Книги по МРТ КТ на английском языке / Functional Neuroimaging in Child Psychiatry Ernst 1 ed 2000

mmol/l for lithium and 1±10±-mol/l for Xuorinated drugs). The validity of 19F MRS as a means for assessing brain drug levels has been validated using a primate model (Christensen et al., 1998). To date, 7Li and 19F MRS have not been extensively used to assess the pharmacokinetics of drugs in the brain of children and adolescents. As can be seen in Tables 4.2 and 4.3 a number of investigators have begun to apply MRS methods to the study of children with psychiatric disorders and neurologic abnormalities. The Wndings to date represent the emergence of unprecedented new data.

Phosphorus

Two diVerent MRS-visible nuclei are evaluated in most studies of brain biochemistry: 31P and 1H. Unlike 7Li or 19F MR spectra, 31P and 1H MR spectra give several resonance lines that arise from well-deWned metabolite pools. Phosphorus MR spectra provide information on the concentration of high-energy phosphate compounds (e.g., PCr and nucleoside trisphosphate (NTP, primarily reXecting adenosine trisphosphate (ATP) in the brain) and phospholipid metabolites, which include phosphomonoesters (PME), phosphodiesters (PDE), and inorganic phosphate (Pi). The functional signiWcance of phosphorous and proton metabolite changes are described in Table 4.4. Information on alterations in brain energy metabolism may be gained by measuring the relative levels of PCr (,1.4 mmol/l), NTP (2.8mmol/l), and Pi (1.4mmol/l) (Buchli et al., 1994). The brain PME resonance, which arises primarily from the phospholipid precursors phosphoethanolamine (PE) and phosphocholine (PC), as well as from sugar phosphates, derives from a total metabolite pool of approximately 3.0mmol/l (Pettegrew et al., 1991). The in vivo PDE resonance has a broad component, arising from membrane bilayers, and a narrow component, which is derived from the phospholipid catabolites glycerophosphocholine (GPC) and glycerophosphoethanolamine (GPE).

At 1.5T, metabolite information can usually be obtained from brain regions as small as 25±50cm3 for phosphorus31; for example, depth-resolved surface-coil spectroscopy (DRESS). MRS data are typically acquired either from single voxels using spatial localization (e.g., Bottomley et al., 1984) image-selected in vivo spectroscopy (ISIS) (Ordidge et al., 1986), or from low-resolution, twoor three-dimensional spectroscopic images (Brown et al., 1982). If desired, the sensitivity of phosphorus spectroscopy can be increased by applying a proton decoupling technique, which requires a special coil for each metabolite of interest (Luyten et al., 1989; Murphy-Boesch et al., 1993). Proton decoupling will produce line narrowing

Fig. 4.3. Proton NMR spectrum taken from a 2cm3 voxel localized over the anterior cingulate cortex. NAA, N-acetylaspartate; Cr(PCr), creatine and phosphocreatine; Cho, cytosolic cholinecontaining compounds; m-Ino, myo-inositol.

eVects for improved resolution of the phospholipid precursors PC and PE, in the PME peak and for the phospholipid breakdown products GPC and GPE in the PDE peak.

Protons

In vivo 1H MRS provides a means to detect and quantify a number of cerebral metabolites, including NAA, Cr, cytosolic Cho, and myo-inositol (m±Ino) (Fig. 4.3). Approximate gray concentrations of these metabolites are 8±11, 6±7, 0.9±1.4, and 4±5mmol/l for NAA, Cr, Cho, and m-Ino, respectively (Pouwels and Frahm, 1998). NAA contributes the largest signal to water-suppressed cerebral spectra and is found primarily in neurons (Birken and Oldendorf, 1989; Tsai and Coyle, 1995). Consequently, the NAA resonance has been viewed as a neuronal marker by a number of investigators. However, the exact role of NAA and other in vivo metabolites is not yet known. Phosphocreatine is a high-energy phosphate and the Cr resonance has been used as a reference standard, reXecting the fact that the total concentration of creatine and Pcr is similar in many brain regions, although slightly higher in cerebral cortex than in white matter (PetroV et al., 1989). Most of the Cho in the brain is incorporated in the membrane lipid phosphatidylcholine, which undergoes a restricted range of motion and, therefore, is largely invisible to in vivo MRS (Miller, 1991). The major contributors to the Cho peak are PC and GPC (Barker et al., 1994). Inositol is involved in phospholipid metabolism as well as in the maintenance of osmotic equilibrium (Moore et al., 1999).

