Three-dimensional Proton MR Spectroscopic Imaging of Premature and Term Neonates

1424

AJNR Am J Neuroradiol 22:1424–1433, August 2001

Three-dimensional Proton MR Spectroscopic Imaging ofPremature and Term Neonates

Daniel B. Vigneron, A. James Barkovich, Susan M. Noworolski, Mary von dem Bussche, Roland G. Henry,Ying Lu, J. Colin Partridge, George Gregory, and Donna M. Ferriero

BACKGROUND AND PURPOSE: Previous studies have primarily used single-voxel tech-niques to obtain MR spectra from the neonatal brain. In this study, we applied 3D MR spec-troscopic imaging techniques to detect the spatial distribution of MR spectroscopic imaging–detectable compounds in premature and term infants. The goals were to test the feasibility ofobtaining 3D MR spectroscopic images of newborns, assess the spatial variations of metabolitelevels, and determine age-dependent differences in MR spectroscopic imaging data.

METHODS: MR spectroscopic imaging data were acquired from nine premature (postcon-ceptional age, 30–34 weeks) and eight term (postconceptional age, 38–42 weeks) neonates, allwith normal clinical and neurologic outcomes. A specialized point-resolved spectroscopy se-quence with very selective saturation pulses was used to select a region encompassing themajority of the brain. Phase encoding in three dimensions was performed in a 17-minuteacquisition time to obtain 3D spectral arrays with a 1.0 cm3 nominal spatial resolution.

RESULTS: This study showed the feasibility of detecting the 3D distributions of choline,creatine, and N-acetylaspartate resonances in the neonatal brain. Significant spectral differ-ences were detected among anatomic locations and between the premature and term groups.

CONCLUSION: This initial study indicates that 3D MR spectroscopic imaging of the neonatalbrain can detect anatomic and age-dependent variations in metabolite levels. This techniqueseems to be a powerful tool to assess the metabolic differences between anatomic regions andto follow the changes in cellular metabolites with brain maturation. This study also indicatesthe need for determining topologic and age-matched normative values before metabolic ab-normalities in neonates can be accurately assessed by MR spectroscopy.

Previous studies have shown that MR spectroscopycan provide a unique and clinically important as-sessment of neonatal brain metabolism (1–12). Inaddition, studies have indicated an important rolefor single-voxel MR spectroscopy in the assess-ment of neonates with neurologic abnormalities, es-pecially premature infants and those with suspectedneonatal asphyxia (4–12). In these studies, spectraobtained from one or two voxels of 5 cm3 or great-er have shown promising results; however, the sin-gle-voxel techniques are very limited in spatialcoverage, spatial resolution, and information con-

Received October 5, 2000; accepted after revision March 17,2001.

From the Departments of Radiology (D.B.V., A.J.B.,S.M.N., M.v.d.B., R.G.H., Y.L.), Pediatrics (A.J.B., J.C.P.,G.G.), Neurology (A.J.B., D.M.F.), and Anesthesia (G.G.),University of California, San Francisco, CA.

This work was supported by National Institutes of Healthgrants PO1 NS35902 and R01 NS40117.

Address reprint requests to Daniel B. Vigneron, MagneticResonance Science Center, Department of Radiology, Box1290, University of California, San Francisco, CA 94143.

q American Society of Neuroradiology

tent. The large voxels average many different areasof the brain, all of which may have different ratiosof metabolites because of differing inherent func-tions and differing levels of maturity. Thus, bio-chemical abnormalities in small regions of thebrain (eg, subthalamic nucleus, caudate nucleus)might be missed because of averaging with sur-rounding normal areas. More importantly, abnor-malities might be overlooked because the normalbiochemical profile of the specific topological re-gion at the specific age in question is not known.Recently, however, techniques for 3D MR protonspectroscopic imaging have been developed that al-low spectra to be obtained from thousands of vox-els throughout the adult brain at 1 cm3 resolution(13, 14). In this study, high-resolution 3D MRspectroscopic imaging techniques were applied tothe neonatal brain studies to assess normal metab-olite levels for premature and term neonates. A de-tailed assessment of the normal distribution of me-tabolite levels in the developing brain is importantto improve our understanding of brain biochemicaldevelopment and to establish a normal referencefor determining abnormal metabolism in pathologic

AJNR: 22, August 2001 PREMATURE AND TERM NEONATES 1425

FIG 1. Square regions of interest, 1.0 cm2 in size, were centered bilaterally on basal ganglia (1), thalami (2), optic radiations (3),calcarine gray matter (4), corticospinal tracts (5), posterior white matter (6), and frontal white matter (7). MR spectroscopic imagingspectra were reconstructed to be centered on each location.

states. Establishing normative MR spectroscopyvalues requires both high spatial resolution and 3Dcoverage because metabolite levels in newbornsvary dramatically with both postconceptional ageand anatomic location.

