Robert

John

J. Sevick, Kucharczyk,

MD PhD

Fumio Kanda, MD, PhD #{149} Jan Mintorovitch, Jay S. Tsuruda, MD #{149}David Norman, MD

#{149}

Cytotoxic Brain Edema: with Diffusion-weighted To determine whether cytotoxic brain edema is associated with a decrease in diffusion, it was induced in rats, in the absence of ischemia, with an established model of acute hyponatremic encephalopathy. Cytotoxic brain edema secondary to acute hyponatremia was induced with intraperitoneal injections of 2.5% dextrose in water and subcutaneous injection of arginine-vasopressin. Coronal spin-echo magnetic resonance (MR) images were obtained with and without strong diffusion-sensitizing gradients before and after induction of acute hyponatremia. The apparent diffusion coefficient (ADC) was measured at two coronal section locations. In hyponatremic rats, the brain ADC was significanfly reduced (P = .0153 and 0001) and was positively correlated with increased total brain water content (P = .0011). Plots of ADC versus total brain water showed a statistically significant inverse linear relationship between ADC and increasing brain water at the anterior coronal section location. The results indicate that the ADC may be a sensitive indicator of cytotoxic brain edema and thus may enable quantitalive evaluation of such edema with diffusion-weighted MR imaging. Index

terms:

Brain, edema, 10.86 #{149}Brain, MR. 10.1214 #{149} Magnetic resonance (MR), diffusion study, 10.1214 #{149} Magnetic resonance (MR), experimental #{149} Magnetic resonance (MR), pulse sequences #{149}Magnetic resonance (MR), technology

Radiology

1992;

From M.E.M.),

185:687-690

PhD #{149} Allen I. Arieff, Michael E. Moseley,

#{149}

D

Assessment MR Imaging’

IFFUSION-WEIGHTED

nance

(MR)

magnetic resoimaging is a rela-

tively new technique diffusion-sensitizing

in which gradients

strong are

added

to conventional spin-echo (SE) The resultant images are extremely sensitive to changes in the microscopic motions of water propulse

sequences.

tons (i,2). Diffusion-weighted imaging has already demonstrated

great

utility

in detection

emic

damage

in animal

stroke.

In several

MR

of early

reported

correlates are poorly

causes decrease tissues, cellular stricted

brain

to arterial

influx

possibility

pulsa-

occlusion,

of water

(4). The

is a particularly

attrac-

evaluate edema ischemia images.

model

in acute objective

the effects not associated

of cytotoxic

was

to

of cytotoxic with cerebral

on diffusion-weighted An established rat

edema

MR brain

was

used

the Departments of Radiology, Section of Neuroradiology (R.J.S., J.M., J.K., J.S.T., D.N., and Medicine, Division of Geriatrics (F.K.), University of California, San Francisco; and San Francisco Veterans Affairs Medical Center, San Francisco (F.K., AlA.). From the 1991 RSNA scientific assembly. Received November 6, 1991; revision requested December 26; final revision received July 6, 1992; accepted July 24. R.J.S. supported in part by a clinical fellowship from the Alberta Heritage Foundation for Medical Research. J.S.T. supported in part by an RSNA Research and Education Fund Seed Grant. Address reprint requests to R.J.S., Department of Radiological Sciences and Diagnostic Imaging, MRI Centre, Foothills Provincial General Hospital, 1403 29th St NW, Calgary, Alta, Canada T2N 2T9. ,-

RSNA,

1992

rats

hyperim-

with

a body

METHODS

Sprague-Dawley

weight

of 220-250

g were

randomly assigned to one of three study groups. Eight rats in the experimental group and six control rats were imaged. Another group of six controls were not imaged. Anesthesia was induced with intraperitoneal injection of 50 mg of ket-

amine hydrochloride zine hydrochloride.

