Introduction

The application of in-vivo 31P MR spectroscopy (31P MRS) in clinical routine is limited by the low signal-to-noise ratio (S/N) at B 0=1.5 T, which leads to long measurement times, poor spatial resolution, and difficult quantitative evaluation of 31P spectra. Additionally, the broad signal of phospholipids, which interferes with the resonances of other metabolites in the in-vivo 31P MR spectrum, complicates the post-processing.

Established techniques to improve S/N and spectral quality of 31P MRS refer to 1H–31P double resonance, i.e., {1H}–31P nuclear Overhauser effect (NOE) and 1H-decoupling. The NOE is bound to dipolar-coupled spins in liquid phase. The signal enhancement for metabolites with scalar 1H–31P couplings can be further increased, e.g. by means of the INEPT (insensitive nucleus enhancement by polarisation transfer) technique. The much higher population differences involved in proton Zeeman energy levels can be transferred to the 31P levels with INEPT by means of a series of broadband RF pulses applied simultaneously with appropriate phases. INEPT achieves a polarisation transfer from 1H to 31P spin states by directly reordering the populations in spin systems with heteronuclear scalar coupling. Thus, only the 31P MR signal of metabolites with scalar 1H–31P couplings [namely, phosphomonoester (PME) and phosphodiester (PDE)] is amplified, while the other metabolite signals in the spectra are suppressed.

INEPT is known in high-resolution MR spectroscopy, but, to our knowledge, has only been applied in two in-vivo 31P MRS studies [1, 2]. The difficulties with this application arise from the weak phosphorus-proton J-couplings (J AK~4–8 Hz), which require long echo times, and the relatively short T 2 relaxation times of 31P metabolites with 31P–1H coupling (11 ms–100 ms) [3, 4]. We therefore explored the theoretical and experimental implications of heteronuclear polarisation transfer (PT) in in-vivo {1H}–31P MRS on a 1.5-T whole-body MR scanner with the ultimate goal of obtaining a most effective refocused INEPT (RINEPT) sequence for 31P signal amplification in MRS studies of the human brain.

PME and PDE are intermediates of membrane phospholipid turnover and thus their resonances are of interest in many brain diseases that involve membrane defects. For the PDE resonance, a correlation with peripheral measures of the highly unsaturated fatty acids docosahexaenoic acid and eicosapentaenoic acid has recently been shown [5].

Theory

The intensities of MR-detectable resonances are proportional to the population differences of the Zeeman energy levels of the observed spin system. In a coupled system of different nuclei, e.g. sensitive nuclei A and insensitive nuclei K (with gyromagnetic ratios γ A>γ K), there are large variations of the population differences of the energy levels depending on the ratio γ A/γ K. Resonant irradiation of one spin species affects the populations of the states of the other. Polarisation transfer can occur when the connectivity of the different spins in a coupled system allows the population differences of the sensitive nuclei to be transferred to the ensemble of insensitive nuclei.

Figure 1 shows the pulse sequence of the {1H}–31P RINEPT experiment [6, 7]. The first part is the "classical" INEPT sequence. The RF pulses applied at 1H and 31P frequencies basically invert the population differences of the Zeeman levels along specific A-nucleus transitions. This is accomplished after the evolution time TE 1=1/(2J) and the simultaneous 90° pulses (Fig. 1, time point 2). The final simultaneous 180° pulses (which expand the INEPT to the RINEPT experiment) refocus transversal magnetisation components depending on the echo time TE 2. TE 2 determines the relative phase of the coupled resonances.