Methods

Study Participants

Our patient population consists of 17 neonates who werestudied by 3D MR spectroscopic imaging. Nine neonates wereborn prematurely and were studied by MR imaging within 21days of birth at postconceptional ages between 30 and 34weeks. Eight term neonates were studied 4 to 10 days afterbirth; their postconceptional ages ranged between 38 and 42weeks. All these neonates were enrolled in studies evaluating

the value of MR imaging in the assessment of neonatal braindamage. All had normal developmental and neurologic out-comes at 12 months of age. The study protocol was approvedby our institutional review board, and parental consent wasobtained for all infants included in this study.

Premature Neonates

All the premature neonates had either grade I or grade IIgerminal matrix hemorrhage (as determined by transfontanelsonography) or were born to a mother with a clinical diagnosisof chorioamnionitis. None of these neonates ever had clinicalevidence of neurologic dysfunction. Those with clinical, lab-oratory, or radiologic evidence of brain injury, inborn error ofmetabolism, or congenital malformation were excluded fromthe study. The MR examination was performed when the infantwas judged by the attending neonatologist to be stable enoughto be transported to and from the MR imager and studied safe-

AJNR: 22, August 20011426 VIGNERON

FIG 2. Representative MR spectroscopic imaging data for a premature and a term neonate.A, Images and spectral array from the brain of a premature, 30-week postconceptional age neonate with normal outcome.B, Images and spectral array from the brain of a term, 41-week postconceptional age neonate with normal neurologic outcome.

ly. Sound protection was provided by placing cotton in theexternal auditory canals and cotton ear muffs over the ears.Warmth was provided by imaging the neonate in a plexiglascylindrical MR imaging–compatible isolette and by placingchemical blankets above and below the neonate to keep a con-stant temperature of 378C. The infants were fed immediatelybefore the study and slept through the procedure; thus, sedationwas not used.

Term Neonates

The term neonates included in this study were a subset ofthe control participants in a study of perinatal asphyxia whomet the inclusion/exclusion criteria for this study and had nor-mal neurologic outcomes. Only those patients with normalclinical findings and normal imaging results at the time of theMR imaging/MR spectroscopic imaging examination and alsonormal developmental and neurologic outcomes at 12 monthsof age were included in this study of normal MR spectroscopicimaging evolution. Only newborn term infants of a gestationalage of $38 weeks were included. Exclusion criteria were ev-idence of in utero or perinatal infection, major anomalies ofthe brain or other major organ systems, and evidence of con-genital metabolic diseases. Although all radiologic and clinicalmeasures indicated that these infants were normal, one cannot

inconclusively rule out metabolic abnormalities undetectableby these clinical assessments.

MR Imaging Protocol

All studies were performed on a 1.5-T Signa EchoSpeedsystem (GE Medical Systems) using an MR imaging–compat-ible isolette that fits inside the head coil. Three MR imagingseries were acquired before the MR spectroscopic imaging wasperformed: T1-weighted sagittal and axial-view spin-echo im-ages (500/11/1 [TR/TE/excitation]; section thickness, 4 mm;acquisition matrix, 192 3 256; acquisition time, 3.5 min 3 2,for a total of 7 min) and T2-weighted spin-echo images (3000/60, 120/1; section thickness, 4 mm; section gap, 2 mm; ac-quisition matrix, 192 3 256; acquisition time, 6 minutes). Thetotal imaging time was approximately 15 minutes.

MR Spectroscopy Protocol

The point-resolved spectroscopy (15–17) technique wasused to excite a selected region with further localization by 3Dchemical shift imaging (18, 19). Optimal uniformity of theselected region was obtained by using shaped RF excitationpulses, designed using the Shinnar-Le Roux algorithm (20), toexcite a slightly larger region than prescribed and then using

AJNR: 22, August 2001 PREMATURE AND TERM NEONATES 1427

FIG 2. Continued.

very selective saturation pulses (21, 22) to provide submilli-meter transition bands for the selected volume. The very se-lective saturation pulses offer five- to 10-fold higher selectivityand 10-fold higher bandwidths than do conventional suppres-sion pulses (21, 22). When used to define the edges of theselected box, they provide very sharp edges and negligiblechemical shift misregistration effects. The point-resolved spec-troscopy-very selective saturation selected region is extremelyuniform with ,5% variation. In this study, the point-resolvedspectroscopy-very selective saturation selected volume includ-ed a region of approximately 60 to 200 cc, which encompassedthe majority of the newborn brains while excluding the spatialinclusion of subcutaneous lipids. Phase encoding in the point-resolved spectroscopy-chemical shift imaging sequence wasused to obtain 8 3 8 3 8, 12 3 12 3 8, or 16 3 8 3 8 3Dspectral arrays with a nominal spatial resolution of 1 cm3. Af-ter a 2- to 3-minute prescan procedure, including autoshim-ming, the 3D MR spectroscopic imaging data were acquiredin a 17- or 19-minute acquisition time with a TR of 1 and aTE of 144.

