Magnetic resonance spectroscopy of the human brainlphe.epfl.ch › ~mtran › phy_gen_III › Cours...

REVIEW ARTICLE

Magnetic Resonance Spectroscopy of theHuman BrainBRIAN ROSS* AND STEFAN BLUML

Magnetic resonance (MR; synonymous with NMR 5 nuclear magnetic resonance) is a universal physical techniquebest known for non-invasive detection and anatomical mapping of water protons (H). MR-spectroscopy (MRS)records protons from tissue chemicals other than water, intrinsic phosphorus containing metabolites, sodium,potassium, carbon, nitrogen, and fluorine. MRS is therefore an imaging technique with the potential to record humanand animal biochemistry in vivo. As a result of wide availability of MRI equipment in research laboratories andhospitals, MRS is a serious competitor with PET to define normal body composition and its perturbation bypharmacological and pathological events. This article describes practical aspects of in vivo MRS with particularemphasis on the brain, where novel metabolites have been described. A survey of these new aspects ofneurochemistry emphasize their practical utility as neuronal and axonal markers, measures of energy status,membrane constituents, and osmolytes, as well as some xenobiotics, such as alcohol. The concept of multinuclearin vivo MRS is illustrated by diagnosis and therapeutic monitoring of several human brain disorders. Although thesemethods are currently most frequently encountered in human studies, as well as with transgenic and knockout mousemodels, MRS adds a new dimension to anatomic and histopathologic descriptions. Anat Rec (New Anat) 265:54–84,2001. © 2001 Wiley-Liss, Inc.

KEY WORDS: biomedical imaging; magnetic resonance imaging; MRI; MRS; spectroscopy; neurochemistry; brain disease;N-acetyl aspartate; myelination; Canavan disease; Huntington disease; Alzheimer disease

Twenty years ago, the optimum tech-niques available to address a meta-bolic question in the brain were con-sidered to be, in order of reliability:brain slices in vitro, arterio-venous

difference in vivo across the jugularbulb and carotid artery, isolated intactbrain perfusion in situ and, the thennewly emerging techniques of isolatedbrain-cell preparation (Ross, 1979).Almost at the same time, however,there appeared a seminal paper de-scribing the transfer, after 25 years, ofthe chemists’ major investigationaltool, nuclear magnetic resonancespectroscopy (MRS; see Box 1 for acomplete list of abbreviations), tothe intact mammalian brain in vivo(Thulborn et al., 1981). Soon there-after followed in vivo MRS of humanmuscle (Ross et al., 1981), and ofneonatal (Hamilton et al., 1986) andadult human brain (Bottomley et al.,1983).

Today, few university hospitals inthe world are without a whole-bodyMR scanner capable of assaying me-tabolites non-invasively in the humanbrain, using robust MRS methods. To-gether with MRS, physiological MRI,fMRI, and PET, the neuroscientist cannow reverse the order of preference

when considering a technique withwhich to address a metabolic questionin the brain. It is almost certainly“easiest” to turn first to the intact hu-man brain with in vivo magnetic res-onance spectroscopy. How best to dothis in practice is the subject of thisarticle.

As a spectroscopy technique, theproduct of MRS is a printout of peaksof different radio-frequency and in-tensity, recording molecules that pos-sess the intrinsic property of the NMRtechnique, nuclear spins, unique res-onance frequencies, spin-couplings,and relaxation properties. Many clas-sical neurochemical events are readilydocumented in the human brain invivo through MRS. But the moleculesthat yield the optimum MR-signalsare not always those of which the neu-roscientist first thinks, or even wants,in her or his pre-conceived experi-ment. As a result, in the first 20 yearsof in vivo brain MRS, some “new”neurometabolites have come to thefore.

Dr. Ross, a physician and neurologist[MD/DPhil(Oxon)], trained at the Univer-sity of Oxford, UK. He is the Director ofthe Clinical MR Unit at Huntington Med-ical Research Institutes (HMRI) in Pasa-dena, CA; Professor of Clinical Medicineat the University of Southern CaliforniaSchool of Medicine in Los Angeles; andVisiting Associate, Chemistry andChemical Engineering at Caltech in Pas-adena. Dr. Bluml received his PhD train-ing in physics at the universities ofFreiburg and Heidelburg, Germany. Cur-rently he is senior physicist in the Clin-ical MR Unit at HMRI, Director of Re-search for Rudi Schulte ResearchInstitutes in Santa Barbara, CA, and wasrecently appointed Associate Professorin Radiology at USC. Grant sponsor:Rudi Schulte Research Institute; Grantsponsor: NIH; Grant sponsor: HMRI.*Correspondence to: Dr. Brian Ross,Clinical Magnetic Resonance Spectros-copy Unit, Huntington Medical Re-search Institute, 660 South Fair OaksAvenue, Pasadena, CA 91105. Fax:(626)397-5889. E-mail: [email protected]

THE ANATOMICAL RECORD (NEW ANAT.) 265:54–84, 2001

© 2001 Wiley-Liss, Inc.

PRINCIPLES OFNEUROSPECTROSCOPY

We define neurospectroscopy as thefield of study resulting from MRS ex-amination of the human brain. Dis-eases and pathologies of the brain arecommonly classified as:

● structural (including degenerative,tumor and embryogenic defects);

● physiological (essentially interrup-tion of blood supply); and

● biochemical or genetic.

Of the latter, some are receptor andneurotransmitter-related (e.g., dopa-

mine in Parkinson disease) but manyare directly or indirectly related to dis-turbances of the pathways of oxida-tive, anabolic and catabolic interme-diary metabolism, the tricarboxylicacid (TCA) or Krebs cycle, glutamine/glutamate turnover, glycolysis, keto-genesis or fatty acid metabolism. PET,and to a lesser extent SPECT (see Fig-ure 1), MRI, fMRI and diffusion-im-aging address blood flow, glucoseturnover and oxygen consumption,and PET and SPECT are uniquely ableto ‘‘image’’ targeted receptor ligands.Until the advent of NMR, however, no

Today, few universityhospitals in the worldare without a whole-

body MR scannercapable of assaying

metabolites non-invasively in the humanbrain, using robust MRS

methods.

TABLE 1. Current methods of magnetic resonance spectroscopy available for the brain

Clinical Method Rf Coil Available

Implementation

on Clinical 1.5T

Localized 1H MRSLong echo * *Short echo * *Quantitation * *Phase-encoded imaging of metabolites, CSI * *Fast metabolite imaging * *Automation * *Osmolality * *Editing * *Functional MRS * *

Localized 31P MRSPulse-acquire *Decoupled 1H–31P *Phase-encoded imaging, CSI * *Fast phosphocreatine imaging * *Magnetization transfer (flux) *

Localized 13C MRSNatural abundance * *13C enriched-flux measures * *1H–13C heteronuclear methods * *

Localized 15N MRS15N enriched-flux measures1H–15N heteronuclear method(s)

Localized 19F MRS19F drug detection19F imaging and blood flow methods * *(19F probes for Ca21 and Mg2 5 determination)a *

7Li MRS23Na MRS

aToxic in vivo.

Box 1. Abbreviations Used in Magnetic Resonance Imaging and Spectroscopy

MRI, magnetic resonance imaging; MRS, magnetic resonancespectroscopy; PET, positron emission tomography; SPECT, singlephoton emission computed tomography; fMRI, functional MRI;MS, multiple sclerosis; AD, Alzheimer dementia; ALD, adrenoleu-kodystrophy; STEAM, stimulated echo acquisition mode; PRESS,point resolved spectroscopy; ISIS, image selected in vivo spec-

troscopy; CSI, chemical shift imaging; VOI, volume of interest;NAA, N-acetylaspartate; Cr, creatinine; mI, myo-inositol; Cho,choline; PCr, phosphocreatine; GPC, glycerophosphocholine;GPE, glycerophosphoethanolamine; PC, phosphocholine; Glx,glutamine; glu, glutamate.

REVIEW ARTICLE THE ANATOMICAL RECORD (NEW ANAT.) 55

direct non-invasive assay of the prod-ucts of gene expression, the cerebralmetabolites, was available. There wasno neuronal marker, no astrocytemarker and no technique to directlydetermine energy metabolism. Thesegaps are now filled by neurospectros-copy and, with increased clinical ex-perience, a diagnostic “need” for mag-netic resonance spectroscopy (MRS)of the brain emerges.

METHODS FOR HUMANNEUROSPECTROSCOPY

At least 20 methods are available forhuman neurospectroscopy. Table 1outlines the currently available meth-ods of MRS as applied to the brain.The first applications of MRS withsurface coils demonstrated the poten-tial of MRS for non-invasive insightsinto brain metabolism (Ackerman etal., 1980). The lack of a proper local-ization, using this easy and straight-forward technique, has been partlyovercome by improvements (Bottom-ley et al., 1984). In the clinical settingof neurospectroscopy, however, moreadvanced localization techniques areused. Currently, ISIS (Ordidge et al.,1986), STEAM (Frahm et al., 1987;Merboldt et al., 1990) or PRESS (Bot-tomley, 1987) sequences are most fre-quently used as single-voxel or CSItechniques for localized MRS. ProtonMRS performed with long or better,short echo times allows the quantita-tion of important metabolites (Frahmet al., 1989; Kreis et al., 1990; Nara-yana et al., 1991; Hennig et al., 1992;Ross et al., 1992; Barker et al., 1993;Christiansen et al., 1993; Kreis etal., 1993a; Michaelis et al., 1993;Danielsen and Henriksen, 1994). Theautomation of the measurement(shimming, water suppression, acqui-sition) (Webb et al., 1994) and of thedata processing (phasing, fitting),have lead to a fully automated exam.Additional information is obtained byspectral editing techniques that arecurrently used for the identification oflow concentration metabolites oroverlapping resonances (Provencher,1993).

Localized 1H MRS includes longecho time; short echo time; STEAM;PRESS; quantitative; chemical shiftimaging (CSI), metabolite imaging;‘fast’ metabolite images; automated;

functional. 31P MRS includes pulse-acquire; DRESS; ISIS; PRESS; protondecoupled 31P MRS; nuclear Over-hauser effect (NOE) enhanced; CSI,fast phosphocreatine metabolite im-aging; magnetization transfer (flux)measurements. 13C includes pulse-ac-quire, CSI, proton decoupled, NOEenhanced, polarization transfer, 13C-labeled glucose or 13C acetate (unpub-lished work in this laboratory) infu-sion flux studies. Each method hascontributed to some aspect of neuro-chemical research or has found clini-cal application. All of the methods of-fer useful and specific information. Anexample would be the renewed inter-est in 31P MRS with proton-decou-pling, to identify separately the com-ponents of the “Choline” peak seen inroutine 1H MRS (see below).

The Appendix to this article offerspractical information and examples ofhow researchers can design protonMRS experiments on typical clinicalscanners.

Milestones ofNeurospectroscopy

The milestones of neurospectroscopyare summarized in Table 2. MRS in

the brain began with 31P spectroscopyin anesthetized rats and other smallanimals. Non-invasive assays of aden-osine triphosphate (ATP) and phos-phocreatine (PCr) (expressed as “me-tabolite ratios”) and of intracellularpH gave exciting new insights. Directmetabolic rate determination in vivo,using 31P magnetization transfer wasamong the first biological applicationsof this now widespread technique.MRS confirmed the dependence of ce-rebral energetics upon oxidative me-tabolism and glycolysis.

A practical future for MRS wasdemonstrated in the gerbil “stroke”model, when carotid ligation wasclearly shown to produce ipsilateralchanges of anaerobic metabolism:loss of PCr and ATP, increase of inor-ganic phosphate (Pi) and acidificationof the affected hemisphere (Thulbornet al., 1981). A satisfying synthesis ofsome of the new neurophysiologywith clinical management comes instudies of a stroke model. Spreadingdepression is the term applied to thedepolarizing condition that is mim-icked by high K1 cell incubation (Ba-dar-Goffer et al., 1992; have exploredthis extensively in tissue slices), andthat likely occurs with energy failureand hypoxia after stroke. Hossman(1994) and Gyngell et al. (1995) havebrought diffusion weighted imaging[DWI; for a basic description, seeMori and Barker, 1999], MRS andelectrophysiology together to providenew insights into growth of infarcts inrats. In short, the depolarization,which in vitro accelerates glycolysis,is detected in regions where excesslactate appears in ’bursts’ within pen-umbra or spreading depression. Theauthors conclude that this physiolog-ical evidence of deterioration can bemonitored by MRS, and segregatedfrom true infarcts by diffusionweighted imaging. Removal of lactateand recovery of the neuronal markerN-acetyl aspartate (NAA; see below)are likely to be excellent end-points inthe newly emerging clinical trials ofbrain salvage post stroke (Gyngell etal., 1995). It is hard now to realize thatbefore MRS, rapid-freezing of wholeanimals, brain-blowing and surgicalbiopsy were the only effective sourceof knowledge of such events.

MRS in the human brain, whichbegan with newborns, verified the

TABLE 2. Milestones of in vivoMRS

Twelve most prevalent uses of

neurospectroscopy, 1984–2001

1. Differential diagnosis of coma:neurodiagnosis of symptomaticpatients

2. Subclinical hepaticencephalopathy and pre-transplant evaluation

3. Differential diagnosis ofdementia (rule out AD)

4. Therapeutic monitoring incancer 6 radiation

5. Neonatal hypoxia6. “Work-up” of inborn errors of

metabolism7. “Added-value” in routine MRI8. Differential diagnosis of white

matter disease especially MS,ALD and HIV

9. Prognosis in acute C.V.A. andstroke

10. Prognosis in head injury11. Surgical planning in temporal

lobe epilepsy12. Muscle disorders

56 THE ANATOMICAL RECORD (NEW ANAT.) REVIEW ARTICLE

predictions of animal studies thathypoxic-ischemic disease of thebrain could be monitored by thechanges in high-energy phosphates,Pi and pH (Cady et al., 1983; Hope etal., 1984; Hamilton et al., 1986). Thepredictive value of MRS has beendemonstrated in several hundrednewborn infants. The outcome aftersevere hypoxic ischemic encepha-lopathy in newborn humans is deter-mined by the intracerebral pH andPi /ATP ratio.

