Main

NMR was first demonstrated in 1946, and over the past half century has been a valuable instrument in research in physics and chemistry. The past two decades have featured increasing biomedical applications of this technique as a noninvasive means of studying anatomy, structure, and metabolism in vivo. 1H MRS provides the specific ability to assess brain biochemistry regionally in a quantitative and longitudinal way. The capacity to begin to carry out 1H MRS studies presently often requires only commercially available software modification of existing widespread clinical MRI technology and an understanding of the technology's physical basis, potential applications(1), and limitations. Application and interpretation of this technique has thus far been largely descriptive in diseases studied. The need for standardized conditions of acquisition and careful design in application to the study of human disease requires the participation of specially trained personnel.

Research in neurology has always been hampered by the need to infer from either static or remote investigative tools what was actually occurring in a highly dynamic tissue. Together with other evolving technologies (functional MRI, positron emission tomography), 1H MRS addresses these deficiencies to provide biochemical and potentially temporal insights. The biochemical data obtained is often disease-nonspecific and potentially needs to be interpreted within a clinical context. This overview is meant to provide an understanding of this emerging technology with particular reference to its future potential to further our understanding of the biochemical markers and mechanisms of highly prevalent pediatric neurologic disorders.

BASIS OF NMR

Many nuclei of potential neurobiologic interest have a property known as magnetic moment. These include the proton (1H), carbon (13C), phosphorus (31P), sodium (23Na), and nitrogen (15N) nuclei. When molecules made up of these nuclei are placed in a magnetic field, the nuclei precess around the axis of the magnetic field at a frequency proportional to the magnetic moment of the nucleus and the strength of the magnetic field. Precession is a rotation of the spin axis of the nucleus produced by a torque applied about an axis mutually perpendicular to the spin axis and the axis of the resulting motion(2). At a field strength of 1 tesla (T), the 1H nuclei precess at 42.6 MHz, which is in the rf range. A tesla is a unit of magnetic flux density or field strength. The earth's magnetic field strength is from 0.05 to 0.7 mT(2). Quantum mechanics dictates that the angular momentum or spin of these nuclei can have only certain discrete values. The quantum number of the 1H nucleus is 1/2, and when placed in a magnetic field it can exist in only two energy states. If we irradiate these nuclei at just the right frequency such that the energy imparted is equal to the separation of these two energy states, nuclei in the higher state will lose energy to fall to the lower energy state, and nuclei in the lower energy state will absorb the energy and transition to the higher energy state. This phenomenon is referred to as a resonant exchange of energy between the nuclei and the applied irradiation(2). Excitation of the nuclei occurs when the rf pulse is at the resonant frequency of the nucleus to be examined and causes these transitions to occur.

To perform the NMR experiment a magnet, a magnetic field gradient, set rf transmitter and receiver designed to transmit and receive at the resonant frequency of the nucleus(i) of interest, and a computer that can precisely control the transmission of the rf pulses and gradients, and can store and analyze the NMR signals from the receiver are all required. Some NMR systems also have a second transmitter that can transmit at the resonant frequency of another nucleus simultaneously. The simplest NMR experiment, known as a pulse-acquire experiment, is performed by placing a sample in the center of the magnetic field, exciting the sample with a short rf pulse, waiting a specified period of time, and then turning on the receiver to acquire the rf signal emitted. The excitation pulse duration is microseconds in duration, and the period that the receiver is on is less than a second. This experiment can be repeated several times over a few minutes to improve the sensitivity of the NMR experiment by signal averaging. The rf signals received are known as FIDs and several FIDs are collected during an experiment. The FID, which represents a decaying of signal with time, from a sample of GABA is shown in Figure 1A.

