J. Fuiham,

Michael Ramesh Andrew

Raman,

J.

FRACP #{149} Alberto Bizzi, MD #{149}Mark J. Dietz, MD #{149}Henry #{149} Geoffrey S. Sobering, PhD #{149}Joseph A. Frank, MD MD #{149} Jeifry R. Alger, PhD #{149}Giovanni Di Chiro, MD

Dwyer,

Mapping ofBrain with Proton MR Imaging: Clinical Brain

metabolism was studied magnetic resonance spectroscopy and positron emission tomography with fluorine-i8 fluorodeoxyglucose in 50 patients. N-acetylaspartate (NAA) was generally decreased in tumors and radiation necrosis but was somewhat preserved at neoplasm margins. Choline was increased in most solid tumors. Solid high-grade gliomas had higher normalized choline values than did solid low-grade gliomas (P < .02), but the normalized choline value was not a discriminator of tumor grade, since

U

recently, the accepted neuroradiological evaluation of brain tumors has included determination of abnormal anatomic features and in-

tumor

high-grade

lesions

had

terms:

tegrity

Radiology

185:675-686

by

reprint

requests

Throughout this article intensity of each metabolite RSNA, 1992 2

barrier

response

with

of tumors

regimens

experienced

ducion

clinicians.

(MRS) signals

mal

brain

(i-3)

has

The

intro-

MR spectro-

techniques of metabolites

and

to

are recognized

of hydrogen-i

scopic of MR

for

intracranial

tantalized

detection in nor-

tumors

investigators

with

the possibility that they might provide novel MR indicators of tumor metabolism that could be useful in clinical management (4-9). Alterations in N-acetylaspartate (NAA), cholinecontaining

compounds,

phosphocreatine,

MD

MRS imaging, spectroscopic ments are far less influenced

measureby par-

tial

single-

volume

effects

than

voxel techniques. In the present 50 brain

tumor

imaging.

Our

with

study,

we evaluated

patients

aims

with

were

the following hypotheses: choline2 reflects degree malignancy, (b) lactate degree of malignancy,

of tumoral correlates with and (c) radia-

tion necrosis is characterized by decreased choline. These hypotheses are relevant to whether H-i MRS imaging is useful

clinically

tumors,

differentiation

in grading

of brain

of recurrent

tumor from radiation necrosis, and monitoring of the effects of therapy. As in our previous report (2), we also

used

positron

emission

tomography

with fluorine-i8 fluorodeoxy(FDG) (i7,i8) to measure glucose metabolism.

(PET)

glucose moral

and

alanine

MATERIALS

AND

tional Institutes of Health (NIH). consent was obtained from each

aging,

before

in which

two

of the

(11-13)

spatial

gradient-based is used

phase to encode

dimensions,

repre-

sent a major improvement. These methods permit the simultaneous acquisition of spectra from a large number of voxels within a slab of tissue and the display of these data in a tomographic format with a volume resolution of about 1 cm3 at an echo time (TE) of 272 msec (14-16). With H-i

10 (M.J.F., A.B., M.J.D., R.R., J.R.A., G.D.C.); the Diagnostic Radiology Research Program and the Department of Diagnostic Radiology of Health, 9000 Rockville Pike, Bethesda, MD 26; revision received July 15; accepted July 27.

to G.D.C. the designations choline, NAA, creatine, at a TE of 272 msec, which is influenced

and

tu-

METHODS

been reported in brain tumors (1-4). These studies, however, mostly used single-voxel techniques, which are limited by partial volume effects (2,4,10). Spectroscopic imaging methods, referred to here as H-i MRS imencoding

MRS

H-i

to investigate (a) elevated

creatine-

lactate,

have

I From the Neuroimaging Branch, Rm 1C451, Bldg National Institute of Neurological Diseases and Stroke, (H.H.L.S.); the in vivo NMR Research Center (G.S.S.); Q.A.F., AID.), Clinical Center; the National Institutes 20892. Received June 2, 1992; revision requested June

Address

blood-brain

and

therapeutic

Brain, effects

1992;

of the

acterislics

re-

of irradiation on, 13.47 #{149} Brain neoplasms, CT, 13.1211 #{149} Brain neoplasms, MR. 13.1214 #{149} Brain neoplasms, therapeutic radiology, 13.47 #{149}Emission CT, 13.1211 . Magnetic resonance (MR), spectroscopy, 13.1219

NTIL

computed tomography (CT) and magnetic resonance (MR) imaging and of tumor vascularity with angiography. The shortcomings of these modalities in assessment of tumor biologic char-

duced choline values. Serial studies in one case showed an increase in choline as the glioma underwent malignant degeneration. Choline values were lower in chronic radiation necrosis than in solid anaplastic tumors (P < .001). In two cases studied before and after treatment, clinical improvement and a reduction in choline followed therapy. Lactate is more likely to be found in high-grade gliomas, but its presence is not a reliable indicator of malignancy.

