J Neurosurg 104:720–730, 2006

Predominance of cellular edema in traumatic brain swelling

in patients with severe head injuries

ANTHONY MARMAROU, PH.D., STEFANO SIGNORETTI, M.D., PANOS P. FATOUROS, PH.D.,

GINA PORTELLA, M.D., GUNES A. AYGOK, M.D., AND M. ROSS BULLOCK, M.D., PH.D.
Departments of Neurosurgery and Radiology, Virginia Commonwealth University Medical College of
Virginia Campus, Richmond, Virginia

Object. The edema associated with brain swelling after traumatic brain injury (TBI) has been thought to be vaso-
genic in origin, but the results of previous laboratory studies by the authors have shown that a cellular form of edema
is mainly responsible for brain swelling after TBI. In this study the authors used magnetic resonance (MR) imaging
techniques to identify the type of edema that occurs in patients with TBI.
Methods. Diffusion-weighted MR imaging was used to evaluate the apparent diffusion coefficient (ADC) in 44
patients with TBI (Glasgow Coma Scale Score , 8) and in eight healthy volunteers. Higher ADC values have been
associated with vasogenic edema, and lower ADC values with a predominantly cellular form of edema. Regional mea-
surements of ADC in patients with focal and diffuse injury were computed. The water content of brain tissue was also
assessed in absolute terms by using MR imaging to measure the percentage of water per gram of tissue. Cerebral blood
flow (CBF) was measured using stable Xe–computerized tomography (CT) studies to rule out ischemia as a cause of
cellular edema.
The mean ADC value in the healthy volunteers was 0.82 6 0.05 3 10–3 mm2/second. The ADC values in the pa-
tients with diffuse brain injury without swelling were close to the mean for the healthy volunteers. In contrast, the pa-
tients with brain swelling had increased brain water content and low ADC values (mean 0.74 6 0.05 3 10–3 mm2/sec-
ond). The ADC values correlated with CT classifications. In all patients with low ADC values, the CBF values were
outside the range for ischemia.
Conclusions. The brain swelling observed in patients with TBI appears to be predominantly cellular, as signaled by
low ADC values in brain tissue with high levels of water content.

KEY WORDS • brain edema • trauma • head injury • brain injury •

apparent diffusion coefficient

contribution of brain edema to progressive brain supports the need for a better understanding of the process-

T
HE
swelling in cases of TBI remains a critical problem
for which there is no effective clinical treatment at
present. Current practice is to utilize combinations of se-
dation, drainage of cerebrospinal fluid, osmotic diuretics,
pressor agents, and hypothermia to suppress high elevations
of pressure, which are a frequent cause of death or very poor
prognosis in survivors.7 The degree of swelling after brain
injury has been classified into four categories of severity
based on earlier studies by the National Institutes of Health
Traumatic Coma Data Bank.32 Using this classification,
Marshall, et al.,32 found that the degree of swelling revealed
on the first CT scan obtained soon after injury was highly
correlated with outcome (p , 0.001). This finding clearly
Abbreviations used in this paper: ADC = apparent diffusion co-
efficient; BBB = blood–brain barrier; CBF = cerebral blood flow;
CT = computerized tomography; DW = diffusion-weighted; GCS =
Glasgow Coma Scale; ICP = intracranial pressure; ICU = intensive
care unit; MR = magnetic resonance; NAA = N-acetyl aspartate;
ROI = region of interest; SEM = standard error of the mean; TBI =
traumatic brain injury.
es leading to brain swelling and of the means by which
these processes can be mitigated.
In earlier studies, it was generally accepted that the swell-
ing accompanying TBI was mainly due to vascular engorge-
ment, with blood volume providing the increase in brain
bulk and subsequent rise in ICP.23–27 Edema was thought to
play a minor role. However, findings from a study in our
laboratory have indicated that edema, and not vascular en-
gorgement, is responsible for brain swelling and that blood
volume is actually reduced after TBI.19 Our previous find-
ings have been confirmed in a study in which both brain
water and blood volume were measured in patients with
head injury.30 In that study, blood volume was reduced, and
it was found that brain edema was responsible for the in-
crease in brain volume. Thus, to better understand the patho-
physiological mechanisms responsible for water movement
into the brain, we shifted our attention to brain edema.
In vivo DW imaging is an MR imaging technique in
which strong magnetic field gradients are used to measure
random molecular translation of water protons. Maps of the
ADC are derived from a series of DW images obtained with
a series of magnetic field gradients. Findings in experimen-