Figure 1
figure 1

1H NMR of GABA. (A) A collection of 128 FIDs of a 10 mM solution of GABA acquired on an 8.5-T spectrometer. The FID is a rf signal that decays over seconds. (B) The proton spectrum of GABA is the result of taking the Fourier transform of this FID after exponential multiplication with 1-Hz line broadening. The scale of the x axis is the frequency difference in Hz between the reference and the resonant frequencies of the three pairs of methylene protons of GABA. (C) The proton spectrum is the result of taking the Fourier transform of this FID after exponential multiplication with 7-Hz line broadening to simulate the line widths expected of the GABA protons in vivo. The scale of the x axis is in parts/million(ppm), which is independent of the field strength of the magnet. The numbers under each peak correspond to the number of the carbons of GABA whose chemical structure is shown at the bottom. [The x axis of the FID is time (seconds or milliseconds) and the x axes of the spectra are expressed in frequency (Hz) or a normalized parameter ppm, which is not a measure of concentration. It is the frequency difference between the peak of interest and a reference frequency divided by the reference frequency. Because the frequency difference is usually on the order of a few hundred Hz and the reference frequency in is the millions of Hz range (MHz), the value is multiplied by a million (106), hence the term ppm. It is therefore a dimensionless parameter, which permits comparison of NMR spectra obtained in spectrometers that are different field strengths. For example the center of the proton peak of the methyl group of lactate is always at 1.33 ppm, whether the spectrum is obtained on a 0.5-, 1.5-, or 12-T NMR spectroscope. The y axis of spectra is usually not shown, because it is a measure of the height of the peak in arbitrary units.]

Full size image

This FID can be converted to a series of peaks (a spectrum) by a mathematical process (Fourier transformation). Fourier transform is a mathematical procedure to separate out the frequency components of a signal from its amplitudes as a function of time. The Fourier transform is used to generate a spectrum from the FID obtained in the NMR experiment(2). Figure 1B shows the signals of the protons of GABA after Fourier transform of the FID in what is referred to as a NMR spectrum. Note that there are three groups of peaks corresponding to the protons attached to the different carbons of GABA separated by a frequency of a few hundred hertz. Each of these groups of peaks consists of smaller peaks that are separated by a few hertz. The complexity of the1 H NMR spectrum of GABA is explained by two NMR properties, chemical shift and spin-spin coupling(3). The frequency at which a particular proton produces a signal is dependent on the local chemical environment of that proton. The protons are influenced not only by nuclei to which they are directly attached, but also by nuclei that are one or two bond lengths away. This frequency effect is known as chemical shift. The NMR signal magnitude is proportional to the number of protons. For example, the three protons attached to a carbon in a methyl group produce a signal that is three times greater than the signal that arises from the single proton on the α carbon of an amino acid. Because protons intrinsically have magnetic properties, they will influence nearby protons, just as two magnets do when they are brought close to each other. If the protons are within three chemical bond lengths of each other they will cause the signal produced by one group of protons (A) to be affected by another group of protons (B). This characteristic is referred to as homonuclear spin-spin coupling if the two nuclei are the same type and heteronuclear coupling if they are different. As can be seen in Figure 1B, the protons of GABA are split into two triplets and a quintet. The quintet occurs because the protons attached to the C3 carbon are independently affected by the two protons on the C2 and C4 carbons.

Because the area of the peak is proportional to the number of protons contributing to the signal, the concentration of the compound can be derived from the intensity (area) of any of the peaks from that compound or expressed as a ratio to another peak or "standard" whose concentration is known. However, to accurately measure concentrations, the peak must be corrected for by the relaxation properties of that signal and optimally must not overlap with any other peaks in the spectrum(4). Each signal in the proton spectrum has properties of relaxation known as longitudinal and transverse relaxation(5). The longitudinal relaxation is referred to as the T1, and the repetition time (TR) in an NMR experiment affects a peak's intensity as a function of the T1 and TR. The transverse relaxation is referred to as the T2, and the total echo time (TE) in an NMR experiment affects a peak's intensity as a function of the T2 and TE. The line width of a peak at its half-height is approximately proportional to the inverse of the T2 of that nucleus. Protons attached to large macromolecules have short T2 values and are observed as broad, short peaks, whereas protons attached to small metabolites have longer T2 values and are observed as narrow, tall peaks in the proton spectrum.