Fig. 1.
figure 1

{1H}–31P INEPT (insensitive nucleus enhancement by polarisation transfer) and refocused INEPT (RINEPT) sequence. Two spin species are excited with a sequence of rectangular RF pulses at 31P and 1H frequencies. The signal of the 31P spins is acquired with WALTZ-4 1H-spin decoupling in the case of RINEPT. In the INEPT experiment, signal acquisition starts immediately after the simultaneous 90° pulses (time point 2). During the evolution time TE 2, gradient pulses for spatial localisation can be applied

Full size image

In the case of a two-spin system, AK, in static magnetic field B 0, the equilibrium density operator in terms of product operators is given by:

$$ \zeta _0 = p\,\left( {{{\gamma _A } \over {\gamma _K }}{\bf{A}}_z + {\bf{K}}_z } \right) $$
(1)

where \( p = \gamma _{\rm{K}} \hbar B_0 {\rm{/}}\left( {{\rm{4}}k_B T} \right) \) is the population difference of K spins at thermal equilibrium (T=temperature, k B=Boltzmann constant). After resonant irradiation of A spins, the density operator equals:

$$ \zeta _1 = {\bf{U}}_{ - {\bf{90x}}}^{\bf{A}} \zeta _0 \left[ {{\bf{U}}_{ - {\bf{90x}}}^{\bf{A}} } \right]^{ - 1} = p\,\left( {{{\gamma _A } \over {\gamma _K }}{\bf{A}}_y + {\bf{K}}_z } \right) $$
(2)

where the operator U A −90x describes a 90°RF pulse at A-spin frequency along the negative x axis (90°–x, Fig. 1, time point 1). Likewise, the density operator at the beginning of the acquisition phase of the INEPT experiment (Fig. 1, time point 2) reads:

$$ {\zeta _{2} {\left[ {TE_{1} = 1/(2J)} \right]} = p\,{\left( {{{\gamma _{A} } \over {\gamma _{K} }}{\bf{2A}}_{z} {\bf{K}}_{x} - {\bf{K}}_{x} } \right)}} $$
(3)

The expectation value of the K-spin transverse magnetisation and hence the INEPT spectrum is then directly obtained by calculating the trace of the product of density (ζ 2), Hamiltonian (U H), and angular momentum operator (I K +=I K x+iI K y):

$$ \displaylines{ \left\langle {\overline {M^K } } \right\rangle (t) = tr\left[ {{\bf{I}}_ + ^{\bf{K}} \left( {{\bf{U}}_{\bf{H}} \zeta _{\bf{2}} {\bf{U}}_{\bf{H}} ^{ - 1} } \right)} \right] \cr = {p \over 2}\left[ {\left( {{{\gamma _A } \over {\gamma _K }} + 1} \right)e^{ - i(\omega _k - \pi J)t} - \left( {{{\gamma _A } \over {\gamma _K }} - 1} \right)e^{ - i(\omega _k + \pi J)t} } \right]. \cr} $$
(4)

The A z K x magnetisation together with K x corresponds to two resonances with relative intensities (γ A /γ K )+1 and (γ A /γ K )−1, demonstrating the signal enhancement of K spins achieved with INEPT (Fig. 2A).

Fig. 2.
figure 2

Calculated 31P MR spectra of {1H}–31P INEPT, {1H}–31P RINEPT, and single-pulse (90°) excitation of the 31P spins in a heteronuclear two-spin system with weak scalar coupling. With INEPT, the lines of the doublet are enhanced in anti-phase configuration with η=γ H/γ P=247%. Both resonances are refocused in the RINEPT experiment

Full size image

The major disadvantage of the INEPT experiment is that proton decoupling cannot be applied because the two K-spin magnetisation components precess out of phase by 180°, hence the resulting signal will be the difference of both amplitudes. This shortcoming is solved by simultaneous irradiation of two 180° RF pulses. In this case, the density operator at echo time TE 2=1/(2J) reads:

$$ {\zeta _{3} {\left[ {TE_{2} = 1/(2J)} \right]} = p\,{\left( {{{\gamma _{A} } \over {\gamma _{K} }}{\bf{K}}_{x} - {\bf{2A}}_{z} {\bf{K}}_{y} } \right)}} $$
(5)

