Magnetic resonance spectroscopy is a noninvasive technique that provides specific biochemical information on numerous intracellular metabolites. Proton MRS has been used to detect several metabolites, such as NAA, choline, creatine, glutamate, glutamine, myoinositol, and glutathione, in the brain of clinically normal subjects.1,2 N-acetyl aspartate, which resonates at 2.01 ppm, is considered a neuronal marker and is present only in neurons, axons, and dendrites. Choline resonates at 3.2 ppm and is involved in membrane synthesis and degradation, whereas creatine resonates at 3.0 ppm and is involved in energy metabolism. Glutamate and glutamine metabolites resonate closely at 3.75 ppm and between 2.1 and 2.5 ppm, respectively. Glutamate is an excitatory neurotransmitter that plays a role in mitochondrial metabolism; glutamine is involved in detoxification and regulation of neurotransmitter activity. Myoinositol is a pentose sugar that resonates at 3.5 to 3.6 ppm and is part of the inositol triphosphate intracellular second messenger system. Glutathione resonates at 3.77 ppm; it is an antioxidant and is essential for maintaining RBC structure and maintaining hemoglobin in a ferrous state. In pathological conditions, these metabolites may be found in abnormal concentrations (absent, lower, or higher concentrations), and other metabolites (eg, lipids or lactate) that typically are not present in a healthy brain may be detected.1
The spectrum of metabolites for 1H MRS is dependent on strength of the magnetic field and the type of sequence and parameters used. Field strengths ≥ 1.5 T are necessary for proper differentiation of the chemical shifts of metabolites in 1H MRS.1–3 Spectral resolution, particularly for coupled spin systems (eg, glutamate, glutamine, and myoinositol), and signal-to-noise ratio are higher at 3.0 than 1.5 T.1,3 Therefore, the use of a 3.0-T field improves reliability of detection of these compounds.
Quantification of the MRS signal amplitude provides a means of estimating the tissue concentration of signal-generating molecules.4 Tissue concentration of a metabolite is related to the integrated amplitude of the MRS signal it generates, which is the area under the 1H MRS signal curve. The integrated signal amplitude generated by the atomic nuclei of a metabolite is directly proportional to the moles of signal-generating molecules in the volume of the brain being examined (typically expressed as millimoles per liter).4 Therefore, use of signal height or even signal height ratio is not an appropriate method for determining the concentration of metabolites responsible for a particular signal.4 Proton MRS signals are usually acquired in the time domain as free induction decays or echoes, but they typically are viewed and analyzed in the frequency domain.1,4 The frequency domain representation is derived from the acquired time domain data by use of Fourier transformation. A variety of methods have been developed for frequency domain analysis.5–10 The simplest approach is to use numerical integration, although this method works poorly when there is spectral overlap. A numerical integration program typically is included in the software of an MRI unit. More sophisticated methods of analysis include parametric curve-fitting programs that use various model functions, such as Lorentzian, Gaussian, Voigt, and fitting algorithms. One of the fitting models widely used in research and human medicine is the so-called linear combination model.8,11 The linear combination model fits the spectrum as a linear combination of the spectra of the pure compounds known to exist in the spectrum.8,11
The signal measured from a sample containing a metabolite is measured by several factors, including 1H MRS voxel size, molar concentration of the metabolite, and pulse sequence depending mainly on repetition and echo times, T1 and T2 relaxation times, strength of the radiofrequency field, number of signal averages, receiver gain, spectrometer operating frequency, and factors related to the geometry and quality of the radiofrequency receiver coil.1 These factors make it challenging or even impossible to calculate absolute metabolite concentrations.1,2,4 The preferred approach for dealing with this complexity has been to use signal calibration. Measured signal amplitudes for a brain metabolite are calibrated against a reference signal generated by a known concentration of the metabolite or some other material. This will lead to a relative or estimated concentration of metabolites. The reference signal may be generated by a metabolite, water, or some other chemical compound in a phantom.1,2,4 The phantom object used to generate an external reference signal may be a small object that fits within the radiofrequency coil along with the patient's head or it may be a larger object for which the signal is measured before or after the patient's examination. The primary problem with phantoms is that they create main and transmit magnetic field inhomogeneities; thus, their use is not widespread.4 Calibration against an internal signal is used far more frequently. Typically, the internal reference signal is for water (4.69 ppm) or creatine (3.00 ppm). If water signal is used as a calibration signal, its amplitude must be measured by performing a separate measurement in the same brain region without the use of water suppression.1,4 The process is to assume the water concentration in a brain and use this to estimate metabolite concentrations in the brain.1,4 Creatine is a metabolite used as a calibration signal in human medicine because it historically has been believed to be the most stable metabolite of the brain; however, this assumption has been questioned.12–14 Furthermore, it has been found that creatine concentrations in humans and experimental animals change on the basis of region of the brain and age.13,15–18
To estimate reliability of obtained quantitative results, values are reported in combination with CRLBs.8,19 The CRLBs represent an estimate for the minimum error associated with fitted values based on the fitting model, noise, and estimated parameters, especially linewidth. When CRLBs are > 50%, a true metabolite concentration of zero would be in the confidence interval of the estimated value ± 2 CRLB; hence, estimated values with this associated error are considered unreliable because they are not significantly different from zero. A CRLB threshold < 20% is often used to include only the most reliable data.8 However, CRLBs do not include information about quality of the spectra, except for noise and linewidth, and rely on the fitting algorithm to provide the global minimum and reasonable estimates of the true parameters. Hence, it is important to visually verify quality of the spectra and data fitting for the presence of artifacts and reasonable convergence of the data fit. In addition to the CRLBs, signal-to-noise ratio as well as linewidth can be used as measures of spectral quality.20
Two general categories of spatial localization techniques are used in 1H MRS: single voxel and multivoxel (also called chemical shift imaging). The advantages of single voxel spatial localization techniques are minimal lipid contamination, better magnetic field homogeneity, improved water suppression and spectral resolution, and short acquisition scan times (typically 8 to 12 minutes). The main disadvantage is that only a small area of the brain is scanned. Multivoxel spatial localization techniques allow detection of metabolic profiles from a multidimensional array of locations. However, there can be several disadvantages, such as considerable field inhomogeneity, spectral degradation attributable to intervoxel contamination, and prolonged acquisition times (typically > 15 minutes).1
Two types of echo times can be used for 1H MRS: long echo times (typically 144 or 288 milliseconds) and short echo times (< 35 milliseconds). With long echo time sequences, NAA, choline, and creatine peaks are identified in brains of clinically normal subjects. Long echo time sequences are associated with a flatter baseline; however, metabolites with a short relaxation time cannot be identified. In addition to NAA, choline, and creatine, short echo time sequences allow identification of the composite peak resulting from overlap of the glutamate and glutamine peaks (glutamine-glutamate complex) and myoinositol. The signal-to-noise ratio is higher in short echo time than in long echo time sequences; however, the baseline is more unstable because of detection of additional macromolecules (eg, proteins and molecules with motion restricted as solids or membranes), which may complicate the analysis. Given the higher signal-to-noise ratio and higher number of metabolites detected, it is generally recommended that short echo times be used.21–23 Ultimately, good shimming, proper water suppression, and absence of unwanted lipid signal from brain boundaries are essential to obtain a spectrum of diagnostic quality.20
Proton MRS is widely used in human medicine for the diagnosis of several intracranial conditions, such as tumors and inflammatory metabolic diseases, as well as psychiatric disorders.21,24–33 To the authors’ knowledge, only anecdotal reports34–37 of the experimental use of 1H MRS in veterinary medicine have been published. Clinical use in dogs has been reported rarely,38,39 and to our knowledge, clinical use has not been reported in cats. Interpretation of spectra from patients with neurologic disorders requires knowledge of normal regional spectral variations. In humans, spectral patterns can depend on the region within the brain and also on age of the patient.15,16,40,41
Recently, relative concentrations and ratios of metabolites for the brain of healthy Beagles have been reported.42,43 Investigators in 1 study42 used single voxel, long echo time sequences at 1.5 T and found significant differences among regions of the brain (cerebellum, occipital lobe, and frontal lobe) and age groups (young, adult, and geriatric). Investigators in the other study43 used short and long echo time multivoxel techniques at 3 T to evaluate the brain. In that study,43 they did not find a normal distribution for the parameters and found large differences from the values reported in humans. Both of these studies42,43 included a qualitative assessment of the data.