The in vivo 1H MRS experiment has three additional features that make it more difficult in actual applications. First, the experiment requires localization, which is accomplished by obtaining a standard MRI and determining the coordinates of the volume from which the biochemical information is desired for the MRS study. A volume of interest of 1-12 mL is the usual volume size for in vivo studies of human brain, and typically it takes 2-10 min to obtain a single spectrum from that region. There are several methods to accomplish localization, each with inherent problems that affect the quality of the spectrum acquired(6). A more recent approach to localization is to identify a larger region of the brain on MRI and acquire 1H NMR spectra from multiple volumes simultaneously. This method is referred to as1 H MR spectroscopic imaging(7,8). The method permits extracting individual spectra from 1-2-mL contiguous volumes and creating a map or image of the distribution of a specific compound over a large region. This study typically takes 20-40 min to complete. Second, the need for water suppression in the in vivo MRS experiment arises from the fact that the concentration of protons of water molecules in a tissue is two to three orders of magnitude higher than the concentration of protons attached to metabolites. To observe the metabolite NMR signals, several methods had to be developed to suppress the signal from the water without affecting neighboring NMR signals. Third, the line width of a peak in an NMR spectrum is dependent both on the intrinsic T2 of that compound and the homogeneity of the magnetic field in the region. The line width of a peak due to its intrinsic T2 is typically less that 1 Hz, whereas the line width from field inhomogeneity may be from 5 to 10 Hz. The term T2* is used to describe the combined effect of the intrinsic T2 of the peak and the magnetic field inhomogeneity that contribute to the line width. In biologic tissues the magnetic field is often affected by several factors that degrade the homogeneity of the magnetic field. Spectrum C in Figure 1 shows the effect of this line broadening on the 1H spectrum of GABA where the fine structure of the spectrum is lost and the smaller peaks from the spin-spin coupling overlap to form one broad peak. The homogeneity can be improved by performing a process referred to as "shimming" where adjustments are made in the magnetic field to maximize the homogeneity over a volume of interest.

Most centers performing 1H MRS use a spin echo NMR experiment to obtain the spectrum from a single volume in a specific region of the brain. The majority of centers use a long total TE, greater than 100 ms, to obtain spectra because this assists in suppressing both the water signal and the signals from the protons attached to large macromolecules, such as proteins and lipids in the tissue. However, because these protons are attached to large molecules, they have very short T2 values and produce broad peaks in the spectrum, which underlie the narrower peaks arising from the smaller metabolites in the tissue(9). Proton spectra obtained with a long TE experiment permit the measurement of three to four compounds present in the brain at relatively high concentrations. As shown in Figure 2, the peaks typically observed in this experiment are from the methyl protons of N-acetyl-containing compounds which in the brain is primarily from NAA, the N-methyl protons of Cr and phosphocreatine that overlap in the proton spectrum, and the N-methyl protons of trimethylamine-containing compounds which in the brain is primarily Cho. The protons from the methyl group of Lac can be observed when it is elevated in pathologic disorders. Most proton spectroscopic imaging studies also use a long TE to obtain the spectra from multiple volumes.

Figure 2
figure 2

Long TE proton spectrum of the occipital lobe in cytochrome oxidase deficiency. The proton spectrum from a 12-mL volume of the occipital lobe of a 18-mo-old child with cytochrome oxidase deficiency on muscle histochemistry. The spectrum was obtained on a 2.1-T Bruker Avance spectrometer using a surface coil. The spectrum took approximately 5 min to acquire, and the TE time was 162 ms. The peaks on the spectrum, from right to left, are: peak from a macromolecule at ∼0.9 ppm(M), the doublet of the methyl of Lac centered at 1.33 ppm(Lac) superimposed on another broad macromolecule peak at ∼1.4 ppm (M1.4), methyl protons of N-acetyl-containing compounds (predominantly NAA) at 2.02 ppm (NA), methylene protons of the aspartyl moiety of NAA at 2.6 ppm (NA), N-methyl protons of both Cr and phosphocreatine at 3.03 ppm (Cr),N-methyl protons of trimethylamine-containing compounds(predominantly Cho) at 3.2 ppm (Cho), protons of myo-inositol at 3.5 ppm (MI), residual water protons at 4.7 ppm (H2O).

Full size image

Several methods have been developed to suppress the water signal without having to use a long TE. Proton spectra obtained with TEs of less than 20 ms permit the observation and quantitation of protons of several other metabolites present in the brain. These include glucose(10), myo-inositol, glutamate(11), glutamine(12), aspartate, GABA(13), scyllo-inositol(14), and homocarnosine, which are of high enough concentrations under normal conditions to be observable on short TE spectra. Several other endogenous compounds have been measured including phenylalanine(15) and branched chain amino acids in disorders where the concentrations are pathologically elevated. Exogenous substances such as ethanol(16) and propanediol(17) have also been observed in the brain.

One of the difficulties in interpretation and analysis of these spectra is the poor spectral resolution or extensive overlap of the peaks from these compounds including the underlying protons from macromolecules(9). Figure 3 shows a method developed to separate these signals where both a proton spectrum of all the proton signals(metabolites and macromolecules) and a T1-weighted spectrum in which the 1H peaks from macromolecules with shorter T1 values that are observed are acquired from the same volume. Subtraction of the two spectra permits accurate measurements of the signals from the metabolites alone. These signals can provide quantitative measurements of the concentrations of several metabolites in the brain when corrections for the overlap, relaxation properties, and other experimental parameters are made(15,18,19).

