ation times was also observed. Key Words: Brain edema. Cerebral ischemia-Nuclear magnetic resonance imaging -Nuclear magnetic resonance spectroscopy.
Journal a/ Cerebral Blood Flow and Metabolism 6:212-221 © 1986 Raven Press, New York
Characterization of Experimental Ischemic Brain Edema Utilizing Proton Nuclear Magnetic Resonance Imaging
Hiroyuki Kato, Kyuya Kogure, Hitoshi Ohtomo, Masahiro Izumiyama, Muneshige Tobita, tShigeru Matsui, tEtsuji Yamamoto, tHideki Kohno, :j:Yoshinori Ikebe, and *Takao Watanabe Department of Neurology, Institute of Brain Diseases, and *Department of Environmental Health, Tohoku University School of Medicine, Sendai, tCentral Research Laboratory, Hitachi Ltd., Tokyo, and tHitachi Naka Works, Katsuta, Japan
Summary: Correlations between Tl and T2 relaxation times and water and electrolyte content in the normal and ischemic rat and gerbil brains were studied by means of both nuclear magnetic resonance (NMR) spectroscopic and imaging methods. In the spectroscopic experiment on excised rat brains, Tl was linearly dependent on tissue water content and T2 was prolonged in edematous tissue to a greater extent than expected by an increase in water content, showing that T2 possesses a greater sensi tivity for edema identification and localization. Changes in Na+ and K+ content of the tissue mattered little in the prolongation of relaxation times. Serial NMR imaging of gerbil brains insulted with permanent hemispheric isch-
emia offered early lesion detection in TJ- and especially T2-weighted images (detection as soon as 30 min after insult). The progressive nature of lesions was also imaged. Calculated TJ and T2 relaxation times in regions of interest correlated excellently with tissue water con tent (r = 0.892 and 0.744 for T 1 and T2, respectively). As a result, detection of cerebral ischemia utilizing NMR imaging was strongly dependent on a change in tissue water content. The different nature of Tl and T2 relax ation times was also observed. Key Words: Brain edema Cerebral ischemia-Nuclear magnetic resonance imaging-Nuclear magnetic resonance spectroscopy Relaxation times.
Proton nuclear magnetic resonance (NMR)
selection of appropriate pulse sequences is neces
imaging enables sensitive detection of cerebral
sary to highlight pathology (Crooks et aI. , 1982;
ischemia within several hours after the onset of dis
Wehrli et aI. , 1984). Therefore, intrinsic NMR pa
ease (Buonanno et aI. , 1983; Mano et aI. , 1983;
rameters such as Tl and T2 relaxation times and
Spetzler et aI. , 1983). T1 (spin-lattice) and T2
proton density (PD) are calculated from acquired
(spin-spin) relaxation times are prolonged in isch
images to obtain indexes that can express an abso
emic tissue, and the degree of damage and extent of
lute degree of tissue damage (Mano et aI. , 1983;
injury can be pictorialized by NMR imaging. Pro
Wehrli et aI. , 1984).
longation of relaxation times is considered to be a
In a previous communication, we reported that
result of the development of brain edema (Go and
proton NMR images of experimental cerebral in
Edzes, 1975; Naruse et aI. , 1982). However, the
farction consistently correspond with retrospective
contrast of the NMR image is arbitrarily controlled
histochemical observations in regard to the degree
by the pulse sequence used. As a consequence, the
and extent of brain edema (Kato et aI. , 1985). To further clarify the mechanisms that cause abnor
Received July 3, 1985; accepted November I, 1985.
mality in the NMR phenomenon pertaining to cere
Address correspondence and reprint requests to Dr. H. Kato
bral ischemia, we performed the following experi
at Department of Neurology, Institute of Brain Diseases, T o
ments in which, using rat and gerbil brains in exper
hoku University School of Medicine, I- I Seiryo-machi Sendai
imental cerebral ischemia, we compared Tl and T2
980, Japan. Abbreviations used: IR, inversion recovery; NMR, nuclear
relaxation times with tissue water and electrolyte
magnetic resonance; PO, proton density; SE, spin echo; SR, sat
content by means of both in vitro spectroscopic and
uration recovery; T E, time to echo; n, time of inversion; T R,
in vivo imaging methods.
time of repetition.