and 2.5 mg of xylaAcute hyponatremia

was induced over 2.5 hours with intrapentoneal injection of 140 mmol/L dextrose in water (12% body weight) and subcutaneous injection of 2 IU of arginine-vasopressin. This was followed 30 minutes later by an additional intraperitoneal injection of 140

mmol/L

dextrose

in water

(6%

body

weight). Body temperature was maintamed by means of a circulating warm water pad while the rats were in the magnet and by means of heat lamps when they

were out of the magnet. All MR images were obtained 2-T

unit

ischemia. of this study

AND

A total of 20 female

live explanation of the temporal sequence of changes observed on MR images The

MATERIALS

studies,

in temperature in ischemic changes in osmolality of extrafluid in the brain, and reproton motion secondary to

intracellular last

reduced

secondary

is correlated with signal intensity on diffusion-weighted ages (4). ADC

of

in acute cerebral ischunderstood. Possible

include

tion

(5-iO). To eliminate Ti and T2 relaxation effects, the apparent diffusion coefficient (ADC) was calculated both before and after the induction of brain edema. Previous studies have demonstrated that a reduction in the

isch-

models

areas of signal hyperintensity were observed on diffusion-weighted MR images well in advance of any abnormalities on T2-weighted SE images (3,4). Although the physical cause of diffusion-weighted signal hyperintensity is known to be associated with slower diffusion of water protons, its physio-

logic emia

MD PhD

#{149}

chemical

shift

(GE Medical

equipped

with

with

a

imager/spectrometer

Systems,

Milwaukee)

self-shielded

gradient

coils

(±15 C/cm, 15-cm bore) (Acustar, GE Medical Systems) (Ii). A low-pass birdcage proton imaging coil with an inner diameter of 5 cm was used to obtain multisection SE MR images in the coronal plane immediately before the induction of hyponatremia and approximately 3 hours later. The diffusion-weighted pulse sequence used was similar to the intravoxel incoherent-motion sequence devised by

Le Bihan

et al (i), which

gradient

pulses

Abbreviations: efficient, ROI

ADC =

places

on either

region

=

side

apparent

of interest,

matched of the

radio-

diffusion co= spin

SE

echo.

687

frequency-refocusing exact manner as the

180#{176} pulse in the original Stejskal-Tan-

ner pulse

The imaging

sequence.

parame-

ters were as follows: repetition time msec/ echo time msec = 1,500/80; field of view, 60 mm; 128 phase-encoding steps; section thickness, 3 mm, without intersection gap;

and four signals

averaged.

The diffusion-

sensitizing gradient was oriented in the y direction (ventral to dorsal) with a pulse duration of 20 msec, gradient separation of 40 msec, and gradient strength of 5.6 C/cm. In aggregate, these factors produce a b value of i,4i3 sec/mm2. The b value was determined by means of careful calibration of several pure liquids for which the diffusion coefficient is known at the temperature of the calibration measurement. The calibration measurements were

with

performed

exactly

the same

parame-

Figure 1. out and

Coronal SE MR images (1,500/80, (right) with strong diffusion-sensitizing

basal ganglia passed most

as well as frontal of one hemisphere,

and temporal excluding

with

four signals averaged) gradients. This anterior

cortex. The subarachnoid

obtained coronal

(left) withlevel includes

ROI used for ADC calculation spaces and ventricles.

encom-

ters as those used for the in vivo measurements. The b value used was approximately equal to the value anticipated from

integration

of the original

Stejskal-Tanner

equations with half-sine shapes for the diffusion-sensitizing gradient pulses and with consideration of potential cross-terms from the inherent MR imaging encodinggradient pulses. Separate images were obtamed with and without the application of strong diffusion-sensitizing gradients.

Region

of interest

(ROI)

performed on the same ages selected from each

before

and after

analyses

were

two coronal multisectional

im-

hyponatremia.

set

The ante-

nor section (Fig i) included the basal ganglia, as well as frontal and temporal cortex, while the posterior section (Fig 2) consisted primarily of thalamus and parietal and posterior temporal cortex. At each of the two levels, ROI measurements coyered most of one hemisphere and encompassed gray and white matter but exduded subarachnoid spaces and ventricles. ADC values were calculated according to the formula ADC

ROI signal

where

S1

fusion

gradients,

without

(lnS2

=

=

S2

=

-

intensity with difROI signal intensity

diffusion gradients, and b in b value between images 1,413 sec/mm2) and without (b

=

diffusion-sensitizing

the

=

with 0)

=

gradients.