Wide-bore high-field magnets per-mitted extension to adults and infantsbeyond a few weeks of age (Bottomleyet al., 1983). Three areas of human neu-ropathology have as a result been exten-sively illuminated by 31P MRS. Usingbrain tumor as a target of newly evolv-ing localization techniques (the earlystudies in infants employed no localiza-tion beyond that conferred by the “sur-face coil”), Oberhaensli et al. (1986) be-gan the slow process of overturningthree decades of thought about brain

tumors in particular, showing their in-tracellular pH to be generally alkaline,not acidic. A new generation of drugs inoncology will be designed to enter alka-line intracellular environments, ratherthan the acidic environment measuredin the interstitial fluid.

The 31P MRS in adult stroke exactlymirrored the findings in hypoxic-ische-mic disease of newborns, and evenseems to offer predictive value throughintracellular pH and Pi/ATP (Welch,1992). This work in turn has provoked

Figure 1. Neurochemical pathways: the new neurochemistry. Reactions and metabolites now readily observed in vivo are showndiagrammatically. A: Proton and phosphorus MRS. B: Choline and ethanolamine metabolites of “myelination” observed through proton-decoupled phosphorus MRS (figure courtesy of Prof. D. Leibfritz). C: Carbon fluxes detected with 13C -glucose enrichment. D: (figurecourtesy of Dr. G. Mason). Nitrogen fluxes detected with 15N-ammonia enrichment (figure courtesy of Dr. K. Kanamori).


a large body of research in experimen-tal models of stroke that now guidesthe human application of MRS.

The third area of work stimulatedby the advent of 31P MRS was that ofneurodegenerative diseases, includingAlzheimer disease (Pettegrew et al.,1984). Commencing with in vitrostudies of tissue extracts, two hithertounrecognized groups of compounds,seen in the 31P spectrum as phos-phomonoesters (PME) and phos-phodiesters (PDE), were empiricallyshown to be altered. The metabolicsignificance of these “peaks” was in-completely understood. Nevertheless,a promising new area of neurochem-istry was opened by 31P MRS, andthen extended to in vivo brain analy-sis. By providing non-invasive assaysof less well-known metabolites andpathways, MRS has identified a “NewNeurochemistry.” A perfect exampleof this new knowledge emerged withthe advent of water suppressed 1HMRS of the brain in vivo.

NAA, a neuronal marker (Tallan,1957), was re-discovered in 1983 (Pri-chard et al., 1983). Of the many ex-pected and new resonances now iden-tified in human neuro-MRS, none hasyielded more diagnostic informationthan NAA. Its identity, concentrationand distribution are now well estab-

lished. Early experiments in animalsshowed loss of NAA in stroke. Verylarge numbers of studies in man showNAA absent, or reduced in brain tu-mor (glioma, see Figure 2), ischemia,degenerative disease, inborn errorsand trauma, so that to a first approx-imation, the histochemical identifica-tion of NAA (and N-acetylaspartyl glu-tamate [NAAG]) with neurons andaxons, and its absence from matureglial cells, is confirmed. The use of 1HMRS as an assay of neuronal “num-ber” seems well justified.

The first generation of human MRSstudies were performed without im-age-guidance. Although MRI is not es-sential to our understanding of neuro-chemistry, the combined use of thesetwo powerful tools permitted the di-rect demonstration that there is oftena dissociation in space, between ana-tomically obvious events in the brainand biochemical changes. Metaboliteimaging has confirmed this importantprinciple in stroke, tumors, multiplesclerosis and degenerative diseases.

A simplified method of localizationpermitted the routine use of MRS toassay neurochemistry in a singleplace, albeit rather large, in the cere-bral cortex, cerebellum or mid-brain.This method, now generally known as“single-voxel MRS” is largely respon-

sible for showing that biochemicaldisorders commonly underlie neuro-logical disease (Prichard et al., 1983;Hanstock et al., 1988; Frahm et al.,1988, 1989, 1990; Kreis et al., 1990;Michaelis et al., 1991; Ross et al.,1992; Stockler et al., 1996). MRS istherefore well poised for “early” diag-nosis. Reversible biochemical changesaccompany several physiologicalevents, and provide a biochemical ba-sis for functional imaging (fMRI)(Merboldt et al., 1992; see also Beck-mann et al., 2001; Zeineh et al., 2001).The inborn errors of metabolism andhereditary diseases, and several of themajor neurological scourges of ourtime, reveal functional biochemicaldisturbance. Neonatal hypoxia, cere-bral palsy, neuro-AIDS, dementias,stroke, epilepsies, neuro-infectionsand many encephalopathies are nowseen to include a biochemical compo-nent.

Automation and quantitation arethe final ingredients required to makeMRS an indispensable tool in humanneuroscience. Automation permitsuniversal access, including urgentMRS in acute, reversible neurologicaldiseases, and large scale clinical trials.Quantitation, a long-overlooked area,gives the precision of measurementthat will be required to conclusively

TABLE 3. Absolute concentrations of brain metabolites in individuals of different age groups in mmol/kg braintissue (mean 6 1 SEM) and significance tests for differences found*

Group

n

(ROIs)

GA

(weeks)

pn age

(weeks) NAA Cr Cho mI

,42 GA 11 38.7 6 0.7 3.9 6 1.6 4.82 6 0.54a 6.33 6 0.32 2.41 6 0.11 10.0 6 1.142–60 GA 8 50.2 6 1.8 9.3 6 1.9 7.03 6 0.41 7.28 6 0.28 2.23 6 0.06 8.52 6 0.92,2 pn 6 40.0 6 1.0 0.8 6 0.2 5.52 6 0.64 6.74 6 0.43 2.53 6 0.12 12.4 6 1.42–10 pn 7 41.9 6 1.5 3.8 6 0.9 5.89 6 0.21 6.56 6 0.44 2.16 6 0.09 7.69 6 0.62Adult 10 1,440 6 68 8.89 6 0.17 7.49 6 0.12 1.32 6 0.07 6.56 6 0.43p , 42 GA

vs. adult,0.0001 ,0.0001 ,0.0001 0.001

p , 42 GAvs. 42–60GA

.0007 0.02 0.07 0.10

p 42–60 GAvs. adult

,0.0001 0.16 ,0.0001 0.0003

p 42–60 2pn vs.adult

,0.0001 0.005 ,0.0001 ,0.0001

p 2 pn vs.2–10 pn

0.7 0.8 0.02 0.004

p 2–10 GAvs. adult

,0.0001 0.001 ,0.0001 0.007

*GA, gestational age; pn, post-natal age; ROI, region of interest.


demonstrate incremental metabolicresponses to intervention or therapy.

CHEMICAL COMPOSITION OFTHE BRAIN

Although obviously an oversimplifica-tion, in MRS-terms brain may be bio-chemically defined as water plus dry-matter.

Brain-Water

The water, as in other tissues, is di-vided into intracellular and extracellu-

lar (about 85% and 15% respectively).Intracellular water, which is furtherdivided into cytoplasmic and mito-chondrial compartments, about 75%and 25% respectively, contains all ofthe important neurochemicals (e.g.,Figures 3–6). These are either uniqueto intracellular water, such as lipids,proteins, amino acids, neurotrans-mitters and low-molecular weightsubstances, or at least have a very dif-ferent concentration from the extra-cellular and cerebrospinal fluid (CSF)compartments. Glucose is an excep-

tion, being found in proportions 5:3:1in blood, CSF and brain-water respec-tively. Amino acids are generally dis-tributed 20:1, brain water:CSF orblood. Brain water and extracellularfluid (ECF) are distinct from the largeCSF compartment, the volume ofwhich depends greatly upon the loca-tion selected, and from the intravas-cular blood, which comprises up to6% of brain water. MRS-assays ofbrain water represent the sum of in-tracellular fluid (ICF) and ECF (Ernstet al., 1993a).

Figure 2. CSI: Heterogeneous metabolism of brain tumor. A: Local concentrations of each of the four principal metabolites to be recordedduring a single MRS-examination. B: Results are displayed as metabolite-images (courtesy of Dr. P.B. Barker).


Brain Dry-Matter

Seen through the MR image and theMRS assays of water, the 20% or so ofbrain that really “matters” is largelyinvisible! Hence the terms “missing”or “invisible” are found occasionallyin the MRS literature. Covered bythese terms are all macromolecules(DNA, RNA, most proteins and phos-pholipids), as well as cell membranes,organelles, including the dry-matterof the mitochondria, the christae, andmyelin. The term is probably equiva-lent to the biochemist’s “dry-weight”and can be used as a more constantunit by which to determine the con-centration of key neurochemicals.This is particularly relevant in pathol-ogies in which brain water (or wet-weight/dry weight) may alter, such asmetabolic disorders, edema, tumors,

inflammation, stroke or infarction.Metabolite concentrations may there-fore more accurately be compared asmmol/g dry weight than by the moreusual mmol/g wet weight, or per ml ofbrain water.

Myelin and Myelination

For the most-part myelin is inaccessi-ble to in vivo MRS (because of themanner in which myelin water con-tributes to the MR signal, the contrastbetween white and gray matter isstriking in MRI). The composition ofmyelin is nevertheless of some inter-est to the in vivo spectroscopist be-cause of the changes that may occurin demyelinating and many other dis-eases. Although major componentslike phosphatidyl choline, phosphati-

dyl-ethanolamine, -serine and -inositolare probably entirely immobile andNMR-invisible, their putative break-down products, such as phosphorylcholine, glycerophosphoryl choline,choline, and myo-inositol, are a normalfeature of the 31P or 1H brain spectrum.These molecules will be frequently en-countered in discussions of clinicalspectroscopy, even though their preciserelationship to myelin is far from clear.This is nowhere more important thanin the detection of developmentalchanges in the brain, which are accom-panied by dramatic changes in the MRspectrum (Bluml et al., 1998).

Edema

This all-important concept in neuro-physiology and in clinical diagnosisby DWI and MRI, has not yet beenclearly defined in MRS assays of brainwater. One possibility is that edema,as seen in MRI, represents less than

1% of total brain water, and fallswithin the limits of error of presentmethods of NMR water assay. Thesemethods rely heavily upon differencesin T2 between water in various states.Although it is T2 that distinguishesedema in MRI, the differences are ei-ther too small or too local to be mea-sured directly with MRS.

Metabolism

Amino acids, carbohydrates, fatty ac-ids and lipids, including triglycerides,form a complex network of biosyn-thetic and degradative pathways. Thenetwork is maintained by the thermo-dynamic equilibrium of hundreds ofidentified enzymes, and relative ratesof flux through the various pathwaysare equally closely controlled. Hence,the concentrations of all but a few keymolecules (messengers and neuro-transmitters) are kept remarkablyconstant (e.g., Table 3, Figure 6, 7 and

Figure 3. Localized in vivo and in vitro 1H MR spectra acquired with a stimulated echosequence at 1.5 T. At the top left is a normal brain spectrum (the sum of results from 10age-matched control subjects). At the top right is a reference spectrum from an aqueoussolution composed of 36.7 mmol/l of N-acetylaspartate (NAA), 25.0 mmol/l of Cr, 6.3 mmol/lof choline chloride, 30.0 mmol/l of glucose (Glu), and 22.5 mmol/l of mI (adjusted to a pHof 7.15 in a phosphate buffer). The remaining spectra were recorded from solutions ofindividual biochemicals. To simulate in vivo conditions, all spectra were subjected to aline-shape transformation yielding gaussian peaks of approximately 4 Hz line width. Theintegration ranges used to detect changes in the cerebral levels of Glx (Glu or glutamine[Gln]) (A1 1 A2) and Glu (A3 1 A4) are indicated. The peaks labeled * originated from glycine,NAA, or acetate, which were added to the various solutions as chemical shift references (themethyl peak of NAA was set to 2.02 ppm). All spectra were scaled individually and cannot beused for direct quantitation (modified from Kreis et al., 1992, with permission of the publisher).

The predictive value ofMRS has been

demonstrated in severalhundred newborn

infants.


8). It is for this reason that a thor-oughly reproducible brain ‘‘spectrum’’can be obtained with MRS. Con-versely, predictable and reversiblechanges, such as increased lactate andglutamate, reduced ATP and in-creased ADP, due to altered redoxstate of the pyridine nucleotide coen-zymes of electron transport do occur.This makes MRS the tool for short-term studies on the brain. A compre-hensive list of brain metabolites stud-ied using MRS is given in Table 4.

Compartmentation

Mitochondrial energetics, the en-zymes of which are controlled by non-Mendelian genetics, consists of theelectron-transport chain and of oxida-tive phosphorylation that provides vir-tually all of the high-energy phosphatebonds to maintain ion pumps, neuro-transmission, cell volume and activetransport of nutrients. ATP, the essen-tial “currency” of this process, is buff-ered in brain by another high-energy

system, that of creatine-kinase, creat-ine (Cr), and phosphocreatine (PCr).These molecules are readily observedin MR spectra. Cytoplasmic enzymescontrol aerobic glycolysis and the for-mation of lactate, which supplementsATP synthesis. Glycolysis is massivelyactivated by the Pasteur effect underhypoxic conditions that obviouslylimit mitochondrial energy produc-tion. Lactate and glutamate are bothformed in excess when the mitochon-drial redox state changes. It is possiblethat a similar activation of glycolysisaccompanies “functional” changes (asin fMRI; see Zeineh et al., 2001) andelectrical activation (in seizures). Itshould be noted, however, that mito-chondrial metabolite pools are to avariable extent NMR-invisible andmay not contribute to the final brainspectrum.