720 J. Neurosurg. / Volume 104 / May, 2006

Apparent diffusion coefficient in patients with head injury
tal ischemia models have suggested that the ADC can pro-
vide early, specific information about tissue damage and
characteristics of edema.8,33,34 Authors of several studies
have adopted this concept in identifying and distinguishing
the type of edema in ischemia and TBI.9,11,22,29,38,39,41,43–45 In
experiments by Ito, et al.,17 cellular edema produced by us-
ing middle cerebral artery occlusion was correlated with the
ADC, which decreased as the cellular edema evolved. Us-
ing the infusion model,31 which produces extracellular ede-
ma, Ito, et al.,17 have shown that ADC increases with the de-
velopment of extracellular edema. Kuroiwa, et al.,22 have
investigated the correlation of ADC with water content after
middle cerebral artery occlusion in cats and have shown that
water content increases linearly with the decrease in ADC.
These observations of a cellular edema were confirmed by
electron microscopy. In a second study, a model of cortical
cold lesion in cats was used. A vasogenic edema was in-
duced in the white matter, and the trace of diffusion tensor
(Trace[D]) obtained from MR imaging was used to measure
water diffusibility, compared with the corresponding tissue
water content determined by gravimetric studies and with
ultrastructural water localization. A linear correlation was
observed between increases in Trace(D) and increases in
extracellular water volume in vasogenic brain edema in vi-
vo, and the authors have concluded that measurement of
Trace(D) is a suitable parameter for the evaluation of vaso-
genic brain edema. These studies have provided compelling
evidence that ADC can be used to identify the predominant
type of edema in brain tissue.
The goal of this study was to utilize these noninvasive
MR imaging measures of ADC to determine the type of
edema that exists in patients with head injury and to quan-
tify the temporal changes that occur during the acute and
late stages of edema development. For measurement of
brain edema, we used our MR imaging method for absolute
measure of brain water.15 For measurement of CBF, we used
stable Xe-CT studies to determine the contribution of isch-
emia to the edematous process.
Clinical Material and Methods
Patient Population
Forty-four patients (nine female and 35 male) with severe
TBI (GCS score , 8) and eight healthy volunteers were
entered into the study (Table 1). The mean age of the pa-
tients was 33 years (range 16–70 years). Based on initial CT
studies, 32 patients were classified as having diffuse injury
and 12 were classified as having focal lesions. Eight pa-
tients with diffuse injury suffered a significant focal injury
and were included in the analysis of patients with contu-
sions.
The patients’ injuries were classified according to the
classification of CT images proposed by Marshall, et al.32
Briefly, the Marshall CT scan classification is used to des-
ignate the intracranial diagnosis with one of six mutually
exclusive categories: Type I, a normal CT scan; Type II, dif-
fuse injury with cisterns present and a midline shift of less
than 5 mm and/or lesion densities present but no high or
mixed density lesions greater than 25 cm3; Type III, diffuse
injury with swelling, cisterns that are compressed or absent,
a midline shift of less than 5 mm but no high or mixed den-
sity lesions greater than 25 cm3; and Type IV, diffuse inju-
J. Neurosurg. / Volume 104 / May, 2006
TABLE 1
Clinical characteristics in 44 patients with severe head injury
Characteristic Value
sex
male 35
female 9
GCS score (range) 3–8
type of injury
diffuse 32
focal 12
contusion* 20
injury classification†
II 20
III 10
IV 2
V or VI 12
* Identification of contusion was based on the patient’s initial CT scan.
† Injuries were classified on the basis of the patient’s initial CT scan ac-
cording to the classification system proposed by Marshall, et al. (see Clini-
cal Material and Methods).
ry with a midline shift of greater than 5 mm but no high or
mixed density lesions greater than 25 cm3. (Note that le-
sions , 25 cm3 may be present in the so-called diffuse in-
jury categories.) The remaining two categories are the fol-
lowing: Type V, any lesion surgically evacuated; and Type
VI, a high or mixed density lesion greater than 25 cm3 that
is not surgically evacuated. Of the 32 patients with diffuse
injury, 20 were classified as having Type II head injury and
12 as having Type III or IV. The 12 patients with focal le-
sions were classified as having Type V or VI head injury.
The age range of the patients with diffuse injury classified
as Type II (range 18–49 years, mean 26 6 10 years) was not
significantly different from the age range of the patients
with more severe edema (range 17–53 years, mean 31.2 6
17.6 years).
Informed consent was obtained from family members or
caregivers to allow the patients to participate in the study.
After stabilization in the ICU and with the approval of the
attending neurosurgeon, the patients were transported to the
MR imaging suites for measurement of brain tissue wa-
ter and DW imaging. The transport team consisted of an
ICU nurse, respiratory technician, and neurosurgical resi-
dent. The research team consisted of an investigator, head
injury fellow, and radiology technician.
All patients in this study were transported safely from the
ICU to the imaging suite and were returned safely without
adverse events after the protocols were completed. Com-
parison of blood pressure and ICP measurements obtained
before transport to the imaging suite and after return to the
ICU indicated no major changes throughout the course of
study (Table 2). The timing of the imaging studies varied
because patients were transported only when they were
considered stable and when imaging suites were available
for use. Within these constraints, we were able to study pa-
tients between 40 and 220 hours after injury, and study time
ranged from 1 to 2.5 hours.
Measurement of Brain Water
After the patient was stabilized in the magnet, the head
was positioned carefully to ensure CT slice compatibility.
First, T1- and T2-weighted pulse sequences were used to
produce images in the axial, coronal, and sagittal planes,
721

