Abstract
Purpose:
To demonstrate J-difference editing of phosphorylethanolamine (PE) with chemical shifts at 3.22 (PE3.22) and 3.98 (PE3.98) ppm, and compare the merits of two editing strategies
Methods:
Density-matrix simulations of MEGA-PRESS for PE were performed at TEs ranging from 80 to 200 ms in steps of 2 ms, applying 20 ms editing pulses (ON/OFF) at: a) 3.98/7.5 ppm to detect PE3.22; and b) 3.22/7.5 ppm to detect PE3.98. Phantom experiments were performed using a PE phantom to validate simulation results. 10 subjects were scanned using a Philips 3T MRI scanner at TEs (90 and 110 ms) to edit PE3.22 and PE3.98. Osprey was used for data processing, modeling, and quantification.
Results:
Simulations show substantial TE-modulation of the intensity and shape of the edited signals due to coupling evolution. Simulated and phantom integrals suggest that TEs of 110 and 90 ms were optimal for the edited detection of PE3.22 and PE3.98, respectively. Phantom results indicated strong agreement with the simulated spectra and integrals. In vivo quantification of the PE3.22/tCr and PE3.98/tCr concentration ratio yielded values of 0.26 ± 0.04 (between-subject coefficient of variation, CV: 15.4%) and 0.18 ± 0.04 (CV: 22.8%), respectively at TE 90 ms, and 0.24 ± 0.02 (CV: 8.2%) and 0.23 ± 0.04 (CV: 18.0%) respectively at TE 110 ms.
Conclusion:
Simulations and in vivo MEGA-PRESS of PE demonstrate both PE3.22 and PE3.98 are potential candidates for editing, but PE3.22 at TE 110 ms yields lower variation across TEs.
Introduction
Phosphorylethanolamine (PE) is a building block of phospholipids and sphingomyelin, present in the brain at millimolar levels (~1.4 mmol/kgww (1)). It serves as a precursor for choline-containing compounds crucial for cell-membrane formation and maintenance (2,3). Ex-vivo studies have demonstrated depleted levels of PE in Alzheimer’s and Huntington’s diseases (4) and ischemia (5), and increases in intracranial tumors (6). Some in-vivo studies have demonstrated that PE decreases with age in neonates (7) and in older adults (8).
Proton (1H) magnetic resonance spectroscopy (MRS) studies have long recognized total choline (free choline+phosphorylcholine+glycerophosphoryl-choline, tCho) as a marker of membrane turnover and neuroinflammation (9). Despite close association with biochemical processes of critical interest, PE is difficult to quantify and has remained largely unstudied by 1H-MRS due to overlap with signals of macromolecules and high-concentration metabolites, such as creatine (Cr), Cho, and myo-inositol (mI). When it is studied, PE is usually quantified on ultra-high-field scanners for better signal-to-noise ratio (SNR) and reduction of spectral overlap (8,10), using proton-decoupled 31P MRS to separate metabolites and afford direct observation of PE (7,11), or ex-vivo measurements (4–6).
J-difference editing is a valuable method for alleviating signal overlap and enhancing the resolution of the 1H spectrum, applied to selective detection of low-concentration brain metabolites including glutathione (GSH), ascorbate (Asc), γ-aminobutyric acid (GABA), aspartate (Asp), and lactate (Lac) (12). The PE spin system, as shown in Figure 1a, consists of two coupled CH2 groups with chemical shifts at 3.22 (PE3.22) and 3.98 (PE3.98) ppm, both of which are potential candidates for edited detection (13–15). To date, there is no study investigating J-difference editing of PE. In this work, we demonstrate J-difference editing of PE using the MEGA-PRESS (16) sequence and compare the merits of two different editing strategies, which are respectively optimized towards the edited detection of one of the two CH2 groups.
Figure 1:
a) The chemical structure for Phosphorylethanolamine (PE). b) TE dependence of the two PE-edited signals at both 3.22 and 3.98 ppm. Spectra from the MEGA-PRESS experiments on PE-only phantom at TE 90, 110, 130, 150, 170, and 190 ms overlaid with simulated spectra. There is a strong agreement between the simulated and phantom spectra. c) PE phantom experiment reveals strong concordance between phantom and simulated PE3.22 and PE3.98 spectral integrals. Phantom and simulations yield optimal TEs at a similar range. d) PE+EA phantom experiment shows the amplitude and form of co-edited EA signals follow the simulated pattern.