MR Spectroscopic Imaging Analysis

The MR images and raw MR spectroscopic imaging datawere transferred off-line to a Sun UltraSparc workstation (SunMicrosystems, Mountain View, CA) for analysis using soft-ware developed at our institution for 3D MR spectroscopicimaging processing. Spectral processing included 4D Fouriertransformations, automatic frequency and phase adjustments,baseline fitting, and peak integration. No k-space filtering wasused. The analysis software also can perform arbitrary post-

processing realignments of the spatial center of the spectralarrays. This function was used to center the MR spectroscopicimaging data on the point-resolved spectroscopy selected re-gions and to position 1 cm3 spectral voxels in specific anatom-ic locations, as defined from the MR images, by using a cus-tom-designed region-of-interest tool. The region-of-interesttool and MR spectroscopic imaging display tool were writtenin the Interactive Display Language (IDL; Research SystemsInc., Boulder, CO).

To assess the MR spectra in various anatomic locations inthe brain, spectral voxels were retrospectively centered in thefollowing regions bilaterally: basal ganglia, thalamus, optic ra-diations, corticospinal tracts, frontal white matter, calcarinegray matter, and parietal white matter. The locations are shownfor a representative case in Figure 1. These seven voxel lo-cations were chosen for several reasons: they include bothmore mature (deep gray matter) and less mature (frontal whitematter) portions of the brain; they involve areas that are im-portant in motor (corticospinal tracts, basal ganglia), visual(calcarine cortex, thalamus), and cognitive (frontal and parietalwhite matter) activities; and they include the regions common-ly affected by neonatal brain injury (deep gray nuclei, frontalintervascular boundary zones).

The areas of the choline, creatine, and N-acetylaspartate res-onances were calculated for each voxel, and ratios were cal-culated between different peaks in the same spectrum and be-tween different regions for each resonance. Because F18fluorodeoxyglucose positron emission tomography and iodine-123-labeled iodoamphetamine single-photon emission CTstudies show that the thalami are the most mature regions of

AJNR: 22, August 20011428 VIGNERON

FIG 3. Representative spectra from 1-cm3 voxels from three regions in the brain of a premature (postconceptional age, 30 weeks) anda term (postconceptional age, 40 weeks) neonate. Note higher relative N-acetylaspartate levels in the thalamus and especially the basalganglia of the term infant. Note also a similar metabolite pattern in the frontal white matter, which is consistent with later maturation ofthis region.

the supratentorial brain in neonates (23–25), the thalamus waschosen as the reference for spectral ratios between anatomicregions.

Statistical Analysis

Statistical analyses were performed using the statistical soft-ware package SAS 7.0 (SAS Institute, Cary, NC). Mean, SD,and data ranges were used to describe data. Mean differencesbetween preterm and full-term babies in MR imaging mea-surements were compared for each anatomic location using at test. Bonferroni P values were also used to adjust for multiplelocations of each MR imaging measurement. Random-effectsmodels were used to study the effect of maturity (prematureversus full term) and anatomic location while controlling forthe effect of multiple locations in the same participant. Partic-ipant effect was the random effect in this analysis. We exam-ined the effects of maturity and anatomic location and theirinteractions. In cases of statistically insignificant interactions(P , .05), the interaction was removed from the final model.Pair-wise comparisons of anatomic locations were performedfor random effects models by the least squared means, and theP values with Tukey-Kramer adjustment for multiple compar-isons were used to evaluate these differences.

ResultsThis study showed the feasibility of acquiring

3D MR spectroscopic imaging data from most ofthe neonatal brain at a 1 cm3 nominal spatial res-olution. Diagnostic quality spectra were success-fully obtained for all the infants studied. Includingautoshimming and set-up time, the entire 3D MR

spectroscopic imaging acquisition typically re-quired 20 to 25 minutes. The number of spectralvoxels obtained during each examination rangedfrom 68 to 130 cm3 in premature infants and 73 to204 cm3 in term babies. Good quality spectra witha signal-to-noise ratio of #41 for the choline res-onance were acquired from each baby; representa-tive MR spectroscopic imaging data are shown fora premature and a term neonate in Figure 2. Notethat these spectral arrays showed great spatial var-iation in choline, creatine, and N-acetylaspartate ra-tios; in addition, lactate/lipid intensities were lowto insignificant in all babies studied.

