Proton MR Spectroscopy for Pediatric Neuroimaging: Key Concepts for Practice

Abstract

Magnetic resonance spectroscopy (MRS) of the brain provides the clinician an in vivo neurochemical assessment within the clinical magnetic resonance imaging (MRI) setting. These neurochemical assessments can yield specificity to the clinical MRI setting when addressing questions pertaining to brain health and metabolism while characterizing disease and injury, evaluating treatment response and prognosticating outcome. Proton MRS techniques can be useful in narrowing the diagnostic differential and capturing time-sensitive information for continually developing pediatric population. This paper provides a review of key proton MRS topics relevant for usage in pediatric populations. We discuss magnetic field strength, pediatric sized head coils, water suppression techniques, localization pulse sequences, post-processing methods, analysis, and interpretation. These elements all require special consideration, particularly for the immature brain. We introduce the fundamentals of spectral editing. Finally, we present Illustrative examples employing proton MRS in clinical practice to begin to synthesize these concepts into practical application.

1. Introduction

Magnetic resonance spectroscopy (MRS) of the brain provides the clinician an in vivo neurochemical assessment within the clinical magnetic resonance imaging (MRI) setting. Commercial MRI scanner systems are equipped with the hardware and software requisite for performing proton MRS sequences. Proton MRS detects signals from the hydrogen atoms of several intracellular neurochemicals with sufficient concentration (e.g., 1–10 millimolar (mM) and parameters for MRS visibility in the brain. The neurochemicals appearing on a proton MRS spectrum provide information about neuronal health, cellular membrane replication, glial cell involvement, cellular energetics, and other conditions such as anerobic glycolysis (see Table 1). These neurochemicals are commonly called metabolites, though not all are substances produced during metabolism. The key neurochemicals typically observed include N-acetyl aspartate (NAA), creatine and phosphocreatine (Cr), and choline-containing compounds (Cho). Newer proton MRS approaches, referred to as spectral editing, yield low concentration neurochemical signals, such as γ-aminobutyric acid (GABA) and glutathione (GSH), which appear in regions of the proton spectrum with more concentration dominant signals. These neurochemical assessments can yield specificity to the clinical MRI setting when addressing questions pertaining to brain health and metabolism while characterizing disease and injury, evaluating treatment response and prognosticating outcome. Proton MRS techniques can be useful in narrowing the diagnostic differential and capturing time-sensitive information for the infant, toddler, child and adolescent.

Table 1.

Neurochemical and Metabolite Features Observed Using Proton MRS

Neurochemical and Metabolite	Abbreviation	Key Resonance (ppm)	Primary Role or Origin
Lipids (CH3)	Lip09	0.9	Membrane Metabolism
β-hydroxybutyrate	β-OHB	1.2	Ketogenic Metabolite
Lipids (CH2)n	Lip13	1.3	Membrane Metabolism
Propylene Glycol*	PG	1.1	Exogenous Compound
Lactate	Lac	1.3	Anerobic Glycolysis
Alanine	Ala	1.4	Meningioma, Abscess, Anerobic Glycolysis
Acetate**	Ac	1.9	Abscess Metabolite
γ-Aminobutryric Acid	GABA	1.9, 2.3, 3.0	Inhibitory Neurotransmitter
N-Acetyl Aspartate**	NAA	2.01	Marker for Integrity of Neurons, Axons
Glutamate	Glu	2.1, 2.3, 3.7	Excitatory Neurotransmitter
Glutamine	Gln	2.1, 2.4, 3.8	Precursor for Neurotransmitters
Glutathione	GSH	2.1, 2.5, 2.9, 3.8	Endogenous Antioxidant
Acetone	Ace	2.22	Ketogenic Metabolite
Acetoacetate	AcAc	2.27, 3.43	Ketogenic Metabolite
Pyruvate***	Pyr	2.35	Abscess Metabolite
Succinate***	Suc	2.39	Abscess Metabolite
Aspartate	Asp	2.6, 2.8, 3.9	Excitatory Neurotransmitter
Citrate	Cit	2.6	Cellular Energetics
Creatine and Phosphocreatine	Cr	3.0, 3.9	Cellular Energetics
Methylsulfonylmethane*	MSM	3.1	Exogenous Compound
Choline**	Cho	3.2	Membrane Metabolism
Scyllo-inositol	sI	3.3	Isomer of Inositol
Taurine**	Tau	3.4	Inhibitory Neurotransmitter, Osmolyte
Glucose**	Glu	3.4, 3.8	Cellular Energetics
Myo-inositol	mI	3.5	Glial Marker, Osmolyte Marker
Glycine	Gly	3.5	Inhibitory Neurotransmitter
Ascorbate	Asc	3.7	Vitamin C; Essential Nutrient; Antioxidant
Mannitol*	--	3.8	Exogenous Compound
Guanidinoacetate	Gua	3.8	Creatine Synthesis Precursor
2-hydroxyglutarate***	2HG	4.0	Oncometabolite found in isocitrate dehydrogenase (IDH) glioma

This listing provides some of the detected neurochemicals and metabolites identified with pediatric brain magnetic resonance spectroscopy listing their primary signal resonances (chemical shift) assignments. The endogenous neurochemicals and metabolites appear at normative levels in healthy conditions; however, abnormal (increased or decreased) levels represent pathologic conditions.

^*Some of the noted neurochemicals and metabolites do not occur naturally but are either exogenous compounds often administered as medications or solvents for medications.

^**Primary resonance assignment observed is listed with secondary resonance assignments omitted for clarity.

^***Pathologic appearing metabolite found in tumor or abscess

Formulating specific medical and biochemical questions for proton MRS to clinically tackle will optimize its utilization in clinical practice. Technical factors as well as some general considerations relevant primarily for pediatric populations guide these questions. This paper provides a review of key proton MRS topics relevant for defining these questions. We discuss magnetic field strength, pediatric sized head coils, water suppression techniques, localization pulse sequences, post-processing methods, analysis, and interpretation. These elements all require special consideration, particularly for the immature brain. We introduce the fundamentals of spectral editing. Finally, we present Illustrative examples employing proton MRS in clinical practice to synthesize these concepts into practical application.

2. General Considerations for Practicing Pediatric Proton MRS

2.1. Questions Proton MRS Can Address

It is important to define specific questions for proton MRS can address within a clinical MRI examination. An initial starting point may simply be raising the question of whether there is evidence of lactate in the spectrum to support diagnosing an anerobic glycolytic process associated with either a tumor, infection or a suspected metabolic disease. Evidence for neuronal and axonal injury from various processes is supported by decreases in the concentration of NAA. Changes in Cho reflect aspects of membrane chemistry with cellular proliferation producing Cho increases, which supports a tumor or destructive processes occurring with demyelination. Another key resonance observed on short echo proton MRS, myo-inositol (mI), can demonstrate increases with either glial proliferation in tumors or gliosis following various insults including demyelination. In contrast, mI decreases when accompanying osmolytic processes. Recognizing individual changes are helpful, however, combining the patterns of increase and decrease signal intensities across metabolites in the composite spectrum becomes an important skill in maximizing the utility of proton MRS in the pediatric clinical setting.