Methods
MRI and MRS scans were performed using a 3T MRI scanner (Philips Healthcare, The Netherlands) with a 32-channel head coil. The study protocol and consent forms were approved by the local Institutional Review Board.
Simulations:
Density-matrix simulations (17,18) of the PE spin system were performed using FID-A (15,19). MEGA-PRESS simulations were performed for two editing strategies, with editing pulses either applied to the PE3.22 or PE3.98 resonances, assuming ideal excitation and shaped refocusing and editing pulses. TEs ranging from 80 to 200 ms in increments of 2 ms were simulated using the following parameters: ideal excitation pulse, 20-ms sinc-Gaussian editing pulse (60.5 Hz bandwidth full-width at half-maximum inversion, FWHM); 21×21 two-dimensional spatial array in the refocusing dimensions; 8192 complex datapoints; 4 kHz spectral width; simulated linewidth of 2 Hz. Simulations were accelerated by intermediate spatial averaging (17) and direct coherence selection filtering of the density matrix (20). Simulations were performed for editing pulses applied at: a) 3.98 and 7.5 ppm (ON3.98 and OFF7.5) to detect PE3.22; and b) 3.22 and 7.5 ppm (ON3.22 and OFF7.5) to detect PE3.98. The integral of the PE3.22 (range: 3.02–3.42 ppm) and PE3.98 (range: 3.78–4.18 ppm) signals were calculated. Another set of ideal pulse‐acquire simulations (simulating “TE of 0 ms”) was performed for PE3.22 and PE3.98 to determine the total available signal (Stotal) in the absence of scalar coupling evolution. Editing efficiency was calculated as the ratio of the integral of the edited signal in the difference spectrum ((ON-OFF)/2) normalized to the Stotal integral of the respective signal.
Phantom Experiments:
MEGA-PRESS experiments to detect PE3.22 and PE3.98 were performed on two separate phantoms: a)PE (10 mM, pH 7.1) and b)PE+ethanolamine (EA) (5 mM each, pH 7.1). The acquisition parameters were: 3×3×3 cm3 voxel size, TR 2000 ms, TE 90/110/130/150/170/190 ms, 64 transients, and excitation water suppression.
Acquisition Protocol:
In vivo experiments were performed in 10 subjects (5/5 females/males; age (mean±standard deviation): 39.2±11.2 years). A 3×3×3 cm3 voxel was placed in the mid-parietal cortex (Figure 2a) and prescribed with the multi-vendor standardized MEGA-PRESS sequence (TR 2000 ms; 2048 datapoints; 2 kHz spectral width; 384 transients; Philips VAPOR water suppression) (16,21). This sequence included standardized slice-selective and frequency-selective RF pulses for PRESS localization and editing. The duration of the slice-selective excitation (asymmetric sinc-Gaussian) and refocusing pulses (amplitude modulated) were 7.2 ms and 7.0 ms, respectively, both with peak B1 13.5 μT. The duration between the point where the excitation pulse yields phase-coherent excited signal and the center of the first refocusing pulse is half of TE1 (6.55 ms) resulting in a fixed first slice-selective echo time (TE1) of 13.1 ms, whereas the second slice-selective echo time (TE2) was adjusted to achieve the desired TE. Symmetric sinc-Gaussian editing pulses, separated by TE/2, were used for editing. The duration and FWHM bandwidth of the editing pulses were 20 ms and 60.5 Hz, respectively. In vivo experiments were performed at TE 90 and 110 ms for PE3.22 and PE3.98 editing.16 averages of interleaved water referencing were acquired to minimize the effects of magnetic field (B0) drift during data acquisition (22). Additional PE3.22 and PE3.98 spectra were acquired in one subject, with metabolite-nulling pre-inversion pulse (TI 600 ms; FWHM 698 Hz).