Index

Shih,

Tumor Metabolites Spectroscopic Relevance’

with hydrogen-i

necrotic

H.-L.

MD

lactate

refer

to the

by T2 and concentration.

signal

Subjects All patients cols

were

approved

by

studied

under

committees

participation

in the

were

Eight

patients

performed

underwent

Na-

Informed patient

studies.

We performed 64 H-I MRS studies in 50 patients. Multiple studies

proto-

at the

imaging imaging

in 11 patients:

two and three

patients underwent three studies. All patients underwent clinical MR imaging and FDG PET examinations. For patients who underwent multiple H-I MRS imaging studies, repeated FDG PET was not always

performed.

23 women 18-76 years).

There

(mean

Abbreviations: tion field, FDG

were

age, 43 years;

The

clinicopathologic

27 men

and

age range, fea-

B,

= constant magnetic inducfluorodeoxyglucose, GRE = gradient echo, CUR = glucose utilization rate, MRS = magnetic resonance spectroscopic, NAA = N-acetylaspartate, NIH = National Institutes of Health, PET = positron emission tomography, ROI = region of interest, SE = spin echo, TE = echo time. =

675

Table 1 Clinical and Pathologic

Features

of Patients Time

Patient No./ Age/Sex 1/27/F 2/33/M

Site Studied

LG astrocytoma Oligo

Frontal Parietal

L L

3/60/M

Frontal

L

4/51/F

Bilateral

5/50/M 6/34/F

Frontal L Frontoparietal

7/29/M 8/36/F

Frontotemporal Parietal L

frontal

11/35/F 12/39/F 13/29/F

Occipital Temporal Pons

14/39/M

Frontotemporal

15/18/F 16/45/Ft 17/56/Ft 18/40/M

thalamus Parietal L Frontoparietal Temporal L Parietal L

19/34/M

Temporal

20/33/M

Bilateral

21/29/F 22/28/F

Frontal Frontal

23/55/Ft

Bilateral frontal thalamus Temporal L Frontal L

27/52/F 28/30/M 29/56/M 30/48/F

NA

NA

LG astrocytoma LG astrocytoma

NA

Gd+,

9, 14 NA

Hemo NA

NA 13, 16

L,

Gliomatosis

cerebri

NA 32, 38 12, 18

48

Method of Diagnosis S biopsy S biopsy

Gd+

Open

Treatment XRT NT

biopsy

(60 Gy)

NT

NA NA

S biopsy S biopsy S biopsy S biopsy Surg resection

NT XRT NT NT XRT

Hemo NA

S biopsy Partial surg resection

XRT (60 Gy) NT

Hemo Gd+ NA

S biopsy NA NA

NT XRT XRT

Gd+

Partial

hemo,

Ca

surgical

(56 Cy)

(60 Gy)

(72 Gy) (50 Gy)

XRT (45 Gy)

resection L

L L

Presumed Presumed Presumed Presumed

LG LG LG LG

Presumed

LG glioma

Ana

glioma glioma glioma glioma

L,

L

Ana

oligo

Ana Ana

astrocytoma astrocytoma

Mixed ana astrocytoma-oligo Ana

L

L

Frontotempcral Occipital L

L

NA NA NA NA

astrocytoma

Gd+ NA NA NA

NA

NA

4

Gd+,

astrocytoma

Ana astrocytoma Ana astrocytoma

L L

Temporal

18 50

LG astrocytoma Pontine astrocytoma Pontine astrocytoma

frontal

Frontoparietal Parietal

Oligo

LG astrocytoma LG astrocytoma L L

MR Imaging Findings NA Mm

Oligo Mixed LG astrocytoma-oligo

L L

from

Treatment (mo)* 15, 23, 30 NA

LG astrocytoma L

Parietal Frontal

26/36/Ft

L

L

9/64/M 10/24/M

24/44/F 25/51/M

Pathologic Diagnosis

NA NA NA NA

NT NT NT NT

NA

NT

hemo

S biopsy

XRT

NT XRT (60 Gy)