The two magnetisation components refocus at time point 3 and the transverse magnetisation of K spins during the acquisition period is given by:

$$ \displaylines{ \left\langle {\overline {M^K } } \right\rangle (t) = tr\left[ {{\bf{I}}_ + ^K \left( {{\bf{U}}_H \zeta _3 {\bf{U}}_H ^{ - 1} } \right)} \right] \cr = - {p \over 2}\left[ {\left( {{{\gamma _A } \over {\gamma _K }} - 1} \right)e^{ - i(\omega _k - \pi J)t} + \left( {{{\gamma _A } \over {\gamma _K }} + 1} \right)e^{ - i(\omega _k + \pi J)t} } \right]. \cr} $$
(6)

Upon 1H-decoupling during the acquisition phase of the K-spin magnetisation, both resonance lines will collapse to a single line and the overall signal enhancement of the RINEPT sequence through polarisation transfer in a weakly coupled system of one 31P nucleus and one proton is given by:

$$ \eta = {{M^K } \over {M_0 }} - 1 = {{\gamma _A } \over {\gamma _K }} - 1 = 147\;\% $$
(7)

with the thermal equilibrium magnetisation M 0 . For complex molecules with more than two interacting spins the theory becomes more complicated. The coherence transfer and the refocusing of the different magnetisation components may not be perfect and the signal enhancements and optimum echo times may vary.

Since the signal obtained in a RINEPT experiment depends on the transfer of coherence rather then on excited magnetisation from thermal equilibrium, the repetition time of the experiment is determined mainly by the T 1 relaxation time of the sensitive spins. In the case of 1H and 31P, T 1 of the less sensitive 31P spins is much longer than that for the protons. This difference can be used to further increase the signal enhancement.

Methods

All experiments were performed on a 1.5-T whole-body MR scanner (Magnetom Vision; Siemens, Erlangen, Germany) equipped with two RF channels and a double-tuned (31P/1H) quadrature birdcage headcoil (∅ 29.2 cm) [8]. An anticipated difficulty of the experiment was the synchronisation of the second RF channel, which must permit simultaneous irradiation of RF pulses with definite phases in good synchronisation with the first RF channel. Because the second RF channel is only specified as a decoupler by the manufacturer, the timing of the RF pulse was verified with an oscilloscope. An unsteady time delay of up to 0.3 ms between both channels was determined.

{1H}–31P RINEPT studies were carried out with model solutions containing 80 mM methylene diphosphonic acid (MDPA) and 80 mM phosphorylethanolamine (PE).

In contrast to the metabolites detectable by in-vivo 31P MRS, which exhibit very weak phosphorus-proton J-couplings (J AK~4–8 Hz, three bond lengths), MDPA is strongly scalar coupled (J AK=21 Hz, two bond lengths). Additionally, MDPA has a quite simple structure, with two protons in symmetric position relative to two 31P nuclei (Fig. 3).

Fig. 3.
figure 3

Chemical structure of methylene diphosphonic acid (MDPA) and phosphorylethanolamine (PE). MDPA is a symmetric molecule with two 31P nuclei that are scalar coupled over two bonds to two 1H spins (J AK=21 Hz). The 31P spin of PE interacts over three bonds with the two protons of the first methylene group (J AK=6.48 Hz). These in turn interact scalar with the protons of the second methylene group (J AA1=6.9 Hz and J AA2=3.25 Hz)

Full size image

PE is an endogenous 31P-containing metabolite with a resolved resonance in the PME region of 1H-decoupled in-vivo 31P MR spectra of human brain [9]. The compound of PE with an additional glycerine group, glycerophosphorylethanolamine (GPE), resonates in the PDE region of these spectra.

The evolution of the polarisation transfer in PE is much more complicated than in MDPA because of the larger set of different J-couplings in PE (Fig. 3). The 31P nucleus interacts with both protons of the adjacent methylene group (J AK=6.5 Hz, three bond lengths). Moreover, the dynamics of the spin system are affected by homonuclear J-couplings of protons in the two methylene groups (J AA1, J AA2).