Figure 2:
a) Voxel was placed in the parietal lobe centered at the midline sagittal slice, as indicated by the box on the structural image. b) In vivo MEGA-PRESS experiments at TE 90 and 110 ms for editing PE3.22 and PE3.98. The saturation range of the editing pulses is overlaid with the ON spectra. At both echo times, the experiments resulted in edited PE3.22 and PE3.98 signals without the overlapping Cr and Cho signals. c) Average in vivo PE3.22 and PE3.98 spectra (normalized to the 3-ppm Cr signals from the OFF7.5 spectra) from all subjects at TE 90 and 110 ms overlaid with the ±1 SD variability (in gray). NAA: N-acetylaspartate; Cho: choline; and Cr: creatine. 1 Hz exponential line-broadening applied to the data for visualization.
Data Processing:
Data were processed and modeled in Osprey (23). Individual in vivo transients were frequency-and-phase corrected (24). The 3.02-ppm Cr signal was used to estimate B0 drift in the in vivo data before frequency-and-phase correction. The averaged spectra were aligned by minimizing the L1 norm between 0.5 and 7.4 ppm of the difference spectrum (25). Finally, the fully processed MEGA-PRESS sub-spectra were subtracted to generate PE3.22 and PE3.98 difference spectra. The residual water signal was removed using an HSVD filter (26).
Data Modeling:
The averaged spectra were modeled using Osprey’s frequency-domain linear-combination algorithm. 2D localized density-matrix simulations were performed using the abovementioned PE3.22 and PE3.98 MEGA-PRESS simulation parameters, except 101 × 101 two-dimensional spatial array. Osprey’s default basis set included the following metabolites: ascorbate, aspartate, Cr, GABA, glycerophosphocholine (GPC), glutathione, glutamate, glutamine, myo-inositol, lactate, NAA, NAAG, phosphocholine (PCh), phosphocreatine (PCr), PE, scyllo-inositol, and taurine. Cr-CH2 is included in the basis set to allow for differences in relaxation weighting between Cr-CH3 and Cr-CH2. For the OFF7.5 spectra, 14 macromolecule and lipid basis functions (MM0.91, MM1.21, MM1.43, MM1.67, MM2.05, MM3.21, MM3.71, MM3.79, MM3.87, MM3.97, MM4.2, Lip0.9, Lip1.3, and Lip2.0) were included, whereas the PE3.22-edited spectra were modeled with 2 macromolecule basis functions (MM1.22 and MM1.38). MMs were simulated as Gaussian peaks with similar chemical shifts, amplitudes, and FWHM as described elsewhere (27). PE estimates obtained with both strategies (PE3.22 and PE3.98) were determined as basis-function amplitude ratios relative to the total Cr (tCr=Cr+PCr) signal (modeled in the OFF7.5 spectra). This analysis was also repeated with a basis set including EA. Osprey has not adapted the Cramer-Rao lower bounds for the modeled metabolites; however, a recent MRS methodological consensus paper (28) also recommended to estimate fit quality (FitQA) for spectral assessment. It is calculated as the relative size of residuals normalized to the standard deviation of noise (29), and the FitQA for PE3.22- and PE3.98-edited spectra at both TE were reported. FitQA values <1 indicate overfitting, equal to 1 indicate perfectly fitted data and >1 indicate that the spectrum was not completely modeled (30,31).
Statistical Analysis:
Between-subject coefficients of variation (CV) were calculated as the mean of PE3.22 and PE3.98 divided by their corresponding standard deviation, separately for TEs of 90 ms and 110 ms. Paired t-tests were performed to test whether PE estimates from each editing strategy (PE3.22/tCr and PE3.98/tCr) differ between the two TEs. A 2-sided p-value<0.05 was taken as statistically significant. PE/tCr ratios are presented as mean or mean±standard deviation (SD), unless otherwise stated.
Results
Density-matrix simulations of PE3.22 and PE3.98 in Figure 1b show substantial TE modulation of the intensity and lineshape of the J-difference-edited multiplets due to coupling evolution. Phantom PE3.22 and PE3.98 spectra are overlaid with the simulated spectra in Figure 1b. The intensity and multiplet shape of edited signals change with TE because of T2 relaxation and coupling evolution. Furthermore, the multiplet shape and integrals of phantom spectra show strong agreement with the simulations (Figure 1c). Both phantom integrals show TE 110 ms as the optimal echo time for the edited detection of both PE signals, closely matching the simulated integrals. Spectra from the phantom containing both PE and EA are shown in Figure 1d. These spectra demonstrate the level of EA contributions in PE-edited spectra, and indicate that the amplitude and shape of co-edited EA signals follow the simulated patterns.