TABLE 2
Physiological measurements in 44 patients with severe head
injury before transport for CT and MR imaging
and on return to the ICU*
Parameter Before Transport After Return
ICP (mm Hg) 17.6 6 5.8 16.1 6 5.5
MABP (mm Hg) 109.5 6 13.6 107.4 6 12.0
PO2 (mm Hg) 154.2 6 34.2 200.5 6 69.4
SaO2 (%) 99.5 6 0.8 99.6 6 0.7
PCO2 (mm Hg) 35.7 6 2.8 35.8 6 3.0
* Values represent means 6 standard deviations. Abbreviations: MABP =
mean arterial blood pressure; SaO2 = arterial saturation of oxygen.
and the traumatic lesions were identified. From these imag-
es, an anatomical slice equivalent to that used in classifica-
tion of head injury type and in CBF measurements was
selected for quantitative brain water measurements. Five
inversion recovery scans of the selected slice were obtained
(TR 2.5 seconds; TE 28 msec; one each at TI 150, 400, 800,
1300, and 1700 msec). From these images, brain water con-
tent (fw; expressed as a fraction of water per gram of tissue)
was derived, as described previously.16 The inversion recov-
ery pulse sequence and the particular values of the param-
eters for measurement of brain water were selected after
extensive studies with calibration standards of known relax-
ation times21 and models of in vivo infusion edema in ani-
mals.13,14,16 This technique has been validated in humans.15
Stated in terms of water content, the percentage of swell-
ing was calculated using the following equation: 100 (fw 2
fwn) / (1 2 fw), where fw represents the water content of the
edematous brain and fwn represents the water content of the
normal tissue. Based on these computations, a fractional in-
crease in water of 2% is equivalent to a volume increase of
8.8%. This equivalence highlights the importance of edema
to the swelling process and subsequent rise in ICP.
Diffusion-Weighted MR Imaging
Diffusion-weighted MR imaging was performed using a
1.5-tesla whole-body clinical imaging unit (Magnetom Vi-
sion MR system; Siemens Medical Solutions, Malvern, PA)
equipped with 15-millitesla/m gradients using spin echo se-
quences with and without diffusion sensitizing gradients.
These pulse sequences generated an ADC trace image us-
ing a single-shot technique with three b values: 0, 500, and
1000 seconds/mm2. Typically, 20 slices were generated with
the following parameters: 5-mm slice thickness with a 1.5-
mm gap, field of view 230 3 230 mm; 96 3 128 image
matrix, TE 100 msec. The ADC trace was generated to
eliminate issues arising from tissue anisotropy and head ori-
entation. Regional measurements of ADC were extracted
from the ADC maps.
Measurement of CBF
After completion of the MR imaging studies, the patient
was transported to the CT suite for measurement of CBF.
After obtaining a diagnostic CT scan, a stable Xe-CT CBF
study was performed on a scanner equipped with a Xe gas
delivery system and a CBF software analysis package (CT/
Plus; Siemens, Erlangen, Germany). Scans were obtained at
four axial planes 15 mm apart with a thickness of 10 mm
each. Two baseline scans were performed at each level fol-
722
A. Marmarou, et al.
FIG. 1. Bar graph depicting the relationship between ICP at the
time of the MR imaging study, brain water content, and head injury
type according to the Marshall CT image classification. Brain ede-
ma in patients with Type III and IV head injury was significantly
elevated, compared with water content in patients with Type II head
injury (p , 0.0001). This increase in edema was equivalent to an
8.16% increase in brain volume. The corresponding ICP increases
were consistent with the level of brain swelling (p , 0.001).
lowed by multiple enhanced scans during which the patient
inhaled a mixture of Xe (30%) and O2 (70%). From mea-
surements of the CT enhancement and the end-tidal curve,
CBF maps were calculated by means of the Kety–Schmidt
equation using a commercially available package (Diversi-
fied Diagnostic Products, Inc., Houston, TX).
Results
Water Content, ICP, and CT Image Classification in
Diffuse TBI
The mean normal hemispheric water content in brain tis-
sue assessed by MR imaging in the healthy volunteers was
76. 6 0.66%. The brain water content of diffusely injured
patients with Type II head injury was close to normal (mean
76.2 6 0.76%) and was consistent with the low ICP in this
group (mean 14.8 6 3.31 mm Hg). In contrast, the brain
water content of diffusely injured patients with Type III or
IV head injury was elevated (mean 77.75 6 1.15%) (p ,
0.001). This increase in edema is equivalent to an 8.16%
increase in brain swelling according to the Elliot–Jasper
equation.12 The ICP at the time of the study for this group
was also elevated (mean 23.2 6 2.48 mm Hg) (p , 0.001).
The increase in edema compared with ICP in patients
grouped according to the Marshall CT scan classification is
shown in Fig. 1. As might be expected, the ICP increased as
more water accumulated in brain tissue (R2 = 0.43) (Fig. 2).
It is interesting to note that the ICP increased to levels of 20
mmHg or higher as the total water content increased high-
er than 77%, which is only 1% above normal. In terms of
swelling, the ICP increased when brain swelling reached
values higher than 4.3% (p , 0.001).
Apparent Diffusion Coefficient in Healthy Volunteers and
Patients With Diffuse Injury
The mean ADC computed from hemispheric ROIs in the
J. Neurosurg. / Volume 104 / May, 2006
Apparent diffusion coefficient in patients with head injury
FIG. 2. Graph depicting the relationship of ICP and tissue water
content in patients with diffuse injury. The ICP increased to levels of
at least 20 mm Hg when water content was greater than 77%, which
was only 1% higher than the normal value. The ICP increased when
brain swelling reached a 4.3% increase in brain volume (p , 0.001).
eight healthy volunteers was 0.82 3 1023 6 0.05 mm2/sec.
Henceforth, we will report ADCs without the common mul-
tiplicative factor (1023 mm2/sec). The ADC values in pa-
tients with diffuse injury without swelling were close to nor-
mal (mean 0.89 6 0.08). This result was not surprising, as
the ICP values of these patients were low. In contrast, in
patients with significant brain swelling, the ADC values
were reduced (mean 0.74 6 0.05) (p , 0.0001), consistent
with a predominantly cellular edema. The distribution of
ADCs among the healthy volunteers, patients with swelling,
and patients without swelling correlated with the CT im-
age classification (Fig. 3). As an example, the brain images
obtained in a 17-year-old girl and the corresponding water
map and ADC are shown in Fig. 4. Based on the admission
CT image, the patient was classified as having diffuse Type
II head injury. The global CBF measured at time of study
was nonischemic (55.6 ml/100 g/min). The mean of the
ADC values shown in the lower right panel of Fig. 4 was
0.72, indicating that a predominantly cellular form of ede-
ma contributed to the 3.9% swelling.
Cerebral Blood Flow in Patients With Diffuse Injury and
Reduced ADC
To confirm that the reductions in ADC seen in patients
with diffuse injury were not caused by frank ischemic dam-
age, we measured CBF at the same time DW imaging was
performed. In patients with diffuse injury in which the ADC
was significantly reduced, the corresponding mean hemi-
spheric CBF was 50.04 6 11.30 ml/100 g/min in the pa-
tients without swelling and 45.33 6 7.99 ml/100 g/min in
the patients with swelling. The reduction of CBF in the pa-
tients with swelling did not reach statistical significance.
The ADC values obtained in patients with diffuse injury and
the corresponding CBF values are shown in Fig. 5. In abso-
lute terms, the CBF values obtained in all patients with re-
duced ADC values were well above the ischemic threshold
J. Neurosurg. / Volume 104 / May, 2006
FIG. 3. Bar graph depicting the relationship of ADC values and
brain water content. In patients with significant brain swelling, the
ADC values were reduced and averaged 0.74 6 0.05 (p , 0.0001),
consistent with a predominantly cellular edema. Brain water content
in patients with Type III and IV head injury was markedly elevated,
compared with that in patients with less diffuse injury (Type II).
(18 ml/100 g/min). When CBF values were corrected and
normalized for a PaCO2 of 34 mm Hg using a 3% change
per mm Hg of PaCO2, the CBF values were 46.37 6 9.63
ml/100 g/min in the patients with swelling and 47.74 6
11.35 ml/100 g/min in the patients without swelling. These
values were not statistically significantly different from the
nonnormalized CBF values.
Water Content, ICP, and CT Image Classification in
Focal TBI
The methods for analysis in patients with focal injury
were different from those in patients with diffuse injury.
First, we selected for analysis only patients with a well-de-
fined unilateral nonevacuated contusion (Type VI head in-
jury in the Marshall CT image classification). Second, we
evaluated water content, ADC values, and CBF in specific
ROIs, including the contusion core, the perilesional area, a
region distant from the contusion on the ipsilateral side, and
a symmetrical site contralateral to the lesion. Of the initial
cohort, 20 patients who met these inclusion criteria were se-
lected for analysis.
In the contusion core, the mean percentage of water con-
tent was 81.81 6 2.36%, well above the brain water content
of the symmetrical site contralateral to the contusion core
(75.83 6 1.44%) (p , 0.0001). (Fig. 6). As water content
increased, the ICP increased exponentially (Fig. 7). Similar
to the findings in patients with diffuse injury, the ICP in-
crease in the patients with focal injury was substantial, high-
er than a level of 77% water. The ICP at this level was 22
mm Hg and corresponded to 4.3% swelling.
Apparent Diffusion Coefficient in Patients With Focal
Injury
The ADC in the contusion core and ring of hyperdensity
surrounding the contusion was high (mean 1.25 6 0.37)
(Fig. 8). This value was well above normal and was con-
723
 A. Marmarou, et al.