Fuels of OxidativePhosphorylation

Glucose dominates the fuel supply forbrain, and its supply via blood flow isstrenuously protected. Vascular oc-clusion, because it brings with it glu-cose deprivation, oxygen lack, andCO2 and H1 accumulation, results inrather different neurochemical insultsfrom that of pure hypoxia, such as isseen in respiratory failure or near-drowning. Thus, hypoxia and ischemiaare different to the spectroscopistwhereas the terms might not need to bedistinguished for the purposes of MRI.

Under severe conditions of starva-tion, when glucose is not available,fatty acids (including acetate) and ke-tone bodies can sustain cerebral en-ergy metabolism, and this may be thenormal state of affairs for the milk-fednewborn. Unlike other tissues, thebrain does not apparently require in-sulin to utilize glucose, so in diabeticsthe marked alterations in cerebral me-tabolism (and in the MR spectrum)are secondary to the systemic meta-bolic disorder. Finally, it is now clearthat ‘two pools’ of cerebral glutamatemetabolism involved in neurotrans-mission represent astrocytes (small-pool) for which acetate is the pre-ferred fuel, and neurones (large-pool)for which glucose metabolism pre-dominates. An exciting opportunitytherefore exists to interrogate thesetwo cell populations independently,

within the living brain, simply by al-tering the fuel supplied in a 13C MRSstudy for example (see below).

Brain Metabolism: A Summary

Figure 1A–D depicts some of the neu-rochemical pathways that have be-come more relevant since the adventof neuro-MRS. The energetic inter-conversion of ATP, PCr and Pi, to-gether with intracellular pH is readilymonitored by 31P MRS. The majorpeaks of the 1H MR spectrum,N-acetylaspartate (NAA), total creat-ine (creatine plus phosphocreatine;Cr), total choline (as reflection ofphosphoryl choline and glycerophos-phoryl choline; Cho), myo-inositol(mI), and glutamate plus glutamine(Glx), were only infrequently encoun-tered in neurochemical discussions of

Figure 4. 1H MR spectra of gray matter ac-quired with PROBE. 1H MRS of gray matteracquired with PROBE. The top spectrum istaken from a healthy volunteer. Spectrafrom global hypoxia due to near-drowning(ND) (spectrum 2), hepatic encephalopa-thy (HE), and probable Alzheimer disease(AD) are shown. All spectra were acquiredon a 1.5 T scanner using stimulated-echoacquisition mode (STEAM) and short echotime TE 5 30 ms; repetition time TR 5 1.5 s.

Figure 5. Typical cerebral MR spectra forsubjects of different age. Typical cerebralproton MR spectra from subjects of differentages. Relative amplitude of the main peaksin STEAM spectra vary drastically with age.Spectra were obtained from a periventricu-lar area in the parietal cortex. Acquisitionparameters: echo time TE 5 30 ms, repeti-tion time TR 5 1.5 s, 144–256 averages, voxelsizes 8–10 cm3 for children, 12–16 cm3 foradults (modified from Kreis et al., 1993b, withpermission of the publisher).


physiology or disease, before the ad-vent of MRS. They now join glucoseuptake and oxygen consumption asthe most easily measured neurochem-ical events, and must become increas-ingly important in neurological dis-cussion. The interconversion ofphosphatidylethanolamine and phos-phatidylcholine (by transmethylation)explains the close links between mye-

lin products now quantifiable throughproton-decoupled 31P MRS (Fig. 1B).

Because of the concentration limit(of protons) at about 0.5–1.0 mM forNMR-detection, virtually all true neu-rotransmitters, including acetylcho-line, norepinephrine, dopamine, sero-tonin (the exceptions are glutamate,glutamine and GABA) are currentlybeyond detection by conventional

neuro-MRS. Similarly, the secondmessenger inositol-polyphosphatesand cyclic AMP are not detected. Thisleaves important gaps in the NewNeurochemistry.

Another evident shortcoming ofNMR is the inaccessibility of mostmacromolecules because of their lim-ited mobility. Accordingly, phospho-lipids, myelin, proteins, nucleosides

Figure 6. Time courses of metabolite peak amplitude ratios vs. gestational age of the subject. Time courses of metabolite peak amplituderatios vs. gestational age of the subject. A,B: Normative curves for the parietal (mostly white matter) and occipital (predominantly graymatter) locations, respectively. The ratios were calculated as detailed in (Kreis et al., 1991). Cho/Cr and mI/Cr were fitted to a mono-exponential, whereas NAA/Cr was fitted to a bi-exponential model function. No developmental curve was calculated for Glx/Cr data,because no clear trend was visible. The curves are well defined for the 1st year of life, where the most dramatic changes take place.Features in the later stages of development are less accurately described by the present data. The normative curves are specific for theacquisition parameters used. Open symbols represent data from parietal cortex, and filled symbols data from occipital cortex.


TABLE 4. Cerebral metabolites measured in vivo by MRS

Name of Metabolite Technique Precision

Acetoacetate 1H 60.5 mMAcetone 1H 60.3 mMAdenosine-triphosphate (ATP) 31P 60.2 mMAspartate 1H 62 mMAtrophy index 1H 61%Beta-hydroxybutyrate 1H 60.5 mMBrain dry-matter 1H 62%Choline 1H 60.1 mMCreatine 1H 61 mMCSF-peak aqueduct flow 1H-Cine 63 ml/minCSF-volume 1H 61%Ethanol 1H 61 mMFluoxetine 19F 61 microg/ml(Tri)fluoperazine 19F 61 microg/mlGABA 1H 61 mMGlucose 1H, 13C 60.5 mM (?)Glucose transport rate (T1/T2) enr13C 61.5 minGlutamate 13C 62 mMGlutamine 1H 61 mMGlycerol 13C 65 mMGlycerophosphorylcholine {1H}-31P 60.2 mMGlycerophosphoethanolamine {1H}-31P 60.2 mMGlycine 1H 61 mMGlycolysis rate en13C 60.37/min/gGuanosine-phosphate 31P 61 mMHydrogen-ion (pH) 31P 60.02 pH unitsInorganic phosphate 31P 60.2 mMInositol-1-phosphate {1H}-31P yes or noIsoleucine 1H 61 mMLactate 1H 60.5 mMLeucine 1H 61 mMLipid 1H; 13C 61 mMLithium 7Li 60.1 mMMacromolecules 1H; 1R 60.1 mMMagnesium (Mg11) 31P 6200 mMMannitol 1H 62 mMMyoinositol 1H; 13C 61 mMNAA 1H 60.7 mMNAAG 1H 60.3 mMOxydized hemoglobin fMRI (1H) 60.2%Phenyl-alanine 1H 62 mMPhospho-choline dc31P 60.2 mMPhosphocreatine 31P 60.2 mMPhosphodiesters 31P 62 mMPhosphoethanolamine dc31P 60.1 mMPhospholipid (membrane) dc31P 630%Phosphomonoesters 31P 62 mMPropane-diol 1H 61 mMPyridine nucleolide(s) (NAD, NADP) dc31P 61 mMScylloinositol 1H 60.2 mMSodium 23Na 6Taurine (see also sI) 1H 61 mMTCA-cycle rate enr13C 60.1 mmol/min/gTransaminase rate enr13C 610 mmol/min/gTriglyceride 13C 65 mMValine 1H 61 mMWater content 1H 63%

enr, enriched; dc, decoupled; 1H-Cine, cine-MRI. Current techniques of MRS on “clinical” MR scanners permit quantifiableassays of each of these metabolites, fluxes or neurochemical events in examination times tolerated by the average volunteeror patient. Not all tests are available on every commercial scanner.


and nucleotides, as well as RNA andDNA are effectively “invisible” to thisfamily of methods. Exceptions may beglycogen, a macromolecule in heart,skeletal muscle and liver, which isreadily detectable in 13C spectra, andthe broad signals from phospholipidsin the 31P spectrum and from low mo-lecular weight proteins in 1H spectrain the brain.

Flux measurements using en-riched stable-isotopes of 13C (Fig.1C) or 15N (Fig. 1D) extend the rangeof neurochemical events accessibleto in vivo MRS. More than 50 metab-olites can now readily be determinedby combination of these techniques(Table 4).

Technical Requirements andMethods

Many years of experience with clinicalMRI have resulted in a widespreadand comfortable understanding of theMR process. The patient (or subject)must be able to lie on a bed, whichenters a confined space where electro-magnetic fields are repeatedlyswitched on and off. The subject musttolerate the accompaniment of con-siderable noise. After an interval ofone to several minutes, all of the ra-dio-signals generated by resonatingprotons are mathematically mappedto produce an image. The anatomicaldisplay in cross-section is now thenorm for all who work in the brain.

MRS is produced in the same way,with three additional steps. Using theimage just obtained, a volume of in-terest (VOI; voxel) is selected for MRSand the field within is further refinedin a process called shimming (shim:old English, a wedge or plough-share).Then, for 1H MRS (but not for MRS ofother nuclei), the protons of H2Owithin the VOI are rendered silent bysuppressing their particular fre-quency band (termed water suppres-sion). Finally, using the same constel-lation of switched magnetic fieldsalready familiar from MRI, a fre-quency profile or spectrum is acquired.Intensity at any given frequency isproportional to concentrations of pro-tons. Frequency is a measure of chem-ical structure; thus the spectrum is atypical output of metabolite composi-tion of the sample.

In MRI, where localization is “ev-erything,” only a single peak (1H ofwater) is mapped. In MRS, localiza-tion must retain chemical shift infor-mation for the acquisition of metabo-lite profiles (Ordidge et al., 1985).Single-volume (or voxel) MRS usesmethods that allow to measure MRsignals originating from one region ofinterest and ensures that unwantedMR signals outside this area are ex-cluded. Alternative strategies exist. Se-lective excitation is analogous to MRI,in which a single metabolite fre-quency is excited and an image is re-constructed. Mapping of cerebral

phosphocreatine (PCr) has employedthis technique (Ernst et al., 1993b).Chemical-shift-imaging (CSI) acquiressimultaneously multiple spectra fromslices or volumes of the brain (Fig. 2A)and metabolite specific images arereadily formed from the resultingpeak-intensities (Fig. 2B). Althoughtheoretically the most time-efficientmethod of in vivo neurochemicalanalysis, in practice CSI brings with itmany unwanted features such as lossof metabolic information, unexpectedquantitative variability across the“slice,” and inconveniently bulky datasets. At this point, single-voxel MRSdominates the field of in vivo brain MRS.

For single-voxel MRS, manufac-tures provide one or more of the fol-lowing capabilities:

● STEAM (stimulated echo acquisi-tion mode),

● PRESS (point resolved spectrosco-py), and

● ISIS (image selected in vivo spec-troscopy).

Technical details, selection criteriaand the necessary physics are exten-sively discussed in Young (2000).

Hardware and Equipment

Relatively few research organizationshave the capability to design and buildhuman MR scanners, and stringentNational and International regula-tions control their use. The MR equip-

Box 2. Representative {1H}-31P MRS Spectra of Diseased Brain in Adults: HepaticEncephalopathy, A Disorder of Cerebral Osmoregulation?

Accumulation of ammonia in the blood causes a series of very specific changes in1H MRS of hepatic encephalopathy (HE). Expected and readily explained by theconversion of ammonia and glutamate into glutamine is an increase of cerebralglutamine. Changes of other metabolites such as a depletion of myo-inositol and areduction in choline, however, are not completely understood. Proton decoupled 31PMRS can significantly improve our understanding of the pathophysiology of HE bymeasuring the Cho constituents separately, quantifying brain phosphoethanolamines,and high energy metabolites. A: In this HE patient a reduction in GPE and therecognized osmolyte GPC can readily be detected whereas PC is unchanged. Whengroups of patients and controls were compared statistically significant re duction ofPE, Pi, and ATP were observed. These finding support the hypothesis of a disturbedosmoregulation. The suggestion that HE is accompanied by cerebral energy failure issupported by the findings of 1–13C glucose MRS, despite the absence of the classicalpattern of reduced PCr and elevated Pi. B: Hyponatremia is recognized to cause adisorder of cerebral osmoregulation. The spectrum from a patient with hyponatremia ofunknown cause (not HE) seems to have a similar pattern as in HE but with theabnormalities even more pronounced. C: Typical spectrum from a young control forcomparison (figure modified from Bluml et al., 1998, with permission of the publisher).


ment is large, heavy and expensive tosite, in magnetically shielded rooms,usually remote from other equipment.Whole-body, or slightly smaller “head-only” units are available. Clinicalequipment is rarely more than 1.5 or2.0 Tesla, but even in the Clinic, 3Tesla, 4 Tesla or 4.7 Tesla are becom-ing commonplace. Seven Tesla, 8Tesla and 9.4 Tesla are in an explor-atory stage but likely to becomeequally indispensable tools in theNeuroscience Research Institute ofthe future.

A Word About Safety

Three different magnetic fields are ap-plied in MRS:

● static magnetic field B0;● gradient fields for localization pur-

poses, and● rf fields to excite the magnetization.

These fields are remarkably safe,with no known biological hazards.Fast switching gradients have beenconsidered as associated with risk,but never more than vaguely identi-fied. Although there exist “exotic”techniques such as echo planar spec-troscopic imaging (EPSI), the vastmajority of MRS techniques switchgradients a magnitude slower than

routinely applied in MR imaging. Pro-longed irradiation of RF is identifiedas hazardous, to the extent that en-ergy is “deposited” in the humanhead. A sensible government limit(SAR) has provided binding guide-lines for nearly 20 years. Provided in-struments are correctly calibrated andfitted with necessary power-monitorand automatic trip, no harm can cometo subjects, voluntary or patients, dur-ing MRS studies. Isolated reports ofburns from faulty electrical equip-ment, home-built RF coils, guidewires and electrodes are rarely seriousbut are a sure sign of sloppy scienceand cannot be tolerated in a first classNeuroscience Institute.