Every participant recruited for this study successfully completed scanning. Figure 2b shows typical data from one subject acquired from the mid-parietal cortex at TEs of 90 and 110 ms. PE3.22 (left) and PE3.98 (right) are well edited, with the overlapping Cho and Cr signals cleanly removed. B0 drift during the ~13-min acquisition and water linewidth were 0.79±0.86 Hz and 8.58±0.72 Hz, respectively, indicating good frequency stability and B0 homogeneity. Figure 2c shows mean (across all subjects) PE3.22 and PE3.98 edited spectra with the ±1 SD variability.
Linear-combination modeling of PE-edited spectra at both TEs are shown in Figure 3. Osprey satisfactorily modeled high-concentration metabolite signals, including mI and Cho (Sup. Figures S1 and S2). Metabolite-suppressed spectra (Sup. Figure S3) indicate that MM contribution to the PE-edited signals was small compared to the edited signals. The FitQA for PE3.22- and PE3.98-edited spectra were 4.4±0.7 and 4.5±2.1, respectively at TE of 90 ms, and 3.9±1.1 and 3.8±1.2, respectively at TE of 110 ms. PE3.22/tCr and PE3.98/tCr in ten subjects yields values of 0.26±0.04 and 0.18±0.04, respectively at TE of 90 ms, and 0.24±0.02 and 0.23±0.04 respectively at TE of 110 ms. Between-subject coefficients of variation for PE3.22/tCr and PE3.98/tCr were 15% and 23% respectively at TE of 90 ms, and 8% and 18% respectively at TE of 110 ms. A paired t-test indicates that the PE/tCr ratios are significantly higher at TE of 110 ms than 90 ms for PE3.98 (p=0.03) but not for PE3.22 (p=0.14). The inclusion of EA in the basis set only resulted in non-zero coefficient in one of the ten subjects, and therefore the results presented are from modeling without an EA basis spectrum. The PE editing efficiency is 36% and 38% for PE3.22 and 37% and 38% for PE3.98 at TEs of 90 and 110 ms, respectively. The respective EA co-editing efficiencies are 21% and 17% for EA3.15 and 32% and 31% for EA3.82.
Figure 3:
Linear-combination modeling of PE3.22 and PE3.98 difference spectra of an example subject with complete fit (red), spline baseline (black), and PE model estimates at both TEs (90 ms and 110 ms). Asp: aspartate, tCho: total choline, tCr: total creatine, Glx: glutamine+glutamate, mI: myo-Inositol, Lac: lactate, MM: sum of macromolecules, Tau: taurine.
Discussion
Quantification of phosphorylethanolamine (PE) is challenging due to overlap with Cho and Cr signals. J-difference editing of the spins at 3.22 ppm and 3.98 ppm can be performed at TE~100 ms, removing these overlying signals. Unlike edited detection of glutathione or lactate, the ‘direction’ of the edited experiment is not immediately apparent – applying editing pulses at 3.98 ppm to detect PE3.22, or at 3.22 ppm to detect PE3.98. Simulations and phantom experiments suggest that both approaches give similar signal yields, with optimal TEs in a similar range. The edited signals are relatively well-resolved, but both still suffer from some degree of overlap with co-edited mI signals. The proposed study has utilized frequency-domain linear-combination algorithm for fitting as it is being recommended by experts for quantification for edited MRS (28). Furthermore, it is considerably better in resolving the overlapped signals of PE and mI compared to simple peak fitting, especially for PE3.98 editing. We have used the FitQA metric to quantify the quality of modelling; there does not appear to be a meaningful difference between PE3.22 and PE3.98 editing or between the TEs. Currently, the modelling residual is ~4 times larger than the noise in these spectra, which we deem to be within acceptable limits for this proof-of-principle study, and which would be reduced further by improving localization and suppressing out-of-voxel echoes at ~4.2 ppm. In vivo data suggest that the PE3.22 signal has lower variance than PE3.98 probably due to greater separation from the water signal, which better avoids baseline perturbations associated with variable water suppression. Furthermore, between-subject variation of PE3.22 is lower at TE of 110 ms compared with TE of 90 ms. As a result, PE3.22 at TE of 110 ms yields reliable detection and consistent measurement between subjects.