Use of 3.0-T scanners in veterinary medicine is increasing. Similarly, interest in and use of 1H MRS in veterinary research subjects and clinical veterinary patients is increasing, and there is a need to standardize single voxel acquisition techniques and brain biomarker values at 3.0 T in clinically normal animals for future use in research and clinical comparisons in veterinary patients. Therefore, the purpose of the study reported here was to establish standard metabolite values for single voxel, short echo time 1H MRS at 3.0 T for various regions of the brain often involved in diseases and analyze the data by means of a linear combination model.8 We hypothesized that single voxel 1H MRS would reveal regional differences among areas of the brain, that no substantial differences would be seen between right and left hemispheres, and that no differences would be seen between male and female dogs.
Materials and Methods
ANIMALS
Ten purpose-bred research Beagles were included in the study. Five dogs were females and 5 were males; all dogs were sexually intact. Age ranged from 39 to 74 months (mean ± SD, 52.9 ± 13.5 months). The study protocol was approved by the Cantonal Veterinary Office of Zürich.
Complete clinical and neurologic examinations were performed on each dog. Blood biochemical analysis, a CBC, and CSF analysis (including total protein concentration, cell count, and differential cell count) were performed. Results were within reference limits in all dogs. Liver biopsy specimens were obtained from all dogs, and histologic evaluation revealed no abnormalities. All dogs were deemed healthy at the beginning of the study and remained healthy at follow-up examination 10 months after termination of the study.
PROCEDURES
Each dog was premedicated with methadone (0.2 mg/kg, IV) and midazolam (0.1 mg/kg, IV). Anesthesia was induced with propofol (4.5 mg/kg, IV). Endotracheal intubation was performed, and the dogs then were mechanically ventilated and anesthesia maintained with sevoflurane in oxygen. These dogs also participated concurrently in a study (unpublished) conducted to evaluate hepatic function during anesthesia. Total amount of time dogs were anesthetized was at least 4 hours (1.5 hours for the 1H MRS study and an additional 2.5 hours for the other hepatic function study [unpublished]). Blood pressure and rectal temperature (evaluated every hour) as well as an ECG (evaluated every 10 minutes) were monitoreda before and during anesthesia; results were within reference limits at all times.
IMAGING PROTOCOL
Magnetic resonance imaging was performed with a 3.0-T MRI scannerb with a 15-channel receive-transmit head coil.c For morphological evaluation of the brain, transverse, dorsal, and sagittal plane turbo spin echo T2-weighted images were obtained before spectroscopy. Scan parameters were as follows: repetition time, 320 milliseconds; echo time, 5,700 milliseconds; field of view, 160; turbo spin echo factor, 16; matrix, 320 × 320; slice thickness, 3 mm; slice gap, 2.5 mm; and number of signal averages, 2. Single voxel 1H MRS was performed by one of the investigators (IC) by use of short echo time point-resolved spectroscopy with voxels graphically prescribed from the T2-weighted images. Parameters for the 1H MRS single voxel technique were as follows: echo time, 32 milliseconds; repetition time, 2,000 milliseconds; number of signal averages, 240 (except for the cerebellum, for which the number of signal averages was 272); and bandwidth, 2,000 Hz. Dual hemispheric volume samples were placed over various regions of the brain parenchyma, including the right and left basal ganglia (voxel of interest, 1.5 cm3), right and left thalamus (voxel of interest, 1.8 cm3), and right and left parietal lobe (voxel of interest, 1.5 cm3). Midline volume samples of the occipital lobe region (voxel of interest, 1.5 cm3) and cerebellum (voxel of interest, 1 cm3) were also obtained. Voxel placement for each of the described regions was verified (Figure 1). The reason for including these anatomic regions was because they are often affected by pathological changes and they provide good placement of the voxel. Dimensions and orientation of the voxels for each anatomic area were adjusted to match the size and shape of each dog. Care was taken to avoid CSF and peripheral soft and bony tissues to prevent lipid contamination. Given that the shapes of the heads of the dogs were similar because of homogeneity of the morphological features of the skull, voxel placements in various regions of the brain were nearly identical. Before each 1H MRS acquisition, field homogeneity was optimized with a second-order automatic pencil-beam shim, which was followed by water suppression techniques (excitation). A water-unsuppressed image was also acquired to serve as a concentration reference for quantifying metabolite concentrations. In addition to a typical shim time of 2 minutes, each examination was acquired in 8 minutes 5 seconds for all regions of the brain, except for the cerebellum, for which the acquisition required 9 minutes 10 seconds. Rejection criteria for the 1H MRS data were presence of an unstable baseline, linewidth > 10 Hz, signal-to-noise ratio < 4, presence of artifacts, or presence of lipid contamination.
The T2-weighted images of the basal ganglia (A), thalamus (B), parietal lobes (C), midline of the occipital lobe (D), and midline of the cerebellum (E) of the brain of a healthy adult Beagle. Each panel represents transverse (left), dorsal (middle), and parasagittal (right; A–C) or transverse (left), dorsal (middle), and sagittal (right; D and E) images. The 1H MRS voxel of interest (white box with circle and line that bisects the circle) is positioned for the various regions of the brain. A—Basal ganglia (voxel centered on the nucleus caudatus and internal capsule; the lateral ventricle is avoided). B—Thalamus (voxel placed immediately lateral to the third ventricle, ventral to the lateral ventricles, and medial to the temporal lobe). C—Parietal lobe (voxel placed approximately midway between the cruciate sulcus and endomarginal sulcus; the lateral ventricles are avoided). D—Midline of the occipital lobe (voxel placed on the midline of the caudal aspect of the cerebral parenchyma immediately dorsal to the tentorium of the cerebellum and caudal to the endomarginal sulcus). Care is taken to avoid the occipital bone and osseous tentorium of the cerebellum. E—Midline of the cerebellum (voxel placed on the center of the cerebellum ventral to the tentorium of the cerebellum; the fourth ventricle is avoided as much as possible). Notice that the left image does not have a line bisecting the circle because it is slightly off the midline. R = Right.
Citation: American Journal of Veterinary Research 76, 2; 10.2460/ajvr.76.2.129
The T2-weighted images of the basal ganglia (A), thalamus (B), parietal lobes (C), midline of the occipital lobe (D), and midline of the cerebellum (E) of the brain of a healthy adult Beagle. Each panel represents transverse (left), dorsal (middle), and parasagittal (right; A–C) or transverse (left), dorsal (middle), and sagittal (right; D and E) images. The 1H MRS voxel of interest (white box with circle and line that bisects the circle) is positioned for the various regions of the brain. A—Basal ganglia (voxel centered on the nucleus caudatus and internal capsule; the lateral ventricle is avoided). B—Thalamus (voxel placed immediately lateral to the third ventricle, ventral to the lateral ventricles, and medial to the temporal lobe). C—Parietal lobe (voxel placed approximately midway between the cruciate sulcus and endomarginal sulcus; the lateral ventricles are avoided). D—Midline of the occipital lobe (voxel placed on the midline of the caudal aspect of the cerebral parenchyma immediately dorsal to the tentorium of the cerebellum and caudal to the endomarginal sulcus). Care is taken to avoid the occipital bone and osseous tentorium of the cerebellum. E—Midline of the cerebellum (voxel placed on the center of the cerebellum ventral to the tentorium of the cerebellum; the fourth ventricle is avoided as much as possible). Notice that the left image does not have a line bisecting the circle because it is slightly off the midline. R = Right.