Abnormal basal ganglia

NAA/Cr increases rapidly during preterm, has a linear relationship with age

Proton MRS is complicated by the fact that the signals from most metabolites of interest are Wve orders of magnitude smaller than the signals arising from tissue water and lipid. However, over the last several years, methods have been developed for the routine suppression of water signals (Ogg et al., 1994) and localized brain spectra do not contain large signals from lipids (Behar et al., 1994). Because of diVerences in the relaxation times of metabolites, spatial localization methods diVer for 31P/1H and MRS studies (Moore et al. 1997). Stimulated echo acquisition and pixel-resolved methods (STEAM and PRESS) are often used to collect 1H MRS data (Moonen et al., 1989) and metabolite information can be obtained from brain volumes in the range 1±10cm3. The relatively high spatial resolution of 1H MRS makes it possible to distinguish metabolite diVerences in gray and white matter (Pouwels and Frahm, 1998), although relatively few studies to date have reported segmented imaging data in conjunction

with metabolite information (Yurgelun-Todd et al., 1996; Renshaw et al., 1997; Lim et al., 1998).

Studies in children with psychiatric disorders

Proton MRS

Childhood-onset schizophrenia/schizophrenia spectrum disorder

Hendren et al. (1995) performed 1H spectroscopy on the frontal lobes of 12 children between the ages of 8 and 12 years who were diagnosed with schizophrenia spectrum disorders and on 13 healthy children who were matched for age, sex, and education. Spectra were obtained from a voxel localized in the left frontal cortex. Although no statistically signiWcant diVerences were found in metabolite concentrations between patients and control subjects, a

Major catabolic products of membrane phospholipid degradation and phospholipids

notable reduction of NAA, as well as a reduction in Cho/Cr was found for this region. The investigators concluded that the Wndings were in agreement with previous Wndings of decreased NAA in the frontal lobes of adult patients with schizophrenia (Yurgelun-Todd et al., 1996) and suggested that the small sample sizes limited the ability to detect signiWcant metabolite reductions. In addition, this study examined neuropsychologic performance and brain morphometric measures based on MR images. Children with schizophrenia spectrum disorders performed more poorly on neurocognitive tests, particularly on measures of verbal ability and verbal memory, and were characterized by smaller amygdala, reduced mesial temporal cortex volume, and smaller callosal area, without ventricular enlargement. The pattern of MRS, MRI, and neuropsychologic Wndings identiWed in the children at risk for schizophrenia led the investigators to propose two etiologic processes as the basis for the cortical changes identiWed. One developmental process may be related to a global cor-

tical eVect that results in alterations of callosal maturation and brain asymmetry. This global process would be associated with a range of neurodevelopmental disorders. In addition, abnormal left and right hemisphere development and changes in the corpus callosum may be related to a number of cognitive and behavioral disorders. The second process would be associated with factors that have greater diagnostic speciWcity and inXuence the development of the temporal lobe and the amygdala. This process would be more directly related to the emergence of schizophrenia.

In a subsequent report from the same investigators, 1H MRS data was described for an independent sample of 16 children with schizophrenia spectrum disorders (nine males and seven females, 8±12 years of age, mean age 11 years) (Brooks et al., 1998). The control group consisted of 12 ageand sex-matched subjects (mean age 10.8 years). Single-voxel spectroscopic measurements of the proton metabolites in the left frontal lobe identiWed a decrease in

the NAA/Cr ratio in children with schizophrenia spectrum disorders compared with levels in the control group. As in the earlier study, the authors concluded that these Wndings provided evidence in support of a neurodevelopmental theory for schizophrenia.