FIG. 4. Computerized tomography (upper) and MR (lower) images obtained in a 17-year-old patient in coma studied
44 hours after diffuse head injury, which was classified as Type II based on the admission CT scan. Global CBF equaled
55.6 ml/100 g/min and water content of the right and left hemispheres measured 76.8 and 77%, respectively. Brain swelling
due to water increase equaled 3.9%. It is noteworthy that the ADC of both the right and left hemispheres was low (mean
0.72 and 0.73, respectively), signifying a predominantly cellular form of edema.

sistent among all of the contusions studied, indicating a

FIG. 5. Graph depicting CBF of patients with diffuse TBI. The
CBF measured well above ischemic levels of 18 ml/100 g/minute
acutely or for as long as 200 hours postinjury. Thus, the ADC val-
ues measured at the time of the study were not influenced by isch-
emic levels of CBF.
predominantly vasogenic edema. Beyond the core, in the
perilesional area, the ADC was below normal (mean 0.72 6
0.14). The value was significantly different from the ADC
in the contused site and in normal brain tissue (p , 0.001),
signifying a predominantly cellular edema. The ADC of the
perilesional area was also low, compared with the ADC in
the symmetrical site in the contralateral hemisphere (p =
0.01). In the region distant from the contusion on the ipsi-
lateral side, the ADC (mean 0.84 6 0.08) was approxi-
mately equal to the ADC observed in the noncontused
hemisphere. In general, the ADC value was high in the site
of contusion and was reduced or near normal in the remain-
ing ROIs, including the perilesional brain tissue, tissue dis-
tant from the lesion on the ipsilateral side, and tissue in the
contralateral hemisphere.
As an example, the brain images obtained 66 hours post-
injury in a 41-year-old man with focal injury are shown in
Fig. 9. At the time of the study, the ICP was 24 mm Hg and
CBF was nonischemic at hemispheric levels of 33.2 ml/100
g/min (injured hemisphere) and 35.3 ml/100 g/min. The
water map in the vicinity of the lesion showed that water
content was high and averaged 82%, compared with 76% in
an equivalent ROI in the contralateral hemisphere. The per-
centage of swelling was 3.97% in the injured hemisphere
and 9.4% in the total brain. The ADC also varied, based on

724 J. Neurosurg. / Volume 104 / May, 2006

Apparent diffusion coefficient in patients with head injury

FIG. 7. Graph depicting ICP at the time of the study, compared

with the percentage of water content. The exponential relationship
shown is similar to that depicted in a pressure-volume curve. The
ICP increase was substantial, higher than a level of 77% water, cor-
FIG. 6. Graph depicting brain water content (%, shown on y axis) responding to an increase in brain volume of 4.3%. Beyond that
measured in patients with focal injury. Brain water content on the level, ICP increased rapidly with relatively small increases in brain
side of the contusion was markedly elevated, compared with the edema.
noninjured hemisphere (p , 0.0001).
matic brain edema is predominantly a cellular phenomenon
distance from the lesion. In the lesion core, the maximum in both focal and diffuse injury. Furthermore, the findings
ADC was 1.03, resulting from effects of a combination of indicate that at the time of the study, when CBF measures
blood, necrotic tissue, and vasogenic edema. This combi- were made in conjunction with MR imaging studies, CBF
nation can be seen as a hypodense region on the CT image measures were well above ischemic thresholds, thereby
obtained at the time of the study (Fig. 9 upper right). How- obviating an ischemia-induced cellular component to the
ever, in the ipsilateral hemisphere, in the perilesional ar- edematous process. The areas of edema were signified by
ea, the ADC was reduced to a level of 0.54, signifying zones of increased water content, measured by our MR im-
a predominant cellular edema. Distal to the lesion and in the aging technique, and, as might be expected, the highest lev-
contralateral hemisphere, the ADC values were nearly nor- els of edema were measured in patients with raised ICP. The
mal. relationship of ICP to the percentage of water content in pa-
In interpreting measures of ADC within the contusion
core, it is important to note that the ADC of hematoma
within the core and the ADC of the brain parenchyma may
not be the same. This situation is illustrated in Fig. 9, where
the ROI of the contusion includes the contused brain paren-
chyma as well as small islands of high contrast that most
probably are small hematomas. In assessing the ADC of the
contusion core, we excluded large hematomas within the
core.
Cerebral Blood Flow in Patients With Focal Injury
The mean CBF in the noninjured hemisphere of the pa-
tients with focal injury was 45.86 6 9.43 ml/100 g/min,
which is in the expected range for comatose patients35 (Fig.
10). In contrast, the CBF in the perilesional area was low
(mean 34.01 6 8.52 ml/100 g/minute), compared with the
noninjured hemisphere (p , 0.001). A similar reduction
was seen when these CBF values were corrected and nor-
malized to a PaCO2 of 34 mm Hg, assuming a 3% change
in CBF per millimeter of mercury of PaCO2, which was al-
so used for normalization of CBF values in the diffuse inju-
ry group. In both cases, the CBF was higher than ischemic FIG. 8. Graph depicting the ADCs (y axis) in four ROIs of con-
tusions: core, perilesional area, tissue distant from the contusion on
threshold levels, supporting the notion that the edema dis- the ipsilateral hemisphere, and equivalent tissue in the contralateral
tant from the contusion core is cellular in nature. hemisphere. Compared with the ADC values in the core, the ADC
values of the remaining ROIs were reduced. The ADC of the perile-
Discussion sional area was reduced, compared with tissue distant from the le-
sion and to the contralateral tissue (p , 0.01). Thus, in focal injury
In this report we provide supportive evidence that trau- the edema surrounding the contusion was predominantly cellular.