(59.4

Gy)

NA 14

Gd+ Gd+,

hemo

Surg resection Partial surg resection

53, 3, 6

Gd+,

Ca

S biopsy

XRT

(60 Gy)

NA 22

Gd+, Gd+,

hemo hemo

Surg resection S biopsy

NA Inter

brachy,

S biopsy

XRT (60 Gy) IA chemo

S biopsy Partial surg resection

XRT (59 Gy) XRT (60 Gy)

Surg Open

resection biopsy

NT NT

Surg

resection

IA chemo,

31, 37, 0.2, 5 77, 82

Ana astrocytoma

48, 52

Ana astrocytoma Mixed ana

NA NA

Gd+,

hemo

Mm Gd+, Mm Gd+

hemo

NA Ca4

astrocytoma-oligo 31/35/M

Temporooccipital

L

Ana

astrocytoma

49

Gd+,

hemo

XRT (60.4 Gy), 32/53/M

Frontal

33/40/F

Frontotemporal

L

Ana

34/44/M 35/47/M

Frontotemporopanetal Temporoparietal

36/50/M 37/56/M

Frontotemporal

L

Ana oligo

Frontotemporal

L

Ana

38/41/M

Bilateral

L

39/40/M 40/57/F 41/52/M

L L L

astrocytoma

Ana astrocytoma Ana oligo Ana

astrocytoma

80

Hemo,

surg

clip

Partial

surg

45

Gd+,

hemo,

surg clip

Partial

surg resection

XRT (60 Gy)

65 51

Gd+, Gd+,

hemo, hemo,

Ca surg

Partial Partial

surg surg

resection resection

Partial

surg

resection

surg

resection

XRT (62 Gy) XRT (50 Gy), IA chemo NT XRT (60 Gy) XRT (56 Gy), system chemo XRT (57 Gy) Inter brachy XRT (72 Gy), surg resection

1

Gd+,

32

Mm

Ana astrocytoma

29

Gd+,

Frontotemporal

Ana

15

Mm

Frontal

Ana astrocytoma

12

Gd+

24

Gd+,

2 5

Gd+ Gd+

frontal

L

Temporoparietal

L

astrocytoma

astrocytoma

Glioblastoma

multiforme

Glioblastoma Glioblastoma

multiforme multiforme

43/46/M

Temporal L Frontoparietal

44/25/M 45/54/Ms

Frontal Bilateral

frontal

L

Glioblastoma Rad necrosis11

multiforme (5CC)

29 15

Gd+, Gd+

46/76/Fl 47/30/M*

Bilateral frontal Frontotemporal

L L

Rad Rad

(AdCa)

96 6

NA Gd+

48/68/M*

Frontoparietal

49/40/M 50/57/F

Frontotemporal Frontal L

42/70/M

L

L

necrosis necrosis)’

(ana astrocytoma) Rad necrosisi (psoriasis)

L L

Lymphoma Breast met

160

29 12

hemo Gd+,

hemo Gd+ hemo

hemo

hemo

clip

resection

system chemo XRT (55 Gy)

S biopsy

Partial S biopsy S biopsy S biopsy

S biopsy Surg resection

XRT (70 Gy) XRT (60 Gy)

Partial surg resection Postmortem examination NA Surg resection

XRT XRT

(69 Gy) (66.35 Gy)

XRT XRT

(80 Gy) (69 Gy)

XRT (dose unknown) XRT (60 Gy) XRT (70 Gy)

Gd+,

hemo

S biopsy

Hemo Gd+,

hemo

S biopsy Surg resection

Note.-AdCa = adenocystic carcinoma of the hard palate, Ana = anaplastic, Ca = calcifications found at CT, Chemo = chemotherapy, Gd+ = enhancement with gadopentetate dimeglumine, Hemo = hemosiderin, IA = intraarterial, Inter brachy = interstitial brachytherapy, L = lobe, LG = low-grade, Met = metastasis, MAn = minimal, NA = not applicable, NT = not treated, Oligo = oligodendroglioma, Rad = radiation, S = stereotactic, SCC = squamous cell carcinoma of the frontal sinus, Surg = surgical, System = systemic, XRT = cranial irradiation. * Time in months from last treatment to each H-i MRS imaging study; for example, patient I underwent three H-I MRS imaging studies performed 15, 23, and 30 months, respectively, after completion of radiation therapy. t Has undergone two H-i MRS imaging procedures but has not been treated. Studied before and after therapy. § Primary diagnosis is in parentheses. Radiation necrosis was confirmed pathologically.