Thomas and colleagues reported similar Wndings in a recent study focused on the investigation of metabolite concentrations in the frontal lobe (Thomas et al., 1996). Thirteen patients with schizophrenia spectrum disorders, including seven males and six females (mean age 14"3 years), and 12 healthy controls, including six males and six females (mean age 11"3 years), were evaluated with single-voxel 1H MRS. The investigators found a 32% reduction of NAA/Cr ratio in frontal gray matter and a 13% reduction of the ratio of glutamine plus glutamate plus gamma-aminobutyric acid (combined as a ªGlxº resonance intensity because of diYculties in resolving these three in spectra) to Cr in occipital gray matter in the children with schizophrenia spectrum disorder. Although not statistically signiWcant, NAA/Cho, Glx/Cho ratios were reduced and Cr/Cho was increased in the frontal lobes of the children with schizophrenia spectrum disorder (Thomas et al., 1996). These Wndings further conWrm that cortical abnormalities are present early in the illness. One unique feature of this study is the quantiWcation of a Glx resonance intensity.

Methods for resolving the resonance lines combined in Glx are currently under active development (Weber et al., 1997; Pouwels and Frahm, 1998). The ability to quantify glutamate, glutamine, and gamma-aminobutyric acid in speciWc brain regions will provide an important tool for the identiWcation of neurotransmitter dysfunction.

Bertolino et al. (1998) applied CSI techniques to evaluate and measure metabolites in multiple brain regions simultaneously. In this study of 13 children with early-onset schizophrenia (mean age 15.9 years) and 13 healthy control subjects (mean age 14.7 years), both the hippocampus and the prefrontal cortex demonstrated reductions in NAA/Cr. Moreover, the eVect size of these metabolite reductions was similar to those reported in previous studies of adult-onset schizophrenia performed by the same group. The consistency in metabolite changes identiWed for these two age groups is suggestive of a developmental continuum for these abnormalities in schizophrenia. One recent study by Buckley and co-workers (Buckley, 1998) compared children with schizophrenia spectrum disorders and those with autism using 1H MRS. The investigators found a reduction in NAA in the frontal lobes in both study groups (Buckley, 1998). These results were interpreted as reXecting the presence of cortical hypoplasias regardless of diagnostic category.

The extent to which age, gender, duration of illness, and medication inXuence 1H MRS Wndings is currently not fully understood. Given that children with schizophrenia spectrum disorders display metabolite changes that are similar to those identiWed in adults with schizophrenia, it seems likely that these alterations in metabolite concentrations are reXective of a neurodevelopmental process associated with the etiology of this illness. A neurodevelopmental hypothesis is further supported by recent 1H MRS Wndings by Renshaw et al., (1995) from a series of consecutively referred patients hospitalized with a Wrst episode of psychosis. Compared with normal control subjects, psychotic patients had a signiWcantly lower NAA/Cr ratio, indicating that abnormalities in temporal lobe NAA concentration are present early in psychotic illness.

Obsessive-compulsive disorder

So far, the focus of this section has been on studies examining children with schizophrenia spectrum disorder; however, several recent investigations have applied 1H MRS methods to other psychiatric disorders. Bartha et al. (1998) have investigated metabolite changes in the left basal ganglia in 13 unmedicated children with obsessivecompulsive disorder (OCD) and 13 matched controls. NAA levels were lower in patients than in controls. Proton MRS has also been used to study treatment response in adolescents with OCD. Rosenberg et al. (1998) investigated the relationship between clinical improvement and metabolite concentration in patients treated with paroxetine. Eleven treatment-naive, nondepressed adolescents between the ages of eight and seventeen who met criteria for OCD were treated with paroxetine for 12 weeks. Subjects demonstrated a signiWcant decrease in the concentration of Glx in the caudate nucleus, supporting theories regarding the relationship between OCD and serotonin pathway disruption (Rosenberg et al., 1998).

