I3C excitation pulse); (B) spectrum acquired using CW proton decoupling, the 'H frequency was set onresonance for the H I proton of glycogen; (C) SINEPT-SP spectrum, the pulse sequence was optimized for the CI resonance of glycogen. Measurements (2048) were performed with a repetition time of 135 ms. A Gaussian filter with 20 Hz line broadening was applied.
69
PROTON DECOUPLING VS POLARIZATION TRANSFER
106 ppm. The signal-to-noise ratio (signal height divided by twice the rms noise) is approximately 4. The other glycogen resonances ( C2 at 72 ppm, C3 at 74 ppm, C4 at 78 ppm, C5at 72 ppm, and C6 at 6 1 ppm, according to ( 9 ) )overlap in part and cannot be identified unambiguously. Spectrum B was obtained using CW proton decoupling. As expected, the C , resonance appears as a singlet, centered at 101 ppm. The signal gain with respect to the reference spectrum is 3.5. A gain by more than a factor of 2 (which could be expected for gated decoupling) was also observed in (20) and explained by the suppression of C-H long-range coupling. We assume that due to the relatively high decoupler duty cycle in our experiment there is also a partial nuclear Overhauser enhancement (NOE). In the SINEPT-SP spectrum (C) the C, resonance of glycogen can be observed as an antiphase doublet centered at 101 ppm. There is positive interference for the right line of the doublet and partial cancellation for the left line. The signal gain with respect to the reference spectrum is 3.5 for the right line. Spectra acquired from the lower leg of a healthy volunteer are shown in Fig. 3. Full-scale spectra are shown on the left, an expanded view on the right. All three spectra are dominated by lipid resonances: carboxyl groups (COOR) at 172 ppm, mono- and polyunsaturated fatty acids (C = C ) at 130 ppm, C2 of glycerol at 69 ppm, C1,3of glycerol at 62 ppm, and (CH,), carbons of fatty acids at 30 ppm (35).The resonance of (Phospho)Creatin at 157 ppm is also visible in all three spectra.
A
1
I
I
I
200
150
100
50
I
0 bpml
I
180
l
l
160 140
I
I
I
120
100
80
bpml
FIG.3. In vivo I3C spectra acquired from the lower leg of a volunteer: (A) one-pulse spectrum (only "C excitation pulse); ( B ) proton decoupled spectrum; ( C ) SINEPT-SP spectrum. Measurements (4096) were performed with a repetition time of 135 ms. A Gaussian filter with 20 Hz line broadening was applied.
70
SANER, MCKINNON, AND BOESIGER
In the reference spectrum ( A ) both lines of the glycogen C I doublet can be observed at 95 and 106 ppm. The C2 - C6 resonances of glycogen are overlapped by the intense glycerol lines. In the proton-decoupled spectrum ( B ) the C I resonance of glycogen shows up as a singlet at 101 ppm. The signal gain with respect to one line of the doublet in Fig. 3A is approximately a factor of 2.3. At 118 ppm the C4 resonance of histidine is visible (22). Interestingly there is also a slight signal enhancement for the quaternary ( P ) C r peak at 157 ppm. Additionally there is an unassigned resonance at 142 ppm which can also be observed in the muscle spectra of a marathon runner in (20) but not in the muscle spectra shown in (22). In the SINEPT-SP spectrum ( C ) resonances appear as antiphase multiplets. They are not point-symmetrical since the signal contributions from initial 13Cmagnetization and polarization transfer are superimposed. For the same reason quaternary resonances (e.g., at 172 ppm) do not vanish. There is positive interference for the right line of the glycogen C I doublet at 95 ppm which can be identified unambiguously. The signal intensity is comparable to the singlet in the decoupled spectrum, the gain with respect to the reference spectrum is approximately 2.5. The left line of the doublet at 106 ppm nearly vanishes, but is positive, as opposed to the phantom spectrum, where it is clearly negative. This indicates that the contribution from polarization transfer is smaller than in the phantom measurement. The massive signal gain for the lipid doublet at 130 ppm indicates the correct setting of the proton transmitter frequency. The corresponding lipid protons have the same chemical shift (5.4 ppm) and approximately the same coupling constant as the H I proton of glycogen ( 3 2 ) . The efficiency of polarization transfer for C , of glycogen is significantly smaller than the maximum value predicted by theory, which has been observed before (2 5 , 3 3 ) . There are two loss mechanisms which contribute to this effect: there is no refocusing of IH magnetization in the SINEPT experiment (as opposed to INEPT (28, 2 9 ) ) . Thus main field inhomogeneities