Citation: American Journal of Veterinary Research 76, 2; 10.2460/ajvr.76.2.129
The T2-weighted images of the basal ganglia (A), thalamus (B), parietal lobes (C), midline of the occipital lobe (D), and midline of the cerebellum (E) of the brain of a healthy adult Beagle. Each panel represents transverse (left), dorsal (middle), and parasagittal (right; A–C) or transverse (left), dorsal (middle), and sagittal (right; D and E) images. The 1H MRS voxel of interest (white box with circle and line that bisects the circle) is positioned for the various regions of the brain. A—Basal ganglia (voxel centered on the nucleus caudatus and internal capsule; the lateral ventricle is avoided). B—Thalamus (voxel placed immediately lateral to the third ventricle, ventral to the lateral ventricles, and medial to the temporal lobe). C—Parietal lobe (voxel placed approximately midway between the cruciate sulcus and endomarginal sulcus; the lateral ventricles are avoided). D—Midline of the occipital lobe (voxel placed on the midline of the caudal aspect of the cerebral parenchyma immediately dorsal to the tentorium of the cerebellum and caudal to the endomarginal sulcus). Care is taken to avoid the occipital bone and osseous tentorium of the cerebellum. E—Midline of the cerebellum (voxel placed on the center of the cerebellum ventral to the tentorium of the cerebellum; the fourth ventricle is avoided as much as possible). Notice that the left image does not have a line bisecting the circle because it is slightly off the midline. R = Right.
Citation: American Journal of Veterinary Research 76, 2; 10.2460/ajvr.76.2.129
DATA PROCESSING
Metabolite concentrations were estimated with an automated data processing spectral fitting algorithm (linear combination model).d The software automatically adjusted the phase and chemical shift of the spectra, estimated the baseline, and performed eddy current correction. Relative metabolite concentrations and their uncertainties were estimated by fitting the spectrum to a basis set of spectra acquired from individual metabolites in solution. Seventeen metabolites were included in the linear combination model (alanine, aspartate, glucose, creatine, phosphocreatine, glutamine, glutamate, glycerophosphocholine, phosphocholine, lactate, lipids, myoinositol, NAA, N-acetylaspartylglutamate, scylloinositol, glutathione, and taurine). Only those metabolites with CRLBs < 20% were evaluated in this study.
One investigator (IC) reviewed MRI and 1H MRS images. The MRI images were assessed for abnormalities such as changes in signal intensity of the gray and white matter, presence of space-occupying lesions, mass effect, or ventriculomegaly. Metabolite concentrations relative to water and metabolite ratios with creatine as a reference were calculated from the 1H MRS data.
STATISTICAL ANALYSIS
Statistical evaluation was performed with the aid of statistical software.e Descriptive results (including mean, median, and SD) were obtained for continuous variables (age and metabolite data). Differences between left and right hemispheres in the basal ganglia, thalamus, and parietal lobes were evaluated by use of the Wilcoxon signed rank test. An independent t test was performed to describe the influence of sex by comparing all metabolites in each area of the brain. Distribution of all metabolites in the various brain regions was evaluated by use of Q-Q plots, which was followed by a 1-way ANOVA with Bonferroni post hoc correction. A 1-way ANOVA was performed on the estimated concentration of metabolites relative to water content as well as on metabolite ratios with creatine as the reference metabolite (denominator). Values of P ≤ 0.05 were considered significant for all tests.
Results
Results of MRI for T2-weighted images were considered normal in all dogs. All spectra obtained were of good quality, with no spectra being rejected. For the 1H MRS data included in the study, signal-to-noise ratios were between 17 and 9 and linewidths were between 3.96 and 5.87 Hz. Representative spectra were obtained from each of the brain regions evaluated (Figure 2). Metabolites with CRLBs < 20% were included in the statistical analysis; this included NAA, total choline (predominantly glycerophosphocholine and phosphocholine), total creatine (the sum of creatine and phosphocreatine), myoinositol, the glutamine-glutamate complex, and glutathione.
Representative 1H MRS spectra of the basal ganglia (A), thalamus (B), parietal lobes (C), occipital lobe (D), and cerebellum (E) of a healthy adult Beagle. The x-axis represents the signature chemical shift of each metabolite concentration, and the y-axis represents the signal intensity. Tissue concentration of a metabolite is related to the integrated amplitude of the MRS signal it generates, which is the area under the 1H MRS signal curve. Both peak area and height are contributing factors for determining the concentration of each metabolite. Cho = Choline. Cr = Creatine. Glx = Glutamine-glutamate complex. MI = Myoinositol.
Citation: American Journal of Veterinary Research 76, 2; 10.2460/ajvr.76.2.129
Representative 1H MRS spectra of the basal ganglia (A), thalamus (B), parietal lobes (C), occipital lobe (D), and cerebellum (E) of a healthy adult Beagle. The x-axis represents the signature chemical shift of each metabolite concentration, and the y-axis represents the signal intensity. Tissue concentration of a metabolite is related to the integrated amplitude of the MRS signal it generates, which is the area under the 1H MRS signal curve. Both peak area and height are contributing factors for determining the concentration of each metabolite. Cho = Choline. Cr = Creatine. Glx = Glutamine-glutamate complex. MI = Myoinositol.
Citation: American Journal of Veterinary Research 76, 2; 10.2460/ajvr.76.2.129
Representative 1H MRS spectra of the basal ganglia (A), thalamus (B), parietal lobes (C), occipital lobe (D), and cerebellum (E) of a healthy adult Beagle. The x-axis represents the signature chemical shift of each metabolite concentration, and the y-axis represents the signal intensity. Tissue concentration of a metabolite is related to the integrated amplitude of the MRS signal it generates, which is the area under the 1H MRS signal curve. Both peak area and height are contributing factors for determining the concentration of each metabolite. Cho = Choline. Cr = Creatine. Glx = Glutamine-glutamate complex. MI = Myoinositol.
Citation: American Journal of Veterinary Research 76, 2; 10.2460/ajvr.76.2.129
Molality of NAA, total choline, creatine, the glutamine-glutamate complex, myoinositol, and glutathione, compared with that of water, was estimated (Table 1). Metabolite ratios with creatine as the reference metabolite were calculated (Table 2).