212
213
NMR IMAGING OF CEREBRAL ISCHEMIA
MATERIALS AND METHODS NMR spectroscopic experiment on excised rat brains Adult male Wistar rats weighing 250-300 g were in sulted with forebrain ischemia by the four-vessel occlu sion method (Pulsinelli and Brierley, 1979). The vertebral arteries were electrocauterized at the first vertebral level while the animal was under pentobarbital anesthesia. On the following day, the animal was paralyzed and mechan ically ventilated under 70% nitrous oxide. 30% oxygen, and 1 % halothane anesthesia. The right femoral artery was cannulated with PE-50 polyethylene tubing to obtain arterial blood samples and to record blood pressure. Body temperature was maintained at 37 ± 0.5°C by using a heat pad. The common carotid arteries were then oc cluded for 60 min, followed by 60 min of blood flow res toration. Those animals whose electroencephalograms became isoelectric within 30 s after occlusion were se lected for continued use in the experiment. At the end of the experiment, decapitation was performed and each brain was quickly removed; the cortical gray matter, sub cortical white matter. hippocampus, and thalamus were dissected out in a humid glove box. Animals treated as above except for the receiving of an ischemic insult served as controls. Three groups (n = 4-6) were used for the measure ment of water, Na + and K + content, and T1 and T2 re laxation times. For water and electrolyte content mea surement, brain tissue was frozen in liquid nitrogen im mediately after dissection, weighed at - 2SOC, and dried at 100°C, thus calculating water content by the wet-dry method. The sample size was 25-75 mg wet weight. The dried samples were resolved in HN03 and HCl04 and the supernatant was used to determine Na + and K + content by the atomic absorption method (Hitachi Atomic Ab sorption Spectrophotometer 208). For relaxation time measurement, samples were transferred to a sealed glass tube. A T1 or T2 value was determined by the inversion recovery and the modified Carr-Purcell-Meiboom-Gill pulse sequences, respectively. The NMR spectrometer (Hitachi R-90H) was operated at 2.1 tesla (proton reso nant frequency of 90 MHz). Samples were kept at O°C until measurement, and were then measured at 35°C. All procedures were completed within at most 2 h after death. Intergroup difference was statistically measured by means of the Student t test for unpaired samples.
tify the area of ischemia. The rest of each block was di vided into cortex, hippocampus, and thalamus (� 1 O mg wet weight) to determine water content by the wet-dry method and the electrolyte content by the atomic absorp tion method, as described in the preceding section. The NMR miniimager (Central Research Laboratory, Hitachi Ltd., Tokyo, Japan) consisted of a supercon ducting magnet operating at 0.5 tesla (proton resonant frequency of 2 1 .3 MHz). Coronal scans 5 mm thick at the thalamus level were performed. The following three modes of pulse sequences were employed: (a) PD weighted saturation recovery (SR) images (TR = 1 .6 s, TE' = 14 ms); (b) T1-weighted inversion recovery (IR) images (TR = 1 .6 s, TI = 300 ms, TE' = 14 ms); and (c) T2-weighted spin echo (SE) images (TR = 1 .6 s, TE = 1 06 ms). The definitions of abbreviations are as follow (as defined by the American College of Radiology): TR, time of repetition; TI, Tl-weighting time or the time of inver sion; TE, T2-weighting time or the time to echo. In modes (a) and (b), signals were collected as SEs (TE' = 14 ms). Each scanning was completed in 1 5 min. In actu ality, in the first set of imagings, SE, IR, and SR pictures were scanned, for example, from between 25-40, 40 - 55, and 55 - 70 min after ischemia, respectively. Increased signal intensity of the image (brightness) corresponded to increased PD and prolonged Tl and/or T2 relaxation times. T1 and T2 values in regions of interest were calcu lated by the pixel-to-pixel computation applying the fol lowing equations: I(IR)
leSE)
=
=
[
[
I(SR) 1 - 2 exp
I(SR)exp
-
( �� )] I
(TE - TEl) T2
]
in which I(SR), I(IR), and leSE) denote signal intensity in SR, IR, and SE images, respectively. Mean values of 25 pixels with a standard deviation were obtained.
RESULTS NMR spectroscopic experiment on excised rat brains The physiological parameters prior to and at the end of ischemia and at the end of reperfusion were as shown in Table I. The ischemic insult (60 min
NMR imaging experiment on ischemic gerbil brains
ischemia plus 60 min reperfusion) to produce max
Adult mongolian gerbils of both sexes weighing 60-80 g were anesthetized by ether and the right common carotid artery was occluded. Experimental protocol was applied only to those animals that exhibited definite hemiparesis such as abnormal limb posture, loss of righting reflex,
imal edema was so severe that some of the animals
circling behavior, etc., within a 5-min observation period (n = 4). Serial NMR scanning was performed under pen tobarbital anesthesia at three points during the ischemic period, i.e., at from 30 to 60 min after occlusion of the carotid artery, at �2 h, and again after �4 h. Two normal gerbils were imaged as controls. After completion of the imaging, brains were frozen in situ with liquid nitrogen by the transcalvarial freezing technique (Ponten et aI., 1 973). From the coronal plane corresponding to the imaged
miyama et aI., 1983; M. Izumiyama et aI., in prepa
level, sections 16 J-Lm thick were cut using a cryostat. These were for use in histochemical studies, i.e., ATP bioluminescence (Kogure et aI., 1980) and potassium staining (Mies et aI., 1 984) so as to histochemically iden-
could not maintain normal blood pressure. The pathophysiology of the Pulsinelli model revealed in our laboratory will be discussed elsewhere (Izu ration). In intact rat brains, TI and T2 relaxation times of gray matter and the hippocampus were longer than those of white matter and the thalamus (Table 2). The difference between them was greater for TI (15.5-24.5%) than for T2 (3.5-7.1%) (Table 3). The difference in water content was 4.7-7.0%. In the ischemia-loaded rat brain, T1 and T2 values of all parts of the brain were significantly prolonged, coinciding with the increase in water content (Table
J
Cereb Blood Flow Metab, Vol.