Phosphorus MR spectra of MDPA and PE aqueous solutions show a triplet with line splitting of 21 Hz and 6.5 Hz, respectively. The 1H MR spectrum of MDPA exhibits the same triplet (chemical shift δ=2.3 ppm) while two multiplets centered at δ=3.27 ppm and δ=4.10 ppm arise from the methylene groups of PE. These multiplets cannot be resolved at 1.5 Tesla. High-resolution MR yields two different homonuclear coupling constants: J AA1=6.9 Hz and J AA2=3.25 Hz (W. Hull, DKFZ, personal communication).

The sequence parameters were optimised by first varying the echo time TE 1 in the INEPT experiment until maximum signal enhancement was obtained. In the second step, the echo time TE 2 of the RINEPT experiment was varied. An appropriate method to quantify the signal amplification with refocused INEPT is to acquire the 31P signal while 1H-decoupling is applied such that the multiplet structure of the differently phased signal components of the coupled spin system is removed.

To validate the measured results, the expected 31P signal enhancement of the refocused INEPT experiment compared to that of 90° pulse 31P excitation was calculated using the theoretical model for each spin system. The numerical calculations were made by programming the INEPT and RINEPT sequences using the GAMMA C++ libraries [10]. All coupling constants displayed in Fig. 3 were considered in the calculations. Relaxation effects were neglected in these calculations.

To study the influence of relaxation on signal enhancement, the relaxation times T 1 and T 2 of the 31P spins were measured for the model solutions. Spectra were acquired with Ernst-angle excitation as well as with the refocused INEPT sequence (32 averages) and varying TR. The spectra were quantified and the ratio of the signal of RINEPT and Ernst-angle excitation (M r /M s ) was calculated.

The RINEPT sequence was then tested in in-vivo studies with the optimised parameters. The required B 1 field for the 180° –31P RF pulses was determined using a 50-ml flask filled with hexamethylphosphotriamide (HMPT) as external reference. The B1 inhomogeneities from the fixed position of the flask to the centre of the coil were measured in phantom studies. HMPT gives a broad 31P signal at 3,400 Hz up-field to the phosphocreatine (PCr) resonance; hence it does not interfere with endogeneous phosphorus resonances and can be employed in relatively high concentrations without saturating the ADC (Analogue-to-Digital Converting). The HMPT signal was also used for intersubject comparison of signal intensities in in-vivo 31P spectra.

All spectra were quantified using time domain fitting with AMARES (MRUI) [11].

Results

Determination of TE 1, TE 2 and signal enhancements

The time parameter TE 1 for maximum polarisation transfer in the MDPA model solution was found to be TE 1=(12±1) ms≅1/(4J). For a system of two spins TE 1=1/(2J) is expected. The outer lines of the triplet are in anti-phase and enhanced depending on TE 1 (Fig. 4C). The centre line did not change compared to the single-pulse spectrum.

Fig. 4A–D.
figure 4

31P MR spectra of a 80mM aqueous solution of MDPA. All spectra were obtained with the same experimental setup, TR=5 s, NEX=2, and the techniques A single-pulse (90°) excitation; B single-pulse excitation with additional NOE 1H-pulse and 256-ms WALTZ-4 1H-decoupling; C INEPT with TE 1=12 ms; D RINEPT with TE 2=10 ms and 256-ms WALTZ-4 1H-decoupling

Full size image

The optimum refocusing time was TE 2=(10±2) ms in the model solution. A refocusing of all three magnetisation components with RINEPT is not possible in this coupled-spin system. Compared to the single-pulse spectrum, a signal enhancement of:

$$ \eta = {{M_r } \over {M_s }} - 1 = (29 \pm \;3)\;\% $$

was measured. The results of the experiments and numerical simulations, in particular the signal enhancements, agree for the MDPA spin system.

The 31P triplet resonance of PE shows the same pattern as that of MDPA when acquired with the INEPT technique, i.e. unaffected centre line and outer lines enhanced and in anti-phase configuration. Owing to the small couplings, maximum polarisation transfer was obtained with long echo time: TE 1=(40±5) ms (Fig. 5).