Metabolite concentrations (mmol/L of brain water) derived from analysis of the 1H MRS spectra with a fitting algorithm (linear combination model) for the brain of 10 healthy adult Beagles.
| Variable | NAA | Total choline | Creatine | Glutamine-glutamate complex | Myoinositol | Glutathione |
|---|---|---|---|---|---|---|
| Left basal ganglia | ||||||
| Mean ± SD | 7.17 ± 0.25 | 2.17 ± 0.24 | 7.12 ± 0.43 | 13.42 ± 0.71 | 7.42 ± 0.39 | 1.95 ± 0.26 |
| Median | 7.18 | 2.17 | 7.17 | 13.51 | 7.41 | 2.05 |
| Right basal ganglia | ||||||
| Mean ± SD | 7.32 ± 0.42 | 2.25 ± 0.15 | 7.30 ± 0.55 | 13.54 ± 0.68 | 7.51 ± 0.45 | 1.89 ± 0.47 |
| Median | 7.28 | 2.27 | 7.22 | 13.54 | 7.50 | 1.88 |
| Left thalamus | ||||||
| Mean ± SD | 7.72 ± 0.34 | 2.304 ± 0.17 | 6.24 ± 0.38 | 11.23 ± 1.05 | 7.29 ± 0.32 | 1.89 ± 0.42 |
| Median | 7.77 | 2.28 | 6.34 | 11.35 | 7.33 | 1.89 |
| Right thalamus | ||||||
| Mean ± SD | 7.85 ± 0.33 | 2.21 ± 0.15 | 6.20 ± 0.41 | 11.20 ± 0.95 | 7.27 ± 0.43 | 1.89 ± 0.22 |
| Median | 7.90 | 2.22 | 6.25 | 11.57 | 7.37 | 1.88 |
| Left parietal lobe | ||||||
| Mean ± SD | 7.70 ± 0.67 | 1.95 ± 0.11 | 5.96 ± 0.48 | 12.39 ± 1.81 | 7.56 ± 0.59 | 1.77 ± 0.23 |
| Median | 7.89 | 1.94 | 6.16 | 12.60 | 7.23 | 1.86 |
| Right parietal lobe | ||||||
| Mean ± SD | 7.87 ± 0.30 | 1.94 ± 0.15 | 6.20 ± 0.33 | 12.12 ± 0.77 | 8.04 ± 0.82 | 1.86 ± 0.28 |
| Median | 7.85 | 1.98 | 6.10 | 12 | 8.05 | 1.85 |
| Occipital lobe | ||||||
| Mean ± SD | 7.28 ± 0.46 | 1.89 ± 0.11 | 6.93 ± 0.39 | 12.37 ± 1.11 | 8.73 ± 0.50 | 1.98 ± 0.58 |
| Median | 7.46 | 1.84 | 7.05 | 12.63 | 8.74 | 2.06 |
| Cerebellum | ||||||
| Mean ± SD | 6.87 ± 0.85 | 2.09 ± 0.20 | 7.040 ± 0.63 | 13.35 ± 3.05 | 10.77 ± 0.866 | 2.28 ± 0.60 |
| Median | 6.85 | 2.04 | 7.13 | 12.71 | 10.626 | 2.32 |
Metabolite ratios calculated for the brain of 10 healthy adult Beagles.
| Variable | NAA-to-creatine ratio | Total choline-to-creatine ratio | Glutamine-glutamate complex-to-creatine ratio | Myoinositol-to-creatine ratio | Glutathione-to-creatine ratio |
|---|---|---|---|---|---|
| Left basal ganglia | |||||
| Mean ± SD | 1.01 ± 0.05 | 0.31 ± 0.02 | 1.89 ± 0.14 | 1.04 ± 0.04 | 0.27 ± 0.03 |
| Median | 0.99 | 0.30 | 1.89 | 1.04 | 0.28 |
| Right basal ganglia | |||||
| Mean ± SD | 1.01 ± 0.07 | 0.31 ± 0.02 | 1.86 ± 0.11 | 1.03 ± 0.06 | 0.26 ± 0.06 |
| Median | 1.00 | 0.31 | 1.92 | 1.04 | 0.25 |
| Left thalamus | |||||
| Mean ± SD | 1.20 ± 0.13 | 0.37 ± 0.04 | 1.80 ± 0.21 | 1.17 ± 0.10 | 0.31 ± 0.04 |
| Median | 1.20 | 0.35 | 1.76 | 1.15 | 0.31 |
| Right thalamus | |||||
| Mean ± SD | 1.27 ± 0.09 | 0.35 ± 0.02 | 1.80 ± 0.18 | 1.17 ± 0.13 | 0.30 ± 0.03 |
| Median | 1.28 | 0.35 | 1.84 | 1.15 | 0.31 |
| Left parietal lobe | |||||
| Mean ± SD | 1.29 ± 0.06 | 0.32 ± 0.02 | 2.07 ± 0.26 | 1.27 ± 0.09 | 0.30 ± 0.04 |
| Median | 1.31 | 0.32 | 2.07 | 1.26 | 0.30 |
| Right parietal lobe | |||||
| Mean ± SD | 1.26 ± 0.08 | 0.31 ± 0.02 | 1.95 ± 0.13 | 1.32 ± 0.09 | 0.29 ± 0.04 |
| Median | 1.28 | 0.31 | 1.93 | 1.33 | 0.31 |
| Occipital lobe | |||||
| Mean ± SD | 1.04 ± 0.04 | 0.27 ± 0.10 | 1.78 ± 0.10 | 1.26 ± 0.06 | 0.31 ± 0.02 |
| Median | 1.02 | 0.27 | 1.76 | 1.25 | 0.32 |
| Cerebellum | |||||
| Mean ± SD | 0.98 ± 0.08 | 0.29 ± 0.03 | 1.78 ± 0.28 | 1.53 ± 0.12 | 0.32 ± 0.09 |
| Median | 1.00 | 0.31 | 1.73 | 1.54 | 0.31 |
Ratios were calculated with the creatine concentration as the reference (denominator) metabolite, which is similar to the convention used for humans.
Comparison of metabolite concentrations between the left and right basal ganglia, thalamus, and parietal lobes revealed 2 significant differences (Tables 3 and 4). Parietal lobe concentrations (relative to that of brain water) for myoinositol (P = 0.047) and the myoinositol-to-creatine ratio (P = 0.021) differed significantly between the right and left hemispheres. No significant differences were detected between the right and left hemispheres for the remainder of the metabolites for all regions evaluated. No significant differences were detected between males and females.
Results of statistical analysis (Wilcoxon signed rank test) of differences in metabolite concentrations between left and right hemispheres of various regions of the brain of 10 healthy Beagles.
| Variable | NAA | Total choline | Creatine | Glutamine-glutamate complex | Myoinositol | Glutathione |
|---|---|---|---|---|---|---|
| Basal ganglia | 0.575 | 0.167 | 0.093 | 0.878 | 0.508 | 0.799 |
| Parietal lobes | 0.508 | 0.799 | 0.093 | 0.241 | 0.047 | 0.646 |
| Thalamus | 0.114 | 0.139 | 0.508 | 0.859 | 0.878 | 0.878 |
Data are P values; values were considered significantly different at P ≤ 0.05.
Results of statistical analysis (Wilcoxon signed rank test) of differences in metabolite concentration ratios between left and right hemispheres of various regions of the brain of 10 healthy Beagles.
| Variable | NAA | Total choline | Creatine | Glutamine-glutamate complex | Myoinositol | Glutathione |
|---|---|---|---|---|---|---|
| Basal ganglia | 0.799 | 0.169 | 1.000 | 0.721 | 0.859 | 0.878 |
| Parietal lobes | 0.285 | 0.241 | 1.000 | 0.203 | 0.021 | 0.859 |
| Thalamus | 0.169 | 0.109 | 1.000 | 0.575 | 0.799 | 0.683 |
Creatine was used as the reference (denominator) metabolite.
See Table 3 for remainder of key.
The 1-way ANOVA with Bonferroni post hoc correction revealed significant regional differences. Comparisons between right and left hemispheres were performed for all brain regions evaluated (Tables 5 and 6).