Though often challenging, neuroradiologists and clinicians tend to first implement proton MRS to discriminate tumors from other neoplasms and conditions such as dysplasia, demyelination, and infection. While this is among the most impactful application of proton MRS, it can be technically difficult due various artifact sources in tumors (e.g., necrosis, increased vascularity, hemorrhage, blood products). When successfully acquired, interpretation and addressing the diagnostic differential requires examination of the composite spectra for the presence and absence of standard metabolites, and their quantification and deviation from normative values. Also, additional metabolites not routinely observed may appear in the spectra. Thus, to recognize a pathologic spectrum, it is first important to recognize a normative spectrum. While a simple prospect, it becomes rather complicated due to a variety of factors.

2.2. Practical Considerations for Pediatric Brain Proton MRS and First Steps in Practice

A pediatric brain spectrum depends on the age of the patient, the location sampled, the tissue composition of that location, and the proton MRS techniques employed (Figures 1 and 2). In pediatrics, the typically developing brain follows a specific time course with changes in water and metabolite concentrations associated with maturation and myelination. Thus, age influences the appearance of the spectrum. The neuroradiologist detection of delays in myelination obtained from the MRI is important information for correlating and interpreting changes within a spectrum. The expected spectral time course follows that mI and Cho begin relatively high at birth, with mI decreasing over the first 4 weeks of postnatal life. Cho shows steep decreases over the first 6 months with slower rates until 3 years of postnatal life. NAA and Cr begin relatively high at birth with steep increases for the first 6 months, again leveling by 3 years of postnatal life.

Figure 1. Infant Age and Basal Ganglia Voxel Placement.

A 24-day old female with a seizure history but unremarkable imaging examination was acquired on a 1.5 T scanner. A) The standard proton MRS protocol samples within the caudate, internal capsule and putamen of the left hemisphere with the voxel location defined on an axial T2-weighted image at the level of the basal ganglia with the white box. B) The basal ganglia short echo proton MRS (35 ms) spectrum was obtained with a repetition time of 2000 ms demonstrated NAA/Cr ratio of 1.28, Cho/Cr ratio of 1.11 and mI/Cr ratio of 0.84 for the 3.9 cc (13 × 15 × 20 mm³) voxel volume. Chemical shift locations for NAA @ 2.0 ppm, Cr @ 3.0 ppm, Cho @ 3.2 ppm and mI @ 3.5 ppm are marked. Abbreviations: NAA-N-acetyl aspartate, Cr-creatine and phosphocreatine, Cho-cholines, mI-myoinositol, ms-milliseconds, cc-cubic centimeters, mm-millimeters, ppm-parts-per-million, T-tesla

Figure 2. Age Effects and Difference of Appearance between Tissue Type.

A male child presenting with a history of developmental delay evaluated at age 6 months (A-D) and 23 months (E-F) with unremarkable imaging examinations acquired on a 1.5 T scanner. A) The standard proton MRS protocol samples within the caudate, internal capsule and putamen of the left hemisphere with the voxel location defined on an axial T2-weighted image at the level of the basal ganglia with the box. B) The basal ganglia short echo proton MRS (35 ms) spectrum was obtained with a repetition time of 2000 ms demonstrated NAA/Cr ratio of 1.27, Cho/Cr ratio of 0.87 and mI/Cr ratio of 0.50 for an 8 cc voxel volume. C) The left hemisphere parietal white matter acquisition voxel location is optimized to maximize the amount of white matter and is defined on an axial T2-weighted image is noted with the box. D) The white matter short echo proton MRS (35 ms) spectrum was obtained on a 1.5 T scanner with a repetition time of 2000 ms demonstrated NAA/Cr ratio of 1.48, Cho/Cr ratio of 1.09 and mI/Cr ratio of 0.70 for an 8 cc voxel volume. Choline assignment is noted in red referring to the relative higher signal areas in white matter compared with gray matter. The follow-up visit at 23 months with (E) corresponding T2-weighted image and (F) short echo proton MRS (35 ms) spectrum was obtained with a repetition time of 2000 ms demonstrated NAA/Cr ratio of 1.81, Cho/Cr ratio of 1.35 and mI/Cr ratio of 0.68 for an 8 cc voxel volume. Abbreviations: NAA-N-acetyl aspartate, Cr-creatine and phosphocreatine, Cho-cholines, mI-myoinositol, ms-milliseconds, cc-cubic centimeters, mm-millimeters, T-tesla

The region of interest, referred to as the volume element (“voxel”) sampled by proton MRS, inevitably has a mixture of gray matter, white matter and cerebrospinal fluid (CSF), which is important to recognize as gray and white matter intrinsically have different metabolite concentrations. Thus, the initial spectra acquired when developing a clinical proton MRS service should be obtained in normal appearing tissue within a region with either pure white matter or pure gray matter. Sampling relatively pure voxels can visually illustrate the differences (Figure 2, e.g., Cho differences). This will also provide the neuroradiologist with a qualitative sense for what the tissue background spectrum should look like especially when sampling a focal lesion. However, to clinically minimize the mixing effect of various gray and white matter tissue mixtures on quantification and interpretation, defining standard locations with fixed voxel sizes and anatomical landmarks will assist in recognition of changes in metabolite concentrations when routinely employed in a clinical proton MRS service. Neuroradiologists should first delineate specific proton MRS protocols for common indications (e.g., suspected metabolic disease, developmental delay). Standard locations for a pediatric population are reported as the basal ganglia, frontal or parietal white matter, and parasagittal cortex either in the parietal or occipital lobe. In children 3 years and older, the typical single voxel dimension is 2 cm per side of the cubic voxel for a total acquisition volume of 8 cubic centimeter (cc). For infants and toddlers, this voxel volume can be too large, so reductions to produce a total 3–4 cc volume are acceptable (Figure 1). While voxel dimensions of 1 cm per cubic side are technically achievable, this selection rarely generates high quality spectra in routine clinical practice. Moving away from a cubic-shaped volume towards a rectangular prism is a way to keep the voxel volume higher with one side approaching 2 cm or more. The voxel dimensions can vary with different lengths to optimize sampling of the pathology or the structure. Ideally, the total volume should remain consistent for relative quantification across patients and over time evaluating the same patient. The specific methods for proton MRS acquisition discussed later in this article need to be consistently applied in a clinical practice to enhance recognition of pathologic metabolite changes. To fully appreciate changes, normative databases based on age and location need to either be developed at a clinical site or shared from another site with replication of voxel locations and acquisition conditions including scanner vendor and model, field strength, water suppression technique, localization technique, repetition time and echo time among the key parameters (Figure 3).