Figure 3
figure 3

Short TE proton spectra from occipital lobe. The proton spectra are from a 12-mL volume of the occipital lobe of a 10-mo-old boy with Wolf-Hirschorn (4p-) syndrome. The spectra were obtained on a 2.1-T Bruker Avance spectrometer using a surface coil. Each spectrum took approximately 5 min to acquire, and the TE time was 16 ms. (A, top) A T1-weighted spectrum obtained using an inversion recovery sequence chosen to obtain the proton signals from macromolecules. The macromolecule resonances are labeled with M and a subscript which indicates their chemical shift in ppm. (B, center) Spectrum containing the proton signals from both the metabolites and the macromolecules in the occipital region. (C, bottom) Calculated spectrum based on the difference between spectra A and B and contains only signals from the protons of metabolites alone. The peaks in the "metabolite" spectrum (C) are (from right to left): the small doublet of the Lac methyl group at 1.30 and 1.36 ppm(Lac), N-acetyl peak at 2.02 ppm (NA), peaks from the C3 protons of both glutamate and glutamine at 2.10 and 2.18 ppm and underlying the 2.02 peak (1) peaks mainly from the C4 protons of glutamate at 2.29 and 2.36 ppm (2), peaks predominantly from the C4 protons of glutamine at 2.36 and 2.45 ppm (3), peaks primarily from the aspartyl moiety of NAA at 2.6 and 2.65 ppm (4), peak mainly from aspartate at 2.75 ppm (5), N-methyl protons of both Cr and phosphocreatine at 3.03 ppm (Cr),N-methyl protons of trimethylamine-containing compounds(predominantly Cho) at 3.24 ppm (Cho), protons of myo-inositol at 3.5 ppm (MI). The residual water protons at ∼4.7 ppm, proton of ribose at 7.05 ppm, and proton of homocarnosine at 7.8 ppm are not shown.

Full size image

The identification and assignment of peaks in the proton spectrum can be accomplished in several ways. First, spectra can be obtained on "phantoms" containing known concentrations of pure compounds thought to be present in the in vivo spectrum using the identical experimental parameters. Many libraries have a large database of NMR spectra accumulated over the last 30 y that provide peak frequencies, intensities, and other parameters on thousands of known compounds. Second, spectra can be obtained using different TEs to observe the behavior of the peaks at different echo times. This provides information about the T2 values of the peaks and the homonuclear spin-spin coupling. Third, more sophisticated NMR experiments such as "editing" experiments(13) and homonuclear or heteronuclear two-dimensional NMR experiments(3) can be performed in vivo in some cases. Finally, obtaining and comparing in vivo spectra before, after, and during experimental protocols that are expected to change the concentration of the compound to which the peaks are being assigned can sometimes be performed(10).

PRACTICAL CONSIDERATIONS

The practicalities of 1H MRS resemble that of more conventional MRI. Immobility is essential for a good quality study, and therefore sedation according to established protocols used in imaging may be necessary for the newborn, infant, and younger child. It is essential in potentially unstable patients that vital signs be monitored while the child is in the magnet. The need for ongoing monitoring and the paraphernalia associated with the critically ill infant may preclude 1H MRS studies in this particular population in many centers.

Given the need for sedation, any site undertaking 1H MRS studies needs to have the capability of carrying out a full pediatric resuscitation should the need arise. Frequently proton MRS study is added onto conventional MRI, and the time required to obtain a spectrum may add 20-30 min to total time in the magnet. The additional time required is progressively being shortened with technologic advances, thus allowing for the potential for multiple spectra acquisition in a single occasion without prolonged sedation. Significant effects of exposure to magnetic field strengths between 1.5 and 4 T have not yet been identified and are largely limited to potential adverse effects on foreign bodies with a metallic component.

NORMATIVE VALUES

Applying 1H MRS to children with neuropathologic disorders obviously requires knowledge of normative data. Given the "immaturity" of the brain at birth and its rapid postnatal development to approximate adult structures, such normative data absolutely need to be age-related and referenced. Not unexpectedly, regional variation (gray/white matter, cortex/cerebellum) in certain metabolites have been demonstrated(20). Thus far, the focus has been on the relative ratios of metabolites as opposed to absolute quantification of metabolite concentration. Extrapolating from animal work, what is missing is careful quantitation, age-specific, reproducible, regional studies in newborns, infants, and children. Adult norms are well established, and it appears that"adult" values, both qualitatively and quantitatively, are achieved by 3 y of age, likely reflecting the successful completion of the majority of central myelination and organization events.