J. Neurosurg. / Volume 104 / May, 2006 725

 A. Marmarou, et al.

FIG. 9. Computerized tomography (upper) and MR (lower) images obtained in a 41-year-old patient with focal injury
measured 5 days postinjury. Water content varied from 82% in the lesion core to 76% in the equivalent ROI of the con-
tralateral hemisphere. The ADC was maximum in the lesion core (1.03 3 1023 mm2/second), signifying a predominantly
vasogenic edema. Proximal to the lesion site the ADC was reduced (0.54 3 1023 mm2/second) over a wide area, consis-
tent with a predominantly cellular edema. Distant from the lesion site and in the contralateral hemisphere, the ADC values
were slightly above normal. The CBF levels in the regions of reduced ADC measured both at 24 hours and at the time of
the study were well above the ischemic threshold, measuring 33.2 ml/100 g/min and 35.3 ml/100 g/min, respectively.

tients with focal injury was exponential, illustrating a sig-

FIG. 10. Graph depicting CBF values in the contused site. The
CBF values were lowest in the core. However, the CBF values in
the perilesional zone and contralateral tissue in regions where the
ADC was reduced were well above the ischemic threshold of 18 ml/
100 g/min. The CBF of the perilesional area was low, compared
with the CBF of the symmetrical site in the contralateral hemisphere
(p , 0.001).
nificantly reduced level of volume reserve in the contused
hemisphere. Taken together, these results suggest that alter-
native therapies must be considered for reducing the cellu-
lar edema that is a component of swelling in both diffuse
and focal injury.
Brain Swelling in TBI
In earlier studies, it was generally accepted that the swell-
ing process accompanying TBI was mainly due to vascular
engorgement and that the increase in blood volume was
mainly responsible for the increase in brain bulk and the
subsequent rise in ICP. Edema was thought to play a minor
role, particularly in contusions, where the BBB was thought
to be transiently compromised. This controversy was re-
solved by studies in patients with head injuries in whom
both tissue water and cerebral blood volume were mea-
sured.30 It was found in both diffuse and focal injury that
edema and not vascular engorgement was responsible for
brain swelling and that blood volume was actually reduced
after severe TBI. The question of which type of edema was
involved in this process remained. By definition, brain ede-
ma is an abnormal accumulation of fluid within the brain
parenchyma that produces a volumetric enlargement of the
brain tissue. The view that the edema occurs secondarily to

726 J. Neurosurg. / Volume 104 / May, 2006

Apparent diffusion coefficient in patients with head injury
BBB opening and that this fluid has a vascular origin led to
its classification as “vasogenic” edema by Klatzo.20 In con-
trast, the water contained in swollen cells was referred to
as “cytotoxic.” The most common cytotoxic edema is ob-
served with cerebral ischemia, where interruption of energy
supply leads to pump failure and an intracellular increase in
sodium and water. To further complicate the classification
process, with reperfusion, the neurotoxic effects of excita-
tory amino acids, particularly glutamate, are considered to
contribute to a “neurotoxic” edema that is characterized by
early astrocytic and especially dendritic swelling. However,
it is possible that all three forms of edema are present, yet
heretofore their relative contribution to swelling and the rise
in ICP has been difficult to quantify.
Laboratory Studies of Edema Classification
Experimental studies in the laboratory directed toward
characterization of the type of edema have used DW im-
aging to identify the extent of the cellular edema present in
an impact-acceleration model of diffuse injury.17 Authors
of these studies have reported that the rise in ICP and the
concomitant reduction in ADC after experimental TBI are
caused by a predominantly cellular edema. When coupled
with secondary insult, ischemia was also considered an im-
portant factor contributing to the cellular swelling. Sim-
ilarly, studies using the cortical contusion model of focal
injury10 and DW imaging have indicated that the type of
edema in the injured area, with or without the superimposed
secondary insult of hypotension, was predominantly cellu-
lar.37 However, it would seem reasonable that the formation
of vasogenic and cellular edema would also be time depen-
dent. Barzo, et al.,6 have used DW imaging and water con-
tent measures to study the type of edema that forms during
both the acute and chronic stages of diffuse impact-acceler-
ation injury. They found a significant increase in ADC dur-
ing the first 60 minutes after injury, which is consistent with
vasogenic edema development secondary to BBB compro-
mise. The transient increase in ADC was followed by a con-
tinuing decrease in ADC that had begun 40 to 60 minutes
after injury and continued for as many as 7 days after inju-
ry. This biphasic change in ADC has also been observed
in fluid-percussion injury.2 The study by Barzo, et al.,6
showed that edema with a vasogenic component can devel-
op soon after injury, but as the BBB closes, cellular edema
predominates. These studies have implied an early closure
of the BBB after TBI. The changes in the BBB have been
measured in a subsequent experimental study by the same
Group in which MR imaging is performed after injection of
a contrast agent.5 The authors found that closed head injury
was associated with a rapid and transient BBB opening that
lasted only 30 minutes. Secondary insult tended to prolong
the duration of the BBB opening. Thus, the increase in bar-
rier permeability in diffuse injury was short-lived, despite
the continued development of brain swelling, further sup-
porting a gradually developing cellular edema. These find-
ings of reduced ADC in experimental cortical contusion in-
jury have been further substantiated by Stroop, et al.,43 who
observed an early rise in ADC followed by ADC reduction
by 90 minutes after injury. Taken together, the experimental
findings provide compelling evidence that the BBB opening
is brief and the subsequent swelling of the brain is due to an
increase in cellular water.
J. Neurosurg. / Volume 104 / May, 2006
The Use of DW Imaging in Patients
The information regarding ADC changes in patients with
severe brain injury is scant. Liu, et al.,28 have studied nine
patients ranging in age from 26 to 78 years who presented
with head trauma. No data were available on the severity of
injury. The authors reported decreased ADC in patients with
diffuse axonal injury in the acute setting, although the num-
ber of patients in their study was small. In some cases the
reduced ADC persisted into the subacute period, beyond the
duration described for cytotoxic edema. The presence of
cytotoxic edema could not be confirmed because no water
measurements were made. The ADC values ranged from a
low of 0.334 at 9 days postinjury to 0.551. A more exten-
sive DW imaging study in patients with head injury has
been conducted by Kawamata, et al.18 They examined 20
patients within 24 to 48 hours posttrauma and determined
the ratio of ADC in the contusion site to that in healthy brain
tissue. The mean ADC value increased in the central area to
1.13 6 0.13 and decreased in the peripheral area to 0.83 6
0.81. The authors concluded that a combination of events
leads to edema fluid accumulation in the central area and
that this edema, together with the cellular edema in the pe-
ripheral area, contributes to the mass effect of contusion
edema. Nakahara, et al.,36 have examined ADC values in
four patients with severe brain injury, compared with four
controls. They found that ADC values in the more severely
injured patients were lower compared with those in the re-
maining patients.
In summary, the findings in investigations of severely
brain injured patients by other researchers are consistent
with our current finding that ADC is reduced after severe
TBI.
Correlation of Brain Edema and ICP With CT Image
Classification
To our knowledge, our study is the first to quantify brain
water using MR imaging methods developed previously.15,30
Using these methods, we could clearly study the regions
of reduced ADC and determine if they were related to in-
creased edema. Most studies have used an increase in T2 re-
laxation times as an indicator of edema. In our study, the
water content was increased in areas of reduced ADC, con-
firming that edema was present. In patients classified as
having Type II head injury according to the Marshall CT
image classification, a relatively low level of edema was
found, consistent with the low ICP. However, as the severi-
ty of diffuse injury increased in patients with Type III or IV
head injury, the water content increased to the equivalent of
an 8.2% swelling (p , 0.001), which was sufficient to raise
the ICP to 20 mm Hg or higher. It is interesting to note that
the ICP rise initiated when edema was only 1% higher than
the normal water content of brain tissue was equivalent to a
4.3% increase in tissue volume. This relationship empha-
sizes the extreme sensitivity of ICP to small changes in in-
tracranial volume, especially in the presence of a diminish-
ing volume reserve.
Apparent Diffusion Coefficient and Brain Edema
As mentioned earlier, researchers in previous studies
have attempted to relate the increased T2 signal relaxation to
increased edema. In our study, in which water was quanti-
727