The relative peak intensities varied between an-atomic regions with the highest choline, creatine,and N-acetylaspartate peak intensities occurring inthe thalamus, which is the most mature area of theneonatal cerebrum. In Figure 3, selected spectrafrom the thalamus, basal ganglia, and frontal wa-tershed white matter for a premature (postconcep-tional age, 30 weeks) and a term (postconceptionalage, 40 weeks) infant are shown. These show lowerlevels of all metabolites and a decreased N-acetyl-aspartate:choline ratio in the basal ganglia as com-pared with the thalamus. The N-acetylaspartate:choline ratio is lowest in the frontal watershedwhite matter, suggesting that this is the least matureregion. A difference was also detected when com-paring the same locations at different postconcep-

AJNR: 22, August 2001 PREMATURE AND TERM NEONATES 1429

TABLE 1: Measured metabolite ratios

NA/Cho NA/Cr Ch/Cr NA/tNA Ch/tCh Cr/tCr

Basal Ganglia

PrematureTerm

0.35 6 0.10*0.58 6 0.31*

0.76 6 0.230.92 6 0.26

2.22 6 0.34*1.59 6 0.25*

0.67 6 0.190.70 6 0.16

0.85 6 0.090.78 6 0.08

1.02 6 0.321.01 6 0.13

Thalamus

PrematureTerm

0.44 6 0.07*0.65 6 0.14*

1.11 6 0.23*1.31 6 0.21*

2.56 6 0.40*2.03 6 0.19*

1.01.0

1.01.0

1.01.0

Optic Radiations

PrematureTerm

0.27 6 0.190.48 6 0.28

0.43 6 0.291.01 6 0.49

1.68 6 0.312.30 6 0.64

0.20 6 0.180.39 6 0.18

0.33 6 0.220.57 6 0.17

0.53 6 0.360.54 6 0.21

Calcarine GM

PrematureTerm

0.36 6 0.140.52 6 0.14

0.60 6 0.310.90 6 0.35

1.66 6 0.391.76 6 0.49

0.40 6 0.240.41 6 0.22

0.49 6 0.290.50 6 0.21

0.81 6 0.570.61 6 0.28

Corticospinal WM

PrematureTerm

0.53 6 0.180.73 6 0.07

1.27 6 0.451.37 6 0.21

2.54 6 0.751.89 6 0.34

0.28 6 0.23*0.45 6 0.28*

0.22 6 0.16*0.40 6 0.24*

0.24 6 0.17*0.42 6 0.25*

Parietal WM

PrematureTerm

0.42 6 0.110.57 6 0.29

1.06 6 0.560.99 6 0.51

2.56 6 1.151.73 6 0.19

0.33 6 0.260.38 6 0.26

0.32 6 0.180.43 6 0.22

0.40 6 0.290.49 6 0.23

Frontal WM

PrematureTerm

0.30 6 0.150.34 6 0.13

0.48 6 0.260.63 6 0.33

1.73 6 0.531.75 6 0.29

0.18 6 0.230.10 6 0.14

0.22 6 0.200.19 6 0.23

0.36 6 0.320.23 6 0.28

* Denotes statistically different (P , .05) values between premature and term neonates.Abbreviations: NA, N-acetylaspartate and other N-acetyl containing compounds; Cho, choline, phosphocholine and other choline-containing

compounds; Cr, creatine and phosphocreatine; NA/tNA, NA value normalized to the thalamus NA value; Cho/tCho, choline value normalized tothe thalamus choline value; Cr/tCr, creatine value normalized to the thalamus CR value; GM, gray matter; WM, white matter.

tional ages; metabolite levels and N-acetylaspartate:choline ratios in the basal ganglia and thalami werelower in the premature infant than in the term in-fant. However, in the immature frontal white mat-ter, where metabolite levels were lowest, the spec-tra were very similar between premature and terminfants.

Table 1 and the graphs in Figure 4 show themean topological variations in N-acetylaspartate,choline, and creatine intensities for both prematureand term infants. Differences in the metabolite dis-tributions and peak area ratios between the termand preterm infants (Figs 5 and 6) show that me-tabolites vary with both topology and brain matu-ration. Despite the small numbers of participants inthis initial study, significant (P , .05) differencesbetween the premature and term groups were ob-served for the N-acetylaspartate:choline and N-ace-tylaspartate:creatine ratios averaged over all re-gions. Regional variations in choline:creatine ratiosalso were found to differ between the prematureand term infants, as shown by a significant (P ,.006) difference in the interaction term. Several ra-tios from specific anatomic locations were also sig-nificantly different between the two age groups(noted by an asterisk in Table 1). These includedN-acetylaspartate:choline and choline:creatine inthe basal ganglia and thalamus. Also of note wasthe significant difference in relative N-acetylaspar-

tate, choline, and creatine levels in the corticospinaltracts between the premature and term infants.