The relative prominence of the major intracranial metabolite peaks, often expressed relative to Cr, is affected by age. All studies have documented an increase in the NAA peak subsequent to birth, with it becoming the dominant spectra peak apparent by 6 mo of age(21–23). Conflicting data exist regarding Cho with reports of increases, no change, and decreases in this spectrum peak. A relative paucity of newborn 1H MRS exists with particular reference to the premature population, limiting in a way the interpretability of published studies on pathologic processes(24–26). Until such time that established consensus data exist, 1H MRS research studies on neonatal and infantile populations should include laboratory-specific age-matched normative data for reference.

1H MRS AND ACQUIRED BRAIN INJURIES

1H MRS may be of particular usefulness in assessing the severity of ischemic or traumatic brain injury(27). Adults with long-term neurologic sequelae after closed head injury, stroke, or global ischemic injury typically show reduced NAA peaks, reduced NAA/Cr ratios, and elevated Lac peaks(28–30). In a study of 30 adults post cardiac arrest, an elevated Lac on spectroscopy together with an absent N20 somatosensory evoked response correlated with significant impairments 4 wk post injury(31). In adults with stroke, increased Lac within the area of ischemia was associated with several standardized quantitative measures of neurologic dysfunction as well as MRI lesion volume and single-photon emission computed tomography measurements of cerebral blood flow(32).

Similar abnormalities have also been observed in neonates, infants, and children(25,26,33–35). In a study of 82 children with a wide variety of acute CNS insults, spectroscopy findings in three age groups (newborns; infants 1-18 mo, and children>18 mo) were compared with 24 control patients(33). Reduced NAA/Cr and NAA/Cho ratios and increased Lac were found in those patients with poor outcomes (severe disability, vegetative state, death) 6-12 mo after injury. In addition, spectroscopy combined with clinical and MRI data predicted outcome correctly in 91% of neonates and 100% of infants and children. Additional studies from the same group compared 36 children with elevated Lac peaks with 61 patients without an identifiable Lac signal(35). Patients with elevated Lac peaks were more likely to have had a cardiac arrest, be hyperglycemic, have lower Glasgow coma scale scores on admission, as well as have abnormal metabolite ratios when compared with age-matched controls or to patients without detectable Lac(Fig. 4). Also, patients with increased Lac were more likely to be severely disabled (39% versus 10%), survive in a persistent vegetative state (13% versus 2%), or have died (39%versus 7%) compared with patients without Lac.

Figure 4
figure 4

Long (TE = 270 ms) and short (TE = 20 ms) echo MRS sequences in three patients with acute brain injury. (A) Spectra at 12 h and 6 d post injury in a 2-y-old girl post cardiac arrest. The initial study showed reduction in the NAA/Cr ratio (1.08) and a small Lac peak. By d 6 there was marked reduction in the NAA peak and NAA/Cr ratio(0.4; i.e. - 2 SD) and an elevated Lac peak. The patient evolved from coma to a vegetative state. (B) Spectra from a 7-mo-old boy with severe closed head injury secondary to nonaccidental trauma. A marked increase in Lac and reduced NAA are both present in both the long (top) and short (bottom) echo sequences. The NAA/Cr (1.07) and NAA/Cho (0.55) ratios are decreased 1-2 SD below age-matched control subjects. The child remains severely disabled.(C) Spectra from a 2-y-old girl with severe closed head injury and coma secondary to a motor vehicle accident. MRI revealed a severe cerebral contusion. Long and short echo sequences revealed high Lac doublets and virtually no NAA. The patient evolved from coma and remains in a vegetative state.

Full size image

1H MRS has also been used to evaluate the prognosis of pediatric near-drowning victims by correlating outcome with reductions in NAA(34). Of 16 children studied with proton spectroscopy after near drowning, those with excess Lac universally had a poor outcome. Lac was also more likely to peak by 4 d rather than immediately after injury.

Limited information concerning 1H MRS and closed head injury in children has been reported. One study of pediatric head trauma found Lac only in regions of contusion and infarction, but did not find Lac in areas of diffuse axonal injury(36). More recent studies with a larger number of patients have shown the importance of Lac as a predictor of poor outcome after pediatric head injury(12). Eight of 24 children with head injury had Lac on spectroscopy, and their outcomes included death (n = 3), severe disability (n = 4), and moderate disability (n = 1). The remaining 16 patients without Lac had better outcomes. Figure 4 demonstrates long and short echo spectra in two children with severe closed head injury and poor outcomes.