fied in absolute terms, the regions of ADC change from the
mean normal value of 0.82 6 0.05 correlated with water in-
crease. For example, in patients with diffuse injury with-
out swelling, the ADC values were close to normal (mean
0.89 6 0.08). However, in patients with major brain swell-
ing, ADC values were reduced to a mean value of 0.74 6
0.05, and this reduction was highly statistically significant
(p , 0.0001) and was consistent with a cellular edema. It
must be emphasized that the measured ADC combines con-
tributions from both the vasogenic and cellular influence on
the diffusion of water through the tissue. If, for example, the
vasogenic and cytotoxic influences were equal, the ADC
would show no change from normal. For this reason, we
stated that cellular edema is the predominant form of ede-
ma present in patients with brain swelling. In summary,
the ADC, brain water content values, and Marshall CT scan
classifications were consistent and changed in the expected
direction. In patients with Type II head injury, where swell-
ing was minimal, the ADC change was also minimal. How-
ever, in patients with Type III and IV head injury with in-
creased brain swelling and increased water content, the
ADC was markedly reduced.
Changes in CBF and ADC
In our study, CBF was measured at the same time that
DW imaging was performed. This strategy was primarily
intended to eliminate frank ischemic damage as a factor
affecting the change in ADC. We found that CBF was re-
duced in patients with brain swelling and lower ADC val-
ues, but this reduction did not reach statistical significance.
More important, all CBF values were higher than the isch-
emic threshold of 18 ml/100 g/min. Thus, the ADC reduc-
tion could not be attributed to reduced CBF at the time of
the study. It does not preclude the occurrence of an ischemic
event in the early period after injury, followed by recovery
to CBF levels higher than the ischemic threshold. Early
CBF measures obtained soon after admission would clearly
help resolve this issue. Another confounding factor is that
CBF is considered to be age-dependent, with younger pa-
tients tending to have higher CBF values after trauma. In a
study by Adelson, et al.,1 of children with head injury, the
mean 6 SEM CBF value on admission was 25.1 6 7.7 ml/
100 g/min and increased to 55.3 6 3.4 ml/100 g/min by
24 hours and as many as 6 days after TBI. These initial val-
ues were not significantly different from those reported by
Schroder, et al.,40 who have found that the mean CBF val-
ue measured less than 4 hours after head injury in adults
was 32 ml/100 g/min in survivors and 20 ml/100 g/min in
nonsurvivors. Zwienenberg and Muizelaar46 have reviewed
available evidence and have concluded that hyperemia may
not be as common in severe pediatric injury as previously
thought and that the treatment of children with head injury
should be similar to that of adults with head injury. In view
of these studies, more work is needed to resolve the issue of
the effects of age and CBF on ADC in children with TBI.
Water Content, ICP, and CT Image Classification in
Focal TBI
The type of edema that surrounds a contusion is of great
interest because the expansion of the lesion leads to brain
shift, ICP, and eventually ischemia. Our study results sug-
728
A. Marmarou, et al.
gest that the contusional area can be divided into three
zones: the lesion core, the perilesional area, and the sur-
rounding tissue. We found that water content and the ADC
were elevated in the lesion core, consistent with the pres-
ence of a zone of necrotic tissue coupled with a mixture of
blood and water. It is interesting to note that the ADC was
reduced in the perilesional area as well as in tissue distant
from the lesion. As the water content increased in the con-
tused hemisphere, the ICP increased exponentially, and the
curve had a shape similar to the well-known ICP-volume
curve. This finding suggests that when the volume reserve
is exhausted, small increases in water content result in dra-
matic increases in ICP.
The CBF changes followed a distinct pattern in patients
with focal TBI. In the center of the lesion, the CBF was
reduced below ischemic levels. In the perilesion area, the
CBF was higher and measured approximately 35 ml/100 g/
min. The CBF in the perilesional area, although higher than
the ischemic threshold, was associated with reduced ADC,
indicating that the edema formed in the area surrounding the
contusion was predominantly cellular. The highest CBF
was found, as expected, in the contralateral hemisphere and
was near expected levels for patients with head injury.
Previously it was thought that changes in density ob-
served in CT images of areas surrounding a contusion were
representative of a vasogenic edema exuding from the le-
sion site and migrating through the tissue. Although a vaso-
genic component cannot be excluded, our results provide
compelling evidence that cellular edema predominates in
both diffuse and focal injury.
Factors Responsible for Cellular Edema
Except for the CBF measures in the lesion core in focal
injury, all CBF measures in this study were higher than the
ischemic threshold of 18 to 20 ml/100 g/min, regardless of
the type of injury. Astrup, et al.,3,4 have reported that flat-
tening of electroencephalography readings in humans oc-
curs when CBF measures 16 to 17 ml/100 g/min. At these
levels, the concentration of cellular potassium in the cortex
remained normal in baboons or only slightly elevated at the
threshold when brain electrical activity ceased. Moreov-
er, an increase in the extracellular potassium concentration,
which indicates pump failure, did not occur until blood flow
was further reduced to 10 ml/100 g/min. Thus, according to
these early studies, electrical activity, energy metabolism,
and ion pumping respond almost immediately to a reduced
supply of O2, whereas the development of infarction ap-
pears to occur over a longer time period. These findings
suggest that brain swelling secondary to pump failure could
coincide with a reduced O2 supply and without frank isch-
emia. If we posit that the majority of the patients in our
study had no frank ischemic episode, how then can cellu-
lar swelling occur? We speculate that mitochondrial dys-
function prevents the restoration of ionic and cell volume
homeostasis. Experimental studies have shown that NAA, a
marker of mitochondrial dysfunction, is reduced in TBI.42
The NAA level does not recover in severely injured ani-
mals, but NAA reduction is reversible in moderately injured
animals. Similarly, the reductions of NAA follow those of
adenosine 59-triphosphate, further implicating mitochondri-
al impairment. More work is required to shed light on the
causes of cell swelling in TBI in humans.
J. Neurosurg. / Volume 104 / May, 2006
Apparent diffusion coefficient in patients with head injury
Conclusions
The brain swelling seen in patients with severe brain in-
jury appears to be cellular, as signaled by a reduction in the