Significant differences in MR spectroscopic im-aging data were also detected between anatomic re-gions within the premature and term groups (Tables1–3). For the term infants, relative N-acetylaspar-tate and relative choline levels were significantly(P , .05) higher in the thalamus than in all otherregions. Basal ganglia values were significantlylower than thalamus values but significantly higherthan the values in the other regions studied. Rela-tive creatine (creatine:total creatine) levels were notsignificantly different between the thalamus andbasal ganglia but were between these two regionsand all other regions. In the premature infants, sim-ilar significant differences were observed betweenthe basal ganglia and other regions; however, therewere more significant differences between differentwhite matter regions than for the term infants (Ta-bles 2 and 3). This suggests a greater variation ofmaturation for white matter tracts in the prematureinfant.

DiscussionThis study showed the feasibility of acquiring

3D MR spectroscopic imaging of the neonatalbrain. The results of this study also indicate thatboth metabolite intensities and ratios of metabolite

AJNR: 22, August 20011430 VIGNERON

FIG 4. Graphic representation of the topological variation of N-acetylaspartate (NA), choline (Ch), and creatine (Cr) peak areasrelative to thalamus values. Higher levels of all metabolites wereobserved in the basal ganglia (BG) and thalamus (TH). N-ace-tylaspartate levels were higher in gray matter regions in bothgroups; however, relative to the calcarine gray matter (CGM), thedifference decreased with age for the optic radiations (ORS), cor-ticospinal tracts (CST), and parietal white matter (PWM), sug-gesting neuronal maturation in these regions. The minor relativechanges in frontal watershed white matter (FWM) metabolite lev-els are consistent with late maturation.

FIG 5. Graphic representation of the differences between pre-mature and term N-acetylaspartate (top), choline (middle), andcreatine (bottom) peak areas for each anatomic region relativeto thalamus values.

intensities in the neonate vary with both anatomiclocation and postconceptional age. These observa-tions are consistent with the fact that different partsof the human brain undergo biochemical matura-tion at different rates (23–25). The results from thisMR spectroscopic imaging study highlight the im-portance of considering the precise anatomic loca-tion and age of the participant when interpretingclinical MR spectroscopy data of neonates.

Relatively few MR spectroscopy studies of thenewborn brain have been reported (1–12). Bothphosphorus and proton MR spectroscopy tech-niques have been applied, with the majority of themore recent studies using proton MR spectroscopybecause of its greater inherent sensitivity. Thesestudies have shown the ability of MR spectroscopyto detect differences in cellular metabolite levelswith developmental maturity, anatomic variation,and brain injury. The neonatal MR spectra werefound to be strikingly different from those of

adults, showing a significant correlation with age.Unlike in the adult brain, the N-acetylaspartate res-onance is much smaller than the choline resonancein the neonatal brain. Also, the choline resonancein the neonate is twice as large as that found in thenormal adult brain. The metabolite concentrationsand ratios in babies change nonlinearly with age,and the changes occur most rapidly in prematurenewborns (1–3). The previous MR spectroscopystudies of the developing brain were acquired fromspecific regions (e.g., occipital cortex, parieto-oc-cipital white matter, and thalamus), with voxel sizesof typically 8 cm3 (2). Although some studies wereacquired with a dual spin-echo acquisition (3, 5),others used stimulated echo acquisitions (1, 2). Byacquiring water-unsuppressed spectra and measur-ing the relaxation times for each resonance, someof the studies measured the absolute quantitation ofeach metabolite relative to water. This is a distinctadvantage of single-voxel acquisitions over the 3DMR spectroscopic imaging, with which the mea-surement of unsuppressed water and relaxationtimes for each voxel would require prohibitivelylong examination times of multiple hours. For com-

AJNR: 22, August 2001 PREMATURE AND TERM NEONATES 1431

FIG 6. Graphic representation of the differences between pre-mature and term N-acetylaspartate:choline (top), N-acetylaspar-tate:creatine (middle), and choline:creatine (bottom) peak arearatios for each anatomic region.

parable regions such as the thalamus, the metabo-lite ratios observed in published single-voxel stud-ies (5) using the similar point-resolvedspectroscopy excitation technique were within 1SD of the values found in this study.