Since 1993, five separate studies, encompassing a total of 83 term and 5 preterm infants, have reported on the proton spectroscopy metabolite changes seen after perinatal HIE(25,26,37–39). In these studies, various definitions of HIE were used, the technical aspects of MRS acquisition were different, and the changes reported in metabolite ratios or in the presence or absence of Lac depended on how soon after insult were studies done. Several important observations are noted: 1) patients studied early within the first 2 d of HIE who ultimately developed severe neurologic deficits were likely to have increased Lac/metabolite ratios such as Lac/Cr or Lac/NAA on long echo sequences but normal conventionally determined ratios such as NAA/Cho, NAA/Cr, and Cho/Cr(37,38); 2) patients who were studied late, 1-2 wk after HIE, and who developed severe neurologic deficits were less likely to have increased Lac and more likely to have metabolite ratios that were reduced (NAA/Cho, NAA/Cr) or increased(Cho/Cr) compared with patients with good outcomes(26,30,39); 3) Lac, in some of the studies, was shown to be present normally in low concentrations, depending on the spectroscopy method used; and 4) the magnitude of change in the metabolite ratios (NAA/Cho, NAA/Cr, Cho/Cr, Lac/Cr, Lac/NAA) appeared to correlate with the severity of subsequent neurologic disability(39). However, because of the small number of asphyxiated neonates studied with spectroscopy and the inconsistencies between studies, there is, as yet, no consensus concerning the timing of study and which specific metabolite ratios and their magnitude of change correlate best with long-term outcome.

Although the available data are still limited, it appears likely that the majority of severe acute focal or global CNS insults will result in major abnormalities in proton spectra in children and that these changes will be best detected several days after injury. As yet, spectroscopy has not been systematically used to help understand the pathogenesis of traumatic and nontraumatic brain injuries in children, in part, because of the logistical difficulties in doing scans early and frequently enough to correlate spectral changes with evolving clinical symptoms or structural changes detected with neuroimaging. It is unlikely that spectroscopy will alter the use of other imaging technologies as the nature of information obtained with spectroscopy is biochemical rather than structural. However, spectroscopy has the potential, particularly concerning its use for outcome prediction or as a marker of the effectiveness of cerebral protective therapy, to supplant currently available electroencephalography and measurements of cerebral blood flow; all have well accepted limited specificity and sensitivity as tools for outcome prediction. Additional studies concerning technical aspects of spectral acquisition including the optimal timing of study, the effects of different etiologies of brain injury on spectral changes, and the influence of development on these spectral changes will be needed before spectroscopy will be clinically useful.

1H MRS AND NEUROMETABOLIC DISORDERS

Studies involving the application of 1H MRS to neurometabolic disorders thus far have been limited by the few numbers of such patients and restricted access to the necessary technology. Spectra obtained are rarely specific to a single disorder; however, principal component analysis of metabolite peaks may permit separation into distinct groups. 1H MRS, together with clinical and traditional imaging data, allows for phenotypic delineation with the degree of metabolic disturbance(s) observed, often providing a predictor for disease severity and eventual outcome. 1H MRS provides the potential for early identification of the as yet asymptomatic at-risk child (e.g. affected sibling) and the means of measuring objectively and serially the success of therapeutic interventions undertaken.

Disorders of oxidative metabolism result in a shift to anaerobic glycolysis with Lac formation a necessary by-product. 1H MRS can detect the abnormal central formation of Lac, which can be separated out from potentially overlapping lipid fractions by its doublet inversion with a longer T2(1). Excellent correlations between CSF Lac measurements and brain Lac determined by 1H MRS in a variety of mitochondrial cytopathies have been demonstrated(40). Regional variations in the degree of metabolic derangement have also been demonstrated, suggesting the metabolic basis for observed phenotypic variation(41). 1H MRS has also allowed for the conclusive demonstration that brain energy failure contributes together with that of muscle to observable symptomatology in glutaric acidemia type II(42). Variations in the degree of oxidative disturbance demonstrated on spectroscopic study has been correlated both with the degree of clinical decompensation within a unitary biochemical defect underlying a geographically restricted Leigh's syndrome(43) and with response to therapeutic interventions(44,45).