At the conclusion of the imaging procedune, a blood sample was obtained by means of cardiac puncture for measurement of plasma sodium; then the rat was decapitated and the brain was removed, as previously described (i2). We analyzed the brain for water content by drying to constant weight triplicate samples of cerebral cortex that weighed approximately 0.2

g each,

as previously

Data were

analyzed

tistics

a paired,

and

described

with

(13).

descriptive

two-tailed

sta-

Student

significantly

plasma

sodium and total and extracellular

measurements

Radiology

#{149}

are

summarized

brain water)

wa-

with

was

increased

brain water. ADC values were consistently lower in the anterior section location (0.63 x i05 cm2/sec ± 0.04 x i0) than those in the posterior section location (0.72 x iO-5 cm2/sec ± 0.04 x 10) in both experimental and control rats. differences existed values at 0 and 3 hours

control group. On the statistically significant

creases

in ADC

values

were

at both

coronal

levels

when

posthyponatremia in Ta-

sodium

correlated

total

in the hand,

RESULTS

688

ble 1. Decreased

No significant between ADC

test.

Plasma ter (intra-

signals averaged) obtained (left) withlevel includes thalamus as well as pan-

lnS1)/b,

difference (b

Figure 2. Coronal SE MR images (i,500/80, with four out and (right) with diffusion gradients. This posterior etal and posterior temporal cortex.

pared

(Table

values

2).

Table 2 ADC Values Section 1 2

in Experimental

Before Hyponatremia 0.61 0.73

Group

After Hyponatremia

±

0.02

0.56

±

±

0.03

0.67

±

0.04 (.0153) 0.03(0001)

Note-Values are the mean x i0 cm2/sec standard deviation; numbers in parentheses are P values. I = anterior, 2 = posterior. ±

other deobserved

prewere

and com-

Figures 3 and 4 illustrate tionship between the ADC water content. Hyponatremia-in-

duced

increases

in total

brain

the and

relabrain

water

December

1992

S S

E

E

06

05

ADC

08

07

(m..n

x

ADC

3.

Os

07

cm’/s.c)

1O

(m..n

x

09

cm’/s.c)

1O.

4.

Figures

3, 4.

ADC ADC

shows versus

(3) Plot

of ADC

an inverse total brain

linear water

crease with increases was not statistically

versus total brain water in anterior coronal section location. The relationship with increasing brain water (R2 = .75). (4) Plot of in posterior coronal section location. A tendency for ADC to de-

in brain significant

water (R2

is evident. =

However,

the

multiple

correlation

change in brain intracellular pH was found, and there was normal resistance to mannitol infusion. The intactness of the blood-brain barrier was confirmed by the finding that administration of gadolinium did not produce signal intensity changes on Tiweighted MR images of the brain in the same animals. Similar results were obtained by Rymer and Fishman (10), who evaluated the integrity of the blood-brain barrier in acutely hyponatremic rats by use of tritiated mannitol. These results indicate that the blood-brain barrier is preserved in the model of acute hyponatremic encephalopathy used in our study and suggest

.16).

gray

lion was found between the ADC and brain water content at the level of the

cerebrospinal fluid into the bloodstream, may be overwhelmed with resultant cellular water expansion or cytotoxic edema. In some instances of hyponatremia, by comparison, the

posterior

intrinsic

correlated

significantly

in the

ADC

3).

statistically

No

in the

with

anterior

decreases

section

significant

section

(Fig

(Fig

correla-

4).

DISCUSSION Diffusion-weighted has

been

the

MR imaging

focus

of a number

of re-

cent studies in animal models of cerebral ischemia (3,4,14). Approximately five explanations have been proposed to account for the areas of early-onset signal hyperintensity sion-weighted MR

seen images

on diffuof acutely

ischemic brain (which appear well in advance of changes in signal intensity on T2-weighted images). One explanation is that signal hyperintensity may

be

secondary

to intracellular

ter accumulation

or cytotoxic

wa-

edema

(4).