Figure 3. Example of Site-Dependent Normative Guidelines.

Vendor (GE-General Electric), field strength (1.5 Tesla), voxel acquisition parameters (echo time 35 milliseconds) and location (BG-basal ganglia) dependent normative values for semi-quantitative metabolite ratios shown by age summaries in table and in graphical form with NAA/Cr, Cho/Cr and mI/Cr normative curves from birth to 3 years of age. Abbreviations: STD DEV-standard deviation, CI-confidence interval, NAA-N-acetyl aspartate, Cr-creatine and phosphocreatine, Cho-cholines, mI-myoinositol, ms-milliseconds

As with MRI, sampling different echo times affords contrast by exploiting metabolite relaxation properties. Technically, the most important feature pertaining to the appearance of metabolite resonances is the sampled echo time. For simplicity, we can define standard proton MRS outputs as short echo, intermediate echo and long echo. These designates refer to how much time is allowed to pass upon exciting the metabolites within the voxel and detecting the signals with short echo referring to 25–35 milliseconds (ms), intermediate with 135–144 ms and long having 270–288 ms.

2.3. Exogeneous Agents

As with any imaging modality, the child must remain motionless for duration of the proton MRS sequences. For the infant, approaches to enhance natural sleep during the session (feed and swaddle) are often successful. However, toddler to preschool aged children (9 months to 5 years) generally require anesthesia or sedation to complete the examination. Children of all ages with cognitive impairments or conditions prohibiting them from holding still will also require sedation or anesthesia.

Solutions associated with anesthetics or sedation, along with medications, diets and dietary supplements may introduce an atypical resonance to the spectrum when presenting with relatively high concentrations (>1 mM). These exogenous agents may interfere with accurately detecting key metabolites such as lactate whose methyl group resonates at 1.35 parts per million (ppm; chemical shift unit for frequence on the x-axis of a spectrum). For instance, the usage of Ringer’s lactate for anesthesia may artificially produce an elevated lactate signal that may be suggestive of a metabolic disease. Also, propylene glycol is used as a solvent in barbiturate medications, however, it metabolizes to lactate, which interferes with the assessment of lactate concentrations occurring from seizure activity alone (Figure 4). Children treated with the ketogenic diet may demonstrate resonances associated with β-hydroxybutyrate at 1.2 ppm, which is close enough to interfere with lactate detection (Figure 5)¹. Additional ketones observed may include acetone with a singlet resonance at 2.2 ppm and acetoacetate with two singlet resonances at 2.27 and 3.45 ppm. These may suggest an elevation of glutamate (Glu) and/or glutamine (Gln). Methylsulfonylmethane, a dietary supplement, can be detected at 3.1 ppm when brain concentrations are within the millimolar range (Figure 6)². Mannitol is used as a therapeutic for treating elevated intracranial pressures, swelling, and as a diuretic. When brain mannitol concentrations exceed 2 mM, a distinct resonance appears at 3.8 ppm from this polyol that persists in signal intensity when sampling from short (30–35 milliseconds) to long echo (270–288 ms) (Figure 7). When atypical resonances appear on a spectrum, a review of the medication record is important for confirmation, though the initial tendency is to consider it a novel feature of a disease.

Figure 4. Exogenous Propylene Glycol.

A four-month-old male evaluated for metabolic disease with imaging consistent for Leigh syndrome and mitochondrial encephalopathy. All proton spectra were acquired within the basal ganglia and display propylene glycol (PG) as a solvent from the treatment with anticonvulsant. A) Axial T2 weighted image. B) Axial Inversion Recovery image. C) Short echo (35 milliseconds (ms) proton spectrum displays a composite of lipids, lactate and PG. PG is simplified with C) and the doublet resonance (arrow) at 1.1 ppm on the D) intermediate echo (144 ms) proton spectrum which appears with lactate (1.3 ppm) as inverted, but both PG and lactate resume positive signals on the E) long echo (288 ms) proton spectrum.

Abbreviations: NAA-N-acetyl aspartate, Cr-creatine and phosphocreatine, Cho-cholines, mI-myoinositol, PG-propylene glycol.

Figure 5. Ketogenic Metabolites Revealed with Multi-echo MRS.

A 5-month-old male with Ohtahara syndrome treated with a ketogenic diet. A) T2-weighted imaging was unremarkable. B) The short echo (35 milliseconds (ms)) proton spectrum acquired within the basal ganglia is displayed from LCModel analysis in light gray with the spectral fitting plotted in red. Originally, the standard fitting for several metabolites was inadequate. Additional parameters were added to the basis set to account for ketogenic metabolites, noted as beta-hydroxybutyrate (β-OHB) at 1.2 ppm, acetone (Ace) at 2.2 ppm and acetoacetonate (AcAc) at 2.27 and 3.45 ppm. C) The intermediate echo (TE 144 milliseconds) spectrum with the Ace and AcAc metabolites marked.

Figure 6. Methylsulfonylmethane Detected with Multi-echo MRS.

An example of spectra displaying Methylsulfonylmethane (MSM) was given as a dietary supplement in a 5-year-old male with autism spectrum disorder. A) Voxel location within parietal white matter (white box) for all spectra shown on an axial fluid attenuated inversion recovery (FLAIR) image. B) Short echo (35 milliseconds (ms)) proton spectrum shows a resonance for MSM at 3.1 ppm (yellow arrow), located between the creatine (Cr) and choline (Cho) resonances. C) This resonance remains present on long echo (288 ms).

Abbreviations: NAA-N-acetyl aspartate, Cr-creatine and phosphocreatine, Cho-cholines, mI-myoinositol and MSM-methylsulfonylmethane

Figure 7. Mannitol Detected with Multi-echo MRS.

A four-year-old male with a myelodysplastic syndrome treated with mannitol to decrease intracranial pressure. A) Voxel location for all spectra shown on an axial FLAIR image within lesion sampling posterior cortex. B) The short echo (35 milliseconds (ms)) proton spectrum demonstrates increased signal at 3.8 ppm for mannitol. There is also increased signal at 1.3 ppm for lactate. C) The resonances for mannitol and lactate persist on the long echo (288 ms) proton spectrum.