1H MRS has also been used to further characterize a distinct clinical phenotype together with clinical evolution and more traditional imaging modalities both in disorders with a known enzymatic defect (pyruvate dehyrogenase deficiency)(46) and a yet unclassified disorder affecting the white matter(47). Indeed the characteristic markedly elevated central NAA peak that results from aspartoacetylase (aspartoacylase) deficiency can be taken as being diagnostic of Canavan's disease before actual enzymatic analysis(48).

Aminoacidopathies have also been studied with 1H MRS thus far in a limited way. In a patient with maple syrup urine disease, central accumulation of branched chain amino acids together with corresponding oxoacids could be consistently demonstrated during episodes of acute metabolic decompensation(49). In studies of adolescents and adults with phenylketonuria(15,50,51), phenylalanine concentrations centrally could be demonstrated.

Patients with X-linked adrenoleukodystrophy, a peroxisomal disorder resulting in perturbed very long chain fatty acid oxidation, demonstrate1 H MRS changes that antedate demonstrable alterations on MRI study. NAA and Cr signals are reduced and decline progressively paralleling disease progression clinically(52,53). The decline in NAA can be construed to reflect a decline in neural integrity and number. In Niemann-Pick type C, a disorder of intracellular cholesterol esterification,1 H MRS demonstrated in a single patient a characteristic central lipid peak that corrected with the use of cholesterol-lowering strategies(cholestyramine, lovastatin, and diet) that mirrored clinical amelioration(54).

Hepatic dysfunction resulting in encephalopathy has also been studied with1 H MRS in adults. Regional lowering of central Cho and myo-inositol levels together with elevations in glutamine/glutamate have been demonstrated to correlate both with basal ganglia changes shown on MRI and with the degree of severity of hepatic coma(55,56). Furthermore, these changes were demonstrated to be reversible with correction of the hepatic dysfunction. Similar findings were demonstrated in a single adolescent with Reye's syndrome(57).

1H NMR STUDIES OF EPILEPSY

Over the last decade significant contributions toward defining the role of1 H NMRS in the investigation of human epilepsy have been made. One of the first demonstrations of changes in the proton spectrum in a subject with epilepsy published in 1990 showed an increase in the Lac signal and decrease in the intensity of the NAA peak at 2.02 ppm(58). Using single volume localization techniques and long TE proton spectra, several studies have demonstrated that, in the epileptogenic region, there is a decrease in the intensity of the NAA peak and an increase in the resonance at 3.2 ppm, which arises from trimethylamine-containing compounds, predominantly Cho(59–61).

Single volume techniques are limited because prior knowledge of the location of the epileptogenic focus must be available to perform the study. More recently at several centers proton spectroscopic imaging has been performed in subjects with epilepsy(61–63). Similar changes in NAA and the trimethylamine resonances have been observed, and the spatial distribution of these abnormalities is beginning to be characterized. These studies have also been performed on high field (4.0 T) systems, and quantitation of these compounds is also being performed(61). Most studies on epilepsy have been on subjects with temporal lobe epilepsy during the interictal period. Several groups have begun investigating focal epilepsies of extratemporal origin as well as examining subjects in the postictal period where localized elevations of Lac have been observed.

The changes in NAA and Cho are suspected to correlate with cell loss and changes in cell type that are observed in the epileptogenic region. Future studies should compare the sensitivity and specificity of NMR spectroscopic studies with volumetric measurements. The compounds observed in long TE proton spectra should also be correlated with the neuroanatomical and histologic changes observed in surgical specimens(64).

Several studies have used proton spectroscopic methods with a short TE time (<20 ms) pulse sequence(65–67). These short TE methods permit measurement of glutamate, glutamine, glucose, GABA, and several other compounds in addition to those that are measured with long TE methods. Some of these metabolites, especially GABA and glutamate, are known to be important in the pathophysiology of epilepsy. More recently changes in brain GABA concentrations have been shown to occur with antiepileptic agents designed to affect GABA metabolism(67,68). One 1H NMR study has demonstrated a correlation between seizure control and brain GABA levels(69). Changes in glutamate have been demonstrated in several studies of human epilepsy and animal models of epilepsy by in vitro methods(70). 1H NMR studies are just now demonstrating changes in both GABA and glutamate in human epilepsy(68,71). Further proton NMR studies of amino acid and glucose metabolism in human epilepsy are likely to provide new insights into the pathogenesis of this disorder.