Our ship sity

study

evaluated

the

between changes on diffusion-weighted

and cytotoxic severe acute

relation-

in signal MR

edema associated with hyponatremic encepha-

that

such

rats

are more

susceptible

hyponatremia-induced

brain

than male rats (5). Acute mia is the most common disorder among hospital has

significant

associated

to

injury

hyponatreelectrolyte patients

and

morbidity

Volume

cells (5-7,15). Adaptive mechaincluding the loss of cellular and increased movement of

185

Number

#{149}

3

in intracellular likely have

an effect as well. Many previous studies have used dilutional hyponatremia (6,7) induced with vasopressin injection and parenteral glucose in water to induce cerebral edema in rats. It is well known vasopressin

retention ponatremia,

can

increase

water

by the kidneys. In acute hysuch an increase could the

osmotic

disequilibrium

between plasma and brain cells. However, the peptide hormone is now also known to significantly increase the permeability of both brain capillaries

and mortality, which are primarily associated with dysfunction of the central nervous system (is). An acute drop in serum sodium causes osmotic disequilibrium and shift of water into brain nisms, solutes

sium pump. Changes viscosity and osmolality

sustain

lopathy. Young adult female rats were used because it has been shown

mechanisms

of the brain are able to reestablish brain cell volume without permanent brain damage (6,7). We can only speculate as to how the ADC is decreased in cytotoxic brain edema. The diffusion of water that is accumulating in cells is slowed. The cause of this slowing is unknown. It could be related to breakdown of the transmembranous sodium-potas-

that

intenimages

compensatory

and

ventricular

ependymal

that

brain water edema. A previous

coefficient

cells

to water (8,16-18), even in the absence of hyponatremia (i9). It is possible that vasopressin may be a hormonal mediator of water movement in the central nervous system. Adler et al (9) evaluated the integ-

and

the

observed

content study

white

increases

indicate has

matter

shown are

in

cytotoxic that

equally

affected by the cytotoxic edema induced in this model (10), a finding which suggests that the differences in the measured ADC values at the antenor and posterior coronal section locations

are

most

likely

attributable

to

regional variations in the relative volumes of gray and white matter in each section. ADC values are known to be substantially different in gray and white matter. For example, the cortical gray matter of cat brain has been reported to have an ADC of 0.80.9 x i0 cm2/sec, whereas ADC values in white matter ranged from 0.3 x io to 1.2 x i05 cm2/sec, depending on which diffusion gradient direction was selected (i4,20,2i). In our study, lower ADC values were consistently observed in the anterior section localion in both control and experimental rats.

In the

anterior

section,

gray

mat-

ter is primarily composed of the thin mantle of cerebral cortex and the large caudate nucleus. White matter at this level includes the corpus cabsum,

anterior

commissure,

internal

capsule, and longitudinally oriented myelinated fiber tracts. In the posterior section, the gray matter is again represented by a thin cerebral cortex but also includes large aggregations of thalamic and hypothalamic neurons that would contribute to an increase in the ADC. On this basis, it seems

possible

that

the

gray

matter!

model of acute hyponatremia induced with water and desmopressin. These workers infused 0.25% HC1 and measured brain intracellular pH by

white matter ratio is higher in the posterior section, and this might account for the higher ADC values. This explanation is speculative, however, because selective ADC measurements were not performed in gray and white matter separately. Other factors to consider regarding the observed differences in ADC be-

means

tween

rity

of the

blood-brain

of MR

spectroscopy.

barrier

in a rat

No

anterior

and

posterior

section

Radiology

689

#{149}

locations include the possibility that the ADC of gray matter and that of white matter differ in various regions of the brain. Finally, it is possible that the observation is artifactual. The results of this study do not prove conclusively that the areas of signal hyperintensity diffusion-weighted

observed MR images

may

prove

valuable

in experi-

mental studies of brain edema and ultimately (in combination with techniques of ultrafast MR imaging) in

690

Radiology

#{149}

of brain

edema

in human

11.

sub-

#{149}

References 1.

Le Bihan

D, Breton

E, Lallemand

D, Gre-

nier P. Cabanis E, Laval-Jeantet imaging of intravoxel incoherent

application neurologic

on in

early cerebral ischemia are due to cytotoxic brain edema. Recent work in our laboratory has shown that the ADC may be reduced by as much as 20% i hour after occlusion of the middle cerebral artery in the rat (Mintorovitch J, unpublished data, i99i). This change in ADC is of greater magnitude than that seen in our experiments (mean decrease, 8%). Because of the sequence of pathophysiologic changes that occur in acute cerebral ischemia, it is likely that early diffusion-weighted signal hyperintensity is multifactorial in origin. The possible quantitative relationship between brain water content and ADC merits further study because this is a potential noninvasive method of quantitating brain edema that has been unavailable until now. In addition, the potential exists to distinguish cytotoxic edema and vasogenic edema, because previous studies have shown an increase in ADC in areas of peritumoral vasogenic edema (i). The technique

studies jects.