2.4. Magnetic Resonance Hardware

The choice of magnetic field strength is usually between 1.5 Tesla (T) versus 3T as these MRI systems are the ones most commonly available in clinical imaging settings. While the issue of scanner field strength selection frequently depends on scheduling availability, there are some factors to consider for using the appropriate field strength for proton MRS. Higher field strengths influence spectra by changing the SNR, chemical shift dispersion, relaxation times, main magnetic field (B₀) homogeneity, excitation radiofrequency (RF) field (B₁) homogeneity, and specific absorption rate (SAR).^3–5 The increased field strength at 3T improves spectral resolution by enhancing chemical shift dispersion a term that relates to the distance between resonances (Figure 8). Improving our ability to note when one resonance begins, and ends can improve metabolite quantification. Higher field strength enables higher spectral quality and improved signal to noise ratio (SNR), but only if higher order shimming is conducted.⁵ Increased chemical shift dispersion afforded from 3T field strength is usually a positive for proton MRS, especially for spectral editing approaches to determine GABA and GSH concentrations.^6–14 In contrast, the reduced chemical shift dispersion afforded from 1.5T makes the detection of the spectroscopic glial marker, myo-inositol (mI), easier. At 1.5T, the observed signal for mI at 3.5 ppm demonstrates higher signal intensity with four protons contributing to the single appearing resonance. In contrast, at 3T, the four protons are split in half appearing with two resonances. From a technical perspective, higher field strength benefits require understanding the factors which are altered in proportion to field strength.

Figure 8. Chemical shift dispersion.

is illustrated using a phantom solution with neurochemicals typically observed on proton MRS. The top inserted spectrum was acquired at 1.5 Tesla, while the bottom spectrum was acquired at 3 Tesla using similar techniques. Both spectra are displayed using a frequency scale on the x-axis instead of the standard parts per million (ppm) with water (not shown) set at 0 Hz. The 300 Hz mark is denoted with dashes. The 3 Tesla spectrum allows the same metabolite resonances to be spread over a larger range.

Higher field strength also requires RF pulses within localization sequences to be modified to overcome metabolite signal misregistration and signal loss. At higher field strength there are shorter T2 and longer T1 relaxation times for the signals combined with increased field inhomogeneities arising from the static magnetic field and the transmit field created by RF pulses.⁵ As these RF pulses deposit energy, higher SAR is produced with increases with the square of the magnetic field strength. Young children are particularly sensitive to RF-induced heating from pulses given their relatively immature systems for controlling body temperature, especially premature neonates and infants.¹⁵ SAR can be minimized for most clinical MRS protocols as there are only a few RF pulses employed over relatively long repetition times (e.g., 1500, 2000 or 3000 milliseconds (ms)). However, the introduction of newer MRS localization techniques with multiple pulses, require temperature monitoring, employing multichannel transmit phased-array coils versus body coils, and maintaining cool scanner rooms to help mitigate the effects of SAR, especially for the swaddled infant.

Commercial head coils are typically designed for adults, including head arrays with 32 elements and quadrature knee coils. Usage of these coils for proton MRS is not optimal for infants, and toddlers.¹⁶ Positioning pediatric heads in adult coils can adversely impact SNR due to the dielectric properties that influence coil loading. The design and development of pediatric-specific head coils would be of great benefit for clinical and research neurological MRI/MRS.

3. Technical Fundamentals for Practicing Proton MRS

3.1. Water Suppression

For pediatric proton MRS, the water suppression approaches do not differ from those employed in adults. However, it is important to consider the effects of maturation and myelination. The greatest degree of myelination occurs in the first two years of life. Water populations in the brain include those within the myelin sheath, possessing short T1 and T2 relaxation times, and those of free water not bound to macromolecules, having longer T1 and T2 relaxation times arising from water outside the myelin sheath in interstitial and intraaxonal compartments.¹⁷ Determining the percentage increase of water for the “wet” brain of the infant relative to the toddler, child or adult is a difficult endeavor. The immature brain may have very low concentrations of select metabolites (especially N-acetyl-aspartate and creatine), thus, optimizing water suppression via the suppression sequence while minimizing inhomogeneity of the static and transmit fields, is essential for obtaining diagnostic quality MR spectra in the infant.¹⁸

The chemical shift selective (CHESS) saturation technique is implemented on commercial MRI scanners.¹⁹ CHESS typically employees three, nominally 90 degree, Gaussian-shaped excitation RF pulses, which are attractive due to their relatively low R value (product of bandwidth and pulse length), and well-defined frequency selectivity.³ The CHESS module with RF pulses and dephasing gradients produce minimal effects on measured metabolite signals distant from the water signal.²⁰ CHESS can be applied before the localization sequence and repeated multiple times to minimize imperfect suppression that arises from T1 relaxation effects, inhomogeneities in the RF magnetic field and the spread of water frequencies. Another water suppression method, variable pulse powers and optimized relaxation delays (VAPOR) demonstrates insensitivity to B₁-inhomogeneity and T1 relaxation with seven (or eight) frequency-selective RF pulses interspersed with optimized T1 recovery delays and providing a range of nutation angles.²¹ Vendors have additional techniques available to implement for water suppression.

3.2. Outer Volume Suppression

Conventional proton MRS localization techniques with voxel selection and placement can avoid in three dimensions the sources of lipid signals, such as the scalp, especially in children. However, we must be sure that the detected lipid signal reflects the pathology and not an artifactual source from outside of the voxel. For example, assessing lesion lipid concentrations is necessary in grading pediatric neoplasms. Outer Volume Suppression (OVS) prevents unwanted signals, primarily lipids, from contaminating the proton MRS. Signals from lipids can distort spectral baselines and interfere with quantification. Clinical MR systems provide several options for implementing saturation pulse modules ranging from single saturation bands to a scheme often used for spectroscopic imaging with a minimum of eight OVS slices positioned in an elliptical shape encompassing the circumference of the skull in the axial plane. These approaches basically excite the lipid rich regions, then dephase the signal.²² OVS is applied prior to voxel localization. The success of OVS depends on accounting for T1 relaxation and RF magnetic field inhomogeneity. The application of the OVS slice order and timing impacts the completeness of the saturation. For infants, the addition of OVS may produce increases in SAR, especially with spectroscopic imaging, so a balance between diagnostic quality and OVS implementation must be considered.