Results of statistical analysis (1-way ANOVA with Bonferroni post hoc correction) of differences in concentrations of metabolite concentrations in various regions of the brain of 10 healthy adult Beagles.
| Variable | Basal ganglia | Cerebellum | Occipital lobe | Parietal lobes | Thalamus |
|---|---|---|---|---|---|
| Creatine | |||||
| Basal ganglia | — | 1.000 | 1.000 | < 0.001 | 0.003 |
| Cerebellum | 1.000 | — | 1.000 | < 0.001 | 0.009 |
| Occipital lobe | 1.000 | 1.000 | — | 0.001 | 0.035 |
| Parietal lobes | < 0.001 | < 0.001 | 0.001 | — | 1.000 |
| Thalamus | 0.003 | 0.009 | 0.004 | 1.000 | — |
| NAA | |||||
| Basal ganglia | — | 1.000 | 1.000 | 0.530 | 0.458 |
| Cerebellum | 1.000 | — | 1.000 | 0.032 | 0.027 |
| Occipital lobe | 1.000 | 1.000 | — | 1.000 | 1.000 |
| Parietal lobes | 0.530 | 0.032 | 1.000 | — | 1.000 |
| Thalamus | 0.458 | 0.027 | 1.000 | 1.000 | — |
| Total choline | |||||
| Basal ganglia | — | 1.000 | 0.001 | 0.017 | 1.000 |
| Cerebellum | 1.000 | — | 0.081 | 0.601 | 0.079 |
| Occipital lobe | 0.001 | 0.081 | — | 1.000 | < 0.001 |
| Parietal lobes | 0.017 | 0.601 | 1.000 | — | < 0.001 |
| Thalamus | 1.000 | 0.079 | < 0.001 | < 0.001 | — |
| Glutamine-glutamate | |||||
| complex | |||||
| Basal ganglia | — | 1.000 | 1.000 | 1.000 | 0.117 |
| Cerebellum | 1.000 | — | 1.000 | 1.000 | 0.141 |
| Occipital lobe | 1.000 | 1.000 | — | 1.000 | 1.000 |
| Parietal lobes | 1.000 | 1.000 | 1.000 | — | 1.000 |
| Thalamus | 0.117 | 0.141 | 1.000 | 1.000 | — |
| Myoinositol | |||||
| Basal ganglia | — | 0.001 | 0.001 | 1.000 | 1.000 |
| Cerebellum | 0.001 | — | 0.001 | 0.001 | 0.001 |
| Occipital lobe | 0.001 | 0.001 | — | 0.001 | 0.001 |
| Parietal lobes | 1.000 | 0.001 | 0.001 | — | 1.000 |
| Thalamus | 1.000 | 0.001 | 0.001 | 1.000 | — |
| Glutathione | |||||
| Basal ganglia | — | 1.000 | 1.000 | 1.000 | 1.000 |
| Cerebellum | 1.000 | — | 1.000 | 0.197 | 0.669 |
| Occipital lobe | 1.000 | 1.000 | — | 1.000 | 1.000 |
| Parietal lobes | 1.000 | 0.197 | 1.000 | — | 1.000 |
| Thalamus | 1.000 | 0.669 | 1.000 | 1.000 | — |
Represents results with the left hemisphere used for comparison; results obtained with the right hemisphere used for comparison were the same (data not shown). Values were considered significantly different at P ≤ 0.05. — = Not applicable.
Results of statistical analysis (1-way ANOVA with Bonferroni post hoc correction) of differences in metabolite ratios in various regions of the brain of 10 healthy adult Beagles.
| Variable | Basal ganglia | Cerebellum | Occipital lobe | Parietal lobes | Thalamus |
|---|---|---|---|---|---|
| NAA-to-creatine ratio | |||||
| Basal ganglia | — | 1.000 | 1.000 | < 0.001 | < 0.001 |
| Cerebellum | 1.000 | — | < 0.001 | < 0.001 | |
| Occipital lobe | 1.000 | 1.000 | — | < 0.001 | < 0.001 |
| Parietal lobes | < 0.001 | < 0.001 | < 0.001 | — | 1.000 |
| Thalamus | < 0.001 | < 0.001 | < 0.001 | 1.000 | — |
| Total choline-to-creatine ratio | |||||
| Basal ganglia | — | 1.000 | 0.098 | 1.000 | 0.001 |
| Cerebellum | 1.000 | — | 0.611 | 1.000 | < 0.001 |
| Occipital lobe | 0.098 | 0.611 | — | 0.002 | < 0.001 |
| Parietal lobes | 1.000 | 0.347 | 0.002 | — | 0.039 |
| Thalamus | 0.001 | < 0.001 | < 0.001 | 0.039 | — |
| Glutamine-glutamate complex-to-creatine ratio | |||||
| Basal ganglia | — | 1.000 | 1.000 | 0.730 | 1.000 |
| Cerebellum | 1.000 | — | 1.000 | 0.058 | 1.000 |
| Occipital lobe | 1.000 | 1.000 | — | 0.057 | 1.000 |
| Parietal lobes | 0.730 | 0.058 | 0.057 | — | 0.106 |
| Thalamus | 1.000 | 1.000 | 1.000 | 0.106 | — |
| Myoinositol-to-creatine ratio | |||||
| Basal ganglia | — | < 0.001 | < 0.001 | < 0.001 | 0.045 |
| Cerebellum | < 0.001 | — | < 0.001 | < 0.001 | < 0.001 |
| Occipital lobe | < 0.001 | < 0.001 | — | < 0.001 | 0.450 |
| Parietal lobes | < 0.001 | < 0.001 | 1.000 | — | 0.243 |
| Thalamus | 0.045 | < 0.001 | 0.450 | 0.243 | — |
| Glutathione-to-creatine ratio | |||||
| Basal ganglia | — | 1.000 | 1.000 | 0.730 | 1.000 |
| Cerebellum | 1.000 | — | 1.000 | 0.058 | 1.000 |
| Occipital lobe | 1.000 | 1.000 | — | 0.057 | 1.000 |
| Parietal lobes | 0.730 | 0.058 | 0.057 | — | 0.106 |
| Thalamus | 1.000 | 1.000 | 1.000 | 0.106 | — |
Creatine was used as the reference (denominator) metabolite.
See Table 5 for remainder of key.
The lowest creatine concentration was detected in the parietal lobes (left hemisphere, 5.96 mmol/L; right hemisphere, 6.10 mmol/L), and the highest was detected in the basal ganglia (left hemisphere, 7.12 mmol/L; right hemisphere, 7.30 mmol/L) and cerebellum (7.04 mmol/L). Generally, the creatine concentration was highest in the basal ganglia, cerebellum, and occipital lobe, whereas the creatine concentration typically was lower in the parietal lobes and thalamus. Comparison of the creatine concentration among regions of the brain revealed several significant differences. Regional creatine concentrations differed significantly between the parietal lobes and the basal ganglia (right and left hemispheres of both structures), cerebellum, and occipital lobe (all comparisons, P < 0.001). Regional creatine concentrations also differed significantly between the right and left hemispheres of the thalamus versus the right and left hemispheres of the basal ganglia (P = 0.003), cerebellum (P = 0.009), and occipital lobe (P = 0.035).
The lowest NAA concentration was detected in the cerebellum (6.87 mmol/L), and the highest concentration was in the parietal lobes (left hemisphere, 7.70 mmol/L; right hemisphere, 7.87 mmol/L). Relative concentrations of NAA were significantly different between the cerebellum versus right and left parietal lobes (P = 0.032) and the cerebellum versus the right and left thalamus (P = 0.027). With regard to the NAA-to-creatine ratio, significant (P = 0.005) differences were detected between the parietal lobes versus the basal ganglia, cerebellum versus the occipital lobe, and thalamus versus the basal ganglia, cerebellum, and occipital lobe.