A different class of safety deals withmagnetic objects. Scissors, knives,scalpels, etc., brought into the vicinityof the magnetic field become fast pro-jectiles and can cause serious injuries.Also implanted metallic objects cancause serious harm to a patient as aconsequence of electrical interaction,torque, or heating before or during anexamination. Therefore, physicalsafety around MR equipment is a mat-ter of great concern. Every unit shouldhave a Safety Officer, a clear code ofconduct and probably metal detectorsat all public entrances. Lax security

results inevitably in “accidents” thatcould have been avoided with consci-entious management. Safety lies in ex-cellence of the design of the MR-Suiteand military-style discipline.

RF Coils and Gradients

MRI and 1H MRS of brain is usuallyundertaken with a standard volumehead coil constructed like a helmet tofit over the entire head, and is pro-vided by the manufacturer. For spe-cialized purposes, and because in gen-eral they furnish much needed extrasignal, surface coils or an assembly ofsurface coils (termed phased-array)are used. Volume coils are more flex-ible and have a special advantagewhen quantitation of cerebral metab-olites is the goal.

Because a 1H head coil is standardequipment on all scanners, protonMRS is in principle available on allclinical scanners and by far the mostwidely used MRS technique. RF coilsare frequency-specific, however, sothat MRS studies with other nucleithen proton demand a different (andcostly) head coil. Often these coils arenot provided directly by the manufac-turer but need to be purchased fromsmaller suppliers. For proton-de-

Figure 7. Age related changes of absolute concentrations of the membrane metabolites PE, PC, GPE, GPC and high energy metabolitesPCr and ATP. [PE] (A) and to a lesser extent [PC] (B) decrease with age. Adult levels are reached at age .12 years. [GPE] (C) and [GPC](D) increase only slightly with age. [PCr] (E) increases whereas [ATP] (F) shows a mild decease with age. Age was corrected for gestationalage where applicable.


coupled MRS, which has increasedsensitivity and other advantages inMRS of nuclei other than proton (X-nucleus), combined RF coils, H 1 X,are the norm. Again, these types of coilsare often not among the options offeredby the manufacturer of the system butneed to be purchased separately.

The rapid technical progress in gra-dient coils in recent years is mainlydriven by MRI applications. In partic-ular functional MRI and diffusionweighted MRI, utilizing echo planarimaging (EPI), requires fast switchinggradients with fast rise times. Al-though in general MRS also benefitsfrom more powerful gradient systems,sometimes larger eddy currents fromspeed and power optimized gradientcoils may affect spectral quality ad-versely on newer systems.

Hetero-Nuclear Detection and1H Decoupling

Even if RF coils tuned to nuclei otherthan 1H are available, not all MRIscanners can progress beyond 1H de-tection. Broadband amplifiers are es-sential if anything other than 1H MRSis to be undertaken. Similarly, RF re-ceivers tuned to the resonance fre-quency of the nucleus to be observedmust be provided along with severalother hardware features that aresometimes difficult to retrofit.

Along with the capability to per-form heteronuclear (other than 1H)MRS comes the need to enhance sen-sitivity and specificity of chemicalanalyses by proton-decoupling. Thisinvolves simultaneous excitation at adifferent frequency, and accordinglyrequires a second RF channel and am-plifier.

Examples of spectra that illustrate1H MRS quantitation, signal advan-tage from a surface coil comparedwith a volume coil, broadband-hetero-nuclear MRS and finally, the effects ofproton-decoupling, all from the sameclinical MR scanner retrofitted withthe above mentioned components, areshown at the end of this chapter.

Pulse Sequences: LocalizationMethods

A necessary step in vivo is the segre-gation of extra-cerebral from intra-ce-rebral metabolites and of one intra-

cerebral location from another.Localization sequences select cubes,rhomboid shapes, “slices” or multipleboxes, none of which confirm to rec-ognizable structures of the humanbrain. The choice of the technique isoften pre-ordained by the manufac-turer and is less important than theneed for absolute consistency withinany research program. Althoughmarked differences in spectral ap-pearance result, data is interchange-able between two different sequences,with some loss of precision.

RESULTS

Neurospectroscopy

Proton MRS is by far the most widelyused spectroscopy technique in thebrain. This is due to the fact that stan-dard MRI hardware components areused, making 1H MRS available, thatthe concentrations of proton are rela-tively high in the brain, and that theMR sensitivity to protons is higherthan the sensitivity to other nuclei.

How to “Read” a ProtonSpectrum

The proton spectrum of the normalhuman brain is most readily under-stood by referring to Figure 3 . Eachmetabolite has a “signature” (Ross etal., 1992), which when added to theother major metabolites results in acomplex spectrum of overlappingpeaks. For all practical purposes, atthe moment, due to ease and universalaccess, proton spectroscopy is synon-ymous with neurospectroscopy. Al-though such spectra are familiar toall, it is crucial to adopt a rigorousapproach to acquiring and interpret-ing spectra.

Figure 4 is composed of 4 spectrafrom “gray matter” acquired by an au-tomated procedure (PROBE™ 5PROton Brain Exam), using a 1.5 Tscanner, STEAM and short echo time(TE 5 30 ms). The equivalent spectraacquired at long echo (TE 5 135 or270 ms) would look substantially dif-ferent, but could be similarly inter-preted by referring to a “normal” spec-trum acquired under identicalconditions.

The top spectrum (Fig. 4, “Norm”)is taken from a healthy volunteer.Reading from right to left there are

two broad resonances that are be-lieved to be due to intrinsic cerebralproteins or lipids. The first and tallestsharp peak, resonating at 2.0 ppm isassigned to the neuronal marker N-acetylaspartate (NAA). The next clus-ter of small peaks consists of the cou-pled resonances of b- and g-glutamineplus glutamate (Glx). The tallest peakof this cluster at approximately 2.6ppm is actually NAA that has three

Figure 8. Normal age related changes inthe spectral appearance of {1H}-31P MRS.Shown are averaged spectra from controlsubjects, each calculated from a group ofsubjects within the age range as indicatedin the figure. Most apparent is the strikingreduction of PE, being the most prominentpeak in a newborn, within the first weeks oflife. This is contrasted by only a small reduc-tion of the second phosphomonoester PC.Inorganic phosphate Pi peaks are observedat 4.8 and 5.1 ppm, representing an inter-cellular compartment and an extracellularor CSF compartment. The two peaks areseparated due to the different pH intra- andextracellular, 7.0 vs. 7.2. The peak originat-ing from extracellular Pi can be readily ob-served in the baby spectra because of theincreased ventricles in hydrocephalus. GPEand GPC show a slight, generally increasingtrend with age. A resonance at 2.2 ppmconsistent with glycerophosphoryl serine orphosphoenolpyruvate was observed only inthe neonate spectrum (age 9 6 7 weeks).PCr increases with age whereas ATP slightlydecreases. All spectra were processed andscaled identically to allow direct compari-son (modified from Bluml et al., 1999, withpermission of the publisher).


peaks, one of which overlaps the glu-tamine resonance. The second tallestresonance (at ;3.0 ppm) is creatineplus phosphocreatine (Cr), and adja-cent to this is another prominent butsmaller peak, assigned to “choline”(Cho). A small peak to the left of Chois that of scyllo-inositol (sI). A promi-nent peak at 3.6 ppm is assigned tomyo-inositol (mI). To the left of mI,two small peaks of the a-Glx triplet areclearly seen, and to the left is the sec-ond Cr peak. Variations in the degreeof water suppression affect the peakintensities of metabolites closest tothe water frequency at 4.7 ppm; i.e.,

the second Cr peak and its immediateneighbors. This effect of water sup-pression, however, has no influenceon the diagnostic value of spectra. Thethree major resonances (NAA, Cr andmI) provide a steep angle up from leftto right in normal spectra acquired atshort TE.

In the near-drowning spectrum(Fig. 4, spectrum 2, “ND”), a lipidpeak and overlying lactate doubletpeak (at 1.3 ppm) replace the nor-mally nearly ‘flat’ baseline. NAA is al-most completely depleted and there isa characteristic pattern of increasedglutamine resonances (2.2–2.4 ppm).

Cho/Cr peak-ratio is apparently in-creased, compared with the normalabove, but in this case the impressionis created by reduction in Cr intensity(that can only be ascertained from aquantitative spectrum).

The patient with acute hepatic en-cephalopathy (Fig. 4, spectrum 3, HE)shows peaks in the lipid/lactate regionthat cannot be reliably interpreted.NAA/Cr is clearly reduced, whereasthe cluster of peaks designated Glx isobviously increased. Cho/Cr is if any-thing slightly less than normal, butthe most striking change is the almostcomplete absence of myoinositol(mI). This also makes the increaseda-Glx peaks to the left of mI moreeasily visible.

The lower spectrum (Fig. 4, spec-trum 4, AD), “probable Alzheimer dis-ease,” also has a characteristic ap-pearance, with NAA much reduced(NAA/Cr close to 1). Glx is if anythingreduced, whereas Cho/Cr is in thiscase slightly higher than the normal.The prominent mI peak is almostequal to Cr and NAA intensities givingthe spectrum its characteristic “flat”appearance.

In each case MRI was essentiallynormal, and gave little or no diagnos-tic information, whereas the spectrumis now well established as character-istic for the disease state described.

Normal Brain Development

The evolution of MRS changes in thenewborn brain, from in utero (in asingle near-term fetus) (Heerschapand van den Berg, 1993) to post-par-tum 300 plus weeks of gestational age,is now well described (Figs. 5 and 6)(Kreis et al., 1993b). The findings addsignificantly to the information rou-tinely obtained in MRI. Van derKnaap et al. (1993) have correlatedthe evolution of changes in the 31Pand 1H MRS with the development ofmyelination. Normative curves fornormal development now establishedfor two cerebral locations (Fig. 6) con-firm earlier published long-echo timeand 31P findings. Myo-inositol domi-nates the spectrum at birth (12 mmol/kg), whereas choline is responsible forthe strongest peak in older infants (2.5mmol/kg). Creatine (plus phosphocre-atine) and N-acetyl groups (NA, ofwhich the major component is NAA)

Figure 9. Representative {1H}-31P and 1H MR spectra of diseased brain in babies andchildren. A: Canavan disease (aspartoacylase deficiency) presents in proton MRS withhighly elevated NAA, elevated mI and reduced total Cr and Cho. Additional informationabout the pathophysiology of this rare inborn error is obtained by {1H}-31P MRS, the mem-brane metabolites PC, GPE, and GPC as well as the high energy metabolites PCr and ATPappearing to be reduced (see Fig. 18B for sequential proton MRS for treatment monitoringin this patient). B: This patient initially diagnosed with and treated for hydrocephalus didpoorly clinically and showed abnormal myelination in follow-up exams. GPC was found tobe reduced whereas PCr appears to be elevated. C: A patient with amino acid inborn errordisease (glutaric aciduria II) showed elevated GPE, GPC, and PCr. D: In consolidated latehypoxic injury, increased GPE, GPC, and PCr may reflect a partial volume effect of in-creased glial cells density. In all patients, separately acquired proton MRS is consistent with{1H}-31P MRS findings insofar as total Cr (5 free Cr 1 PCr) and total Cho (GPC 1 PC 1 minorcontributions from other metabolites) parallels changes in PCr or GPC and PC (right panel).Inset (left panel) are the corresponding 1H MRS for patients A–D (from above).


are at significantly lower concentra-tions in the neonate than in the adult(Cr ;6 and NAA ;5 mmol/kg). NAAand Cr increase, whereas Cho, andparticularly mI decrease during thefirst few weeks of life (Table 3) (Kreiset al., 1993b). Increased NAA and Crare determined by gestational age,whereas the falling concentration ofmI correlates best with postnatal-age.

Absolute metabolite concentrationsdepend upon metabolite T1 and T2 re-laxation. Although T1 values alter sig-nificantly with age for the metabolitesNAA, Cr and mI, that for Cho is notaltered. T2 of NAA does seem to showimportant changes between newbornsand adults whereas those of Cr, Choand mI seem to be unimportant.

Quantitative 1H MRS is expected to

be of particular value in diagnosis andmonitoring of pathology in infants(Van der Knaap et al., 1993), becausemetabolite ratios are often mislead-ing. This may be particularly useful inthat period before myelination is ap-parent in the developing brain.

Nonspecific cerebral damage in in-born errors of metabolism may con-sist of disturbance of brain matura-

TABLE 5. Differential diagnostic uses of magnetic resonance spectroscopy

Metabolite

(normal cerebral

concentration) Increased Decreased

Lactate (Lac) (1 mM; notvisible)

oftenHypoxia, anoxia, near-

drowning, ICH, stroke,hypoventilation (inbornerrors of TCA, etc),Canavan, Alexander,hydrocephalus

unknown

N-acetylaspartate(NAA) (5, 10, or 15 mM)

rarelyCanavan

oftenDevelopmental delay,

infancy, hypoxia, anoxia,ischemia, ICH, herpes II,encephalitis, near-drowning, hydrocephalus,Alexander, epilepsy,neoplasm, multiplesclerosis, stroke, NPH,diabetes mellitus, closedhead trauma

Glutamate (Glu) orglutamine (Gln) (Glu 5 ?10 mM; Gln 5 ? 5 mM)

Chronic hepaticencephalopathy (HE),acute HE, hypoxia, near-drowning, OTC deficiency

unknownPossibly Alzheimer disease

Myo-inositol (mI) (5 mM) Neonate, Alzheimer disease,diabetes mellitus,recovered hypoxia,hyperosmolar states

Chronic HE, hypoxicencephalopathy, stroke,tumor

Creatine (Cr) 1phosphocreatine (PCr)(8 mM)

Trauma, hyperosmolar,increasing with age

Hypoxia, stroke, tumor, infant

Glucose (G) (;1 mM) Diabetes mellitus, ? parentalfeeding (G), ? hypoxicencephalopathy

Not detectable

Choline (Cho) (1.5 mM) Trauma, diabetes, “white”vs. “gray”, neonates, post-liver transplant, tumor,chronic hypoxia,hyperosmolar, elderlynormal, ? Alzheimerdisease

Asymptomatic liver disease,HE, stroke, nonspecificdementias

Acetoacetate; acetone;ethanol; aromatic aminoacids; xenobiotics(propanediol; mannitol)

Detectable in specificsettings

Diabetic coma; ketogenicdiet (Seymour et al., 1998)etc.