Fig. 5.
figure 5

Signal intensity of PE with INEPT as a function of TE 1 (0–80 ms) from A numerical simulation of the spin dynamics and B experiment. The polarisation transfer is maximum at TE 1=(40±5) ms

Full size image

As in the case of the MDPA coupled-spin system, it was not possible to refocus the three magnetisation components completely with RINEPT. In the 1H-decoupled RINEPT experiment with the PE model solution, a maximum 31P signal enhancement of:

$$ \eta = {{M_r } \over {M_s }} - 1 = (22 \pm 5)\;\% $$

relative to the single-pulse spectrum was observed at TE 2=(32±5) ms (Fig. 6).

Fig. 6A–D.
figure 6

31P MR spectra of a 80 mM aqueous solution of PE. All spectra were obtained with the same experimental setup, TR=8 s, NEX=4, and the techniques A single-pulse excitation; B single-pulse excitation with additional NOE 1H-pulse and 256-ms WALTZ-4 1H-decoupling; C INEPT with TE 1=40 ms; D RINEPT with TE 2=32 ms and 256-ms WALTZ-4 1H-decoupling

Full size image

The numerical calculations for a spin system with the J-coupling constants valid for PE yielded TE 1=37 ms, TE 2=32 ms, and η=23%, which all are within the error range of the experimental results (Fig. 5).

Relaxation effects

The measurements of the phosphorus relaxation times in the model solutions yielded T 1 MDPA=(5.380±0.001) s, T 2 MDPA=(383.34±2.01) ms, T 1 PE=(8.184±0.023) s, and T 2 PE =(657.2±1.41) ms. Accordingly, the signal loss of 3.9% in MDPA and 4.7% in PE due to T 2 relaxation during the echo time TE 2 is smaller than the error range of the measured signal enhancement. The effect of T 1 relaxation on the signal enhancement in the RINEPT experiment compared to Ernst-angle excitation was estimated using the Ernst-angle for the measured T 1 of the PE model solution and repetition times TR in the range of 1–20 s.

The expected 31P signal intensity after single-pulse excitation with the Ernst-angle is given by:

$$ M_s = M_0 {{1 - e^{ - TR/T_1^P } } \over {\sqrt {1 - e^{ - 2TR/T_1^P } } }} $$
(8)

where M 0 is the magnetisation in thermal equilibrium and T 1 P the 31P longitudinal relaxation time. The experiments showed that the intensity of the middle line of the triplet was independent of the excitation mode. This suggests that the RINEPT signal enhancement depends not only on T 1 H, but also on T 1 P. Accordingly, the RINEPT signal function reported in [1] had to be extended to:

$$ M_r = d_1 M_0 (1 - e^{ - (TR - \delta )/T_1^H } ) + d_2 M_0 (1 - e^{ - TR/T_1^P } ) $$
(9)

where d 1, d 2 quantify both the enhancement through polarisation transfer as predicted by theory (d 1+d 2=M r/M0 =η+1 for TR→∞) and the signal contribution of the magnetisation components with different phases which depend on T 1 H or T 1 P. The 1H-decoupling time is taken into account by δ. Finally, with the ratio M r/M s from Eq. 8 and Eq. 9 we obtain the signal enhancement with RINEPT as a function of TR:

$$ {{M_r } \over {M_s }}(TR) = {{d_1 (1 - e^{ - (TR - \delta )/T_1^H } ) + d_2 (1 - e^{ - TR/T_1^P } )} \over {1 - e^{ - TR/T_1^P } }}\sqrt {1 - e^{ - 2TR/T_1^P } } $$
(10)

Figure 7 shows ratios of measured 31P MR signal intensities of the PE model solution from RINEPT and single-pulse experiments as a function of TR (NEX=32). A fit of Eq. 10 to these data with use of the parameters T 1 P=8.184 s and δ=0.5 s yielded M r/M 0=1.219±0.050 (η=22%) and T 1 H=(2.019±0.089) s. The plot shows, that for short TR, i.e. TR≅(1.2–2.3)×T 1 H, M r/M s exceeds the theoretically predicted enhancement through polarisation transfer M r/M 0. The fit gives a maximum at M r/M s=1.560±0.001 (η=56 %).