3.3. Shimming

Shimming is an essential element of proton MRS as it optimizes the homogeneity of the magnetic field around the brain. For proton MRS, water suppression, SNR, and metabolite resonance line shapes depend upon the level of homogeneity created by shimming. A poorly shimmed voxel will have incomplete water suppression and lower SNR. If homogeneity is poor, this will adversely impact defining where signal from one metabolite begins and ends and inherently quantification of the signals. Within the static magnetic field, interfaces between brain tissue, air, eyes and bone within the human head distort the magnetic field and produce susceptibility artifacts. Due to the relatively spherical shape of the skull and the lack of developed bony and air-filled structures such as the frontal sinus and nose, pediatric brain homogeneity is typically better than that of adults. However, reduced voxel sizes (<1.5 cm³) prescribed in the frontal lobe can be problematic in terms of SNR and the placement proximity to the orbits produces susceptibility artifacts within the spectrum. Iron deposition within the globus pallidus does not occur until later childhood, thereby, allowing MRS sampling within the basal ganglia without metabolite and water signal broadening as observed in adults.

3.4. Localization

A combination of slice-selective RF pulses applied with gradients are referred to as a localization method. Different approaches are used to isolate the metabolite signals in a selected region of the brain. MRS localization of a single voxel is commonly referred to as single-voxel spectroscopy (SVS). Multiple voxels simultaneously localized are referred to as multi-voxel spectroscopy or magnetic resonance spectroscopic imaging (MRSI). Because MRS is used for quantification of metabolite signals that are 1–10 mM in concentration, an optimized localization sequence providing good SNR and linewidth is essential for optimal post-processing and interpretation.

3.4.1. Single Voxel Spectroscopy

SVS acquires metabolite signals from the intersection of three orthogonal spatially selective slices, producing a voxel with spatial resolution equal to the thickness of all three slices. For most commercial scanner systems, a user-friendly software interface for positioning a three-dimensional cubic voxel on anatomical images guides the acquisition site of the vendor-tailored SVS sequences. Some commonly used sequences are described below.

3.4.2. Localization with STEAM

STimulated Echo Acquisition Mode (STEAM) uses three 90-degree pulses to excite three orthogonal slices and localize the signal for the spectrum to a cubic region of interest as prescribed by the voxel placement software. This sequence allows for very short TE acquisitions to acquire signal from metabolites with short relaxation times, however with relatively low SNR compared to other methods. Increasing the number of acquisitions can increase the signal, however this will increase acquisition time, which may nullify the potential benefits for improved precise localization when employed for fetal MRS or in patients with subtle movement. Historically, the 90-degree pulses in this localization sequence demanded less power from the RF amplifier. For modern MR systems, STEAM is typically used for higher field (e.g., 7T) MRS acquisitions due to SAR considerations and demand for precise localization.

3.4.3. Localization with PRESS

The Point-RESolved Spectroscopy (PRESS) localization method applies one 90-degree RF pulse followed by two 180-degree RF pulses to produce higher SNR than STEAM.²³ Due to the longer frequency-selective RF pulses, PRESS cannot achieve as short echo times as those acquired using STEAM, but the increase in metabolite signal benefits processing. However, as a spin-echo sequence with two 180-degree RF pulses, PRESS suffers from chemical shift displacement artifact (CSDA), which impacts the composition of tissue sampled with a contribution from outside of the voxel and thus, the interpretation of results.

3.4.4. Localization with LASER and Semi-LASER

The CSDA from PRESS localization can be diminished using adiabatic pulses. Adiabatic pulses are a series of shorter RF pulses with larger bandwidths that provide essentially the same localization effect as a single pulse.²⁴ LASER (Localization by Adiabatic Selective Refocusing) localization replaces the three pulses in PRESS with six adiabatic pulses, however, with the adverse effect of longer acquisition times. To minimize scan time, another sequence, known as semi-LASER, only replaces the two 180 degree pulses with four adiabatic pulses to provide the advantages of LASER, yet without as large an impact on acquisition time.²⁵ Semi-LASER is more precise with localization and robust to patient movement.²³ Due to these advantages, semi-LASER has been recommended by an MRS consensus group for localization in all research MRS studies.²³ Commercial MRS vendors offer semi-LASER options for localization. However, SAR increases with semi-LASER may not be appropriate for some pediatric populations and may lead to element failure in some head coils with short echo times.

3.4.5. Improving SNR when Employing SVS

While the selection of the localization sequence can influence SNR, increasing the voxel size effectively increases the amount of signal obtained during the acquisition. Also, collecting more spectral averages allows for more information, which can improve SNR to allow for more reliable separation of metabolite signal from the noise, but with increased acquisition time.

3.5. Multi-Voxel Spectroscopy/MRSI

If a clinician is interested in multiple proton MRS regions of interest in the brain, SVS can be sequentially acquired in several locations at the expense of increasing overall examination time. Multivoxel spectroscopy or MRSI, provides a localization method to simultaneously acquire metabolite signals from many regions in a single acquisition. Effectively, MRSI is the combination of a SVS sequence with phase-encoding. This method is more challenging than SVS because shimming will need to address magnetic field inhomogeneity across the entire region sampled. MRSI may also have intervoxel contamination due to imperfect RF excitation. Additionally, traditional MRSI has longer acquisition times than SVS with a larger amount of data for analysis and interpretation. However, fast MRSI techniques have been adapted to incorporate echo-planar, spiral, turbo and parallel imaging techniques and reduce acquisition time.^26–28

3.6. Spectrum Postprocessing

Given the needs for immediate post-processing in a clinical environment, metabolite signal ratios can be generated at the scanner for both SVS and MRSI. This is referred to as semi-quantitative analysis. Clinically acquired proton spectra can be post-processed with vendor-supplied scanner-installed software by the MR technologist with appropriate training. The software packages range from fully automated with no selection of parameters or semi-automated requiring an operator to select a post-processing protocol with parameters that mathematically treat the raw spectra and are readily modified by the technologist, physicist or neuroradiologist based upon their preferences. These software packages are suitable for an individual patient examination producing a basic integration of the signal area for each metabolite resonance. These signal areas can be referenced to one another to form ratios, which are often reported by the neuroradiologist within their overall interpretation of an MRI examination for a patient. Typically, the Cr signal area is regarded as the reference metabolite, as both creatine and phosphocreatine are measured as a single resonance on proton MRS at typical MR field strengths of 1.5 and 3T. However, many pathologies (neoplasms, metabolic diseases, hypoxia and ischemia) can impact the composite Cr signal and concentration. Yet, from among the three observed metabolites (NAA, Cr and Cho), Cr is the most constant for many conditions due to the measured pool combining creatine and phosphocreatine. The generation of metabolite concentrations is the more accurate post-processing approach. However, it requires a post-processing pipeline to send the data (raw or DICOM) through software to quantify the metabolite resonances, making assumptions about the referenced water concentration, corrections for partial volume effects from tissue/CSF mixtures, and accounting for signal characteristics due to water and metabolite relaxation properties.