ICH, intracerebral hemorrhage; TCA, tricarboxylic acid cycle; NPH, normal pressure hydrocephalus; OTC, ornithinetranscarbamylase.


Figure 10. Time course of development of bilateral fetal grafts in a patient with Huntington disease. In ascending order, sequential MRIstudies demonstrate the increasing volume of bilateral grafts placed in the putamen and caudate. In the latest examination, a cyst hasdeveloped on the left. Localized 1H MR spectra are depicted for each examination of left and right sided grafts. Peaks identified as NAA,Cr, Cho and mI reflect a near-normal adult (rather than fetal) neuro-chemistry for those well-established neuro-transplants. Correspondingto the developing cyst, lactate is noted in the latest 1H MR spectrum (top left), and may be an early indicator of graft “failure” (modifiedfrom Hoang et al., 1998, with permission of the publisher).


tion, demyelination, or neuronaldegeneration. In demyelinating disor-ders, it is primarily the myelin sheaththat is lost; secondarily, axonal dam-age and loss occurs.

In demyelinating disorders, the rar-efaction of white matter implies thatthe total amount of membrane phos-pholipids per volume of brain tissuedecreases. Myelin sheaths consist ofcondensed membranes with a highlipid content.

Myelination in normal brain com-mences in the 6th month of fetal de-velopment and continues to adultyears. Peak myelin production, how-ever, occurs from 30 weeks gestationto 8 months postnatal developmentwith young adult-like myelination ob-served at age 2 years (Brody et al.,1987). Phospholipids containing etha-nolamine (E) and phosphoglyceridescontaining choline (Cho) are constit-uents of sphingomyelin and lecithinrespectively, both of which are com-ponents of the myelin sheath. 1H de-coupled 31P MRS ({1H}-dc31P) is ableto separate the complex phospho-monoester and diester peaks intotheir components of phosphoethano-lamine (PE), glycerophospho-ethanol-amine (GPE), phosphatidylcholine(PC) and glycerophosphatidycholine(GPC). Quantitative {1H}-dc31P can beused to investigate and quantify age-related changes in the choline and

ethanolamine constituents of normal,developing and diseased human brainin vivo (Bluml et al., 1998a,b) (see dis-cussion of hepatic encephalopathy inBox 2).

Using a modified PRESS (TE 5 12ms, TR 5 3 s) sequence, and [PCr]

obtained from each quantitative non-decoupled 31P MRS as an internal ref-erence, quantification of [PE], [PC],[GPE], [GPC] shows the age-relatedchanges of membrane metabolites invivo. The {1H}-dc31P spectra fromcontrols of different ages show clear

Figure 11. The creatine pool. Creatine (Cr)synthesis requires participation of kidneyand liver. Tissues may express creatine ki-nase, in which case phosphocreatine (PCr)will be present. Other tissues lack PCr.

Figure 12. Creatine: Gibb-Donnan Equilibrium. Modified from R.L. Veech, with permission.


differences related to age. The largestchange is observed in [PE] that is ini-tially high and decreases to adult lev-els by about age 10 years. [PC] is alsohigh in the neonate and decreases toyoung adult concentrations by aboutage 4 years. Phosphocreatine (PCr)concentrations increase and reachadult levels at age 4. [GPE] and [GPC]show a slight generally increasingtrend with age (Figs. 7 and 8)

Pathology

Comparing MR spectra with thosefrom relevant age-matched normalsubjects, a number of examples of in-appropriate development have beenidentified. Canavan disease (Fig. 9left), a disorder arising from asparto-acylase deficiency. It is presented inproton MRS (see Fig. 18B for sequen-tial proton MRS for treatment moni-toring in this patient) with highly ele-vated NAA, elevated mI and reducedtotal Cr and Cho. Additional informa-tion about the pathophysiology of thisrare inborn error is obtained by {1H}-31P MRS, the membrane metabolitesPC, GPE, and GPC as well as the highenergy metabolites PCr and ATP ap-pearing to be reduced.

A patient (Fig. 9B) initially diag-nosed with and treated for hydroceph-alus did not do well clinically and

showed abnormal myelination in fol-low-up exams. GPC was found to bereduced whereas PCr appears to beelevated.

A patient (Fig. 9C) with amino acidinborn error disease (glutaric aciduriaII) showed elevated GPE, GPC, andPCr.

In consolidated late hypoxic injury,increased GPE, GPC, and PCr may re-flect a partial volume effect of in-creased glial cells density (Fig. 9D).

In all patients information sepa-rately acquired with proton MRS isconsistent with {1H}-31P MRS findingsinsofar as total Cr (5 free Cr 1 PCr)and total Cho (GPC 1 PC 1 minorcontributions from other metabolites)parallels changes in PCr or GPC andPC (see Fig. 9 left panel).

Rapid changes in [PE] and [PC],which are important precursors ofphospholipids, may be related to thehigh rate of synthesis of membranesand myelin in the young developingbrain.

WHAT HAVE WE LEARNEDFROM NEUROSPECTROSCOPY?

Table 5 summarizes abnormalitiesobserved by 1H MRS in diseases.When observing neuropathologicalevents through MRS, there seems tobe a rather limited range of metabolic

changes in response to disease. Tu-mor, MS, stroke, inflammation, andinfections may produce very similarpatterns of change. This is not surpris-ing and should not be condemned aslack of specificity. Rather, it shouldteach us more about the brain’s re-sponse to injury and its preventionand repair.

N-Acetylaspartate (NAA)

Most observations with 1H MRSstrongly support the original formula-tion of NAA as a “neuronal marker.”This simple conclusion, however,must be modified in some particulars.In addition to neurons, there is evi-dence that NAA is found in a precur-sor cell of the oligodendrocyte. Thetime-course of appearance of NAA inhuman embryology remains un-known, but the best estimate is thatNAA biosynthesis may begin in themiddle trimester, i.e., it is not depen-dent upon the existence of MRI-visi-ble myelin, which is only slowly addedto the brain in the months after birth.Furthermore, the finding of approxi-mately equal concentrations of NAAin white and gray matter of the hu-man brain makes it inescapable thatNAA is also a component of the axonor the axonal sheath in man. In addi-tion to NAA, there is good evidencenow for the existence of NAAG in hu-man (as well as animal) brain, withthe preponderance in white matter,and posterior and inferior regions ofadult brain, especially the cerebellum.

Human pathobiology also supportsthe idea of NAA as a neuronal marker,loss of NAA being generally an accom-paniment of diseases in which neuro-nal loss is documented. Glioma,stroke, the majority of dementias, andhypoxic encephalopathy all show lossof NAA.

That NAA is an “axonal marker”too, is supported by the loss of NAA inmany white matter diseases (leu-kodystrophies of many kinds havebeen studied), in MS plaques and inwhite matter in hypoxic encephalopa-thy.

If NAA is a neuronal marker, can weever expect to see recovery of NAA inpractice? A clear example of neuronalrecovery or regeneration is providedby the fetal neural transplant intoadult human brain (Hoang et al.,1997, 1998). Convincing evidence of

Figure 13. Ultra high field resolves “choline” region in vivo. 1H NMR spectrum acquired froma 1 ml volume lateral to the ventricle in dog brain at 9.4 Tesla. Processing consisted ofzero-filling, 3 Hz Lorentz-to-Gauss lineshape conversion and FFT. Peaks were tentativelyassigned based on dominant constituent and published chemical shifts. Courtesy of Drs. R.Gruetter and I. Tkac.


the presence of NAA in the graftedregion of the putamen is provided bysequential examinations in such a pa-tient, by means of localized 1H MRS(Fig. 10). Most other examples pro-posed are perhaps best understoodnot as evidence of neuronal recovery(still to be viewed as unlikely), but asone of five possible alternatives:

Axonal recovery, after a lessthan lethal insult to the neuron,as for example in MS plaques, orin the rare MELAS syndrome.

An “artifact” of cortical atro-phy. Thus, as neuronal death oc-curs, the consolidation of surviv-ing brain tissue is well

documented by neuropathologicalstudies, and by cortical atrophyon MRI. MRS that determineslocal ratios or even concentra-tions of NAA will record a realincrease in the local concentra-tion of this metabolite.

Survival (or recovery) of a peakat 2.01 ppm in the proton spec-trum may not be due to NAA orNAAG, but to several other me-tabolites that contribute to thisspectral region. A small decreasein the NAA peak in diabetes melli-tus may be explained better interms of another metabolite.

NAA appears also to be a re-

versible cerebral osmolyte, in-creasing or decreasing in responseto hyper-osmolar states, and de-creasing noticeably in hypo-osmo-lar states, such as sodium deple-tion and possibly hydrocephalus(Bluml et al., 1997).

Slow NAA Resynthesis (Morenoet al., unpublished observations).

Creatine (Cr) andPhosphocreatine (PCr)

We know that, in human brain atleast, these two compounds, whichare in rapid chemical, enzymatic ex-change, represent a single T2 species.MRS estimates 8 mM Cr 1 PCr inhuman gray matter, compared withpublished values of 8.6 mM for rap-idly frozen rat brain. Cr concentrationin human gray matter significantly ex-ceeds that measured in white matter,in contrast to the results of tissue cul-ture studies, in which Cr seems to bemore related to astrocytes than toneurones (Flogel et al., 1995).

As with NAA, MRS studies havethrown very interesting light upon thefactors that might control Cr 1 PCr inthe human brain. In addition to thewell-known regulation by enzymeequilibrium that permits a presum-ably crucial role of PCr in energeticsof ATP synthesis, two new conceptshave emerged. The first is that cere-bral Cr is controlled by distant events,due to the complex biosynthetic path-way through liver and kidney enzymes.Before Cr can be available for transportto the brain it must be synthesized (Fig.11). Absolute cerebral [Cr] falls inchronic liver disease, and recovers afterliver transplantation. Even more strik-ing is the recent discovery of a new hu-man inborn error of Cr biosynthesisthat manifests as absence of cerebral Crfrom the proton spectrum, which canbe corrected by dietary administrationof creatine (Hanefeld et al., 1993).

The third method of regulation ofcerebral Cr content is surprising, inview of the crucial nature of cerebralenergy conservation. This is themarked modification of cerebral Cr byosmotic (Donnan) forces, increased inhyperosmolar states and decreased inthe very common setting of hypo-os-molar states due to sodium depletion.The explanation for this apparent

Figure 14. Reactions involving myo-Inositol (mI).


over-riding of the all-important en-zyme equilibrium (Gibbs) forces isprobably the same as that recently dis-covered for the mammalian heart andfor cancer cells. Namely, Gibbs equi-librium and Donnan equilibrium arevery closely linked. When all equilib-ria are interdependent, then the total[Cr 1 PCr] may rise or fall to maintainthe osmotic equilibrium. We pre-sume, but cannot tell from 1H MRSalone, that even under these circum-stances, the ratio of PCr/Cr continuesto comply with the over-riding re-quirements of the thermodynamicequilibrium between PCr and ATP(Fig. 12).

An interesting example of this com-plexity is the observation that the con-centration of [Cr] and of [PCr] isincreased in the late-hypoxic-enceph-alopathy brain (Fig. 9D). This second-ary effect is presumably a reflection ofa new steady state, in which creatinekinase equilibrium is maintained, butthe residual cell-population (“gliosis”)is defined by a higher total creatinecontent (Bluml et al., 1998a).

Cholines (Cho)

A number of new ideas concerning thecholine resonance and its constituentmetabolites have emerged from clini-cal studies. Although theoretically as-sociated with myelin, the choline con-centration in cerebral white matter isnot much higher than that in graymatter, even though this is the im-

pression one gains from constantlyseeing 1H spectra, in which Cho/Cr ismuch higher and nearer to 1.0 inshort echo-time spectra of white mat-ter. The explanation lies in the differ-ence of [Cr] concentration betweenthe two locations, the [Cr] concentra-tion being approximately 20% higherin gray matter. Thus, the [Cho] is onlya little higher, 1.6 mM in white matterand 1.4 mM in gray matter. The cho-line head-groups of phosphatidyl-cho-line contribute hardly at all to the pro-ton spectrum of the human brain invivo since the total of free choline,plus phosphoryl choline plus glycero-phosphoryl choline determined bychemical means in human brain biop-sies and post-mortem samples is veryclose to 1.5 mM.

As with Cr, osmotic events areamong the many local and systemicevents that alter its concentration inbrain. The finding that many focal,inflammatory and hereditary diseasesresult in increased choline concentra-tion has lead to the speculation thatthese metabolites represent break-down products of myelin. Conversely,the finding that several systemic dis-ease processes also modify cerebralcholine indicates that biosynthesisand hormonal influences outside thebrain, possibly in the liver, can mark-edly alter the composition and con-centration of the choline peak. Theseremain to be elucidated. Althoughproton spectroscopy offers little hope

of distinguishing the different compo-nents, (in vivo 1H MRS at very highfield (e.g., 9 Tesla) is showing somepromise (Fig. 13), proton decoupledphosphorus spectroscopy undoubt-edly can do so, giving the opportunityto use disease processes to further un-derstand these interesting metabo-lites. Figure 1B (modified from Prof.D. Leibfritz) integrates a number ofthese ideas concerning cerebral cho-line and ethanolamine metabolites,which are directly accessible throughMRS.