Fig. 7.
figure 7

31P MR signal enhancement of PE with RINEPT as a function of repetition time (TR). RINEPT (TE 1=40 ms, TE 2=32 ms, NEX=32) and single-pulse spectra (Ernst-angle excitation, NEX=32) were obtained with 500-ms WALTZ-4 1H-decoupling. The fit of Eq. 10 to the measured data points is shown

Full size image

Refocused INEPT in vivo

In-vivo whole-head 31P spectra of a healthy control (informed consent) were acquired using 64 averages with Ernst-angle excitation plus NOE enhancement and with RINEPT. The repetition time was set to 1.2 s, which is the minimum allowed within SAR limits when using 150 ms WALTZ 1H-decoupling. With estimated 31P relaxation times T 1 P of about 1.7–2.1 s of the phosphomono- and phosphodiesters [3, 4], the Ernst angle is 60°. The time parameters of the RINEPT sequence were set to the optimum values determined in experiments with the PE model solution: TE 1=40 ms and TE 2=32 ms.

Figure 8 shows in-vivo 31P MR spectra from the brain of a volunteer. In comparison to the single-pulse spectrum (Fig. 8A), the spectrum obtained with the RINEPT sequence (Fig. 8B) is strongly simplified. The resonances of metabolites with scalar 1H–31P coupling (PE, GPE, GPC) are amplified while the other resonances are largely suppressed. The broad phospholipid signal has disappeared. As a consequence, post-processing of RINEPT spectra is easier, also because linear phase correction is unnecessary owing to the spin-echo character of the sequence. The signal acquired with Ernst-angle excitation always needs to be corrected for linear phase because of the hardware-dependent delay between excitation and acquisition. The delay becomes longer when phase encoding gradients have to be inserted for spatial localisation.

Fig. 8A, B.
figure 8

Whole-head 1H-decoupled in-vivo 31P MR spectra (1.5 T, TR=1,200 ms, NEX=64) acquired in the same session. A Ernst-angle excitation (α=60°, NOE); B RINEPT (TE 1=40 ms, TE 2=32 ms). Data post-processing in the time domain with MRUI software (AMARES). Signal enhancement: η PE=(0±14)%, η GPE=(163±66)%, η GPC=(54±22)%

Full size image

In spectra acquired with Ernst-angle excitation, the evaluation of signal enhancements is complicated by the broad phospholipid signal overlapping with the resonances of interest. The variance of the quantified broad resonance band from phospholipids which interferes both with PDE and PME signals is in the order of magnitude of the quantified PE, GPE and GPC resonances.

Line fitting in the time domain with AMARES (MRUI) yielded signal enhancements in the range of η=(0±14)% (PE) to η=(163±66)% (GPE).

The residual signal in the RINEPT spectra of metabolites without scalar 31P–1H-coupling can be completely eliminated by implementing additional phase cycling of the second 90° 1H pulse and the receiver channel. Since only the phase of the signals corresponding to coupled metabolites is changed, with each following acquisition the signals of uncoupled metabolites cancel out. This is demonstrated in Fig. 9.

Fig. 9A–C.
figure 9

Whole-head 1H-decoupled in-vivo 31P MR spectra (1.5 T, TR=1,200 ms, NEX=64) acquired in the same session. The scaling of the vertical axes is identical. A Ernst-angle excitation (α=60°, NOE); B RINEPT (TE 1=40 ms, TE 2=32 ms). Signal enhancement: ηPE=(55.6±8.6)%, ηGPE=(97.6±38.0)%, ηGPC=(72.6±14.1)%; C Experiment of B but with additional phase cycling. ηPE=(21.4±6.9)%, ηGPE=(79.6±33.2)%, ηGPC=(61.4±12.3)%

Full size image

Even though the RINEPT spectrum acquired with phase cycling (Fig. 9C) is of good quality, this method has some disadvantages for our purposes. As shown in the previous sections, the MR signals of molecules with more than two coupled spins still have components that are not enhanced by polarisation transfer when acquired with RINEPT. These signal components still contribute to the overall signal when acquired with proton decoupling but cancel out when using phase cycling. Thus, the signal enhancement of refocused INEPT is about 10–30% lower when phase cycling is used (Fig. 9B, C). Since the residual signal of the uncoupled metabolites in spectra acquired without phase cycling does not disturb the post-processing, this sequence is preferred due to the higher signal yield.