Depression

In the Wrst application of 1H MRS to the study of depression in adolescents, Steingard et al. (1998) evaluated 14 adolescents diagnosed with depression (mean age 15.6 years) and 26 psychiatrically healthy adolescents (mean age 14.3 years). Using 1H MRS and a 5 in (13 cm) surface coil, investigators collected metabolite data from a 3.4cm3 voxel in the orbitofrontal cortex (Fig. 4.4). Depressed subjects demonstrated an increase in the ratio of Cho/Cr in the orbitofrontal cortex compared with the healthy adolescents, while no statistically signiWcant diVerences were noted for NAA/Cr between the groups. These Wndings are similar to those reported in adult patients with depression and are suggestive of alterations in cholinergic neurotransmission

in the orbitofrontal cortex in patients with aVective disorders (Charles et al., 1994). An abundant literature exists on the relationship between cholinergic neurotransmission and the expression of clinically signiWcant mood disorders (e.g., as reviewed by Dilsaver (1986) and Fritze (1993)). However, the relationship between cytosolic Cho as detected by 1H MRS and cholinergic activity is by no means direct as Cho is used for multiple purposes within cells and the total cellular Cho pool is quite large.

Phosphorus MRS

Phosphorus-31 MRS has proven useful in the investigation of phosphorus-containing metabolites that are associated with energy and lipid metabolism.With the application of in

vivo 31P MRS, detectable phospholipid metabolites include both PME and PDE. As described above, the PME peak consists primarily of PC/PE and metabolites known to be precursors to membrane phospholipid synthesis. The PDE resonance is reXective of breakdown products of membrane phospholipids. Three metabolites are associated with highenergy intracellular metabolism: PCr, Pi, and ATP.

To date, relatively few studies of children and adolescents have utilized 31P MRS. Minshew et al. (1993) used 31P MRS to investigate brain membrane phospholipid and energy metabolism in a region encompassing approximately 15±20cm3 of the dorsal prefrontal cortices of 11 adolescents and young adults diagnosed with autism and 11 control subjects matched for age, IQ, race, gender, and socioeconomic status. Subjects in the autistic group demonstrated

decreased levels of PCr and esteriWed ends of ATP, ADP, dinucleotides, and diphosphosugars compared with the control group. The authors hypothesized that this decrease in PCr was likely reXective of an increased utilization of PCr to maintain ATP levels, or a hypermetabolic state (Minshew et al., 1993). Additionally, signiWcant correlations were found for PME, PDE, and PCr levels and measures of cognitive performance. As performance decreased in the autistic subjects, PCr and PME levels fell while PDE levels rose, a pattern that was not detected in the control subjects. The authors concluded that the correlations of phospholipids and declining test performance in autistic subjects were consistent with undersynthesis and enhanced degradation of brain membranes in these subjects.

In a study of children at risk for developing a substanceuse disorder, Moss et al. (1997) used CSI (31P MRS) to examine four cortical regions, which included the frontal, occipital, and right and left parietal regions. Subjects at the highest risk for developing this disorder showed a signiWcantly reduced right parietal PDE concentration compared with other voxel locations. In explanation, the authors proposed a heightened degree of right parietal synaptic pruning since the PDE signal is reXective of membrane catabolism.

Moss and Talagala (1997) applied CSI (31P MRS) in a study aimed at examining normal development, with 18 male and 11 female children (mean age 13.4 years), and reported a sex-related diVerence in the metabolite concentrations of the frontal and occipital lobes (Moss and Talagala, 1997). Male children had a higher concentration of (-NTP in their frontal lobes and lower concentrations in their occipital lobes, while female children showed the reverse pattern. In addition, males had a higher concentration of PDE in their occipital lobes and lower concentrations in frontal lobes (Moss and Talagala, 1997). These results underscore the importance of using ageand sexmatched controls in studies of children and adolescents.