will cause a signal loss by dephasing. However, t l is only 2.5 ms and a water linewidth of 14 Hz corresponds to a decay time constant of 1/xAu = 23 ms (assuming a Lorentzian lineshape). Thus, dephasing during t l only accounts for 10% of the signal loss. We therefore conclude that the principal loss is caused by TT relaxation of the proton magnetization between the two 90" pulses. As an estimate of the relaxation time constant T ; , a value of 8 ms was found, based on a series of SINEPT-SP experiments with variable t l ( 3 3 ) .Using this value, the efficiency of the t l period, E l = sin( x J t l ) exp( -rl / T ; ) , can be calculated. A value of 71% is obtained. For the signal decay during the f 2 interval, where the spin system is in a state of longitudinal two-spin order. a time constant of 50 ms was found. With t2 = 8 ms the efficiency in this interval is 85%. Thus an overall efficiency of polarization transfer of 60% can be expected. corresponding to a signal gain by a factor of 2.4. If the contributions from initial 13C magnetization and polarization transfer are superimposed, this results in a doublet intensity ratio of - 1.4:3.4. While the phantom measurement is in good agreement with these values, the in vivo experiment shows an even smaller con-
-
PROTON DECOUPLING VS POLARIZATION TRANSFER
71
tribution from polarization transfer, resulting in a doublet intensity ratio of approximately 0:2. We assume that the T ; of the glycogen H I proton may be even shorter than 8 ms under in vivo conditions. DISCUSSION
The two methods investigated in this study, proton decoupling and polarization transfer using SINEPT, both produce a signal gain by more than a factor of 2 for the C, resonance of glycogen under in vivo conditions. This corresponds to a reduction in measurement time by more than a factor of 4. Therefore we consider decoupling and SINEPT to be alternatives for this application. The actual acquisition time for a spectrum from the human calf can be made as small as 10 min for either one of the two methods. Thus dynamic studies with a reasonably fine temporal resolution become feasible. The major advantages of proton decoupling are the simplification of the spectra by the conversion of multiplets into single lines and the robustness of the method. However. decoupling has also two major drawbacks: there is an inherent danger of tissue damage by excessive absorption of R F power and the technique requires a double channel transmitter system which is not available on the vast majority of installed whole-body MR systems. In contrast, due to the sequential arrangement of excitation pulses, SINEPT-SP can be performed on any single-channel MR scanner which allows a fast change of the system frequency. The hardware requirements are minimal: either a double-tuned coil or two coils ( ‘H and 13C)and a frequency-selective splitter/combiner (36) are required but no additional transmitter hardware. As opposed to decoupling there is no extended period of ‘H irradiation in the SINEPT-SP pulse sequence. The two discrete 90” proton pulses do not produce any significant tissue heating. Therefore short repetition times in the order of 100 to 150 ms can be used. The minimum is not dictated by the need to reduce the duty cycle of the decoupler (20, 22). The major drawbacks of SINEPT-SP are: ( i ) the method can be optimized for one resonance only; (ii ) the splitting of resonance lines into multiplets is not removed, and therefore crowded spectra are obtained; (iii) the method produces so-called antiphase multiplets, i.e., there are resonance lines with negative intensities in the spectrum and there is a possibility for signal cancellation; (iv) SINEPT is not a robust technique since polarization transfer is highly dependent on the ‘H transmitter frequency. However. for the specific application of glycogen detection, the consequences of these properties are not too severe: ( i ) there is only one resonance of interest, the C , line at 101 ppm; (ii) the CI doublet of glycogen is located in a clear “spectral window,” relatively far away from intense lipid resonances; (iii) the only resonance which could possibly interfere with the C, resonance of glycogen centered at 101 ppm is the histidine resonance at 1 18 ppm (20,22).However, if the right line (at 95 ppm) of the glycogen doublet is selected for enhancement, any cancellation of the left line becomes irrelevant; (iv) by localized shimming over a relatively small volume and monitoring the lipid resonance at 130 ppm the correct setting of the ‘H transmitter frequency can easily be verified. In conclusion, we have shown that both proton decoupling and polarization transfer by SINEPT-SP can be used for the in vivo detection of glycogen. For the C1resonance