Although the previous single-voxel MR spec-troscopy studies clearly showed significant meta-bolic changes with maturation, the normative datathat they provide is of limited use for clinical as-sessments of neonates. The reasons for this are lim-ited coverage and that the coarse spatial resolutionof previous studies resulted in the averaging ofspectral intensities from tissues in contiguous butdifferent anatomic locations. Therefore, the thala-mi, basal ganglia, and adjacent white matter are allaveraged together. Subtle changes involving onlythe thalamic nuclei might be missed because of av-eraging with unaffected adjacent tissue. Accurate

localization of a voxel to a specific anatomic lo-cation without spectral contamination from adja-cent tissues is especially difficult in the prematurebrain because of its very small size. A single-voxel5- to 8-cm3 MR spectroscopy study of the prema-ture brain averages signals from a large portion ofthe brain, limiting the study to one or two regionsof interest, both of which include multiple anatomicstructures. The 1-cc spatial resolution of our 3DMR spectroscopic imaging data is fivefold smallerthan the previous singe voxel studies, markedly di-minishing the averaging of adjacent structures.With this higher spatial resolution, normative datawere acquired for the maturation of specific regionswithin relatively small structures, such as the basalganglia and thalami, which are areas that can beselectively affected by neonatal hypoxic-ischemicinjury. The ability to better localize the biochemi-cal changes will allow more accurate assessment ofbrain maturation. More importantly, it may allowmore sensitive identification of small areas of ab-normality. This increased sensitivity may allow ear-lier diagnosis of diseases that cause shifts in thelevels of metabolites. We have already had anec-dotal experience in detecting localized elevationsof lactate in the white matter of premature neonateswho went on to develop periventricular leukoma-lacia of prematurity (unpublished results).

In addition to providing multiple individual spectrathroughout the volume of interest, 3D MR spectro-scopic imaging has the advantage that the experi-mental conditions stipulate only the size of individualvoxels. The spatial position of the spectra can be se-lected retrospectively via ‘‘voxel shifting,’’ using theappropriate mathematical weighting of the raw dataon the basis of the translation property of the Fouriertransform (26, 27). In addition to removing the guess-work from the location of abnormal metabolite levels,this feature allows retrospective alignment of thespectra with specific anatomic locations. Althoughsingle-voxel MR spectroscopy studies have shownsignificant anatomic differences at an 8-cc spatial res-olution (2), a more complete assessment of the dis-tribution of metabolite levels in the developing brainis provided by 3D MR spectroscopic imaging. The3D MR spectroscopic imaging technique used in thisstudy also allows the acquisition of large arrays ofspectral data that can be analyzed in a number ofdifferent ways, as illustrated in Figures 4 through 6.Relative metabolite levels can be measured betweendifferent resonances and for the same resonance be-tween regions. Topological variation of different me-tabolites within age groups can be assessed (Fig 4).Also, metabolite differences in specific topologicallocations can be compared between age groups (Fig5). The plotting of ratios (Fig 6) provides compari-sons of metabolite variations with respect to one an-other. Information such as this will be invaluable bothin assessing normal biochemical development of thebrain and in detecting, localizing, and potentiallyspecifying disorders of the brain that affect metabo-lite concentrations.

AJNR: 22, August 20011432 VIGNERON

TABLE 2: Premature neonates—MR spectroscopy imaging ratios significantly (P , .05) different between anatomic regions

BG TH ORS CGM CST PWM FWM

BG NA/tNA NA/tNA,Ch/tCh

Ch/tCh Ch/tCh,Cr/tCr

Ch/tCh,Cr/tCr

NA/tNA,Ch/tChCr/tCr

TH NA/tNA NA/tNACh/tChNA/Cr

NA/tNACh/tCh

NA/tNACh/tChCr/tCr

NA/tNACh/tChCr/tCr

NA/tNACh/tCh

ORS NA/tNACh/tCHCr/tCr

NA/tNACh/tChCr/tCrNA/Cr

NA/Cr

CGM Ch/tChCr/tCr

NA/tNACh/tChCr/tCr

Ch/tChCr/tCr

Cr/tCr

CST Ch/tChCr/tCr

NA/tNACh/tChCr/tCr

NA/Cr Ch/tChCr/tCr

NA/Cr

PWM Ch/tChCr/tCr

NA/tNACh/tChCr/tCr

Cr/tCr

FWM NA/tNACh/tChCr/tCr

NA/tNACh/tChCr/tCr

NA/Cr

Abbreviations: NA, N-acetylaspartate and other N-acetyl containing compounds; Cho, choline, phosphocholine and other choline-containingcompounds; Cr, creatine and phosphocreatine; NA/tNA, NA value normalized to the thalamus NA value; Cho/tCho, choline value normalized tothe thalamus choline value; Cr/tCr, creatine value normalized to the thalamus CR value; BG, basal ganglia; TH, thalamus; ORS, optic radiations;CGM, calcarine gray matter; CST, corticospinal tracts; PWM, parietal white matter; FWM, frontal watershed white matter.