The lowest choline concentration was detected in the occipital lobes (1.89 mmol/L), and the highest was in the basal ganglia (left hemisphere, 2.17 mmol/L; right hemisphere, 2.25 mmol/L). Comparison of metabolite concentrations relative to water revealed significant differences between the occipital lobe versus basal ganglia (P = 0.001), parietal lobes versus basal ganglia (P = 0.017), thalamus versus occipital lobe (P < 0.001), and thalamus versus parietal lobes (P < 0.001). Significant regional differences were evident for choline-to-creatine ratios. These included thalamus versus basal ganglia (P = 0.001), thalamus versus cerebellum (P < 0.001), thalamus versus occipital lobe (P = 0.001), thalamus versus parietal lobes (P = 0.039), and parietal lobes versus occipital lobe (P = 0.002).
The lowest concentration for the glutamine-glutamate complex was detected in the thalamus (left hemisphere, 11.23 mmol/L; right hemisphere, 11.20 mmol/L), and the highest concentration was in the basal ganglia (left hemisphere, 13.42 mmol/L; right hemisphere, 13.54 mmol/L). There were no significant differences with regard to the glutamine-glutamate complex concentrations or corresponding ratios of the glutamine-glutamate complex to creatine.
The lowest myoinositol concentration was detected in the thalamus (left hemisphere, 7.29 mmol/L; right hemisphere, 7.27 mmol/L), and the highest was detected in the cerebellum (10.77 mmol/L). Relative concentrations of myoinositol were significantly (P < 0.001) different between the cerebellum versus the basal ganglia, occipital lobe, parietal lobes, and thalamus; occipital lobe versus the basal ganglia, cerebellum, parietal lobes, and thalamus; parietal lobes versus the cerebellum and occipital lobe; and thalamus versus the cerebellum and occipital lobe. Comparison of metabolite ratios with creatine as the reference metabolite revealed significant differences between the basal ganglia versus the cerebellum, occipital lobe, and right and left parietal lobes (P < 0.001); basal ganglia versus the thalamus (P = 0.045); and cerebellum versus the occipital lobe, right and left parietal lobes, and thalamus (P < 0.001).
The lowest glutathione concentration was detected in the parietal lobes (left hemisphere, 1.77 mmol/L; right hemisphere, 1.86 mmol/L), and the highest was in the cerebellum (2.28 mmol/L). No significant differences among regions were detected for glutathione concentrations relative to the glutathione-to-creatine ratio.
Discussion
In the present study, we found metabolite differences among various regions of the brain but no substantial differences between the right and left hemispheres and between females and males. The lowest concentration of total choline was in the occipital lobe, the highest concentration of creatine was in the cerebellum and basal ganglia, the lowest concentration of creatine was in the parietal lobes, and the highest concentration of NAA was in the parietal lobes. These findings are in agreement with those in humans.15,44,45 Gray and white matter are easily differentiated in humans.15,44,45 Because of the smaller size of the canine brain, single voxel differentiation of gray and white matter is challenging. Optimization of the technique with a smaller voxel size, multivoxel technique, and segmentation techniques may yield better differentiation between gray and white matter of dogs in the future.
Some differences were evident between regional concentrations and metabolite concentrations for the dogs reported here, compared with results for a study15 in humans that involved use of a similar magnetic field strength (3.0 T), short echo time point-resolved spectroscopy sequence, and spectral editing program (linear combination model). However, these differences should be interpreted cautiously given that corrections for T1 and T2 relaxation and CSF partial volume were used in the human study.15 We did not use T1 and T2 corrections and CSF corrections because it would have required an increased time for image acquisition and necessitated performing segmentation, respectively. We attempted to minimize CSF contamination for the voxel placement in all regions. The most challenging region was the cerebellum because of invaginations of the folia.
Comparisons between dogs and humans for a single region of the brain (ie, the thalamus) revealed interesting results. The mean ± SD apparent concentration for dogs and humans,15 respectively, of myoinositol was higher (7.29 ± 0.32mM vs 3.53 ± 0.52mM), NAA concentration was lower (7.72 ± 0.34mM vs 13.56 ± 0.71mM), choline concentration slightly higher (2.3 ± 0.17mM vs 1.89 ± 0.21mM), concentration of the glutamine-glutamate complex slightly higher (11.23 ± 0.95mM vs 10.33 ± 1.40mM), and creatine concentration lower (6.24 ± 0.38mM vs 9.22 ± 1.16mM).
In the present study, the NAA-to-creatine ratio was lower in the cerebellum than in other regions of the brain. This finding is in agreement with results of another study42 in which metabolite concentrations in the brain of clinically normal Beagles was evaluated by use of long echo time, single voxel techniques. The authors of that study42 used manual integration of peaks and reported a higher choline-to-creatine ratio in dogs (1.18 to 2.39), compared with that in humans (0.26 to 0.50). Findings for the study reported here were similar to those reported in humans,42 and we suspect that the reason for this is that we used the same spectral fitting program (linear combination model). We presumed this was the reason that we found higher ratios, compared with those for humans.
Quantification of spectral peaks plays an important role in MRS, whereas pure visual examination of spectra is less appropriate.1,4 Peak height alone does not represent metabolite concentrations. The appropriately quantified concentration is represented by the area under the peak. There are 2 main methods for determining area under the peak: manual or automatic integration and fitting model.1,2 For the 2 methods, an investigator or a computer program selects 2 frequency points (one to the left and the second to the right of a metabolite peak). The amplitude at these 2 points is assumed to define the spectral baseline and is set to zero. The computer then proceeds to integrate the area under the peak between the 2 frequency points in accordance with a numerical algorithm. Manual peak integration is extremely intuitive, but it has substantial limitations. First, in vivo spectra are commonly crowded and contain many overlapping metabolite peaks; therefore, it may be difficult to distinguish between adjacent peaks, which results in cross-talk. For instance, choline compounds and total creatine are separated by 0.2 ppm, and there is no clear boundary between them. Of even greater concern is the overlap between some singlet resonances and broader multiplets, such as NAA resonating at 2.02 ppm and the broad multiplet of glutamine and glutamate resonating between 2.0 and 2.4 ppm.
Problems of numerical peak integration can be resolved with sophisticated spectral fitting programs, such as linear combination of model spectra.8 This program models in vivo spectra as a linear superimposition of the basis spectra, each of which represents a specific metabolite in the organ of interest. The basis spectra are generated by quantum mechanical simulations or are acquired from model solutions of individual metabolites. The fitting program determines the contribution of each basis spectrum to a given in vivo spectrum and thus determines the relative concentration of each metabolite set.8 Benefits of the linear combination model are that it exploits all resonances in the molecule; it is fully automated and user independent, with baseline and phase line corrections; and, with appropriate calibration data, it can provide absolute metabolite concentrations and the estimate of uncertainty (CRLBs).