676

Radiology

#{149}

December

1992

tures, the regions imaging, interval

(if any) features

examined from the

to the H-i at clinical

MRS imaging MR imaging,

and

of treatment

diagnosis, in Table

type

1. Pathologic

obtained

reotactic

tion

of gliomas

this

paper,

into

just

of

the

groups:

low-grade

and

classifica-

(20), so for

divided

tumors,

on

or ste-

The

is problematic

three

listed

based

resection

in 42 cases.

we

anaplastic

are

of

classification examination

at surgical

biopsy

MRS

study, method

diagnosis,

the Kernohan four-tiered (19), was made from the tissue

with H-i last treatment

proved

cases

tumors,

glioblastoma

multi-

formes cerebri

(21). The single case of gliomatosis was considered separately. Seven patients had clinical histories, neurologic signs, and neuroradiological findings compatible with a diagnosis of low-grade tumor but did not have pathologic confirmation.

Two

diological mas;

had

typical

findings

the

five

ra-

of presumed low-grade glioma. tients showed effects of radiation In three, a diagnosis of radiation

Four padamage. necrosis

confirmed

histologically.

The

in each of these patients 1. One patient had been

other

than

gliomas

is listed treated

to determine

if

they exhibited different patterns of metabolites. Most patients had been treated with combinations of surgery, irradiation, and chemotherapy, treated.

but

15 had

not

was

been

Imaging

Pulse

Sequence

The point-resolved spectroscopy-chemical shift imaging pulse sequence used to obtain the H-I spectroscopic images has been described recently in detail by Moonen et al (14). This sequence is similar to that used by Luyten et al (15) and Herholz et al (16). It employs three sectionselective pulses that cause the spins in the volume located at the intersection of the three selected sections to form a spin echo at the chosen TE. Section selection widths are narrow in the superior-inferior dimension

and

wide

in the

left-right

and

antero-

the spin echo (SE) arises from an axial slab of tissue. Unwanted signals from lipids within the diploic space are partially suppressed posterior

by

dimensions,

section

selection

so that

in the

anteroposterior

and left-right dimensions. A TE of 272 msec was used, because any lactate signal present is completely rephased in an SE acquisition nals from

at this TE, and water and lipid

manageable spectra

in long are

also

Volume

185

TE spectra.

more

automated fashion, fewer overlapping behaved baselines.

easily

sigmore

Long

analyzed

TE in an

because they contain signals and have well-

Number

#{149}

unwanted become

studies

20 mm

chosen

for the field of view

and section thickness. This nominal volume serves only as an approximation to the actual volume that may contribute to the generation of a given signal, as is the for

all Fourier

imaging

methods.

A

more exact specification requires a definition of what constitutes a significant contribution to signal generation; the actual volume may therefore be larger or smaller than the nominal volume given alone. The water signal is suppressed prior to the localization pulses by using frequency-selective radio-frequency and gradient pulses (22).

tories,

ages

Cedar Knolls, were obtained

studies

were

NJ) clinical MR imin all patients. Most T (Vista

HP;

Picker, Highland Heights, Ohio), and some were performed at 1.5 T (Signa, Medical Systems). These studies were

GE per-

formed,

performed

with

MR

Data

Collection

The MRS data

were

collected

with

a

1.5-T unit Milwaukee)

(Signa; GE Medical Systems, equipped with shielded gradients and a quadrature head coil. After the patient was positioned in the unit,

gradient-recalled

echo

(GRE)

at 0.5

the orientation

lions in the same second appointment prior to or in the MRS imaging.

Imaging

of the sec-

plane

as for PET, at a at least 72 hours afternoon following H-i

Data

Processing

Raw MRS imaging data (an array of 1,024 SEs) were transferred to a Sun workstation (Sun Microsystems, Mountain View, Calif) and reconstructed by using software developed at our institution (14). The k-space data array (1,024 time domain points, 32 x 32 spatial k-space domains) was baseline corrected, and the spatial kspace domain was filtered by using a sine-

bell function.