When more rigorous determinations of metabolite concentration are desired in either the clinical or research setting, there are several widely recognized software packages that read in raw data from the major clinical MRI vendors (LCModel²⁹, jMRUI³⁰,Tarquin^31–34, Osprey³⁵). These software packages allow concentration estimates to be generated but often require extra off-line processing at a separate workstation. Within an institution, these methods can be reliably used for clinical purposes; however, further corrections (which may not be included in the software packages) are necessary, such as accounting for water and metabolite signal relaxation rates, tissue and CSF composition within the voxel, and the internal water concentration to convert the institutional unit to a measure of concentration (e.g., mM) useful for comparisons across investigators using various scanner systems at multiple institutions. Details regarding these and additional corrections for proton MRS quantification have been reported.^36–40

3.7. Interpretation

It is very important to recognize that for a pediatric clinical MRS service employing ratios, the actual values obtained are not only age, sampling location, scanner type (as a vendor may have variation within their models or software platform), and field strength specific as mentioned earlier (Figure 3), but also factors associated with the technique including localization methodology, water suppression technique, acquisition with or without OVS, echo time, repetition time, number of averages and voxel volume. Consistency in approach and voxel placement can alleviate technical variations allowing for the desired developmental and pathological features to be appreciated in the interpretation.

4. Fundamentals for Spectral Editing Approaches

4.1. Spectral Editing Concepts

Standard proton magnetic resonance spectroscopy is a robust tool for measuring high-concentration metabolites with strong signals, such as NAA, Cr, Cho, and mI. In vivo measurements of low-concentration metabolites, such as the inhibitory neurotransmitter GABA and the antioxidant GSH, are approximately one order of magnitude smaller than those of high-concentration metabolites. In vivo, proton MRS spectra suffer from the incomplete resolution of metabolite signals due to limited chemical dispersion along the chemical shift dimension (x-axis), broad signal linewidth, and low signal intensities. Consequently, the signals of low-concentration metabolites significantly overlap with larger signals. For instance, GABA exhibits low-intensity signals across the MRS spectrum, with the 3-ppm signal overlapping with the Cr signal, the 2.28-ppm signal overlapping with Glu, and the 1.9-ppm signal overlapping with the NAA signal. Such overlaps hinder direct measurements and result in quantification with low certainty. Spectral editing (also known as J-difference editing) addresses this challenge by utilizing the intrinsic scalar coupling properties of metabolites to selectively edit the chemicals of interest and eliminate the overlapping signals of more concentrated metabolites.⁴¹

Spectral editing comprises two sub-experiments. In the first sub-experiment, a frequency-selective, narrow-band editing pulse is applied to the target metabolite to refocus the coupling evolution of the signal of interest (edit ON step); in the second sub-experiment (edit OFF step), no editing pulse is applied to allow free coupling evolution. Subtracting the two sub-experiments results in the difference (DIFF=ON–OFF) spectrum with the elimination of the overlapping signal since they are unperturbed in both sub-experiments and contain contributions only from the signals affected by the editing pulse. The most widely used spectral editing method is Mescher-Garwood point resolved spectroscopy (MEGA-PRESS)⁴², primarily due to the relative ease of implementation of MEGA editing pulses in the PRESS⁴³ sequence. PRESS is the localization sequence and largely independent from MEGA editing. Therefore, MEGA pulses can be incorporated into other spin-echo-based localization sequences, such as SPECIAL^44–46, semi-LASER^47,48, and LASER^49,50. MEGA-PRESS provides experimental flexibility to assess different metabolites in the brain, including measurements in various brain regions and adjusting acquisition parameters, such as echo time, to optimize for a particular target. Some common metabolites edited using MEGA-PRESS are briefly reviewed below.

4.2. GABA Editing

GABA is arguably the most edited metabolite using MEGA-PRESS⁵¹ due to its crucial roles in brain development and function, including cortical myelination, synaptic pruning^52,53, learning⁵⁴, and spatial mapping⁵². Additionally, GABA is implicated in various brain health and disease conditions, including healthy aging^55,56, autism spectrum disorder (ASD)^57–59, and epilepsy⁶⁰. In the MEGA-PRESS experiment for GABA, the edit ON step involves applying a frequency-selective pulse at 1.9 ppm. The edit OFF step applies the editing pulse at 7.5 ppm (instead of no pulse), which is far from any signal of interest to avoid interference with GABA’s free signal evolution. Subtracting the two editing steps results in a GABA-edited (or difference) spectrum containing the GABA-edited signal at 3 ppm without overlapping with the Cr signal. However, the 1.9-ppm editing pulse also impacts the macromolecule (MM) at 1.7 ppm, which couples with the 3-ppm MM signal. Consequently, this approach produces a difference spectrum where the GABA and MM signals overlap at 3 ppm, commonly referred to as GABA+ in the literature. To reduce MM contamination, the editing pulse is applied at 1.5 ppm during the edit OFF step, symmetrically about 1.7 ppm, without affecting the 1.9 ppm GABA signal.⁶¹ The 1.5-ppm pulse refocuses the coupling evolution of the MM at 3 ppm in the same manner as the 1.9-ppm pulse in the edit ON step. Thus, subtracting the two steps reduces MM contamination and yields a cleaner GABA signal.⁶²

An added benefit of GABA editing is the co-editing of the oncometabolite 2-hydroxyglutarate (2HG), which accumulates in the brains of patients with isocitrate dehydrogenase (IDH) glioma, making it a sensitive and specific biomarker for IDH mutation.^63,64 2HG has a signal at 4.02 ppm coupled to another signal at 1.9 ppm, which can also be utilized during GABA editing. Thus, applying an editing pulse at 1.9 ppm for GABA concurrently detects 2HG at 4.02 ppm without interference from overlapping Cr, Cho, Lac, and mI signals around 4 ppm.^63,64

4.3. Glutamate Editing

Glu is an excitatory neurotransmitter and plays a crucial role in brain development. Glu is essential for neurodevelopmental processes, including neuronal migration, which allows proper nervous system functioning, such as language comprehension.^65,66 Similar to NAA, Glu can indicate neuronal integrity, with alterations observed in brain conditions such as HIV⁶⁷, ASD^57–59, ALS⁶⁸, and traumatic brain injury^69,70. In a standard MRS or GABA-edited spectrum, Glu overlaps with Gln, commonly called ‘Glx.’ The 1D J-PRESS (or echo time averaging) method remains the standard for separating Glu from Gln by leveraging the coupling evolution of both chemicals across various echo times. By summing (or averaging) data acquired at different echo times, Glu is isolated from Gln⁷¹. Glu can also be co-edited and separated from Gln using the sum spectrum of the MEGA-PRESS GABA data (sum = ON at 1.9 ppm + OFF at 7.5 ppm), showing a strong association with the echo time-averaged Glu.⁷² Since the editing efficiency of Glu can vary depending on the pulse width at 1.9 ppm, further research is required to establish the precision of Glu measurements using edited MRS.

