Observation of 31P magnetization transfer at 3 Tesla using asymmetric adiabatic inversion and two different fitting strategies.
Bertrand Pouymayou1, Tania Buehler1, Roland Kreis1, and Chris Boesch1

1Depts. Radiology and Clinical Research, University of Bern, Bern, Switzerland


31P-MR spectroscopy inversion transfer (IT) is increasingly investigated as a complementary method to study ATP-synthesis and creatine kinase in vivo. Three aspects of the IT experiment are studied here, in a test-retest design (12 volunteers, resting vastus muscle): the ability to produce an efficient half band inversion in vivo with a short asymmetric adiabatic pulse, the repeatability of the kinetic parameters estimation at 3T and the impact of two different fitting strategies (individual spectrum vs. two-dimensional fitting). As a result, k[Pi>γ-ATP] can be reliably estimated within cohorts while k[PCr>γ-ATP] is accurate enough to be distinguished between individuals.


31P-MR spectroscopy allows monitoring exchange rates (k) between ATP and inorganic phosphate (Pi, ATP-synthesis in mitochondria and glycolysis) and phosphocreatine (PCr, creatine kinase reaction) in vivo. In parallel to saturation transfer (ST) methods, inversion transfer (IT) is increasingly investigated as a complementary method, e.g. as comprehensive (Magnetization Transfer MT and nuclear Overhauser enhancement NOE) analysis of multi-site reactions at 7T1. While ST using long saturation pulses has proven to be robust, the use of short pulses in IT has some advantages which should be evaluated, e.g. reduced influence from small metabolite pools, easier implementation of short pulses on clinical MR scanners, and the opportunity to observe multiple paths simultaneously. A recent study2 showed that the inversion should be as short as possible to allow better observation of the early MT phase and thus to determine the time development of the involved resonances more accurately using an appropriate fitting strategy. Therefore, three aspects of the IT experiment were studied:

-The ability to produce an efficient half-band inversion in vivo with a short asymmetric adiabatic pulse3.

-The reproducibility of the kinetic parameters estimation at 3T with a multi-sites formalism1.

-The impact of two different fitting strategies4,5.


A 3T MR system (VERIO, SIEMENS) with a 1H/31P flexible surface coil (RAPID) was used for a test-retest design (12 examinations of right vastus muscle in healthy volunteers at rest (6m/6f, age=35±13, BMI=24±3), 14 recovery times, 8 averages, TR=20s). The pulse sequence consisted of an asymmetric adiabatic pulse (22ms)3 prior to an adiabatic excitation pulse (2.56ms). Two fitting strategies were applied to the spectra. The first approach fits each spectrum independently with AMARES (jMRUI4) and uses soft constraints (line width and chemical shift) to allow maximum flexibility. The second uses FiTAID5 which takes advantage of the two-dimensional character of the IT experiment. In this approach the peak positions are optimized over the full experiment while line-widths can be tuned on individual spectra. To judge the benefit of simultaneous fitting of all spectra in FiTAID, the Cramer-Rao Minimum Variance Bounds (CRMVB) with and without simultaneous fitting were evaluated and compared to the variances found in the repeated exams. Once the magnetization was estimated, MT was quantified in MATLAB using a matrix calculation based on the Bloch-McConell-Salomon equation1. ANOVA tests were performed in SPSS (IBM, Version 22).


Fig.1 illustrates the time evolution and typical SNR of the acquired spectra. Fig.2 illustrates one dataset (fitted by jMRUI) and the MATLAB fit of the IT effect. The time courses of the 5 fitted resonances (Pi, PCr, γ-ATP, α-ATP, β-ATP) are evaluated simultaneously by the Bloch-McConell-Salomon equations, leading to the parameter estimation as listed in Table 1 (assuming steady state connecting forward and reverse k's as well as NOE's). The hypothesis-driven ANOVA tests for the exchange rates show a significantly different k[PCr>γ-ATP] between subjects which is not the case for k[Pi>γ-ATP]. In turn, the fitting strategy has no influence on k[PCr>γ-ATP] while k[Pi>γ-ATP] is different between jMRUI and FiTAID. Surprisingly, the retest session shows significantly different k[Pi>γ-ATP] as compared to the immediately preceding test session. Regarding the CRMVB values, the uncertainty for the Pi area was found to be ~40% higher if individual spectra were fitted compared to the two-dimensional alternative. The coefficient of variation for the repeated exams (4.1%) was very similar to the CV expected based on CRMVB (3.6%).


According to both spectral fitting strategies, the applied pulse preserves the Pi and PCr peaks (>90% of initial magnetization) and efficiently inverts the ATP side (>75%). The fact that inter-individual differences in k[PCr>γ-ATP] can be observed indicates that the measurement is robust and it can be expected that the methodological error is small compared to expected effect sizes in physiological studies. In addition, the average value is in agreement with literature (summary in 2). In turn, k[Pi>γ-ATP] don't show significant differences between subjects which can be explained by the low SNR of the Pi peak. This is also supported by the CRMVB analysis. Nonetheless the cohort values for both k's (FiTAID and JMRUI) are in good agreement with literature2.


-The proposed sequence performs an efficient and short enough inversion that allows extended chemical exchange rate quantification in vivo at 3T.

-The k[PCr>γ-ATP] exchange rate estimation is robust enough to distinguish even between individuals while k[Pi>γ-ATP] gives a reliable cohort value.

-Both fitting strategies show similar results but the two-dimensional strategy using FiTAID leads to more stable results.


This research was supported by the Swiss National Science Foundation #310030-149779 and #320030-156952.


1. Ren J et al. Exchange kinetics by inversion transfer: Integrated analysis of the phosphorus metabolite kinetic exchanges in resting human skeletal muscle at 7 T. Magn Reson Med. 2015; 73:1359-1369. ·

2. Buehler T, Kreis R, Boesch C. Comparison of 31P saturation and inversion magnetization transfer in human liver and skeletal muscle using a clinical MR-system and surface coils. NMR Biomed 2015; 28(2):188-199. ·

3. Hwang TL et al. Asymmetric adiabatic pulses for NH selection. J Magn Reson. 1999; 138:173-177.

4. Vanhamme L, vandenBoogaart A, Van Huffel S. Improved method for accurate and efficient quantification of MRS data with use of prior knowledge. J Magn Reson 1997; 129:35-43. ·

5. Chong DGQ, Kreis R, Bolliger CS, Boesch C, Slotboom J. Two-dimensional linear-combination model fitting of magnetic resonance spectra to define the macromolecule baseline using FiTAID, a Fitting Tool for Arrays of Interrelated Datasets. Magn Reson Mater Phy 2011; 24:147-164.


Table 1: Parameter estimation by two different fitting strategies (jMRUI, FiTAID). The ANOVA tests are hypothesis-driven for k's (significance level p<0.05) while the values in italic are exploratory only (conservative correction for multiple testing would require significance levels of p<0.001).

Figure 1: Acquired spectra at different recovery times (from 0 to 19.5s).

Figure 2: Measured (blue dots) and fitted (red line) time evolution of the IT experiment, (example fitted with jMRUI).

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)