A two-dimensional

Fourier

transform was then applied to the spatial domain to produce a 32 x 32 array contaming the time domain data from the volume elements. Residual water signals were

filtered

from these

time domain

data

by using a previously described algorithm (23,24). Zero filling of the time domain in each element to 2,048 points and Fourier transformation yielded a 32 x 32 array of spectra. Further analysis was performed on magnitude-corrected

MRS

spectral

data.

It was

necessary to bring the frequency axes of the various spectra into registration because of B0 variation over the imaged region. This was done by using a combina-

tion of automated

and interactive

algo-

rithms

for identifying the location of the NAA, choline, or other signals in the relevant spectra. Spectra in which these signals could not be identified (eg, outside the selected volume and voxels located on or near the skull surface) were nulled so that these data did not appear in the final spectroscopic images. The magnitude of each remaining spectrum was computed, and the signal strength 0.1 ppm on each side of the choline, creatine, NAA, and lactate signal positions was integrated to produce four 32 x 32 arrays showing spa-

(600/30 [repetition time msec/TE msec]; flip angle, 10#{176}) and/or SE data were collected. The axial section location for the H-I MRS image acquisition and the extent of delimitations in the left-right and anteroposterior directions were chosen by using MR imaging and PET data. If the z-axis extent of the lesion was greater than the thickness of the H-i MRS imaging seclion, the MR images and PET scans were useful aids in selection of the section considered to best reflect the nature of the lesion. This policy meant that in some cases, regions of hemorrhage and calcification were included in the section of interest. The constant magnetic induction field

interpolated to 256 x 256 and transferred to a Macintosh computer where they were displayed in color and subjected to further region of interest (ROI)

(B0) homogeneity

analysis

by

the NIH Services

(Image, Branch,

over

the chosen

slab

was adjusted by using an automated routine provided in the Signa research package. The necessary transmitter level, signal gain, and water suppression adjustments were made. The H-I MRS imaging data were collected by using a repetition time of 2,000 msec and one acquisition per phase-encoding step (34.1-minute acquisition time). The entire examination was usually completed in 70 minutes or less. TI-weighted (600/16), T2-weighted (2,583/40,

um-enhanced

3

in most

15 mm;

and

multiplanar MRS

thickness

12 mm was used in two studin four. Thus, the “nominal volume” (defined as section thickness x field of view/resolution) from which the spectra arose ranged from 0.675 to 2.0 cm3 (with the majority at 0.8 cm3), depending ies

case

primary

for unilateral scalp psoriasis with local irradiation (dose unknown) delivered in an uncontrolled fashion. Single cases of primary cerebral lymphoma (not associated with acquired immunodeficiency syndrome) and cerebral metastasis were confirmed pathologically. These two cases were included as examples of brain tumors

The section

on the values

as cases

diagnosis in Table

were

and

astrocyto-

included

was

other

clinical

of pontine

Many additional gradient pulses are incorporated into the pulse sequence to suppress unwanted echoes arising from outside the selected volume. Phase-encoding gradient pulses (11-13) are included in the sequence to permit the determination of the location of the signal sources within the axial plane. Phase encoding is performed with a resolution of 32 x 32 over a field of view of either 24 or 32 cm. Thus, the sequence ultimately yields an array of 1,024 H-I spectra, where each spectrum is produced by the nuclei in a small volume element within the selected slab.

100),

and

Ti-weighted

(Magnevist,

gadolini-

Berlex

Labora-

tial variation of the strengths of the signals in these regions. These metabolite maps were

using

software

version National

Mental Health, NIH). the mulliplanar GRE into this environment lation of the H-i MRS software

permits

ROIs

developed

at

1.35; Research Institute of

Relevant sections of data were imported for anatomic correimaging data. This drawn

on one

map

to be transferred to the identical location on other maps and onto the multiplanar GRE images. The metabolite maps were first analyzed qualitatively. In the region of the lesion, increases or decreases in the signal inten-

Radiology

677

#{149}

Table 2 Lesions with

Lactate:

Site of Lactate

and Relation

to Glucose

Metabolism

0.1

Metabolic Characteristic 0

Patient

Pathologic

Site of

No.