2.

to diffusion and perfusion in disorders. Radiology 1986; i6i: 13.

401-407. Le Bihan

D, Breton

ML, Vignaud tion

3.

5.

6.

in cats:

Fraser

Separa-

in intra-

14.

of diffusion-

Arieff

Al, Rollin

1983;

women. i6.

17.

18.

Am J Physiol

1987;

19.

252:F66i-F669.

Doczi T, Szerdahelyi P. Gulya K, Kiss J. Brain water accumulation after the central of vasopressin. 11:402-407.

20.

S, Williams D, VerbalisJC. Effect of and chronic hyponatremia on the blood brain barrier. Am J Physiol (in press). Rymer MM, Fishman BA. Protective adaptation of brain to water intoxication. Arch Neurol 1973; 28:49-54.

21.

transport

at arachnoid

villi of cats.

EndocrinolJpn 1979; 26:239-244. Raichle ME, Grubb RL. Regulation of brain water permeability by centrally-released vasopressin. Brain Res 1978; 143: 191-198. Rosenberg GA, Kyner WT, Estrada E, et al. Effect of vasopressin on ependymal and capillary permeability to tritiated water in cat. Ann N Y Acad Sri 1986; 481:383-387.

Rosenberg

GA, Estrada

E, Kyner

WT.

Va-

sopressin-induced brain edema is mediated by the Vi receptor. Adv Neurol 1990; 52: 149-154. Moseley ME, Kucharczyk J, Asgani H, et al. Anisotropy in diffusion-weighted MRI.

Magn

Neurosur-

N EngI J Med 1986; 314:1529-1535.

Noto T, Nakajima Y, Saji Y, Nagawa Y. Effects of vasopressin and cyclic AMP on

water

244:R724-R732.

Melton JE, Patlak CS, Pettigrew KD, Cserr HF. Volume regulatory loss of Na, Cl, and K from rat brain during acute hyponatre-

Adler

15.

C,

Sarnacki P, Norman D. Sex differences result in increased morbidity from hyponatremia in female rats. Am J Physiol 1989; 256:R880-R885. Melton JE, Nattie EE. Brain and CSF water and ions during dilutional and isosmotic hyponatremia in the rat. Am

acute

10.

comparison

CL, KucharczykJ,

administration gery 1982;

9.

M.

perfusion

and T2-weighted MRI and spectroscopy. Magn Reson Med i990; 14:330-346. Moseley ME, KucharczykJ, Mintorovitch J, et al. Diffusion-weighted MR imaging of acute stroke: correlation with T2-weighted and magnetic susceptibility-enhanced MR imaging in cats. AJNR 1990; 11:423-429.

mia.

8.

and

D, Aubin

voxel incoherent motion MR imaging. Radiology 1988; 168:497-505. Moseley ME, Cohen Y, Mintorovitch J, et al. Early detection of regional cerebral

Physiol

7.

E, Lallemand

J, Laval-Jeantet

of diffusion

ischemia

4.

12.

M. MR motions:

Mintorovitch J, Moseley ME, Chileuitt L, Shimizu H, Cohen Y, Weinstein PR. Cornparison of diffusionand T2-weighted MRI for the early detection of cerebral ischemia and reperfusion in rats. Magn Reson Med 1991; 18:39-50. Fraser CL, Sannacki P, Aneff Al. Abnormal sodium transport in synaptosomes from brain of uremic rats. J Clin Invest 1985; 75:20i4-2023. Arieff Al, Kleeman CR, Keushkerian A, Bagdoyan H. Studies on mechanisms of cerebral edema in diabetic comas: effects of hyperglycemia and rapid lowering of plasma glucose in normal rabbits. J Clin Invest 1973; 52:571-583. Moseley ME, Sevick RJ, Wendland MF, et al. Ultra-fast MRI: diffusion and perfusion. J Can Assoc Radiol 1991; 42:31-38. Arieff Al. Hyponatremia, convulsions, respiratory arrest, and permanent brain damage after elective surgery in healthy

Reson

Med 1991; 19:321-326.

Moseley ME, Cohen Y, KucharczykJ, et al. Diffusion-weighted MR imaging of anisotropic water diffusion in cat central nervous system. Radiology 1990; 176:439-445.

December

1992