In summary, several studies have shown that 1H NMRS can demonstrate alterations in cerebral metabolism in epileptogenic regions of the human brain. Some studies have suggested that these biochemical changes are more sensitive than conventional imaging. Before these methods can be useful clinically, further studies are needed to correlate these NMR spectroscopic findings with MRI, quantitative relaxation and volumetric MRI studies, histologic and molecular studies of human tissue, and ultimately the clinical outcomes of the subjects studied. 1H NMRS will undoubtedly play an important role in both the diagnosis and accurate localization of subjects with focal epilepsy who are surgical candidates and in increasing our understanding of the pathophysiology of human epilepsy. These 1H NMR techniques also show promise in studies of the generalized and genetically determined epilepsies. Proton NMR studies of medical therapy where serial measurements can be obtained to investigate the changes in cerebral GABA and glutamate metabolism in response to specific therapies are just now being performed. Newer NMR methods such as spectroscopic imaging of glutamate(72) and measuring the turnover of glutamate and glutamine pools in vivo(73) will greatly enhance our ability to elucidate the underlying mechanisms of human epilepsies and improve treatment.

FUTURE DIRECTIONS

1 H NMRS is just beginning to provide new information on the changes in brain metabolism during normal development and in certain pathologic conditions. The techniques are constantly changing and different types of 1 H NMRS methods are required to obtain measurements of GABA, glucose, glutamate, NAA, Cr, and other compounds. Some NMRS studies have used only qualitative analysis or relative measurements using different NMR methods. This has made comparison of results from different groups problematic. Rigorous, quantitative NMRS studies and comparison of various techniques are still needed to permit widespread clinical use of the many advantages that this noninvasive method has to offer.

The major benefits of 1 H NMRS are: 1) it provides noninvasive, regional measurements of several metabolites in vivo; 2) it is capable of studying normal control subjects;3) it can provide serial, longitudinal measurements on subjects to follow disease progression or normal changes with development and aging; and 4) it can furnish serial measurements of metabolism after administration of medications or with other experimental paradigms. Examples of studies that have taken advantage of these characteristics of 1H NMR are studies of normal development in children(22,23). Glucose transport into the brain has been measured by performing 1 H NMRS during a glucose infusion(72). Changes in both cerebral glucose(73) and Lac(74) in the occipital lobe in response to visual stimulation have been identified using 1H NMRS and suggest that there is a mismatch between glycolysis and oxidative glucose metabolism during normal physiologic stimulation. Changes in amino acids, Lac, and other metabolites have been monitored with 1H NMRS in treatment of mitochondrial diseases(75,76), epilepsy(77), and multiple sclerosis(78). Several studies are ongoing using this technique to identify the biochemical changes and monitor the effect of treatment in brain tumors(79,80).

Other recently developed 1H NMRS techniques take advantage of the interaction of the 1H nucleus with other nuclei with magnetic properties such as the 13C or 15N nuclei. The heteronuclear coupling permits observation of compounds "tagged" with these nuclei with the sensitivity of the proton nucleus. Several compounds can be tagged with these stable isotopes to permit the observation of their transport and metabolism in the brain. Studies of glucose and amino acid metabolism in the human brain have been performed using 13C-labeled glucose(81,82), permitting measurements of Krebs cycle, glutamine synthesis, and glycolytic fluxes in vivo. Similar studies have been performed in animals with CNS gliomas(83,84). Magnetically tagging other metabolites of interest and measuring their transport and metabolism in the brain will undoubtedly provide new insight into the mechanisms of neurologic disorders and diseases in man.

Currently, most 1H NMRS studies in humans are being performed on 1.5 T MR systems. A significant improvement in spatial resolution and sensitivity is being observed in 1H NMRS studies being performed at higher magnetic fields such as 4.1 T. The high resolution images obtained at these higher fields are impressive(85,86). Regional multivolume measurements of glutamate and glutamine(87,88) are readily performed, as are measurements of the gray/white matter metabolite differences(20) at this higher field strength. Functional MRI performed at both high and conventional magnetic fields also permits exquisite localization of language(89), visual(90), and motor(91) functions. A discussion of functional MRI is beyond the scope of this review, and readers are referred to recent reviews(92–94). All three NMR techniques (MRI, MRS, and functional MRI) can be performed concurrently on the same subject to permit measurements of anatomy, metabolism and function using this noninvasive methodology. These methods can now readily be used on children and will assuredly provide new insights into the neurobiology of normal development and the specifics of acquired neurologic disorders.