Diagnosis*

Lactate

0.

z

:

8

8

0.2

0

-

LESION

Figure

1.

values

Scatterplot

AR

=

for each

TYPE

of normalized

type.

Data

necrotic

anaplastic

glioma

with

L & Met

=

lymphoma

and

metabolite

were

as lactate.

breast

neme-

verified

were

GRE, PET, and clinical also noted. analysis for choline and performed with a large ROI that was placed in the

(size

and

measured,

shape)

2.20

31 NA NA 65 1 i2

Hyper Hyper Hyper Hyper Hyper Hyper

1.96 2.78 2.67 4.29 6.76 5.14

24

Hyper

rim

4.25

2 5 29 6 160

Hyper Hyper Hyper Hypo Hypo

rim

3.74 3.75 3.83 1.89 0.53

part of tumor part of tumor

of tumor of tumor

part part

and solid part

cyst

surgical

cyst

Necrotic

of tumor cyst

Solid part of tumor Solid part of lesion and lat-

3.97 0 1.18 2.50 3.01 3.62 4.85

Hypo

Hypo Hyper Hyper Hyper

rim

Hyper

rim

rim

ventricle

Note-Only

the initial study is listed. oligodendroglioma, LGA = low-grade A = mixed anaplastic astrocytoma-oligodendroglioma, glioblastoma multiforme, RN = radiation necrosis. t Indicates the time in months from last treatment : Hetero = predominantly hypometaboliclesion per = hypermetabolic, hypo = hypometabolic. 4

Oligo

=

astrocytoma,

AA

=

anaplastic

astrocytoma,

mixed

AO = anaplastic oligodendroglioma, GBM NA = not applicable. to the initial H-i MRS imaging study. with focal areas of increased glucose utilization,

=

hy-

by

and a control value was obtained by placing the same ROI in the corresponding area in the opposite normal hemisphere. dimensions

Hyper

eral

mine

administration.

these

lesions

lesions, that ily including

We chose

were

to analyze

in the same way as the solid is, with a large areas of tumor

ROI necessarand necrosis,

rather than with a much smaller ROI placed on only part of the lesion. The necrotic anaplastic tumors are thus pre-

a marked

ROl

22

Necrotic

GBM RN RN

nals as exhibiting lactate. The location of abnormal metabolite signals with respect

was

cyst

Solid part

Apparently

intensity

A

GBM CBM

as a separate

mean

cyst

cyst Necrotic

group.

the

criteria

Necrotic Necrotic

Cyst Solid Solid Solid Solid

GBM

sented

lesion;

A

Hetero

Normalized CUR

NA 13 50 4 NA 14 NA

cavity

of tumor Necrotic and/or

used to avoid designating regions of the brain that were contaminated by lipid sig-

to the multiplanar MR images was Quantitative NAA maps was circular or oval

These

Solid part of tumor Solid part of tumor Necrotic cyst Solid part of tumor

of Tumor at PETS

=

consultation with hard copies of the relevant parts of the 32 x 32 spectral arrays. Lactate was deemed present by the appearance of a doublet at 1.3 ppm with a line separation of 7-8 Hz. The doublet was not always completely resolved, but the valley separating the peaks had to be at least 30% of the height of the doublet to be accepted

LGA AA AA AA AA AA Mixed AA Mixed AO AO AA

42 43 44 47 48

tastasis.

sity of each

Surgical

of

a rim

solid anaplastic glioma, GBM multiforme, RN = radiation

=

Solid part of tumor

Mixed

41

normalized NAA values and site for each patient

tissue, AS glioblastoma crosis,

lesion

Oligo

8

NAA

points repfor a single study with two exceptions: all three lesions for the case of gliomatosis cerebri (CC) (three lesions) and both frontal lobes for patient 46. Multiple studies are not included. LG = low-grade glioma,

resent

3 10 20 21 22 24 25 26 29 30 34 36 40

0.4 z

Time from Treatmentt

creatine

were

category

of this

artifactual

sometimes

increase

seen

in

if there

was

because

part

of the large choline “skirt” was included by the program in integration of the creat-

me peak. curve-fitting creatine

Therefore, was

in the absence

algorithms, performed

of

quantitation only if there

of was

clear separation of the choline and creatme peaks. A large ROI was placed in the lesion,