Summary

Interpretation of the Wndings described above is preliminary because investigations so far have only included small numbers of subjects and results of the spectral data have not been adjusted for the eVects of tissue heterogeneity within the brain regions examined. Furthermore, none of the studies utilized absolute quantiWcation of phosphorus or proton metabolites, or adequately determined whether the observed alterations were associated with structural changes in the frontal, parietal, occipital, or subcortical regions. In order to deWne homogeneous subgroups better within children diagnosed with psychiatric disorders, the

associations between alterations in brain metabolite measures and speciWc impairments in behavior, neuropsychologic functioning, and clinical symptomatology must be identiWed. Medication status and duration of illness are important factors in the interpretation of metabolic changes in psychiatric illnesses, although these confounds are minimized in studies of children relative to studies of adults.

Additional studies are needed to examine whether alterations in brain metabolite measures are clinical state or trait markers, and whether they are associated with unique subgroups of patients. Determining how diVerent metabolic alterations relate to speciWc clinical features and whether these patterns remain stable over the course of development will be an essential part of understanding the pathophysiology underlying the varied clinical presentations.

Neurologic conditions

MRS Wndings in children with acute or chronic neurologic insults are important because they may clarify the signiWcance of biochemical changes identiWed in children with psychiatric diagnoses. Furthermore, many etiologic models of psychiatric illness have proposed that multiple neural processes are responsible for the manifestation of psychiatric disorders (Kinney et al., 1994; Brooks et al., 1998). Therefore, metabolite abnormalities, secondary to cortical insults during infancy and in childhood, may represent either risk factors or primary determinants for some behavioral disorders.

In infants, 1H MRS has been useful both as a diagnostic tool and as a predictor of the outcome of injuries sustained. For example, in a study by Holshouser et al. (1997), the outcome of central nervous system (CNS ) injury was predicted using 1H MRS in 23 neonates (#1 month), 31 infants (1±18 months), and 28 children ($18 months). A total of 82 patients, all of whom had sustained acute CNS injuries including head injury, hypoxic ischemic encephalopathy, near drowning, and CNS infections, were imaged along with 24 control subjects. Measurements were taken from the occipital region using 8cm3 single-voxel 1H MRS. The authors found that ratios of NAA/Cr were lower in infants and children with poor predicted clinical outcomes, as were ratios of NAA/Cho in neonates and infants with poor predicted clinical outcomes. The reported level of accuracy in prediction of outcome in the neonates was 91%, rising to 100% in the infants and children when 1H MRS and MRI data were combined. This technique may be helpful for clinicians and psychiatrists in predicting clinical outcome of post-CNS injury.

Similar results were found in a study by Auld et al. (1995), who applied single-voxel 1H MRS to examine metabolite ratios in an 8cm3 area of occipital and parietal white matter in 30 infants with acute CNS injury (mean age 38"52 months). The authors found that the ratios of NAA/Cr and NAA/Cho were signiWcantly lower in the group of infants with poor neurologic outcomes compared with the infants exhibiting good-to-moderate outcomes. Spectroscopic data, including ratios of NAA/Cr, NAA/Cho, and Cho/Cr, correctly classiWed 81% of patients. The combination of 1H MRS and clinical variables classiWed all patients into their correct outcome group. These Wndings suggest that spectroscopic data may be clinically helpful in predicting the neurologic outcome in children who have experienced a neurobiologic insult.

In a study by Lee et al. (1997), 1H MRS was used to examine metabolite ratios of 12 preterm infants, all of a gestational age of 28±35 weeks. Data acquired from parietal white matter was compared with a normal adult spectra. Preterm infants demonstrated lower NAA/Cr ratios and higher Cho/Cr and m-Ino/Cr ratios compared with the adult spectra. Further, the NAA/Cr ratio was found to increase linearly with age, while the m-Ino/Cr ratio decreased linearly with age. The authors concluded that the NAA/Cr ratio increases rapidly during the preterm period when maturation of the brain occurs.