ADC in regions with increased water content. The cellular
form of edema predominates in both focal and diffuse brain
injury. Cerebral blood flow in these regions is above the
threshold for ischemia, indicating that other mechanisms re-
sulting in a tissue energy crisis must be involved.
References
1. Adelson PD, Clyde B, Kochanek PM, Wisniewski SR, Marion
DW, Yonas H: Cerebrovascular response in infants and young
children following severe traumatic brain injury: a preliminary re-
port. Pediatr Neurosurg 26:200–207
2. Albensi BC, Knoblach SM, Chew BG, O’Reilly MP, Faden AI,
Pekar JJ: Diffusion and high resolution MRI of traumatic brain in-
jury in rats: time course and correlation with histology. Exper
Neurol 162:61–72, 2000
3. Astrup J, Siesjo BK, Symon L: Thresholds in cerebral ischemia-
the ischemic penumbra. Stroke 12:723–725, 1981
4. Astrup J, Symon L, Branston NM, Lassen NA: Cortical evoked
potential and extracellular K1 and H1 at critical levels of brain
ischemia. Stroke 8:51–57, 1977
5. Barzo P, Marmarou A, Fatouros P, Corwin F, Dunbar J: Magnetic
resonance imaging-monitored acute blood-brain barrier chang-
es in experimental traumatic brain injury. J Neurosurg 85:
1113–1121, 1996
6. Barzo P, Marmarou A, Fatouros P, Hayasaki K, Corwin F: Con-
tribution of vasogenic and cellular edema to traumatic brain swell-
ing measured by diffusion-weighted imaging. J Neurosurg 87:
900–907, 1997
7. Becker DP, Miller JD, Ward JD, Greenberg RP, Young HF, Saka-
las R: The outcome from severe head injury with early diagnosis
and intensive management. J Neurosurg 47:491–502, 1977
8. Benveniste H, Hedlund LW, Johnson GA: Mechanism of detec-
tion of acute cerebral ischemia in rats by diffusion-weighted mag-
netic resonance microscopy. Stroke 23:746–754, 1992
9. Dijkhuizen RM, de Graaf RA, Tulleken KA, Nicolay K: Changes
in the diffusion of water and intracellular metabolites after exci-
totoxic injury and global ischemia in neonatal rat brain. J Cereb
Blood Flow Metab 19:341–349, 1999
10. Dixon CE, Clifton GL, Lighthall JW, Yaghmai AA, Hayes RL: A
controlled cortical impact model of traumatic brain injury in the
rat. J Neurosci Methods 39:253–262, 1991
11. Ebisu T, Naruse S, Horikawa Y, Ueda S, Tanaka C, Uto M, et al:
Discrimination between different types of white matter edema
with diffusion-weighted MR imaging. J Magn Reson Imaging 3:
863–868, 1993
12. Elliott KAC, Jasper H: Measurements of experimentally induced
brain swelling and shrinkage. Am J Physiol 157:122–129, 1949
13. Fatouros PP, Inao S, Marmarou A, Kraft KA: In vivo character-
ization of infused edema by magnetic resonance imaging. Adv
Neurol 52:533–537, 1990
14. Fatouros PP, Marmarou A: Experimental studies for use of mag-
netic resonance in brain water measurements. Acta Neurochir
Suppl 51:37–38, 1990
15. Fatouros PP, Marmarou A: Use of magnetic resonance imaging
for in vivo measurements of water content in human brain: meth-
od and normal values. J Neurosurg 90:109–115, 1999
16. Fatouros PP, Marmarou A, Kraft KA, Inao S, Schwarz FP: In vivo
brain water determination by T1 measurements: effect of total wa-
ter content, hydration fraction, and field strength. Magn Reson
Med 17:402–413, 1991
17. Ito J, Marmarou A, Barzo P, Fatouros P, Corwin F: Characteriza-
tion of edema by diffusion–weighted imaging in experimental
traumatic brain injury. J Neurosurg 84:97–103, 1996
J. Neurosurg. / Volume 104 / May, 2006
18. Kawamata T, Katayama Y, Aoyama N, Mori T: Heterogeneous
mechanisms of early edema formation in cerebral contusion: dif-
fusion MRI and ADC mapping study. Acta Neurochir Suppl 76:
769–12, 2000
19. Kita H, Marmarou A: The cause of acute brain swelling after the
closed head injury in rats. Acta Neurochir Suppl 60:452–455,
1994
20. Klatzo I: Evolution of brain edema concepts. Acta Neurochir
Suppl 60:3–6, 1994
21. Kraft KA, Fatouros PP, Clarke GD, Kishore PR: An MRI phan-
tom material for quantitative relaxometry. Magn Reson Med 5:
555–562, 1987
22. Kuroiwa T, Nagaoka T, Ueki M, Yamada I, Miyasaka N, Akimoto
H: Different apparent diffusion coefficient: water content correla-
tions of gray and white matter during early ischemia. Stroke 29:
859–865, 1998
23. Langfitt TW: Etiology and management of acute brain swelling.
W V Med J 62:49, 1966
24. Langfitt TW, Marshall WJ, Kassell NF, Schutta HS: The patho-
physiology of brain swelling produced by mechanical trauma and
hypertension. Scand J Clin Lab Invest Suppl 102: XIV, 1968
25. Langfitt TW, Tannanbaum HM, Kassell NF: The etiology of acute
brain swelling following experimental head injury. J Neurosurg
24:47–56, 1966
26. Langfitt TW, Weinstein JD, Kassell NF: Cerebral vasomotor pa-
ralysis as a cause of brain swelling. Trans Am Neurol Assoc 89:
214–215, 1964
27. Langfitt TW, Weinstein JD, Sklar FH, Zaren HA, Kassell NF:
Contribution of intracranial blood volume to three forms of exper-
imental brain swelling. Johns Hopkins Med J 122:261–270,
1968
28. Liu AY, Maldjian JA, Bagley LJ, Sinson GP, Grossman RI: Trau-
matic brain injury: diffusion-weighted MR imaging findings.
AJNR Am J Neuroradiol 20:1636–1641, 1999
29. Loubinoux I, Volk A, Borredon J, Guirimand S, Tiffon B, Seylaz
J, et al: Spreading of vasogenic edema and cytotoxic edema as-
sessed by quantitative diffusion and T2 magnetic resonance imag-
ing. Stroke 28:419–427, 1997
30. Marmarou A, Fatouros PP, Barzo P, Portella G, Yoshihara M,
Tsuji O, et al: The contribution of edema and cerebral blood vol-
ume to traumatic brain swelling in head-injured patients. J Neuro-
surg 93:183–193, 2000
31. Marmarou A, Takagi H, Shulman K: Biomechanics of brain ede-
ma and effects on local cerebral blood flow. Adv Neurol 28:
345–358, 1980
32. Marshall LF, Marshall SB, Klauber MR, von Berkum Clark M,
Eisenberg HM, Jane JA, et al: A new classification of head injury
based on computerized tomography. J Neurosurg 75 (Suppl):
S14–S20, 1991
33. Mintorovitch J, Moseley ME, Chileuitt L, Shimizu H, Cohen Y,
Weinstein PR: Comparison of diffusion- and T2-weighted MRI
for the early detection of cerebral ischemia and reperfusion in rats.
Magn Reson Med 18:39–50, 1991
34. Moseley ME, Cohen Y, Mintorovitch J, Chileuitt L, Shimizu H,
Kucharczyk J, et al: Early detection of regional cerebral ischemia
in cats: comparison of diffusion- and T2-weighted MRI and spec-
troscopy. Magn Reson Med 14:330–346, 1990
35. Muizelaar JP, Fatouros PP, Schroder ML: A new method for
quantitative regional cerebral blood volume measurements using
computed tomography. Stroke 28:1998–2005, 1997
36. Nakahara M, Ericson K, Bellander BM: Diffusion-weighted MR
and apparent diffusion coefficient in the evaluaion of severe brain
injury. Acta Radiol 42:365–369, 2001
37. Portella G, Beaumont A, Corwin F, Fatouros P, Marmarou A:
Characterizing edema associated with cortical contusion and sec-
ondary insult using magnetic resonance spectroscopy. Acta Neu-
rochir Suppl 76:273–275, 2000
38. Rumpel H, Nedelcu J, Aguzzi A, Martin E: Late glial swelling
after acute cerebral hypoxia-ischemia in the neonatal rat: a com-
729