72
SANER, MCKINNON, AND BOESIGER
both methods produce an increase in signal strength by more than a factor of 2. We believe that proton decoupling is the method of choice, provided that the necessary hardware is available and that excessive RF power deposition can be excluded. Polarization transfer by SINEPT-SP is a powerful alternative. It is an inherently safe technique which is equivalent with respect to sensitivity, but less demanding in terms of hardware requirements. ACKNOWLEDGMENTS We thank Dr. Nicolau Beckmann from the Biozentrum, Basel, and Beat Werner from the University Children’s Hospital, Zurich, for helpful discussions. Support through the Swiss National Science Foundation (NFP-18) is gratefully acknowledged. REFERENCES
E. HULTMAN AND J. BERGSTROM, Acra Medica Scand. 182, 109 ( 1967). E. HULTMAN, Scand. J. Clin. Lab. Invest. 19, 209 ( 1967). L. NILSSONAND E. HULTMAN, Scand. J. C h . Lab. Invest. 32, 325 ( 1973). L. NILSSONAND E. HULTMAN, Scand. J. Clin. Lab. Invest. 33, 5 ( 1974). J. R. ALGER,L. 0. SILLERUD, K. L. BEHAR,R. J. GILLIES,AND R. G. SHULMAN, Science 214, 660 (1981). 6. P. CANIONI, J. R. ALGER, AND R. G. SHULMAN, Biochemistry 22,4974 ( 1983). 7. N. V. REO, B. A. SIEGFRIED, AND J. J. H. ACKERMAN, J. Biol. Chem. 259, 13664 ( 1984). 8. J. R. ALGER, K. L. BEHAR,D. L. ROTHMAN, AND R. G. SHULMAN, J. Magn. Reson. 56, 334 (1984). 9. L. 0. SILLERUD AND R. G. SHULMAN, Biochemistry 22, 1087 ( 1983). 10. R. A. SHALWITZ, N. V. REO, N. N. BECKER,AND J. J. H. ACKERMAN,Magn. Reson. Med. 5, 462 (1987). 11. T. JUE,A. B. LOHMAN,R. J. ORDIDGE,AND R. G. SHULMAN, Magn. Reson. Med. 5, 377 (1987). 12. M. J. AVISON,D. L. ROTHMAN, E. NADEL,AND R. G. SHULMAN, Proc. Natl. Acad. Sci. USA 85, 1634 (1988). 13. T. JUE, M. AVISON,D. L. ROTHMAN, E. NADEL,J. HAMM,AND R. G. SHULMAN, “Abstracts of the Sixth Meeting of the SMRM, New York, 1987,” p. 584. 14. H. BOMSDORF, T. HELZEL,D. KUNZ,P. ROSCHMANN, 0. TSCHENDEL, AND J. WIELAND, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 22. 15. T. JUE,D. L. ROTHMAN, B. A. TAVITIAN, AND R. G. SHULMAN, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 248. 16. A. HEERSCHAP, P. R. LUYTEN,J. I. VAN DER HEIJDEN, AND J. A. DEN HOLLANDER, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 326. 17. T. JUE,B. A. TAVITIAN, D. L. ROTHMAN, AND R. G. SHULMAN, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 356. 18. T. JUE,D. L. ROTHMAN,G. I. SHULMAN, B. A. TAVITIAN, R. A. DEFRONZO,AND R. G. SHULMAN, Proc. Nut1 Acad. Sci. USA 86,4489 ( 1989). 19. P. A. BOTTOMLEY, C. J. HARDY,P. B. ROEMER,AND 0. M. MUELLER, Magn. Reson. Med. 12, 348 (1989). 20. A. HEERSCHAP, P. R. LUYTEN,J. I. VAN DER HEYDEN.L. J. M. P. OOSTERWAAL, AND J. A. DEN HOLLANDER, NMR Biomed. 2, 124 ( 1990). 21. G. I. SHULMAN, D. L. ROTHMAN, T. JUE,P. STEIN,R. A. DEFRONZO,AND R. G. SHULMAN, N Engl. J. Med. 322,223 (1990). 22. N. BECKMANN, J. SEELIG,AND H. WICK,Magn. Reson. Med. 16, 150 ( 1990). 23. S. D. SWANSON AND H. N. YEUNG,“Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 485. 24. M. SANER,G. MCKINNON, AND P. BOESIGER, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 65, (Works in Progress) ( 1988). 25. H. BOMSDORF AND J. WIELAND, “Abstracts of the Eighth Meeting of the SMRM, Amsterdam, 1989,” p. 184. I. 2. 3. 4. 5.
PROTON DECOUPLING VS POLARIZATION TRANSFER
73
26. A. HEERSCHAP,P. LUYTEN,R. VAN STAPELE, AND J. DEN HOLLANDER,“Abstracts of the Eighth Meeting of the SMRM, Amsterdam, 1989,” p. 1016. 27. N. BECKMANN, S . MOLLER,AND J. SEELIG,“Abstracts of the Eighth Meeting of the SMRM, Amsterdam, 1989,” p. 250. 28. G. A. MORRISAND R. FREEMAN, J. Amer. Chem. Soc. 101,760 ( 1979). 29. D. P. BURUM A N D R. R. ERNST,J. Mugn. Reson. 39, 163 (1980). 30. H. J. JAKOBSEN, 0. W. WRENSEN, A N D H. BILDS~E, J. Mugn Reson. 51, 157 (1983). 31 D. M. DODDRELL,D. T. P E W , A N D M. R. BENDALL,J. Mugn. Reson. 48,323 (1982). 32. M. J. AVISON, D. L. ROTHMAN,E. NADEL,G. I. SHULMAN, AND R. G. SHULMAN, “Abstracts of the Seventh Meeting of the SMRM, San Francisco, 1988,” p. 503. 33. M. SANER,G. MCKINNON,AND P. BOESIGER,“Abstracts of the Eighth Meeting of the SMRM, Amsterdam, 1989,” p. 602. 34. N. MOLLERAND A. BAUER,J. Mugn. Reson. 82,400 (1989). 35. CH. T. MOONEN,R. J. DIMAND,AND K. L. Cox, Mugn. Reson. Med. 6, 140 ( 1988). 36. M. SANER,G. MCKINNON,AND P. BOESIGER,“Abstracts of the Eighth Meeting of the SMRM, Amsterdam, 1989,” p. 624.