as for quantitation

of choline

NAA, and was compared on the contralateral side. A Kruskal-Wallis

test

with was

and

a large

used

ROI

to corn-

PET PET

Scanning scanners

Forty-eight

were scanning

used

in this

procedures

were performed with a 15-section tomograph with 6-mm in-plane resolution and 6-mm section thickness (PC 2048-i5B; Scanditronix, Uppsala, Sweden), and 12 were performed with a 7-section tomograph with 6-7-mm in-plane resolution and 10.5-mm section thickness (PC 10247B, Scanditronix). Both scanners use a measured method of attenuation correclion. Our FDG PET technique has been described in detail elsewhere (25). Patients fasted for at least 6 hours before the study.

Blood

was sampled

from

a radial

artery

pare the normalized low-grade tumors,

used to compare the levels of glucose uptake between the cases with and those without lactate, and also to compare GUR for low- and high-grade tumors. A x2 test

plastic mask molded to the contours of the head. For all subjects, the scanning plane was parallel to the canthomeatal line, and

by the control value from the uninvolved site to give a normalized value. Some of the high-grade lesions were

for differences nuity correction)

necrotic, enhanced

grade

gliomas

lence

in the high-grade

patched throughout the study, which was performed in a dimly lit room. Emission scans were obtained after a 30-minute uptake period and were interleaved to cover the entire brain. Typically, 30 axial sections

678

Radiology

#{149}

that usually dimeglu-

values for the anaplastic tu-

study.

the metabolite maps were by using Image software, for as illustrations in this article.

identical for each map in a study. In cases with bilateral lesions, an uninvolved site (eg, the occipital lobe if the lesion was bifrontal) was used for the control value. In the patient with the pontine astrocytoma, the lesion involved half the pons and middle cerebellar peduncle; an ROI including the contralateral pons and middle cerebellar peduncle therefore was used as the control. The lesion value was then divided

with a rim of tissue after gadopentetate

choline the solid

FDG Two

increases

in choline,

Finally, smoothed, presentation

mors, and the cases of radiation necrosis; the same test was also used to compare glucose utilization rates (CURs) in these groups. A Wilcoxon rank sum test was

observed

in probabilities was used

prevalence

and

(with contito compare the

of lactate

the observed

gliomas.

in the

preva-

low-

catheter in most catheter placed hand (26). FDG was injected via

patients or from a venous in the dorsum of a heated (dose range, 185-370 MBq) a peripheral intravenous

line in the opposite subjects

were

arm. The heads

immobilized

with

the eyes and ears of the subjects

of the

a thermo-

were

December

1992

b.

a.

C.

I

Figure

2. Patient

17.

Images

obtained

in a

56-year-old woman with presumed lowgrade glioma. (a) Ti-weighted (left) and gadolinium-enhanced (right) MR images (600/ 16) obtained at 0.5 T show a nonenhancing hypointense mass in the right temporal lobe and insula. (b) Middle image from multiplanar GRE (600/30; flip angle, 10#{176}) localizing sequence. Hyperintense signal is seen in the right superior temporal lobe and insula. The box outlines the axial dimensions of the section examined. (c) Transaxial FDG PET scan. Hypometabolic eas are blue-green,

d.

glucose

were obtained with the and 28 with the 7-section study

duration

metabolic

was

rates

modification compartment rate constant

70 minutes.

(28)

value

was

the

visually relative

normal

section

imum

in the

CUR

=

a

matter

0.1300,

The

lumped

or hypomemetabolism

white was

in

matter.

performed ROI placed

in the

at the area of the maxlesion. Glucose utiliza-

normalized

rate

was recorded. in the lesion

to this

white

The was

matter

glioblastoma

3i%

reduction

patient to the

who was appearance

gluthen

lient

i4).

and presumed modest reductions

In the

tended

to be seen

and

and

10) with

Volume

185

two

value.

solid

#{149} Number

tumors.

3

The

outly-

extends NAA

glioalso

at the periphery

necrotic

of low metabolite and intensity

of increased beyond

signal

metabolism

Table

2. Lactate

tients: grade

in three tumors

been grade

tumors

Of the 7

choline

seen

seven in

in 20 pa-

of which

treated;

differed

three

present

two

had

of which and in two cases of The presence of and high-grade

at the

patients

ii

P with