TABLE 3: Term neonates—MR spectroscopy imaging ratios significantly (P , .05) different between anatomic regions

BG TH ORS CGM CST PWM FWM

BG NA/tNA,Ch/tCh

NA/tNA,Ch/tChCr/tCr,Ch/Cr

Ch/tCh,Cr/tCr

Ch/tCh,Cr/tCr

Ch/tCh,Cr/tCr

NA/tNA,Ch/tChCr/tCr

TH NA/tNA,Ch/tCh

NA/tNACh/tChCr/tCr

NA/tNACh/tChCr/tCr

NA/tNACh/tChCr/tCr

NA/tNACh/tChCr/tCr

NA/tNACh/tChCr/tCr

ORS NA/tNACh/tChCh/CrCr/tCr

NA/tNACh/tChCr/tCr

CGM Ch/tChCr/tCr

NA/tNACh/tChCr/tCr

CST Ch/tChCr/tCr

NA/tNACh/tChCr/tCr

PWM Ch/tChCr/tCr

NA/tNACh/tChCr/tCr

FWM NA/tNACh/tChCr/tCr

NA/tNACh/tChCr/tCr

Abbreviations: NA, N-acetylaspartate and other N-acetyl containing compounds; Cho, choline, phosphocholine, and other choline-containingcompounds; Cr, creatine and phosphocreatine; NA/tNA, NA value normalized to the thalamus NA value; Cho/tCho, choline value normalized tothe thalamus choline value; Cr/tCr, creatine value normalized to the thalamus CR value; BG, basal ganglia; TH, thalamus; ORS, optic radiations;CGM, calcarine gray matter; CST, corticospinal tracts; PWM, parietal white matter; FWM, frontal watershed white matter.

AJNR: 22, August 2001 PREMATURE AND TERM NEONATES 1433

This study supports the important concept thatmetabolite concentrations vary with both age andlocation in the developing brain and that 3D MRspectroscopic imaging can play in important role inassessing these changes. In both the preterm andterm groups, significant differences were noted be-tween anatomic regions; the highest levels of cho-line, creatine, and N-acetylaspartate were in thethalamus, followed by the basal ganglia and thenthe other regions (Tables 1–3). Significant differ-ences also were detected between the prematureand term groups for MR spectroscopic imagingdata averaged over all locations and for specificanatomic locations (Table 1). These results are inagreement with the previous single-voxel MR spec-troscopy studies of brain maturation (1–3) but in-clude greater anatomic coverage and much higherspatial resolution.

The SD of the N-acetylaspartate:choline, N-ace-tylaspartate:creatine, and choline:creatine ratios(Fig 6) showed topological and age-dependent var-iations and were highest in the white matter regionsof the premature infants. This may be artifactualbecause of the small sample size and lower signal-to-noise ratio in these locations but may also rep-resent greater individual variation in maturation inthese specific white matter fascicles. Further studieswith much larger numbers at each age are requiredto address this question and to better define the nor-mative metabolite distributions at each topologicallocation and for each week of brain maturation.This study indicates that the creation of such a nor-mative database is critical to allow the reliable de-termination of metabolic abnormalities by MRspectroscopic imaging in clinical assessments ofthe neonatal brain.

ConclusionWe have applied 3D MR spectroscopic imaging

techniques to study the anatomic distributions ofcholine, creatine, and N-acetylaspartate in a smallgroup of premature and term infants. Reproducible,good-quality spectra were obtained in a clinical set-ting by using a clinical MR imager and a reason-able imaging time. The spectra that were obtainedindicate that the biochemical maturation of thebrain varies both temporally and topologically. Anunderstanding and mapping of the variations asso-ciated with brain maturation are critical for the useof MR spectroscopic imaging as a tool to diagnosedisease in the neonatal brain and to study normalbrain development.

References1. Huppi PS, Posse S, Lazeyras F, Burri R, Bossi E, Herschkowitz

N. Magnetic resonance in preterm and term newborns: 1H-spectroscopy in developing human brain. Pediatr Res 1991;30:574–578

2. Kreis R, Ernst T, Ross BD. Development of the human brain:in vivo quantification of metabolite and water content withproton magnetic resonance spectroscopy. Magn Resos Med1993;30:424–437

3. Cady EB, Penrice J, Amess PN, et al. Lactate, N-acetylaspartate,choline and creatine concentrations, and spin-spin relaxationin thalamic and occipito-parietal regions of developing humanbrain. Magn Reson Med 1996;36:878–886

4. Shu SK, Ashwal S, Holshouser BA, Nystrom G, Hinshaw DB Jr.Prognostic value of 1-H MRS in perinatal CNS insults. PediatrNeurol 1997;17:309–318