4.4. Glutathione Editing

There is a growing interest in measuring GSH due to its role in protecting the brain from harmful radicals and maintaining the redox state. GSH is the brain’s most abundant antioxidant, safeguarding cells from reactive oxygen compounds, and is considered a marker of oxidative stress. Lower levels of GSH are implicated in various scenarios indicative of oxidative stress, including COVID-19⁷³, substance use⁷⁴, and epilepsy.⁷⁵ The MEGA-PRESS technique for GSH involves similar steps, with the editing pulse in the edit ON step applied at 4.56 ppm to refocus the coupling evolution of the GSH signal at 2.95 ppm. Subtracting the edit ON and edit OFF steps produces an edited spectrum with the GSH signal at 2.95 ppm without interference from overlapping signals Cr.⁷⁶

4.5. Lactate Editing

Although lactate is considered a byproduct of anaerobic metabolism, it plays a key role in brain energy metabolism and mediating brain function in healthy and diseased states.^77,78 Lactate has a doublet signal at 1.3 ppm coupled to a quartet signal at 4.1 ppm. Lactate has been measured using MEGA-PRESS by applying the editing pulse at 4.1 ppm to detect the signal at 1.3 ppm. However, lactate at 1.3 ppm significantly overlaps with the MM signals at 1.24 and 1.43 ppm⁷⁹, rendering precise quantification difficult. Recent works on optimizing editing pulses have demonstrated lactate editing with minimal contamination.^80,81

4.6. Other Considerations for Spectral Editing

In addition to editing pulse frequencies, selecting the optimal echo time that maximizes the editing efficiencies is important. Signals with a triplet-like shape, such as GABA, are optimally edited at the echo time of 1/2J (~70 ms), whereas signals with a doublet-like shape, such as GSH and Lac, are optimally edited at the echo time of 1/J (~140 ms). In vivo experiments of GABA, echo time values can range between 68–80 ms, and the higher echo time (80 ms) is usually selected to accommodate more selective (narrower bandwidth) editing pulses for MM suppression⁶². In vivo GSH editing is generally performed at a substantially lower echo time, but reduced T₂ relaxation effects mitigate the resulting editing efficiency losses due to the lower echo time. Studies have demonstrated that GSH editing at echo time ranging from 68 to 140 ms does not substantially change GSH signal intensity⁸². Other metabolites with more complex shapes would benefit from simulated and in vivo echo time series (a range of echo times) of MEGA-PRESS experiments to determine optimal echo times based on editing efficiencies (incorporating T₂ effects), as previously done for other chemicals.^82–84

4.7. Future Directions for Spectral Editing

Spectral editing represents an important methodological advancement that effectively removes overlapping signals by selectively resolving the signal of interest from one specific metabolite at a time. Recent developments in the field of edited MRS have facilitated significant acceleration in ‘metabolite space.’ Multiplexed editing with HERMES (Hadamard Encoding and Reconstruction of Mega-Edited Spectroscopy) enables the simultaneous editing of two metabolites (e.g., GABA and GSH^85,86, N-acetyl aspartate [NAA], and N-acetyl aspartyl glutamate [NAAG]⁸⁷) or three metabolites (e.g., NAA, NAAG, and aspartate Asp⁸⁸), substantially enhancing acquisition efficiency. A more advanced approach, HERCULES (Hadamard Editing Resolves Chemicals Using Linear Estimation of Spectra), allows the editing of seven relevant metabolites (GSH, ascorbate, GABA, NAAG, aspartate, NAA, lactate) in a single 11-minute experiment.⁸⁹ These methods are also standardized across four major MRI vendors⁹⁰, enabling impactful multi-site clinical investigations of brain diseases or disorders, such as those conducted in HIV.^91,92 Further acceleration can be achieved by incorporating the HERMES-editing scheme into localizers, such as parallel reconstruction in accelerated multivoxel (PRIAM), allowing simultaneous editing of multiple metabolites in various brain locations.^86,93

In vivo accelerated editing can facilitate innovative investigation and provide potential biomarkers to monitor disease progression and treatment. The immediate impact of these accelerated approaches can enhance understanding of the pathophysiological relationship between oxidative stress (e.g., depleted levels of GSH and ascorbate), neuronal damage (e.g., alterations in NAA, NAAG, GABA, and Glu), neuroinflammation (e.g., alterations in Cho and mI), and cognitive impairment. These measurements may also guide the development of new medicines to mitigate oxidative stress and neuronal damage.

Although edited MRS represents substantial methodological progress, acquiring data with a sufficient signal-to-noise ratio requires relatively long scan times, especially for pediatric populations. Therefore, it is sensitive to patient motion⁹⁴, which may adversely affect the voxel localization accuracy, spectral quality^95,96, and bias metabolite estimations.^46,50 Coupling edited MRS with prospective motion-correction methods will make the editing scheme less sensitive to subject motion, resulting in more precise measurements.^46,50

5. Pediatric Proton MRS Applied in the Clinic

5.1. Canavan’s Disease-Rare but Unique MRS Presentation

Canavan’s disease is a rare genetic disorder that typically presents in the first 6 months of life and is characterized by a defect in aspartoacylase (ASPA), which is responsible for hydrolyzing NAA into acetate and aspartic acid. As NAA accumulates it is accompanied by vacuolization in the lower layers of the cerebral cortex and subcortical white matter with intramyelinic swelling and myelin loss⁹⁷. Proton MRS acquired in the cerebral white matter of patients with Canavan’s disease, regardless of echo time, will feature a distinctive elevation of the NAA resonance located at 2.0 ppm (Figure 10). This elevation should be qualitatively recognized and semi-quantitative analyses with ratio values to Cr should be dramatically elevated. With the appropriate acquisition technique, elevations of myo-inositol (on short echo) and lactate can also be detected in patients with Canavan’s. Plasma, CSF and urinary concentrations of NAA also support the diagnosis. Relatively small increases of NAA clinically observed on proton MRS are not suggestive of Canavan’s, especially without the noticeable changes in white matter observed on brain MRI. Other leukodystrophies or disorders should be considered in that scenario.

Figure 10. Canavan Disease.

A 12-month-old female presenting with Canavan’s disease. A) An axial T2-weighted image. Sampling within the frontal white matter, B) short echo (35 milliseconds (ms)), C) intermediate echo (144 ms) and D) long echo (288 ms) spectra demonstrate a dramatic elevation of the signal for NAA at 2.0 ppm. The mI signal also demonstrates increased signal area on the short echo proton spectrum.