The PE spin system consists of two mutually coupled CH2 groups and a phosphorus spin. Thus, PE3.22 consists of two doublets of doublets with the same chemical shift, and PE3.98 consists of two doublets of doublets of doublets with the same chemical shift (having the additional coupling to phosphorus). Four different vicinal coupling constants interact with the editing pulses: 3.2 Hz; 7.2 Hz; 6.7 Hz; and 3.0 Hz (28). Editing doublets of doublets involves a compromise between the fast rate at which the outer peaks acquire phase and the slow rate at which the center peaks evolve. The calculated editing efficiency of ~40%, while not as high as one might hope, is consistent with these realities.
One notable issue for editing PE3.22 and PE3.98 is the limited chemical shift difference between the spins (~100 Hz at 3T). Editing pulse selectivity is a key constraint in optimizing the experiment; inversion of both frequencies with an insufficiently selective editing pulse would result in a loss of editing efficiency. For this reason, we developed the experiment with 20-ms sinc-Gaussian editing pulses with 60.5 Hz bandwidth that result in 0.5% inversion of the detected signal. The duration of the editing pulse limited the minimum TE to 80 ms for MEGA-PRESS with a B1 limit of 13.5 uT. The TE modulation seen in simulations shows local maxima at 90 and 110 ms, each of which might be the global maximum depending on the precise value of T2 – this selectivity constraint rules out other possible TE options <80 ms. MEGA-PRESS relies on subject and scanner stabilities for optimal editing. Integration of prospective motion correction with our real-time frequency correction would further improve measurement efficacy and reproducibility (32).
Given their similar spin systems, one must consider the co-editing of EA in PE-edited spectra. The in vivo concentration of EA is not well characterized, but appears to be substantially lower than PE (1,15). EA has a similar proton spin system to PE, with mutually coupled methylene signals at 3.15 and 3.82 ppm (15). These are, respectively, 0.07 and 0.16 ppm from the analogous PE resonances. PE-directed editing will therefore also substantially co-edit EA, to the extent that it is present, because these shift differences are much lower than the editing selectivity. The detected signals are resolved to differing degrees at in vivo linewidths (with PE3.22 partially resolved from EA3.15 and PE3.98 largely resolved from EA3.82), so the fact that PE-only modeling yields larger values for PE3.22/tCr than PE3.98/tCr would be consistent with some EA contributions. Failure to model EA with an appropriate basis function suggests that in vivo levels of EA are substantially lower than those of PE (as supported by a good level of agreement between the PE+EA phantom/simulation spectra and the form of the in vivo spectra) and not reliably modellable at the selected echo times. PE also overlaps with MM signals, which are substantially reduced at TE~100ms, but not absent. Metabolite-nulled PE-edited spectra indicate that MM contributions to the edited spectrum and negligible, in line with prior work characterizing the macromolecular background (summarized in Cudalbu et al. (33)).
The estimated in vivo PE3.98/tCr ratio was significantly higher at TE of 110 ms than 90 ms which was an unexpected result. These ratios represent the amplitude factors applied to basis functions modeling the PE difference signal and the tCr editing-off signal – which should address any differences due to coupling modulation and isolate relaxation effects. This suggests that the in vivo T2 of PE3.98 is longer than that of the tCr singlet (~160 ms (34)). Applying editing pulses as 3.98 ppm to detect PE3.22 also partially inverts the 4.1 ppm Lac signal (to co-edit Lac1.3). The Lac1.3 signal is a doublet, which optimally edited at a TE of ~140 ms (1/J). Editing Lac at TE of 110 ms is sub-optimal, but only results in editing efficiency losses of ~12%. Hence, TE of 110 ms could be a reasonable compromise for simultaneous editing of Lac1.3 and PE3.22, potentially useful for studying membrane turnover and lactate build-up in tumors, or tracking cancer treatments (35,36). The PE3.22 editing scheme can be further extended to a 4-step (A, B, C, D) Hadamard-encoded editing scheme (37). For instance, simultaneous editing of GABA and PE3.22 (A: dual-lobe editing pulse, ON1.9/ON3.98; B: ON1.9; C: ON3.98; D: OFF7.5) or glutathione (GSH) and PE3.22 (replace ON1.9 with ON4.56) would be possible. PE3.22 is edited with negative polarity and reasonable efficiency in “Hadamard Editing Resolves Chemicals Using Linear-combination Estimation of Spectra” at TE of 80 ms (37). Such progressions of the PE3.22-editing scheme provide an avenue for studying multiple neurochemical systems and their roles in the healthy and diseased brain.