5. Penrice J, Cady EB, Lorek A, et al. Proton magnetic resonancespectroscopy of the brain in normal preterm and term infants,and early changes after perinatal hypoxia-ischemia. PediatrRes 1996;40:6–14

6. Hanrahan JD, Sargentoni J, Azzopardi D, et al. Cerebral metab-olism within 18 hours of birth asphyxia: a proton magneticresonance spectroscopy study. Pediatr Res 1996;39:584–590

7. Leth H, Toft PB, Peitersen B, Lou HC, Henriksen O. Use of brainlactate levels to predict outcome after perinatal asphyxia. ActaPaediatr 1996;85:859–864

8. Holshouser BA, Ashwahl S, Luh GY, et al. Proton MR spec-troscopy after acute central nervous system injury: outcomeprediction in neonates, infants, and children. Radiology 1997;202:487–496

9. Groenendaal F, Veenhoven EH, van der Grond J, Jansen GH, Wit-kamp TD, de Vries L. Cerebral lactate and N-acetyl-aspartate/choline ratios in asphyxiated full-term neonates demonstratedin-vivo using proton magnetic resonance spectroscopy. PediatrRes 1994;35:148–151

10. Peden CJ, Cowan F, Bryant KJ, Cox IJ, Menon DK, Young IR.Proton and phosphorus MR spectroscopy of infants with hyp-oxic-ischemic brain injury. Radiology 1990;177P:123

11. van der Grond J, Veenhoven RH, Groenendaal F, de Vries LS,Mali WP. MR spectroscopy in full-term infants with perinatalasphyxia (abstr). Radiology 1992;185(P):185

12. Barkovich AJ, Baranski K, Vigneron D, et al. Proton MR spec-troscopy for the evaluation of brain injury in asphyxiated,term neonates. AJNR Am J Neuroradiol 1999;20:1399–1405

13. Duijn JH, Matson GB, Maudsley AA, Weiner MW. 3D phaseencoding 1H spectroscopic imaging of human brain. Magn Re-son Imaging 1992;10:315–319

14. Vigneron D, Nelson S, Kurhanewicz J. Proton chemical shiftimaging of cancer. In: Hricak H, Higgins CB, Helms CA, eds.Magnetic Resonance Imaging of the Body. New York: RavenPress; 1997:205–220

15. Bottomley PA. Spatial localization in NMR spectroscopy invivo. Ann N Y Acad Sci 1987;508:333–348

16. Luyten PR, Marien AJ, den Hollander JA. Acquisition and quan-titation in proton spectroscopy. NMR Biomed 1991;4:64–69

17. Frahm J, Merboldt K-D, Hanicke W. Localized proton spectros-copy using stimulated echoes. J Magn Reson Imaging 1987;72:502–508

18. Brown TR, Kincaid BM, Ugurbil K. NMR chemical shift im-aging in three dimensions. Proc Natl Acad Sci U S A 1982;79:3523–3526

19. Maudsley AA, Hilal SK, Simon HE, Wittekoek S. In vivo MRspectroscopic imaging with P-31: work in progress. Radiology1984;153:745–750

20. Pauly J, Le Roux P, Nishimura D, Macovski A. Parameter re-lations for the Shinnar-Le Roux selective excitation pulse de-sign algorithm. IEEE Trans Med Imag 1991;10:53–65

21. Le Roux P, Gilles RJ, McKinnon GC, Carlier PG. Optimized out-er volume suppression for single-shot fast spin-echo cardiacimaging. J Magn Reson Imaging 1998;8:1022–1032

22. Tran T-K, Vigneron D, Sailasuta N, et al. Very selective sup-pression pulses for clinical MRSI studies of brain and prostatecancer. Mag Res Med 2000;43:23–33

23. Chugani HT, Phelps ME, Mazziotta JC. Positron emission to-mography study of human brain functional development. AnnNeurol 1987;22:487–497

24. Tokumaru AM, Barkovich AJ, O’uchi T, Matsuo T, Kusano S. Theevolution of cerebral blood flow in the developing brain: eval-uation with iodine-123 iodoamphetamine SPECT and MR im-aging correlation. AJNR Am J Neuroradiol 1999;20:845–852

25. Kato T, Okuyama K. Assessment of maturation and impairmentof brain by I-123 iodoamphetamine SPECT and MR imagingin children. Showa University J Med Sci 1993;5:99–115

26. Bracewell RN. The Fourier Transfrom and Its Applications. NewYork: McGraw-Hill; 1978:104–107

27. Mareci TH, Brooker HR. High-resolution magnetic resonancespectra from a sensitive region defined with pulsed field gra-dients. J Magn Reson 1984;57:157–163