Abbreviations: NAA-N-acetyl aspartate, Cr-creatine and phosphocreatine, Cho-cholines, mI-myoinositol

5.2. Creatine Deficiency Syndromes – Rare but Unique MRS Presentation

In humans, there are three recognized cerebral creatine deficiency syndromes (CCDS) linked to defects in creatine metabolism⁹⁸. Two CCDS are related to creatine synthesis and the other to transport. Creatine is synthesized in a two-step reaction occurring predominately within the liver. Creatine circulating in the blood stream is transported into cells using the creatine transporter protein (a transmembrane protein referred to as: SLC6A8). There are at least two genes for the creatine transporter, but the one most widely expressed in the human body is localized to the X chromosome (Xq28). The synthesis disorders (arginine:glycine amidino transferase (AGAT) and guanidinoacetate methyltransferase (GAMT)) are autosomal recessive. When SLC6A8 is defective, transport of creatine into the brain is impaired resulting in creatine transporter deficiency (CTD). The three CCDS present with a striking absence of creatine and phosphocreatine at 3.0 ppm on proton MRS at all echo times (Figure 11). It is important the recognize the typical decrease in signal intensity for Cr signals at intermediate and long echo times, so as not to confuse the finding with CCDS in a child with intellectual disabilities, autism spectrum disorders or speech delay. Employing short echo SVS or MRSI is the most appropriate for making the diagnosis of a CCDS. Proton MRS cannot reliably distinguish CTD from a creatine synthesis defect, so its role is a functional confirmatory test when a suspected pathogenic variant is identified on exosome screening.

Figure 11. Cerebral Creatine Deficiency Syndromes.

An 11-year-old female with Guanidinoacetate Methyltransferase (GAMT) deficiency (Figures 10 A-B-C) and a 4-year-old male with Creatine Transporter Deficiency (CTD) (Figures 10 D-E-F). A) An axial T2-weighted image demonstrating abnormal T2 signal prolongation within the globus pallidus. B) Short echo (35 milliseconds (ms)) proton spectrum was acquired within the basal ganglia. The signal area for creatine and phosphocreatine (Cr) is severely diminished with questions persisting whether this is Cr signal or some other neurochemicals such as GABA or glutathione. At 3.8, the signal area may represent the Cr precursor, guanidinoacetate (GAA), or the typical composite of glucose, glutamate and glutamine. C) Only NAA and Cho are detectable on the long echo (288 ms) proton spectrum due to low SNR. D) An axial T2 image showing the voxel placement within the basal ganglia. E) Short echo (35 milliseconds (ms)) proton spectrum was acquired within the basal ganglia. The signal area for Cr is also severely diminished. F) Improved SNR allows for a small resonance at 3.0 to be detected, suggestive of a small Cr resonance.

Abbreviations: NAA-N-acetyl aspartate, Cr-creatine and phosphocreatine, Cho-cholines, mI-myoinositol, SNR-signal-to-noise ratio.

5.3. Proton MRS Examples

Pediatric neuroradiologists encounter a variety of conditions each year, month and week. A brief survey of images and spectra that illustrate various masses (Figures 12, 13, and 14) and signal abnormalities (Figure 15) are presented to integrate their features with technical MRS aspects.

Figure 12. Medulloblastoma.

An 8-year-old male presenting with esotropia and papilledema with a right 6^th nerve palsy. A) Sagittal T1-weighted image with voxel location noted with the white box. B) Post-contrast Sagittal T1-weighed image. C) Axial T2-weighted image. D) Short echo (35 milliseconds (ms)) proton spectrum demonstrates elevated Cho, mI, taurine and lactate with reduced NAA and Cr upon comparison with normative semi-quantitative ratios. E) Long echo (288 ms) confirms the elevated Cho and lactate with reduced creatine. These findings are supportive of medulloblastoma due to a significant Cho concentration, the presence of taurine and lactate.

Abbreviations: NAA-N-acetyl aspartate, Cr-creatine and phosphocreatine, Cho-cholines, mI-myoinositol

Figure 13. Abscess.

A 13-year-old female presented with a right frontal mass. A) Axial FLAIR image and B) axial post-contrast T1-weighed image demonstrating vasogenic edema, thin rim peripheral enhancement and mass effect. C) Short echo proton spectrum sampled within the core demonstrates the absence of standard metabolites but an elevation of cytosolic amino acids, acetate and succinate. D) Long echo proton spectrum sampled within the core clarifies the cytosolic amino acids with discrete resonances for valine (0.9 ppm), lactate (1.3 ppm) and alanine (1.4 ppm). The detection of these metabolites was thought to reflect a mobile core corresponding to a relatively early presentation of the abscess.

Figure 14. Abscess.

A 15-year-old male presented with a left frontal mass. A) Axial FLAIR image and B) axial post-contrast T1-weighed image demonstrating vasogenic edema, thin rim peripheral enhancement, mass effect and destructive changes within the left frontal bone. C) Short echo proton spectrum sampling within the core demonstrates the absence of standard metabolites but only an elevation of lipids and lactate due to the chronic nature of the lesion.

Figure 15. Infection.

A 12-month-old female presenting with acute neurological deterioration following a systemic febrile illness. A) Axial T2-weighted image. B) Axial FLAIR image with voxel location noted with the white box. C) Short echo (35 milliseconds (ms)) proton spectrum demonstrates elevated glutamate and glutamine (GLX) and mI with the absence of lactate. Serologic studies reported elevated acute Epstein-Barr viral (EBV) antibody titers and EBV IGM antinuclear antigen levels.

6. Summary and Future Directions

When incorporated into clinical neuroradiology practice, proton MRS provides a distinct tool for providing metabolic and neurochemical support for a differential diagnosis. Understanding the principles behind proton MRS will enable the neuroradiologist to use this tool in an informed manner. While currently within the research setting, spectral editing approaches will reveal a greater number of dynamic metabolites that eventually can be integrated into understanding disease and injury further in the clinical setting.

Figure 9. Example of Spectral Editing Output.

A) Axial, Sagittal and Coronal T1 images illustrating the 27 cc voxel placement for a hadmard encoding and reconstruction of mega-edited Spectroscopy (HERMES) spectral editing acquisition within the supplementary motor cortex of a 17-year-old female. B) The composite output of the two edited spectra expanded in D) and E). The γ-aminobutyric acid (GABA) and glutathione (GSH) resonances are referenced to water shown in (C) with an insert to illustrate their chemical shift location co-resonates with Cr. D) Fitting of the edited spectrum for GSH. E) Fitting of the edited spectrum for GABA and glutamate and glutamine (GLX). The fits provide concentration estimates in institutional units.