Conclusion
J-difference editing of PE has been demonstrated, with detection of PE3.22 at TE of 110 ms recommended for future applications.
Supplementary Material
Figure S1: Spectral modelling of phantom PE3.22 (left) and PE3.98 (right) difference spectra to assess modelling of high-concentration metabolites NAA, Cr, Cho, Lac, and mI at TE 90 and 110 ms. Subtraction artifacts are present in the edited data and impact modelling accuracy. Modelling of PE3.98 difference spectra yields low residuals from signals of high-concentration metabolites. Modelling of PE3.22 difference spectra yields high residuals in the region above 3.5 ppm with structured signals from free choline at ~4.2 ppm and Cr at ~4 ppm. Data were acquired from a Braino phantom without PE with the following parameters: 3×3×3 cm3 voxel size, TR 2000 ms, TE 90 and 110 ms, 128 transients, and excitation water suppression.
Figure S2: Spectral fitting of mI for in vivo and phantom data (without PE) acquired at TE 90 and 110 ms. mI is satisfactorily modelled in both PE difference spectra suggesting good agreement among in vivo, phantom, and basis signals. The Braino phantom data were acquired with the following parameters: 3×3×3 cm3 voxel size, TR 2000 ms, TE 90 and 110 ms, 128 transients, and excitation water suppression.
Figure S3: Contribution of macromolecule signals to the edited PE3.22 and PE3.98 signals. Pre-inversion nulling of PE signals indicates negligible MM contributions.
Acknowledgement
This work was supported by NIH grants R01 EB016089, R01 EB023963, K99/R00 AG062230, K99 DA051315, P41 EB015909, P41 EB031771, and S10 OD021648.
References
- 1.Govindaraju V, Young K, Maudsley AA. Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed 2000;13(3):129–153. [DOI] [PubMed] [Google Scholar]
- 2.Perry T, Berry K, Hansen S, Diamond S, Mok C. Regional distribution of amino acids in human brain obtained at autopsy. Journal of neurochemistry 1971;18(3):513–519. [DOI] [PubMed] [Google Scholar]
- 3.Li Z, Vance DE. Thematic review series: glycerolipids. Phosphatidylcholine and choline homeostasis. Journal of lipid research 2008;49(6):1187–1194. [DOI] [PubMed] [Google Scholar]
- 4.Ellison DW, Beal MF, Martin JB. Phosphoethanolamine and ethanolamine are decreased in Alzheimer’s disease and Huntington’s disease. Brain Res 1987;417(2):389–392. [DOI] [PubMed] [Google Scholar]
- 5.Smart S, Fox G, Allen K, et al. Identification of ethanolamine in rat and gerbil brain tissue extracts by NMR spectroscopy. NMR Biomed 1994;7(8):356–365. [DOI] [PubMed] [Google Scholar]
- 6.Kinoshita Y, Yokota A, Koga Y. Phosphorylethanolamine content of human brain tumors. Neurologia medico-chirurgica 1994;34(12):803–806. [DOI] [PubMed] [Google Scholar]
- 7.Blüml S, Seymour KJ, Ross BD. Developmental changes in choline‐and ethanolamine‐containing compounds measured with proton‐decoupled 31P MRS in in vivo human brain. Magnetic Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance in Medicine 1999;42(4):643–654. [DOI] [PubMed] [Google Scholar]
- 8.Marjańska M, McCarten JR, Hodges J, et al. Region-specific aging of the human brain as evidenced by neurochemical profiles measured noninvasively in the posterior cingulate cortex and the occipital lobe using 1H magnetic resonance spectroscopy at 7 T. Neuroscience 2017;354:168–177. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zahr NM, Mayer D, Rohlfing T, Sullivan EV, Pfefferbaum A. Imaging neuroinflammation? A perspective from MR spectroscopy. Brain Pathol 2014;24(6):654–664. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lu M, Chen W, Zhu XH. Field dependence study of in vivo brain 31P MRS up to 16.4 T. NMR Biomed 2014;27(9):1135–1141. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Bluml S, Zuckerman E, Tan J, Ross BD. Proton‐decoupled 31P magnetic resonance spectroscopy reveals osmotic and metabolic disturbances in human hepatic encephalopathy. Journal of neurochemistry 1998;71(4):1564–1576. [DOI] [PubMed] [Google Scholar]