Myo-Inositol (mI) and Scyllo-Inositol (sI)

Some remarkable facts have emergedconcerning this simple sugar-alcoholthat was rediscovered with the adventof short TE in vivo human brain spec-troscopy. Its concentration fluctuatesmore than any of the other majorcompounds detected in the protonspectrum, over 10-fold, from thethree-times adult normal values innewborn infants and hypernatremicstates, to almost zero, in hepatic en-cephalopathy. mI has been recognizedas a cerebral osmolyte since 1990, andits cellular specificity is believed to beas an astrocyte ‘marker.’ Like Cho, mIhas been labeled as a breakdownproduct of myelin (because it is seenat apparently increased concentrationin MS plaque, HIV infection andmetachromatic leukodystrophy). But

Figure 15. Natural abundance proton decoupled 13C MRSm. Shown is a comparison of localized (top) and unlocalized (bottom) naturalabundance 13C MRS at 4 Tesla. Peaks from myo-inositol at 72.0, 73.0, 73.3, and 75.1 ppm, as well as the C2 glutamate at 55.7, C2 glutamineat 55.1 ppm, and C2 NAA at 54.0 ppm can readily be detected. Peaks from creatine and choline overlap at 54.7 ppm and are eliminatedfrom the localized spectrum by the polarization transfer technique used to acquire the localized spectrum (reproduced from Gruetter etal., 1996 with permission of the publisher).


the evidence is particularly indirect onthis point. Despite attempts to confinethe role of mI to that of a chemicallyinert osmolyte or cell marker, it is im-portant to remember that mI is atthe center of a complex metabolicpathway that contains among otherproducts the inositol-polyphosphatemessengers, inositol-1-phosphate, phos-phatidyl inositol, glucose-6-phosphateand glucuronic acid (Fig. 14). Any orall of these products may be involvedin diseases that result in marked alter-ations in mI or sI concentration. Thedifferentiation of inositol phosphate

from mI that is difficult to achievewith 1H MRS, is likely to be achievedby a combination of proton decoupled31P MRS and natural abundance 13CMRS (Ross et al., 1997).

Glutamine (Gln) andGlutamate (Glu)

Provided care is taken, and the appro-priate sequences applied, even at 1.5T, the two amino acids that contributeto the spectral regions 2.2–2.4 and3.6–3.8 ppm can be separated. Glu-tamine, particularly when present at

elevated concentrations can be deter-mined with some precision. Even bet-ter separation is achieved at 2.0 T,when glutamate can be unequivocallyidentified and quantified. It is glu-tamine concentration, rather thanthat of glutamate that seems to re-spond to disease. Increased cerebralglutamine concentration occurs inmany settings, from Reyes syndrome,and hepatic encephalopathy to hy-poxic encephalopathy. It is the lattercase that seems contrary to popularneurochemical theory.

The determination of in vivo cere-bral glutamate and glutamine (Glx)concentrations using 1H MRS is com-promised, however, by the complexspectral appearance of glutamate/glu-tamine due to J-coupling. Further,other metabolites contributing to thesignal at the chemical shift of gluta-mate/glutamine render their quantita-tion difficulty. Studies demonstratedthe potential of natural abundance invivo 13C for direct determination ofcerebral metabolites at 2.1 and 4 Texperimental systems (Fig. 15) (Gru-etter et al., 1994, 1996). In a recentstudy, however, it was shown thateven on a 1.5 Tesla clinical scanner,glutamate and glutamine can be sep-arated from each other, and naturalabundance 13C MRS provides enoughS/N ratio for their in vivo quantitation(Fig. 16) (Bluml, 1998).

A much closer look at glutamateturnover is achieved through the useof either 13C or 15N MRS (the formerin human brain). Mason et al. (1995)and Gruetter et al. (1994) determinedthe rates of the TCA cycle, glucoseconsumption, glutamate formationfrom 2-oxoglutarate (Fig. 17) and fi-nally the rate of glutamine synthesis(GS) in vivo. Their data is consistentwith the long held view of two glu-tamate compartments. By selectingthe appropriate starting substrate,acetate vs. glucose (Bluml et al.,2001), 13C MRS permits direct assayof the in vivo astrocyte and neuronalglutamate turnover rates in the hu-man brain.

Although no explanation is yetavailable for the accumulation of glu-tamine rather than glutamate in hy-poxic brain, the work of Kanamoriand Ross (1997) with 15N MRS offerssome clues (Fig. 1D). Thus, the rate ofPAG, the sole pathway of glutamine

Figure 16. Natural abundance proton-decoupled 13C MRS in Canavan disease acquiredon a Clinical Scanner at 1.5 T. Details of in vivo {1H}-13C spectra from a child diagnosed withCanavan Disease (A), 7 months and 3 years old controls (children with unrelated diseases)(B,C) aligned with a spectrum from a model solution (D). mI at 72.1, 73.3, and 75.3 ppm andNAA at 40 and 54 ppm can be clearly identified. Peaks at 54.6 and 55.2 ppm are consistentwith Cr/Cho and Gln. Note that Glu at 55.7 ppm is obviously depleted in Canavan diseasewhen compared with control spectra. NAA and mI resonances appear to be elevated inCanavan disease, confirming 1H MRS results. Glycerol peaks at 62.9 ppm and 69.9 ppm arewell decoupled (reproduced from Bluml, 1998 with permission of the publisher).


breakdown to glutamate, is undertight metabolic control in the (rat)brain. Because there is a cycle con-verting glutamate to glutamine andback, it may be that PAG holds theanswer to the regulation of cerebralglutamate concentration in hypoxia.On the horizon are new insightsthrough 13C and 15N; Figures 1C,D il-lustrate the present state of ourknowledge of in vivo flux rates, mea-sured in the brain.

MULTINUCLEAR MRS

Solving a Problem:Multinuclear MRS of the BrainIn Vivo

With the extraordinary family of MRStechniques now available to the hu-man neuroscientist, we await an ex-plosion of new knowledge. MRS is sorich in information, and convenientlycorrelated with MRI, fMRI and re-lated procedures, that the concept of“one-stop-shopping” is close at hand.Within MRS, combining carefullyquantified multi-nuclear studies canexpand the utility of the equipment.

This approach is illustrated by theexample of Canavan disease, a defi-

ciency of aspartoacylase that resultsin hypomyelination with megaloceph-aly, blindness and spasticity, anddeath within the first few years of life.The potential of multi-nuclear in vivoMR spectroscopy is illustrated on asingle patient who underwent a seriesof noninvasive MRS examinations(see Fig. 18 and also Fig. 2 of the Sup-plementary Material [INSERT URL]).Regions of different cell type compo-sition (e.g., gray vs. white matter)were readily identified on MRI. Local-ized 1H MRS provided quantitation ofbrain water and the neuronal markerNAA (Fig. 18B). Reduced aspartoacy-lase activity in this disease is expectedto result in an elevation of NAA,readily observed by 1H MRS. Sequen-tial 1H MRS further offers the possi-bility of monitoring therapy. Otherabnormalities such as low cerebralCho, high mI, excess of sI, and a sub-tle reduction of total Cr are observedas well.

Additional information about thepathophysiology of this rare inbornerror was obtained by proton-de-coupled 31P MRS insofar as the mem-brane metabolites PC, GPE, and GPCappear to be reduced (Fig. 18C). Alter-

ations in those putative membranemetabolism markers may reflect de-layed or abnormal myelination. TotalCho from 1H MRS can be correlatedwith myelin metabolite and osmolytedeterminations from proton-de-coupled 31P MRS. A reduction in ce-rebral PCr confirms low total Cr (5free Cr 1 PCr) from 1H MRS. Thesignificance of the slightly decreasedATP is uncertain.

Natural abundance {1H }-13C MRSconfirmed elevated NAA and mI anddetected a striking reduction of gluta-mate (Fig. 18D). This may be a resultof the sequestration of aspartate inNAA and the reduction of free aspar-tate. 1-13C-glucose MRS demonstrateda 60% reduced rate of NAA synthesisin Canavan disease compared to con-trol (not shown). In summary, as dem-onstrated in this rare inborn error dis-ease, in vivo MRS can be used toquantify and monitor metabolic ab-normalities and may provide signifi-cant contributions to our understand-ing of human neuropathophysiology.

Phenotyping Knock-Out MouseModels of NeurologicalDisease by In Vivo MRS

Although our emphasis has been onhuman neurochemistry (an obviousneed, given the inaccessibility of thebrain for in vivo analysis) there hasbeen no shortage of applications ofMRS to experimental animals. Per-haps the most important in the futurewill be studies of transgenic andknock-out mice. This rapidly expand-ing research tool poses the broadproblem of phenotyping to confirmthat the deleted gene really does reg-ulate the anticipated metabolic path-way. MRS studies of knock-outs ortransgenics energy metabolism, dia-betes, and neurological disorders suchas Huntington and Alzheimer diseaseare to be found in the literature; re-cent advances are reviewed in Beck-mann et al. (2001).

CONCLUSIONS

Studies in neuroscience have gener-ated myriad biochemical questions,many of which are beyond the scopeof in vivo MRS. The ideal questioninvolves global (or at least “million-neuron”) events and millimolar

Figure 17. In vivo proton-decoupled 13C MRS with 13C-labeled glucose infusion at 4T. 13Cenriched glucose infusion studies allow the determination of glycolysis and TCA cycle fluxrates in vivo. Direct 13C NMR detection of label accumulation in a 22.5 ml volume in thevisual cortex was measured by Greutter et al. (1996) at 4 T. A: Stack plot with 3 min timeresolution of the region containing the Glu C4 resonance at 34.2 ppm. B: Pre-infusionspectra with data accumulation extended to 12 min and spectrum acquired within 30 minafter start of infusion. C: Time course of the C4 glutamate resonances after infusion of1-13C-labeled glucose. Courtesy of Dr. G. Mason.


Figure 18. Multi-nuclear MRS of the brain in vivo: Canavan disease. A Canavan disease patient underwent 1H, {1H}-31P, and {1H}-13C MRSat a standard clinical 1.5 T scanner equipped with a second rf channel. Quantitative information is transferable from one assay to the next,greatly enhancing the study. A: Standard MRI to detect anatomical abnormalities and to identify regions or volumes of interest (VOI). B:Sequential 1H MRS of white and gray matter, from 6 to 18 months. C: By {1H}-31P MRS the membrane metabolites PC, GPE, and GPC appearto be reduced. A small reduction in cerebral PCr and ATP was also detected. D: Natural abundance {1H}-13C MRS shows elevated NAAand mI, and reduction of glutamate.


changes in concentration or flux, ontime-scales of minutes rather thanseconds. By applying these three fil-ters at the outset (volume, concentra-tion, and time), the likelihood of asuccessful outcome for the early stud-ies is increased.

As a new technology, MRS is per-haps less bound by rules of “hypothe-sis-driven” research. Witness theunexpected inventions of fMRI, diffu-sion tensors, magnetization transfer,relaxation time, and nuclear-Over-hauser-effect determination. In vivoMRS brings great benefits to neuro-science, not least because studies pre-viously only conceivable in experi-mental animals, become extremelyinviting in man.

ACKNOWLEDGMENTSThanks to Ms. Mary Munoz whotyped the manuscript and Jeannie Tanwho prepared many of the figures. Weare grateful to our colleagues, Drs. Ro-land Kreis, Thomas Ernst, ElseRubæk Danielsen, Kay Seymour,Jong-Hee Hwang, Alex Lin, FrederickShic, Angel Moreno and Cat-HuongNguy for permission to quote recentor unpublished work. The MRS Unitat HMRI received financial supportfrom the Rudi Schulte Research Insti-tute and from NIH and the Board ofHMRI, without which this manu-script would not have been possible.Figures are reproduced with permis-sion. B.D.R. and S.B. are also VisitingAssociates at the California Instituteof Technology.

LITERATURE CITED

Ackerman JJH, Bore PJ, Wong GG, GadianDG, Radda GK. 1980. Mapping of me-

tabolites in whole animals by 31P NMRusing surface coils. Nature 283:167–170.

Badar-Goffer RS, Ben-Yoseph O, Bach-elard HS, Morris PG. 1992. Neuronal-glial metabolism under depolarizingconditions. Biochem J 282:225–230.

Barker PB, Soher BJ, Blackband SJ,Chatham JC, Mathews VP, Bryan RN.1993. Quantitation of proton NMR spec-tra of the human brain using tissue wa-ter as an internal concentration refer-ence. NMR Biomed 6:89–94.

Beckmann N, Muggler T, Allegrini PR,Laurent D, Rudin M. 2001. From anat-omy to the target: contributions of mag-netic resonance imaging to pre-clinicalpharmaceutical research. Anat Rec (NewAnat) 265:85–100.

Bluml S. 1999. In vivo quantification ofcerebral metabolite concentrations us-ing natural abundance 13C MRS at 1.5Tesla. J Magn Reson 136:219–225.

Bluml S, McComb JG, Ross BD. 1997. Dif-ferentiation between cortical atrophyand hydrocephalus using 1H MRS. MagnReson Med 37:395–403.

Bluml S, Moreno A, Ross B. 2001. 1-13Cglucose magnetic resonance spectros-copy of pediatric and adult brain disor-ders. Proc Int Soc Mag Reson in Med (inpress).

Bluml S, Seymour K, Ross BD. 1999. De-velopmental changes in choline and eth-anolamine containing compounds mea-sured with proton decoupled 31P MRS invivo human brain. Mag Reson Med 42:643–654.

Bluml S, Zuckerman E, Tan J, Ross BD.1998. Proton-decoupled 31P magneticresonance spectroscopy reveals osmoticand metabolic disturbances in humanhepatic encephalopathy. J Neurochem71:1564–1576.

Bottomley PA. 1987. Spatial localization inNMR spectroscopy in vivo. Ann NY AcadSci 508:333–348.

Bottomley PA, Foster TB, Darrow RD.1984. Depth-resolved surface spectros-copy (DRESS) for in vivo 1H, 31P and 13CNMR. J Magn Reson 59:338–342.

Bottomley PA, Hart HR, Edelstein WA,Schenck JF, Smith LS, Leue WM, Muel-ler OM, Redington RW. 1983. NMR im-aging/spectroscopy system to study bothanatomy and metabolism. Lancet 1:273–274.