Discussion/conclusion

This report describes theory and experimental observations of 31P signal enhancements through heteronuclear polarisation transfer with the RINEPT experiment. It demonstrates that RINEPT amplifies the signal of coupled 31P–1H spin systems in aqueous model solutions and in vivo. The measured signal enhancements of metabolites with heteronuclear J-coupling in vivo varied from 0% (PE) to 163% (GPE) and differed between the experiments (Figs. 8 and 9). The reason for this variance is the difficult post-processing of the spectra obtained with Ernst-angle excitation, particularly in the range of the phospholipid signal. In contrast, the variation between different refocused INEPT experiments is small. Longitudinal measurements of the same volunteer over 4 weeks with the RINEPT sequence showed variations <5% of the metabolite signal intensities (PE, GPE, GPC).

We explain the relatively small enhancement of PE by the short T 2 (31P) or an unfavourable combination of T 1 (1H) and T 1 (31P) relative to the applied sequence parameters. This will possibly improve when more precise data on in-vivo 31P relaxation times of the human brain are available.

The enhancement with RINEPT is remarkable for all metabolites with scalar 13P–1H coupling when comparing the spectra to those in which the signal is enhanced only by NOE (Fig. 9). The signal amplification with refocused INEPT exceeds that of NOE when using the typical parameters for in-vivo measurements (short TR, large NEX).

The determination of the echo times TE 1 and TE 2 in the experiments with solutions containing MDPA and PE as well as the observed signal enhancement with the RINEPT sequence were in good agreement with the calculations of the theoretical model for these spin systems. This confirmed that the poor synchronisation of the second RF channel has no detectable effect on the double-resonance experiments when using 1-ms rectangular pulses.

One theoretical argument against the use of RINEPT in vivo are the assumed short T 2 times of the endogenous 31P spins. The values of these constants in the literature range from 10 ms to more than 80 ms [1, 3, 4, 12]. Our results show that even though the required echo times for the RINEPT experiment are quite long, there is still a large signal enhancement compared to conventional methods. From the observation of enhancement with TE=40 ms and of narrow spectral line widths of GPE, GPC, and PE resonances (range: 4.9–7 Hz) we conclude that the in-vivo T 2 times of these metabolites must be larger than expected.

Besides the enhancement through polarisation transfer, an additional signal amplification is obtained for short TR in the order of T 1 H of the protons. The maximum of this relaxation-dependent enhancement was found with TR=1.2–1.7 s, which is nearly the minimum repetition time the SAR monitor permits for in-vivo 31P MR spectroscopy with 1H-decoupling. Nevertheless, these values need further investigation through measurement of T 1 H in subsequent studies. Another advantage of RINEPT is that the long echo times allow application of phase encoding gradients for MR spectroscopic imaging.

A limitation of the RINEPT method is that only information on metabolites with scalar 1H–31P coupling can be acquired. However, there is strong interest to detect and quantify in-vivo 31P MR signals of PME and PDE. Altered concentrations of these compounds have been hypothesised and observed in many 31P studies, e. g., of schizophrenic patients [13, 14, 15, 16, 17], depressive/bipolar patients [18, 19], Alzheimer patients [20, 21], and in patients with tumours [22, 23]. The simplified RINEPT spectra are easier to quantify than 31P spectra obtained with more conventional methods which facilitates the analysis and comparison of longitudinally acquired intra-individual spectra as well as inter-individual comparisons of theses signals.