- 12.Star-Lack J, Spielman D, Adalsteinsson E, Kurhanewicz J, Terris DJ, Vigneron DB. In vivo lactate editing with simultaneous detection of choline, creatine, NAA, and lipid singlets at 1.5 T using PRESS excitation with applications to the study of brain and head and neck tumors. J Magn Reson 1998;133(2):243–254. [DOI] [PubMed] [Google Scholar]
- 13.Pfeuffer J, Tkac I, Provencher SW, Gruetter R. Toward an in vivo neurochemical profile: quantification of 18 metabolites in short-echo-time (1)H NMR spectra of the rat brain. J Magn Reson 1999;141(1):104–120. [DOI] [PubMed] [Google Scholar]
- 14.Ala-Korpela M, Posio P, Mattila S, Korhonen A, Williams SR. Absolute quantification of phospholipid metabolites in brain-tissue extracts by 1H NMR spectroscopy. J Magn Reson B 1996;113(2):184–189. [DOI] [PubMed] [Google Scholar]
- 15.Govind V, Young K, Maudsley AA. Corrigendum: proton NMR chemical shifts and coupling constants for brain metabolites. Govindaraju V, Young K, Maudsley AA, NMR Biomed. 2000; 13: 129–153. NMR Biomed 2015;28(7):923–924. [DOI] [PubMed] [Google Scholar]
- 16.Mescher M, Merkle H, Kirsch J, Garwood M, Gruetter R. Simultaneous in vivo spectral editing and water suppression. NMR Biomed 1998;11:266–272. [DOI] [PubMed] [Google Scholar]
- 17.Zhang Y, An L, Shen J. Fast computation of full density matrix of multispin systems for spatially localized in vivo magnetic resonance spectroscopy. Med Phys 2017;44(8):4169–4178. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Bak M, Rasmussen JT, Nielsen NC. SIMPSON: a general simulation program for solid-state NMR spectroscopy. J Magn Reson 2011;213(2):366–400. [DOI] [PubMed] [Google Scholar]
- 19.Simpson R, Devenyi GA, Jezzard P, Hennessy TJ, Near J. Advanced processing and simulation of MRS data using the FID appliance (FID‐A)—An open source, MATLAB‐based toolkit. Magn Reson Med 2017;77(1):23–33. [DOI] [PubMed] [Google Scholar]
- 20.Landheer K, Swanberg KM, Juchem C. Magnetic resonance Spectrum simulator (MARSS), a novel software package for fast and computationally efficient basis set simulation. NMR Biomed 2019:e4129. [DOI] [PubMed] [Google Scholar]
- 21.Saleh MG, Rimbault D, Mikkelsen M, et al. Multi-vendor standardized sequence for edited magnetic resonance spectroscopy. Neuroimage 2019;189:425–431. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Edden RA, Oeltzschner G, Harris AD, et al. Prospective frequency correction for macromolecule‐suppressed GABA editing at 3T. J Magn Reson Imaging 2016;44(6):1474–1482. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Oeltzschner G, Zollner HJ, Hui SCN, et al. Osprey: Open-source processing, reconstruction & estimation of magnetic resonance spectroscopy data. J Neurosci Methods 2020;343:108827. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Mikkelsen M, Tapper S, Near J, Mostofsky SH, Puts NA, Edden RA. Correcting frequency and phase offsets in MRS data using robust spectral registration. NMR Biomed 2020;33(10):e4368. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Cleve M, Kramer M, Gussew A, Reichenbach JR. Difference optimization: Automatic correction of relative frequency and phase for mean non-edited and edited GABA (1)H MEGA-PRESS spectra. J Magn Reson 2017;279:16–21. [DOI] [PubMed] [Google Scholar]
- 26.Barkhuijsen H, De Beer R, Van Ormondt D. Improved algorithm for noniterative time-domain model fitting to exponentially damped magnetic resonance signals. Journal of Magnetic Resonance (1969) 1987;73(3):553–557. [Google Scholar]