Brody BA, Kinney HC, Kloman AS, GilesFM. 1987. Sequence of central nervoussystem myelination in human infancy:an autopsy study of myelination. J Neu-ropathol Exp Neurol 46:283.

Brown TR, Kincaid BM, Ugurbil K. 1982.NMR chemical shift imaging in three di-mensions. Proc Natl Acad Sci USA 79:3523–3526.

Cady EB, Costello AM, Dawson MJ, DelpyDT, Hope PL, Reynolds EO, Tofts PS,Wilkie DR. 1983. Non-invasive investiga-tion of cerebral metabolism in newborninfants by phosphorus nuclear magneticresonance spectroscopy. Lancet 1:1059–1062.

Christiansen P, Henriksen O, StubgaardM, Gideon P, Larsson HBW. 1993. In

vivo quantification of brain metabolitesby means of 1H MRS using water as aninternal standard. Magn Reson Imaging11:107–118.

Danielsen ER, Henriksen O. 1994. Quanti-tative proton NMR spectroscopy basedon the amplitude of the local water sup-pression pulse. Quantification of brainwater and metabolites. NMR Biomed7:311–318.

Ernst T, Hennig J. 1994. Observation of afast response in functional MR. MagnReson Med 32:146–149.

Ernst T, Kreis R, Ross BD. 1993a. Absolutequantitation of water and metabolites inthe human brain. Part I: compartmentsand water. J Magn Reson 102:1–8.

Ernst T, Lee JH, Ross BD. 1993b. Direct31P imaging in human limb and brain.J Comput Assist Tomogr 17:673–680.

Flogel U, Niendorf T, Serkowa N, Brand A,Henke J, Leibfritz D. 1995. Changes inorganic solutes, volume, energy state,and metabolism associated with osmoticstress in a glial cell line: a multinuclearNMR study. Neurochem Res 20:793–802.

Frahm J, Bruhn H, Gyngell ML, MerboldtK-D, Hanicke W, Sauter R. 1989. Local-ized proton NMR spectroscopy in differ-ent regions of the human brain in vivo.Relaxation times and concentrations ofcerebral metabolites. Magn Reson Med11:47–63.

Frahm J, Merboldt K-D, Hanicke W. 1987.Localized proton spectroscopy usingstimulated echoes. J Magn Reson72:502–508.

Frahm J, Michaelis T, Merboldt K-D,Bruhn H, Gyngell ML, Hanicke W. 1990.Improvements in localized proton NMRspectroscopy of human brain. Watersuppression, short echo times, and 1 mlresolution. J Magn Reson 90:464–473.

Frahm J, Michaelis T, Merboldt KD, Han-icke W, Gyngell ML, Chien D, Bruhn H.1989. Localized NMR spectroscopy invivo: progress and problems. NMRBiomed 2:188–195.

Greenman RL, Axel L, Lenkinski RE. 1998.Direct imaging of phosphocreatine inthe human myocardium using a RAREsequence at 4.0 T. Proceedings of the 6thInternational Society for Magnetic Res-onance in Medicine. Sydney, Australia:1922.

Gruetter R, Adriany G, Merkle H, Ander-son PM. 1996. Broadband decoupled,1H-localized 13C MRS of the humanbrain at 4 Tesla. Magn Reson Med 36:659–664.

Gruetter R, Mescher M, Kirsch J, et al.1997. 1H MRS of neurotransmitterGABA in humans at 4 Tesla. Proceedingsof the 5th International Society for Mag-netic Resonance in Medicine. Vancou-ver, Canada: 1217.

Gruetter R, Novotny EJ, Boulware SD, Ma-son GF, Rothman DL, Shulman GI, Pri-chard JW, Shulman RG. 1994. Localized13C NMR spectroscopy in the humanbrain of amino acid labeling from D-[1-13C]glucose. J Neurochem 63:1377–1385.

In vivo MRS brings greatbenefits to

neuroscience, not leastbecause studiespreviously onlyconceivable in

experimental animalsbecome extremely

inviting in man.


Gyngell ML, Busch E, Schmitz B, KohnoK, Back T, Hoehn-Berlage M, HossmannKA. 1995. Evolution of acute focal cere-bral ischemia in rats observed by local-ized 1H MRS, diffusion weighted MRIand electrophysiological monitoring.NMR Biomed 8:206–214.

Hamilton PA, Hope PL, Cady EB, DelpyDT, Wyatt JS, Reynolds EOR. 1986. Im-paired energy metabolism in brains ofnewborn infants with increased cerebralechodensities. Lancet 1:1242–1246.

Hanefeld F, Holzbach U, Kruse B, Wili-chowski E, Christen HJ, Frahm J. 1993.Diffuse white matter disease in threechildren: an encephalopathy withunique features on magnetic resonanceimaging and proton magnetic resonancespectroscopy. Neuropediatrics 24:244–248.

Hanstock CC, Rothman DL, Prichard JW,Jue R, Shulman R. 1988. Spatially local-ized 1H NMR spectra of metabolites inthe human brain. Proc Natl Acad SciUSA 85:1821–1825.

Heerschap A, van den Berg P. 1993. ProtonMR spectroscopy of the human fetus inutero. Proceedings of the 12th Society ofMagnetic Resonance in Medicine. NewYork: 318.

Hennig J, Pfister H, Ernst T, Ott D. 1992.Direct absolute quantification of metab-olites in the human brain with in vivolocalized proton spectroscopy. NMRBiomed 5:193–99.

Hetherington HP, Newcomer BR, Pan JW.1998. Measurements of human cerebralGABA at 4.1 T using numerically opti-mized editing pulses. Magn Reson Med39:6–10.

Hoang T, Dubowitz D, Bluml S, KopyovOV, Jacques D, Ross BD. 1997. Quanti-tative 1H MRS of neurotransplantationin patients with Parkinson and Hunting-ton disease. 27th Annual Meeting, Soci-ety for Neuroscience. New Orleans, LA.Oct. 25–30: 658.9.

Hoang TQ, Ross BD, Bluml S, et al. 1998.Viability and neural proliferation in hu-man fetal tissue transplants. Proceed-ings of the 6th International Society forMagnetic Resonance in Medicine. Syd-ney, Australia, April 18–24:483.

Hope PL, Costello AM, Cady EB, Delpy DT,Tofts PS, Chu A, Hamilton PA, ReynoldsEO, Wilkie DR. 1984. Cerebral energymetabolism studied with phosphorusNMR spectroscopy in normal and birth-asphyxiated infants. Lancet 2:366–370.

Hossman K-A. 1994. Viability thresholdsand the penumbra of focal ischemia.Ann Neurol 36:557–565.

Hurd RE, Gurr D, Sailasuta N. 1998. Pro-ton spectroscopy without water suppres-sion: the oversampled J-Resolved exper-iment. Magn Res Med 40:343–347.

Kanamori K, Ross BD. 1997. In vivo nitro-gen MRS studies of rat brain metabo-lism. In: Bachelard H, editor. Advancesin neurochemistry, Vol. 8. New York:Plenum Publishing. p 66–90.

Kreis R. 1997. Quantitative localized 1HMR spectroscopy for clinical use. ProgNucl Magn Reson Spectrosc 31:155–195.

Kreis R, Ernst T, Ross BD. 1993a. Absolutequantitation of water and metabolites inthe human brain. Part II: metaboliteconcentrations. J Magn Reson 102:9–19.

Kreis R, Ernst T, Ross BD. 1993b. Devel-opment of the human brain: in vivoquantification of metabolite and watercontent with proton magnetic resonancespectroscopy. Magn Reson Med 30:1–14.

Kreis R, Farrow NA, Ross BD. 1990. Diag-nosis of hepatic encephalopathy by pro-ton magnetic resonance spectroscopy.Lancet 336:635–636.

Kreis R, Farrow NA, Ross BD. 1991. Local-ized 1H NMR spectroscopy in patientswith chronic hepatic encephalopathy.Analysis of changes in cerebral glu-tamine, choline and inositols. NMRBiomed 4:109–116.

Kreis R, Ross BD, Farrow NA, AckermanZ. 1992. Metabolic disorders of the brainin chronic hepatic encephalopathy de-tected with 1H MRS. Radiology 182:19–27.

Kumar A, Welti D, Ernst RR. 1975. NMRFourier zeugmatography. J Magn Reson18:69–83.

Mason GF, Gruetter R, Rothman DL, Be-har KL, Shulman RG, Novotny EJ. 1995.NMR determination of the TCA cyclerate and a-ketoglutarate/glutamate ex-change rate in rat brain. J Cereb BloodFlow Metab 15:12–25.

Merboldt K-D, Bruhn H, Hanicke W,Michaelis T, Frahm J. 1992. Decrease ofglucose in the human visual cortex dur-ing photic stimulation. Magn Reson Med25:187–194.

Merboldt KD, Chien D, Hanicke W, Gyn-gell ML, Bruhn H, Frahm J. 1990. Local-ized 31P NMR spectroscopy of the adulthuman brain in vivo using stimulated-echo (STEAM) sequences. J Magn Reson89:343–361.

Michaelis T, Merboldt K-D, Bruhn H, Han-icke W, Frahm J. 1993. Absolute concen-trations of metabolites in the adult hu-man brain in vivo: quantification oflocalized proton MR spectra. Radiology187:219–227.

Michaelis T, Merboldt K-D, Hanicke W,Gyngell ML, Bruhn H, Frahm J. 1991.On the identification of cerebral metab-olites in localized 1H NMR spectra ofhuman brain in vivo. NMR Biomed4:90–98.

Moreno A, Ross B, Bluml S. 2001. Directdetermination of NAA synthesis rate ofhuman brain in vivo. J Neurochem (inpress).

Mori S, Barker PB. 1999. Diffusion mag-netic resonance imaging: its principleand applications. Anat Rec (New Anat)257:102–109.

Narayana PA, Johnston D, Flamig DP.1991. In vivo proton magnetic resonancespectroscopy studies of human brain.Magn Reson Imag 9:303–308.

Oberhaensli RD, Hilton Jones D, Bore PJ,Hands LJ, Rampling RP, Radda GK.1986. Biochemical investigation of hu-man tumours in vivo with phospho-rus-31 magnetic resonance spectros-copy. Lancet 2:8–11.

Ordidge RJ, Bendall MR, Gordon RE, Con-nelly A. 1985. Volume selection for invivo biological spectroscopy. In: Govil G,Khetrapal CL, Saran A, editors. Mag-netic resonance in biology and medicine.New Delhi: Tata McGraw-Hill. p 387–397.

Ordidge RJ, Connelly A, Lohman JAB.1986. Image-selected in vivo spectros-copy (ISIS). A new technique for spa-tially selective NMR spectroscopy. JMagn Reson 66:283–294.

Pettegrew JW, Minshew NJ, Cohen MM,Kopp SJ, Glonek T. 1984. P-31 NMRchanges in Alzheimer and Huntingtondiseased brain. Neurology 34:281.

Prichard JW, Alger JR, Behar KL, PetroffOAC, Shulman RG. 1983. Cerebral met-abolic studies in vivo by 31P NMR. ProcNatl Acad Sci USA 80:2748–2751.

Provencher SW. 1993. Estimation of me-tabolite concentrations from localized invivo proton NMR spectra. Magn ResonMed 30:672–679.

Ross BD. 1979. Techniques for investiga-tion of tissue metabolism. In: KornbergHC, Metcalf JC, Northcote D, Pogson CI,Tipton KF, editors. Techniques in meta-bolic research, part I. Vol. B203.Elsevier. p 1–22.

Ross BD, Bluml S, Cowan R, DanielsenER, Farrow N, Gruetter R. 1997. In vivomagnetic resonance spectroscopy of hu-man brain: the biophysical basis of de-mentia. Biophys Chem 68:161–172.

Ross BD, Kreis R, Ernst T. 1992. Clinicaltools for the 90’s: magnetic resonancespectroscopy and metabolite imaging.Eur J Radiol 14:128–140.

Ross BD, Narasimhan PT, Tropp J, DerbyK. 1989. Amplification or obfuscation: islocalization improving our clinical un-derstanding of phosphorus metabolism?NMR Biomed 2:340–345.

Ross BD, Radda GK, Gadian DG, RockerG, Esiri M, Falconer-Smith JC. 1981. Ex-amination of a case of suspectedMcArdle syndrome by 31P nuclear mag-netic resonance. N Engl J Med304:1338–1342.

Rothman DL, Petroff OAC, Behar KL,Mattson RH. 1993. Localized 1H NMRmeasurements of g-aminobutyric acid inhuman brain in vivo. Proc Natl Acad Sci90:5662–5666.

Ryner LN, Sorenson JA, Thomas MA.1995. Localized 2D J-Resolved 1H MRspectroscopy: strong coupling effects invitro and in vivo. Magn Reson Imag 13:853–869.

Seymour K, Bluml S, Sutherling J, Suther-ling W, Ross BD. 1998. Identification ofcerebral acetone by 1H MRS in patientswith epilepsy, controlled by ketogenicdiet. (submitted).

Seymour KJ, Bluml S, Sutherling J, Suth-erling W, Ross BS. 1999. Identificationof cerebral acetone by 1H MRS in pa-tients with epilepsy controlled by keto-genic diet. MAGMA 8:33–42.

Shen J, Novotny EJ, Rothman DL. 1998. Invivo lactate and b-hydroxybutyrate edit-ing using a pure phase refocusing pulsetrain. Proceedings of the 6th Interna-


tional Society for Magnetic Resonancein Medicine. Sydney, Australia: 1883.

Stockler S, Hanefeld F, Frahm J. 1996. Cre-atine replacement therapy in guanidino-acetate methyltransferase deficiency, anovel inborn error of metabolism. Lan-cet 348:789–790.