The neurobiologic underpinnings of certain rare childhood diseases have been further clariWed through the application of MRS technology. Vucurevic and colleagues (1997) investigated infants with Canavan's and Alexander's diseases, which are typically diYcult to diVerentiate. Both diseases are characterized by megaencephaly and massive leukodystrophy. The study included three patients with Canavan's disease, aged 11, 20, and 21 months, one patient with Alexander's disease, aged 4.5 years, and Wve normal subjects with a mean age of 5.5 years. It reported an increase in NAA and a decrease in Cho in the subjects with Canavan's disease compared with healthy controls. In the gray matter of the child with Alexander's disease, there was twice the normal amount of NAA and half the amount of Cho; in white matter there was a third less NAA and half the amounts of Cho and Cr (Vucurevic et al., 1997). Although the eVects of age have not been fully accounted for in this study, the results highlight the metabolic changes found in the two forms of leukodystrophy.

MRS technology has also been applied to the study of hydrocephalus, another rare childhood disease. Whereas researchers originally hypothesized that the increase in cerebrospinal Xuid associated with this disorder produced a decrease in blood Xow and a change in energy metabolism, MRS technology has revealed that no metabolic change

occurs in this disorder (Bluml et al., 1996). Two other diseases that have been further clariWed by the application of MRS technology are adrenoleukodystrophy (Rajanaygam et al., 1997) and Sturge±Weber syndrome (Moore et al., 1996). These diseases, which are progressive, are currently being investigated with 1H MRS. Longitudinal studies will provide objective evidence of the progression and severity of these diseases. Finally, the sensitivity of MRS is illustrated in a study of hypoxic ischemic encephalopathy, where a diagnostic abnormality was detected in the basal ganglia of a 1- day-old infant, whereas structural MR scanning did not detect the abnormality until day 11 (Kim et al., 1998).

Discussion

The interpretation of MR study results must be considered within a neurobiologic framework for understanding psychiatric illness and with an awareness that study methods are often not directly comparable. In general, cortical abnormalities, including MRS Wndings in psychiatric disorders, are viewed as nonprogressive and are, therefore, interpreted as the result of a neurodevelopmental process (Brooks et al., 1998; Thomas et al., 1998). The direct application of MRS methods to young children and adolescents has allowed this hypothesis to be tested and reWned. Future studies will be able to document what brain changes occur in children with schizophrenia after the illness, and the extent to which metabolic alteration is associated with treatment. At present the results of cross-sectional brain spectroscopy studies in children have produced promising results, although only a limited number of nuclei and psychiatric diseases have been examined so far.

One important methodologic constraint in interpreting MRS studies is the presence of confounding variables. For example, in studies based on MRI, intracranial volume in patients is frequently compared with that of nonpsychiatric control subjects; however, this comparison is valid only when controls and patients are carefully matched for relevant variables. SigniWcant correlations have been reported between gender and overall head size; consequently, it is necessary to match control subjects and patients for gender. Overall head size has also been correlated with socioeconomic status as well as with level of educational attainment. Studies applying MRS methods will be subject to similar sources of error. Although there has been some dispute regarding these results, matching patients and control subjects for socioeconomic status and educational level when possible is appropriate. Finally, investigators must be cautious in interpreting changes associated with various psychiatric disorders, as there will be abnormal-

ities with etiologic signiWcance and others that may be related to the course of the illness.

Use of adequate sample size and associated statistical power is an important consideration for interpreting past studies as well as in designing new ones. At present, most MRS studies of children with psychiatric illness have been based on small sample size. It is likely that the brain abnormalities in children are subtle, particularly if assessed as a ratio of the concentration of two metabolites. Consequently, the mean diVerences in regional cerebral metabolites between children with psychiatric disorders and control subjects may well be small relative to the background of normal variation in the general population. In this situation, investigators using small or modest sample sizes risk a high rate of failures to detect true diVerences between patients and controls.