bined magnetic resonance and histochemical study. Pediatr Res
42:54–59, 1997
39. Schlaug G, Siewert B, Benfield A, Edelman RR, Warach S: Time
course of the apparent diffusion coefficient (ADC) abnormality in
human stroke. Neurology 49:113–119, 1997
40. Schroder ML, Muizelaar JP, Kuta AJ, Choi SC, et al: Thresholds
for cerebral ischemia after severe head injury: relationship with
late CT findings and outcome. J Neurotrauma 13:17–23, 1996
41. Sevick RJ, Kanda F, Mintorovitch J, Arieff AI, Kucharczyk J,
Tsuruda JS, et al: Cytotoxic brain edema: assessment with diffu-
sion-weighted MR imaging. Radiology 185:687–690, 1992
42. Signoretti S, Marmarou A, Tavazzzi B, Lazzarino G, Beaumont A,
Vagnozzi R: N-Acetylaspartate reduction as a measure of injury
severity and mitochondrial dysfunction following diffuse traumat-
ic brain injury. J Neurotrauma 18:977–991, 2001
43. Stroop R, Thomale UW, Pauser S, Bernarding J, Vollmann W,
Wolf KJ, et al: Magnetic resonance imaging studies with cluster
algorithm for characterization of brain edema after controlled cor-
tical impact injury (CCII). Acta Neurochir Suppl 71:303–305,
1998
730
A. Marmarou, et al.
44. Tsuura M: [Diffusion/perfusion MRI study on cerebral ischemia in
a rat embolism model.] No To Shinkei 48:659–666, 1996 (Jpn)
45. Yanaka K, Shirai S, Kimura H, Kamezaki T, Matsumura A, Nose
T: Clinical application of diffusion-weighted magnetic resonance
imaging to intracranial disorders. Neurol Med Chir (Tokyo) 35:
648–654, 1995
46. Zwienenberg M, Muizelaar JP: Severe pediatric head injury: the
role of hyperemia revisited. J Neurotrauma 16:937–943, 1999
Manuscript received November 23, 2004.
Accepted in final form January 11, 2006.
This work was partially supported by National Institutes of Health
Grant Nos. NS12587 and NS19235 (Anthony Marmarou, Ph.D, prin-
cipal investigator).
Address reprint requests to: Anthony Marmarou, Ph.D, Depart-
ment of Neurosurgery, Virginia Commonwealth University Medical
College of Virginia Campus, 1001 East Broad Street, Suite 235, P.O.
Box 980508, Richmond, Virginia 23298–0508. email: amarmaro@
vcu.edu.
J. Neurosurg. / Volume 104 / May, 2006