Spectral fitting programs aid in calculation of metabolite concentrations in 2 main ways: by providing metabolite concentrations relative to water and by allowing for the calculation of metabolite ratios (with a metabolite used as a reference). The convention for metabolite ratios is to use creatine as the reference (ie, denominator). Ratios adjust for unknown or uncontrollable experimental conditions that may affect spectral accuracy, such as poor magnetic field homogeneity, instrumental gain drifts, differences in image acquisition and localization methods, and voxel partial volume contamination of the voxel of interest with CSF. The use of creatine as the reference metabolite has been questioned in human medicine.12–14 As indicated in the present study of dogs, creatine concentrations can differ among regions of the brain in humans and rats.13,15–18 Furthermore, creatine concentration increases with age, which also raises questions as to the validity of the conversion.18 In dogs of the study reported here, we found higher creatine concentrations in the basal ganglia and cerebellum and lower creatine concentrations in the parietal lobes, occipital lobe, and thalamus. Consequently, care must be taken when interpreting ratio calculation data with creatine as a reference metabolite in dogs. This has also been suggested elsewhere in the human and experimental animal literature.12,14,17 Contrary to the present study in which the concentrations of creatine were similar in the basal ganglia and cerebellum, creatine concentrations are higher in the cerebellum than the basal ganglia in humans and rats.15,17
Glutathione resonates at 3.77 ppm, which is extremely close to the glutamine-glutamate complex peak at 3.75 ppm. Linear combination model analysis of glutathione concentrations revealed no significant differences among the various regions of the brain that were evaluated. However, these results should be interpreted with caution because detection of glutathione is improved with special spectral editing techniques or stronger magnetic fields (eg, 7.0 T).46–48
In clinical brain spectroscopy examinations of humans, it is a common strategy to analyze a region of interest in comparison to the contralateral, presumably healthy side. Studies41,49 of humans have revealed minimal to no hemispheric asymmetry of brain metabolites. In the present study, only the concentration of myoinositol in the parietal lobes was significantly different when compared between hemispheres. We do not have a definitive reason to explain why that one metabolite had left-to-right hemispheric variation, whereas the other metabolites did not. Investigators of a recent multivoxel study43 of the brain of clinically normal Beagles reported asymmetry between the left and right hemispheres for all metabolites evaluated. The discrepancy of these results could be explained by differences in technique (multivoxel vs single voxel; long echo time vs short echo time). The possibility of artifacts also must be considered. Artifactual spectral asymmetry includes inhomogeneities of the static and radiofrequency magnetic fields or pulse sequence–related factors.49 For instance, point-resolved spectroscopy chemical shift imaging causes substantial left-to-right differences in metabolite ratios as a result of the interaction between the limited bandwidth of the slice-selective 90° and 180° refocusing pulses and the chemical shifts of individual resonances.50
We did not detect any differences in brain metabolite concentrations between males and females in the dogs of the present study. This is in accordance with results for studies49,51 of humans.
Age-related variation in metabolite concentrations in canine brains have been reported.42 Age-related differences were beyond the scope of this investigation, but the dogs of the present study were all considered adults (age range, 39 to 74 months); therefore, variances with regard to age were not assessed. Future studies that involve use of single voxel, short echo time acquisition at 3.0 T to investigate effects of aging in the canine and feline brain are necessary.
Apparent brain metabolite concentrations measured with different scanners and techniques have been reported for humans.40 In general, they indicate similar patterns, but absolute values can differ among scanners.40 Therefore, the values reported in the present study may serve as a reference, but probably not as a direct comparison. It is recommended that investigators collect data for individual control subjects with specific scanners and protocols for use in direct comparison with values in patients with intracranial disease.
Limitations of the study reported here included the fact that complete MRI morphological evaluation of the brain in accordance with a standard protocol was not performed. Morphological evaluation was limited to T2-weighted evaluation because of time constraints. However, there was no history of previous intracranial disease, and results of neurologic examination, hematologic analysis, blood biochemical analysis, CSF analysis, and histologic evaluation of liver biopsy specimens were within reference limits or negative for pathological changes. Thus, the presence of intracranial disease was deemed highly unlikely. Another limitation was that T1 and T2 relaxation times were not determined because image acquisition times were excessive. Therefore, no corrections for T2 and T1 effects were applied and absolute metabolite concentrations were not determined. Other limitations were that CSF corrections were not applied and we were unable to obtain pure spectra for white and gray matter separately, mainly because of the smaller size of the canine brain in comparison to that of humans. Future studies with segmentation may help to isolate white and gray matter spectra.
To the authors’ knowledge, the study reported here was the first in which regional metabolite concentrations were compared in various regions of the brain of clinically normal dogs and variability of the conventionally used reference metabolite creatine was evaluated. Establishing reference limits for metabolite concentrations and a standardized protocol with technical and postprocessing approach will aid diagnosticians, clinicians, and researchers. Furthermore, these metabolite concentrations will assist with comparisons when dual hemisphere spectroscopy is not possible, such as for mass lesions and structures located on the midline or widespread systemic disease.
Acknowledgments
This manuscript represents a portion of a thesis submitted by Dr. Carrera to the Bern University Graduate School for Cellular and Biomedical Sciences as partial fulfillment of the requirements for a Doctor of Philosophy degree.
The authors thank Drs. Chris Boesch, Ronald Kreis, and Fraser McConnell for technical assistance.
ABBREVIATIONS
| CRLB | Cramer-Rao lower bound |
| 1HMRS | Proton magnetic resonance spectroscopy |
| NAA | N-acetyl aspartate |
Footnotes
N-MRI2-01 Datex Ohmeda, Anandic Medical Systems AG, Diessenhofen, Switzerland.
Philips Ingenia scanner, Philips AG, Zurich, Switzerland.
dStream HeadSpine coil solution, Philips AG, Zurich, Switzerland.
LCModel, version 6.3, S Provencher, Oakville, ON, Canada.
IBM SPSS statistics, version 21.0.0.0, 64-bit edition, IBM SPSS, Chicago, Ill.
References
1. Barker PB, Bizzi A, Stefano ND, et al. Introduction to MR spectroscopy. In: Clinical MR spectroscopy. Cambridge, England: Cambridge University Press, 2010; 1–19.
2. Graaf RAD. In vivo NMR spectroscopy-static aspects. In: In vivo NMR spectroscopy: principles and techniques. 2nd ed. Chichester, West Sussex, England: John Wiley and Sons Ltd, 2007; 43–111.
3. Gruetter R, Weisdorf SA, Rajanayagan V, et al. Resolution improvements in in vivo 1H NMR spectra with increased magnetic field strength. J Magn Reson 1998; 135: 260–264.
4. Alger JR. Quantitative proton magnetic resonance spectroscopy and spectroscopic imaging of the brain: a didactic review. Top Magn Reson Imaging 2010; 21: 115–128.
5. Gillies P, Marshall I, Asplund M, et al. Quantification of MRS data in the frequency domain using a wavelet filter, an approximated Voigt lineshape model and prior knowledge. NMR Biomed 2006; 19: 617–626.
6. Laudadio T, Selen Y, Vanhamme L, et al. Subspace-based MRS data quantitation of multiplets using prior knowledge. J Magn Reson 2004; 168: 53–65.
7. Mierisová S, Ala-Korpela M. MR spectroscopy quantitation: a review of frequency domain methods. NMR Biomed 2001; 14: 247–259.
8. Provencher SW. Automatic quantitation of localized in vivo 1H spectra with LCModel. NMR Biomed 2001; 14: 260–264.
9. Soher BJ, Young K, Govindaraju V, et al. Automated spectral analysis III: application to in vivo proton MR spectroscopy and spectroscopic imaging. Magn Reson Med 1998; 40: 822–831.
10. Vanhamme L, Van Huffel S, Van Hecke P, et al. Time-domain quantification of series of biomedical magnetic resonance spectroscopy signals. J Magn Reson 1999; 140: 120–130.
11. Provencher SW. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med 1993; 30: 672–679.
12. Jansen JF, Backes WH, Nicolay K, et al. 1H MR spectroscopy of the brain: absolute quantification of metabolites. Radiology 2006; 240: 318–332.
13. Li BS, Wang H, Gonen O. Metabolite ratios to assumed stable creatine level may confound the quantification of proton brain MR spectroscopy. Magn Reson Imaging 2003; 21: 923–928.
14. Miller BL. A review of chemical issues in 1H NMR spectroscopy:N-acetyl-l-aspartate, creatine and choline. NMR Biomed 1991; 4: 47–52.