Tallan HH. 1957. Studies on the distribu-tion of N-acetyl-L-aspartic acid in brain.J Biol Chem 224:41– 45.

Thulborn KR, Ackerman JJH. 1983. Abso-lute molar concentrations by NMR ininhomogeneous B1. A scheme for analy-sis of in vivo metabolites. J Magn Reson55:357–371.

Thulborn KR, du Boulay GH, Radda GK.1981. Proceedings of the Xth Interna-tional Symposium on Cerebral BloodFlow and Metabolism.

Van der Knaap MS, Ross BD, Valk J. 1993.Uses of MR in inborn errors of metabo-lism. In: Kucharcyzk J, Barkovich AJ,Moseley M, editors. Magnetic resonanceneuroimaging. Boca Raton: CRC Press,Inc. p 245–318.

Webb PG, Sailasuta N, Kohler S, Raidy T,Moats RA, Hurd RE. 1994. Automatedsingle-voxel proton MRS: technical de-velopment and multisite verification.Magn Reson Med 31:365–373.

Welch KMA. 1992. The imperatives ofmagnetic resonance for the acute strokeclinician (Plenary Lecture). 11th Societyof Magnetic Resonance in Medicine.Berlin: 901.

Young IR. 2000. Methods in biomedicalmagnetic resonance imaging and spec-troscopy. New York: John Wiley andSons.

Zeineh MM, Engel SA, Thompson PM,Bookheimer SY. 2001. Unfolding the hu-man hippocampus with high resolutionstructural and functional MRI. Anat Rec(New Anat) 265:111–120.


APPENDIX

Appendix:How to Make MR Spectroscopy Measurements

APPLIED MRS-SINGLE-VOXEL1H MRS

When expertly performed, quantita-tive MRS brain studies provide re-sults with a variance between 5 and10% across large populations of nor-mal subjects. The variance in singlesubjects may be as little as 63%, al-though measurements of complex-coupled peaks do not achieve betterthan 610 –15%. The single greatestvariable is not, we believe, true bio-logical or diet-imposed variabilitybut inaccuracy in positioning the pa-tients’ head or the VOI of localizedMRS.

Below the reader can find a fewsuggestions on how to set up single-voxel proton MRS on a clinical scan-ner. These suggestions are heavilybiased by the authors’ experiencewith a General Electric 1.5 T Signascanner. Nevertheless, the followingsection may be useful for users onany system because the underlyingprinciples are the same on any scan-ner. Focusing for the moment on 1HMRS, simple protocols with fixedecho time (TE) and repetitive time(TR) are best embedded in “macros,”to prevent acquisition errors and torigorously standardize techniquesthat may be required for longitudi-nal studies over many years. STEAM20 ms (MPI, Gottingen), STEAM 30ms (HMRI, Pasadena) and PRESS30 ms have been “safe” choices andcan be recommended. STEAM (orPRESS) 135 ms and 270 ms (or morerecently 144 ms and 288 ms) are sat-isfactory long-echo times. Short-echo times TE give the smallest T2

losses and therefore the best signal-to-noise (S/N) ratio and also thesmallest susceptibility to T2 changesin pathology. Background signals,

however, may cause considerablebaseline distortions at extremelyshort TEs. To ensure reliability andquality we recommend performing atest as described below for single-voxel proton MRS based on axial lo-calizers.

IDENTIFYING THE VOLUME OFINTEREST (VOI)

A few simple preparations should bedone before starting the examination.We recommend to obtain a film withaxial T1 weighted images and to try to

Figure 1S. MRI for voxel placement.

80 THE ANATOMICAL RECORD (NEW ANAT.) APPENDIX

identify standard grey matter (GM)and white matter (WM) locations asused in the literature (see Fig. 1S anddescription below for HMRI defini-tion). Mark the voxels with a pen onthe film. Do the same on a set of T2-weighted images. Accuracy in pre-scribing the voxel is of great impor-tance! Remember, the slice thicknessof an MRS voxel is typically 20 mmwhereas the slice thickness of MRI istypically 5 mm. Therefore slices belowand above the center slice for MRSneed to be reviewed. Also check slicesoutside the range of the MRS voxel forbeing not too close to the skull. Con-

firm the voxel location with the fol-lowing prescription:

Gray Matter

Center across falx at level 1 cm aboveposterior commissures of corpus cal-losum. Start inferior one slice (5 mm)above the slice with the a) internalcapsule, b) angular artery in sylvianfissure, c) occipitoparietal fissure, d)vein of Galen, e) internal cerebralvein, f) frontal horn of lateral ventri-cle. Note: If the head is tilted not all ofthese landmarks may be visible. Bychoosing a voxel shape of 27 mm inA/P and 21 mm in R/L, most of the

tissue within the voxel is grey matter.The recommended slice thickness (S/Idimension) is 20 mm. Review aboveand below center position at least twoslices (for 5 mm MRI slice thicknessand 20 mm voxel slice thickness).

White Matter

Center left (or right) in parietal cortex.Stay in largely white matter, but allowup to 25% of grey matter. Landmarksare the center posterior rim of left(right) lateral ventricle ;1–1.5 cmabove posterior commissures of cor-pus callosum. Review above two and

Figure 2S. Positioning subject for MRS ofbrain.

TABLE 1S. Locations for single-voxel MRS in various diseases

Diagnosis Preferred location Information needed

Global hypoxia Gray matterTrauma Far away from blood/lesions. Gray matter (1st

choice) to rule out hypoxic injury, white matter is2nd choice. If there is no suspicion of hypoxicinjury white matter is 1st choice.

Anticonvulsants?Conscious/unconsciousDate of Injury?

Stroke (chronic) Center (1st choice) and rim (2nd choice) Date of CVALiver disease; hepatic

encephalopathyGray matter (1st choice), however earliest

changes in liver disease are in white matter (2ndchoice, Cho reduction)

Lactulose? Neomycin?

Dementia Gray matter Clinical Dx? Symptoms?MS No evidence of lesions: White matter (1st choice).

Lesions: Lesion (1st choice) and contra lateralside (2nd choice)

HIV Lesions: Lesion (1st choice), contra lateral side(2nd choice). No lesions: White matter

Medication, clinicaldiagnosis of lesionAIDS 1, CD-4

HIV (AIDS dementia) Through lesion/if no lesion gray matter: as abovedementia

Tumors, rule out tumors Center of lesion suspicious region, sometimessmaller voxel necessary (adjust (increase)number of total scans); contralateral side ascontrol. Repeat rim of lesion.

Type of tumor, chemo-,radiation therapy,Post-contrast o.k.

Unknown not focaldisease

Gray matter

APPENDIX THE ANATOMICAL RECORD (NEW ANAT.) 81

below at least two slices (for 5 mmMRI slice thickness and 20 mm voxelslice thickness). The S/I center posi-tion should be not more than 0–5 mm(superior) off from the grey mattercenter.

There is an element of subjectivityand “in-house” training in voxel place-ment. Landmarks are defined on thesubject’s MRI. Voxel dimensionsshould be held constant for a givenlocation. Here we encounter the firstserious variable, which is the size ofthe head. Dimensions are tailored ac-cordingly. Signal from outside theVOI is included in the final spectrum,because to the extent that pulse im-perfections occur, the limits of the

VOI noted on the screen are inaccu-rate. The skilled spectroscopist avoidsthe worst problem, excess lipid signal,by selecting a VOI 5-7 mm away fromthe outer borders of the brain.

Finally, the best results should notdepend on endless compliance of thesubject. Protocols that include bothMRI and MRS would do well to com-plete MRS and scout MRI if possiblebefore, rather than after the bulk ofthe MRI protocol.

Subject Positioning

Plan the first tests at a time whenthere is no pressure for fast work. Thesubject is always supine. The RF head

coil center is always “landmarked”within 1 mm of the same position onthe subjects’ head. The head is accu-rately positioned in all three planes.Sagittal: bore center at the mid-pointof brow, nose and chin. Coronal (thecommonest source of error) is se-lected only after defining the angle atwhich the subject is lying. Use a rulerto define the distance tip of chin tosternal notch and record this value forfuture MRS examinations. Axial: se-lect a bony landmark, usually the su-praorbital ridge, and drop a verticalthrough two other landmarks, sayouter angle of the eye and tragus ofthe ear lobe (Fig. 2S). For even themost trivial procedure, it is recom-

Figure 3S. Stability and reproducibility of sin-gle-voxel MRS in controls and patients. Re-peated single-voxel MRS (STEAM, TE 5 30ms, TR 5 1.5 s, TM 5 13.7 ms) in healthycontrols and seventeen single-voxel MRScarried out in a patient with clinically sus-pected Alzheimer disease over a period of13 months. All spectra can be distinguishedfrom controls by elevated mI/Cr and re-duced NAA/Cr, a feature of probable Alz-heimer disease.


mended that the head be fixed by Vel-cro straps. If an external reference isused for quantitation, it is importantfor automation and reproducibilitythat the vial is placed at the same po-sition in all experiments.

Data Acquisition

Perform axial MRI and display imageappropriate for center position. After

the scout MRI, angulation errorsshould be noted. Unless these are se-vere, the MRS can proceed.

MRS is prescribed in different wayson different instruments. They allhave in common, however, the need todefine coordinates orthogonal to themagnet bore (this explains the need tominimize angulation errors in initialpositioning of the patient; they cannot

be accurately corrected despite theavailability of all the desired land-marks on MRI).

Start with either a standard greymatter or white matter location. Donot forget how important it is to checkthe location in all slices covering thevolume of interest (VOI). The MRIslice thickness is usually 5 mm,whereas the MRS slice thickness is 20mm. Write down voxel position andvoxel size.

The next step is the shimming andthe adjustment of the scan parame-ters. A well-shimmed VOI is a prereq-uisite for good MRS, as resolution andlineshape have significant impact onthe accuracy of the quantitation. Mostscanners provide automated shim-ming that is generally faster thanmanual shimming and is the methodof choice. Automated adjustment ofRF transmitter gain, receiver gain,transmitter frequencies is standard onall modern scanners. Also the adjust-ment of the water suppression is au-tomated on most systems. On a GEscanner using the PROBE™ (5 PRO-ton Brain Exam), all of the abovesteps are fully automated and can bedone by pushing one button. Acquirethe spectrum and print the spectrumor document it on a film. Save thespectrum file for later off-line process-ing.

Processing and Quantitation

Determine peak ratios and comparethe spectrum with literature spectra.Stability and reproducibility are thekey to successful longitudinal studiesin neurophysiology and neuropathol-ogy. Figure 3S shows the resultsachieved in single subjects, volunteersand a patient with probable Alzhei-mer disease, illustrating the robust-ness of 1H MRS over a 13-month pe-riod in normal and diseased humansubjects. Dramatic neurochemicalevents, with a time constant of weeksor months, can also be readily ob-served (Ross and Michaelis, 1994).Two examples are shown in Figure 4Swhere 1H MRS monitors the restora-tion of biochemical abnormalities af-ter liver transplantation, and the ap-pearance of ketone peak in the brainof a epileptic patient undergoing a ke-togenic diet treatment (Seymour etal., 1999).

Figure 3S. (Continued)

APPENDIX THE ANATOMICAL RECORD (NEW ANAT.) 83

Where to Place the Voxel inClinical Exams

The clinical question determines voxellocation and size. In global diseasesstandard locations should be selected.The decision whether grey or whitematter depends on the questionasked. For example to predict out-come in head trauma, it is most im-portant to rule out global hypoxic in-jury. For this question the grey matterlocation is more sensitive and wouldbe the first choice. In tumors the voxelshould be placed in the center of the

suspicious region minimizing partialvolume with apparently normal tis-sue. In case of a small lesion two vox-els at the same center position butwith different volumes can be mea-sured to estimate partial volume.There is some controversy about MRSafter contrast agent. Because forshort-echo time MRS in particular,there is no evidence for a significantimpact of contrast agents on the spec-tral quality, the improved informationabout the region of interest after con-trast agent may be favorable whenMRI from a separate examination

with contrast agent is not available.For suggestions where to place thevoxel in a clinical situation see Ta-ble 1S.

LITERATURE CITED

Ross BD, Michaelis T. 1994. Clinical appli-cations of magnetic resonance spectros-copy. Magn Reson Quarterly 10:191–247.

Seymour, K, Bluml S, Sutherling J, Suth-erling W, Ross BD. 1999. Identificationof cerebral acetone by 1H MRS in pa-tients with epilepsy, controlled by keto-genic diet. MAGMA 8:33–34.

Figure 4S. The potential of MRS in diagnosis and treatment monitoring. A: Restoration of biochemical abnormalities of the brain post-livertransplant. The patient is a 30-year-old man with acute-on-chronic hepatic encephalopathy (HE) secondary to hepatitis and subsequentlysuccessfully treated by liver transplantation. Spectra were acquired 6 months apart from the same parietal white matter location, (15.0 ccstimulated-echo acquisition mode (STEAM) TR 1.5 s, TR 30 ms; NEX 128) and scaled to the same creatine (Cr) intensity for comparison. Theobvious abnormalities before liver transplantation; increased a, b, and g-glutamine; and reduced choline (Cho)/Cr and myo-inositol(mI)/Cr (upper spectrum) were completely reversed 3 months after transplantation and Cho/Cr exceeded normal (lower spectrum). B:Comparison of 1H MRS in pre- and post-ketogenic diet in the same child. A 1H MR spectrum from occipital grey matter of a patient beforeinitiation of ketogenic diet reveals normal proton MRS (middle trace). Four days after starting the diet the 1H spectrum acquired at the samelocation shows the presence of a single peak at 2.2 ppm (lower trace). The difference spectrum is shown in the upper trace (Seymour etal., 1999).


Magnetic resonance spectroscopy of the human brainlphe.epfl.ch › ~mtran › phy_gen_III › Cours...

Documents

Transcript of Magnetic resonance spectroscopy of the human brainlphe.epfl.ch › ~mtran › phy_gen_III › Cours...