- 27.Birch R, Peet AC, Dehghani H, Wilson M. Influence of macromolecule baseline on (1) H MR spectroscopic imaging reproducibility. Magn Reson Med 2017;77(1):34–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Wilson M, Andronesi O, Barker PB, et al. Methodological consensus on clinical proton MRS of the brain: Review and recommendations. Magn Reson Med 2019;82(2):527–550. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Pedrosa de Barros N, Slotboom J. Quality management in in vivo proton MRS. Anal Biochem 2017;529:98–116. [DOI] [PubMed] [Google Scholar]
- 30.Slotboom J, Nirkko A, Brekenfeld C, van Ormondt D. Reliability testing of in vivo magnetic resonance spectroscopy (MRS) signals and signal artifact reduction by order statistic filtering. Measurement Science and Technology 2009;20:104030. [Google Scholar]
- 31.Wilson M, Reynolds G, Kauppinen RA, Arvanitis TN, Peet AC. A constrained least-squares approach to the automated quantitation of in vivo (1)H magnetic resonance spectroscopy data. Magn Reson Med 2011;65(1):1–12. [DOI] [PubMed] [Google Scholar]
- 32.Saleh MG, Edden RAE, Chang L, Ernst T. Motion correction in magnetic resonance spectroscopy. Magn Reson Med 2020;84(5):2312–2326. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Cudalbu C, Behar KL, Bhattacharyya PK, et al. Contribution of macromolecules to brain (1) H MR spectra: Experts’ consensus recommendations. NMR Biomed 2021;34(5):e4393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Traber F, Block W, Lamerichs R, Gieseke J, Schild HH. 1H metabolite relaxation times at 3.0 tesla: Measurements of T1 and T2 values in normal brain and determination of regional differences in transverse relaxation. J Magn Reson Imaging 2004;19(5):537–545. [DOI] [PubMed] [Google Scholar]
- 35.de la Cruz-Lopez KG, Castro-Munoz LJ, Reyes-Hernandez DO, Garcia-Carranca A, Manzo-Merino J. Lactate in the Regulation of Tumor Microenvironment and Therapeutic Approaches. Front Oncol 2019;9:1143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Osawa T, Shimamura T, Saito K, et al. Phosphoethanolamine Accumulation Protects Cancer Cells under Glutamine Starvation through Downregulation of PCYT2. Cell Rep 2019;29(1):89–103 e107. [DOI] [PubMed] [Google Scholar]
- 37.Oeltzschner G, Saleh MG, Rimbault D, et al. Advanced Hadamard-encoded editing of seven low-concentration brain metabolites: Principles of HERCULES. Neuroimage 2019;185:181–190. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure S1: Spectral modelling of phantom PE3.22 (left) and PE3.98 (right) difference spectra to assess modelling of high-concentration metabolites NAA, Cr, Cho, Lac, and mI at TE 90 and 110 ms. Subtraction artifacts are present in the edited data and impact modelling accuracy. Modelling of PE3.98 difference spectra yields low residuals from signals of high-concentration metabolites. Modelling of PE3.22 difference spectra yields high residuals in the region above 3.5 ppm with structured signals from free choline at ~4.2 ppm and Cr at ~4 ppm. Data were acquired from a Braino phantom without PE with the following parameters: 3×3×3 cm3 voxel size, TR 2000 ms, TE 90 and 110 ms, 128 transients, and excitation water suppression.
Figure S2: Spectral fitting of mI for in vivo and phantom data (without PE) acquired at TE 90 and 110 ms. mI is satisfactorily modelled in both PE difference spectra suggesting good agreement among in vivo, phantom, and basis signals. The Braino phantom data were acquired with the following parameters: 3×3×3 cm3 voxel size, TR 2000 ms, TE 90 and 110 ms, 128 transients, and excitation water suppression.
Figure S3: Contribution of macromolecule signals to the edited PE3.22 and PE3.98 signals. Pre-inversion nulling of PE signals indicates negligible MM contributions.