15. Baker EH, Basso G, Barker PB, et al. Regional apparent metabolite concentrations in young adult brain measured by 1H MR spectroscopy at 3 Tesla. J Magn Reson Imaging 2008; 27: 489–499.
16. Jacobs MA, Horska A, van Zijl PC, et al. Quantitative proton MR spectroscopic imaging of normal human cerebellum and brain stem. Magn Reson Med 2001; 46: 699–705.
17. Mayer D, Zahr NM, Sullivan EV, et al. In vivo metabolite differences between the basal ganglia and cerebellum of the rat brain detected with proton MRS at 3T. Psychiatry Res 2007; 154: 267–273.
18. Reyngoudt H, Claeys T, Vlerick L, et al. Age-related differences in metabolites in the posterior cingulate cortex and hippocampus of normal ageing brain: a 1H-MRS study. Eur J Radiol 2012; 81: e223–e231.
19. Cavassila S, Deval S, Huegen C, et al. Cramer-Rao bounds: an evaluation tool for quantitation. NMR Biomed 2001; 14: 278–283.
20. Kreis R. Issues of spectral quality in clinical 1H-magnetic resonance spectroscopy and a gallery of artifacts. NMR Biomed 2004; 17: 361–381.
21. Kousi E, Tsougos I, Tsolaki E, et al. Spectroscopic evaluation of glioma grading at 3T: the combined role of short and long TE. ScientificWorldJournal 2012; 2012: 546171.
22. Mekle R, Mlynarik V, Gambarota G, et al. MR spectroscopy of the human brain with enhanced signal intensity at ultrashort echo times on a clinical platform at 3T and 7T. Magn Reson Med 2009; 61: 1279–1285.
23. Soher BJ, Vermathen P, Schuff N, et al. Short TE in vivo 1H MR spectroscopic imaging at 1.5 T: acquisition and automated spectral analysis. Magn Reson Imaging 2000; 18: 1159–1165.
24. Caivano R, Lotumolo A, Rabasco P, et al. 3 Tesla magnetic resonance spectroscopy: cerebral gliomas vs. metastatic brain tumors. Our experience and review of the literature. Int J Neurosci 2013; 123: 537–543.
25. Chang L, Munsaka SM, Kraft-Terry S, et al. Magnetic resonance spectroscopy to assess neuroinflammation and neuropathic pain. J Neuroimmune Pharmacol 2013; 8: 576–593.
26. Jaskólski DJ, Fortuniak J, Stefaczyk L, et al. Differential diagnosis of intracranial meningiomas based on magnetic resonance spectroscopy. Neurol Neurochir Pol 2013; 47: 247–255.
27. Mader I, Rauer S, Gall P, et al. 1H MR spectroscopy of inflammation, infection and ischemia of the brain. Eur J Radiol 2008; 67: 250–257.
28. Möller-Hartmann W, Herminghaus S, Krings T, et al. Clinical application of proton magnetic resonance spectroscopy in the diagnosis of intracranial mass lesions. Neuroradiology 2002; 44: 371–381.
29. Morita N, Harada M, Otsuka H, et al. Clinical application of MR spectroscopy and imaging of brain tumor. Magn Reson Med Sci 2010; 9: 167–175.
30. Panigrahy A, Nelson MD Jr, Bluml S. Magnetic resonance spectroscopy in pediatric neuroradiology: clinical and research applications. Pediatr Radiol 2010; 40: 3–30.
31. Rossi A, Biancheri R. Magnetic resonance spectroscopy in metabolic disorders. Neuroimaging Clin North Am 2013; 23: 425–448.
32. Sailasuta N, Ross W, Ananworanich J, et al. Change in brain magnetic resonance spectroscopy after treatment during acute HIV infection. PLoS ONE 2012; 7(11):e49272.
33. Tartaglia MC, Arnold DL. The role of MRS and fMRI in multiple sclerosis. Adv Neurol 2006; 98: 185–202.
34. Anderson JH, Strandberg JD, Wong DF, et al. Multimodality correlative study of canine brain tumors. Proton magnetic resonance spectroscopy, positron emission tomography, and histology. Invest Radiol 1994; 29: 597–605.
35. Barker PB, Blackband SJ, Chatham JC, et al. Quantitative proton spectroscopy and histology of a canine brain tumor model. Magn Reson Med 1993; 30: 458–464.
36. Lee SH, Kim SY, Woo DC, et al. Differential neurochemical responses of the canine striatum with pentobarbital or ketamine anesthesia: a 3T proton MRS study. J Vet Med Sci 2010; 72: 583–587.
37. Magnitsky S, Vite CH, Delikatny EJ, et al. Magnetic resonance spectroscopy of the occipital cortex and the cerebellar vermis distinguishes individual cats affected with alpha-mannosidosis from normal cats. NMR Biomed 2010; 23: 74–79.
38. Carrera I, Kircher PR, Meier D, et al. In vivo proton magnetic resonance spectroscopy for the evaluation of hepatic encephalopathy in dogs. Am J Vet Res 2014; 75: 818–827.
39. Vite CH, Cross JR. Correlating magnetic resonance findings with neuropathology and clinical signs in dogs and cats. Vet Radiol Ultrasound 2011; 52: S23–S31.
40. Haga KK, Khor YP, Farrall A, et al. A systematic review of brain metabolite changes, measured with 1H magnetic resonance spectroscopy, in healthy aging. Neurobiol Aging 2009; 30: 353–363.
41. Komoroski RA, Heimberg C, Cardwell D, et al. Effects of gender and region on proton MRS of normal human brain. Magn Reson Imaging 1999; 17: 427–433.
42. Ono K, Kitagawa M, Ito D, et al. Regional variations and age-related changes detected with magnetic resonance spectroscopy in the brain of healthy dogs. Am J Vet Res 2014; 75: 179–186.
43. Warrington CD, Feeney DA, Ober CP, et al. Relative metabolite concentrations and ratios determined by use of 3-T region-specific proton magnetic resonance spectroscopy of the brain of healthy Beagles. Am J Vet Res 2013; 74: 1291–1303.
44. Michaelis T, Merboldt KD, Bruhn H, et al. Absolute concentrations of metabolites in the adult human brain in vivo: quantification of localized proton MR spectra. Radiology 1993; 187: 219–227.
45. Pouwels PJ, Frahm J. Regional metabolite concentrations in human brain as determined by quantitative localized proton MRS. Magn Reson Med 1998; 39: 53–60.
46. Choi C, Dimitrov IE, Douglas D, et al. Improvement of resolution for brain coupled metabolites by optimized 1H MRS at 7T. NMR Biomed 2010; 23: 1044–1052.
47. Emir UE, Raatz S, McPherson S, et al. Noninvasive quantification of ascorbate and glutathione concentration in the elderly human brain. NMR Biomed 2011; 24: 888–894.
48. Godlewska BR, Near J, Cowen PJ. Neurochemistry of major depression: a study using magnetic resonance spectroscopy [Published online ahead of print Jul 31, 2014]. Psychopharmacology (Berl) doi: 10.1007/s00213-014-3687-y.
49. Nagae-Poetscher LM, Bonekamp D, Barker PB, et al. Asymmetry and gender effect in functionally lateralized cortical regions: a proton MRS imaging study. J Magn Reson Imaging 2004; 19: 27–33.
50. Nelson SJ. Analysis of volume MRI and MR spectroscopic imaging data for the evaluation of patients with brain tumors. Magn Reson Med 2001; 46: 228–239.
51. Charles HC, Lazeyras F, Krishnan KR, et al. Proton spectroscopy of human brain: effects of age and sex. Prog Neuropsychopharmacol Biol Psychiatry 1994; 18: 995–1004.