One important goal of imaging studies is to identify abnormalities central to the development of psychiatric disorders, including the neural substrates of vulnerability or risk factors, as well as the core psychopathology. Identifying brain pathology that distinguishes speciWc illnesses involves detecting deviances against a background of extensive normal variation in the general population. Given that a considerable portion of this normal variation in brain structure and function results from genetic and other familial factors, choosing appropriate comparison groups to reduce the sources of this normal, familial variance should help to enhance the signal-to-noise ratio. One important comparison group for psychiatric probands is a group of sibling controls of the same sex and similar age. These subjects will reduce multiple familial sources of nonpathologic variance in brain parameters by providing a partial control for genetic sources of variability. Moreover, these subjects control for demographic factors such as race, ethnic background, and parental socioeconomic status, as well as environmental experience.

MR technology, in combination with other brain imaging techniques, oVers great promise to the understanding of psychiatric disorders. The clinical utility of MRS studies of children with psychiatric illness has so far been limited as no metabolic Wndings are diagnostic for a speciWc illness. Recent Wndings have related diVerences in metabolic concentrations to neurobehavioral symptoms, and to treatment response and course of illness in adult psychiatric subjects (D. Yurgelun-Todd, S. Gruber, A. Sherwood, C. Moore, B. Cohen, and P. Renshaw, unpublished data). Similar patterns are beginning to emerge from MRS studies in children. One diYculty with single-voxel spectroscopic studies has been the search for metabolite changes corresponding to a speciWc location that is presumed to underlie the neuropathology. To address this

diYculty, recent investigations have interpreted results in the context of inter-related systems of brain function rather than as isolated focal brain abnormalities.

The application of MRS techniques will facilitate this theoretical perspective, as its noninvasive nature provides the methods necessary for repeated systematic exploration of both brain structure and neurochemistry. Multivoxel methods such as the MRS methods applied by Bertolino (1998) have allowed investigators to examine multiple brain regions simultaneously. The ability to document metabolite changes prospectively will allow investigators to diVerentiate signiWcant risk factors from CNS changes associated with normal development. In addition, recent techniques have demonstrated that MRS methods may be used to estimate brain concentrations of neurotransmitters. Investigations of this type have been carried out with neurotransmitters such as GABA and glutamate. Proton resonances that arise from GABA and glutamate are diYcult to resolve because of their complexity, and spectral editing methods must often be used for this purpose (Keltner et al., 1997; Weber et al., 1997). In addition, the brain concentration of GABA is quite low, approximately 1 mmol/l; consequently, the resonance lines often have limited signal-to-noise ratios at 1.5T. Although glutamate is present in the brain at much higher concentrations than GABA, only a small fraction of brain glutamate participates in neurotransmission. It is also quite diYcult to resolve resonances that arise from glutamate from those that arise from glutamine. Despite these limitations, the increasing availability of high-Weld MR scanners makes it likely that these methods will be applied to the study of individuals with psychiatric illnesses (Hetherington et al., 1997).

To date, the application of MRS methods to children with psychiatric disorders have been most useful in clarifying processes with etiologic signiWcance in children with schizophrenia. The patterns of decreased NAA concentrations in the dorsolateral prefrontal cortex and mesial temporal lobe have been identiWed for both adults and children, suggesting a developmental continuum for schizophrenia. The rapid growth in the application of MRI/MRS is beginning to yield new insights into aVective illness and OCD. In addition to clarifying the neurobiologic underpinnings of psychiatric illness, MRS methods will ultimately assist in clinical intervention and treatment planning for children with psychiatric disorders.

Acknowledgement

The authors would like to thank Ms Staci Gruber and Ms Norah Janosy for their assistance in preparing this manuscript.