6,
No.2,
1986
214
H. KATO ET AL.
had no consistent relation with tissue water con
TABLE 1. Physiological variables
MABP (mm Hg)
Pa02 (mm Hg) PaC02 (mm Hg) pH
Before ischemia
Ischemia
Reperfusion
129± 3 140± 12 37± I 7.41±0.03
101± 22 115± 9 37± 2 7.36±0.02
110± 16 112± 5 39± 2 7.35±0.01
Values are means± SE for six experimental animals.
tent: K + content was decreased by ischemia in ac cordance with the initial K + content in each region. This resulted in poor correlations on the whole be tween Tl or T2 and K + content (Fig. 2).
NMR imaging of the ischemic gerbil brains Typical NMR images of a normal gerbil brain are iIIustrated in Fig. 3. In the SR image, the brain was homogeneously imaged. Intracerebral structures, i.e., the cortex, white matter, hippocampus, and
2). The difference of Tl, T2, or water content be
thalamus, were clearly delineated in the T 1-
tween gray matter or the hippocampus and white
weighted IR image and less clearly in the T2-
matter or the thalamus of ischemic brains was al
weighted SE image.
most noted. However, differences between normal
A typical sequential change in NMR pictures ob
and i s c h e m i c t i s s u e w e r e 7.1 -9.2% for Tl,
tained from a gerbil brain that had undergone per
8.8-13.0% for T2, and 2.4-3.4% for water content
manent hemispheric ischemia is shown in Fig. 4. IR
(Table 3). When T1 and T2 values were plotted
and especially SE images clearly depict the growing
against water content as illustrated in Fig. 1, T1
abnormality in the affected hemisphere, the first
was linearly dependent on it. In contrast, T2 values
abnormality having been observed in a T2-weighted
of each brain part independently exhibited prolon
image 30 min after ischemia. The right-to-left ratios of signal intensities in the
gation. Sodium and potassium content in normal rat
regions of interest (the cortex and thalamus) in SR,
brains showed differences between structures (gray
IR, and SE images are tabulated in Table 4. A sig
hippocampus> thalamus"" white matter)
nificant increase in signal intensity was recognized
as well as in water content (Table 2). In ischemic
in the thalamus in SE images 30 min after ischemia.
tissue, Na + content was increased and, coinciding
Calculated Tl and T2 values in regions of interest
with this, the K + content was decreased (Table 2).
as well as water and electrolyte content are tabu
Correlation between Na + and water content was
lated in Table
recognized but not linear (Fig. 2): Na + accumula
were progressively prolonged. A significant differ
tion compared with an increase in water content
ence was observed at 2 and 4 h after occlusion of
matter
=
5. Tl and T2 values in each region
was different between gray matter or the hippo
the carotid. Prolongation of T1 after 4 h of isch
campus and white matter or the thalamus. As a
emia, as compared with the normal value, was
consequence, Tl or T2 was correlated with tissue
and 12.2% for the cortex and thalamus, respec
Na + content, but the slope of change was different
tively. The values for T2 were 13.3 and 13.0%, re
between gray matter or the hippocampus and white
spectively.
matter or the thalamus (Fig. 2). Tissue K + content
6.5
The area of ischemia and brain edema was con-
TABLE 2. TJ and T2 relaxation times and water, Na + , and K+ content of gray matter, white matter, hippocampus, and thalamus of normal and ischemia-injured (60 min of forebrain ischemia followed by 60 min of reperfusion) rat brain
T2 (ms) (n = 5)
HP(%)
1.19± 0.01 0.98 ± 0.02 1.22± 0.02 1.03± 0.03
77.5±1.6 74.9± 3.4 78.4± 1.9 73.2± 2.3
79.9±0.2 74.4± 0.6 79.8± 0.3 75.9±1.0
225±12 165± 7 212±16 184± 12
485± 7 345 ± 20 465± 24 409± 33
1.30± 0.02a 1.05± O.03b 1.31± 0.02a 1.11± 0.05b
86.0± 3.1a 81.5± 4.4b 88.6± 2.0" 82.6± 4.1"
81.5± 0.4" 76.2± 0.8" 81.6±0.4" 78.4±1.2"
295± 6" 183±12" 274± 23a 218± 20"
431± 13a 308±11a 434 ± lOb 373 ± 16b
TI (s) = 4)
(n
(n
=
6)
Na � (mEq/kg) (n
=
6)
K - (mEq/kg) (n
=
6)
Normal Gray matter White matter Hippocampus T halamus Ischemia Gray matter White matter Hippocampus T halamus
Values are means ± SD. T I and T2 relaxation times were prolonged, water and Na+ contents were increased, and K + content was decreased by the ischemic insult, all with significance by Student's t test.
a p < 0.01 versus control values. b p < 0.05 versus control values.
J
Cereb Blood Flow Metab. Vol. 6, No.2, 1986
NMR IMAGING OF CEREBRAL ISCHEMIA
215
TABLE 3. Ratio (%) of re/axation times or water
immediately after the onset of ischemia in the pres
content between regions
ence of residual tissue perfusion, which supplies
T2
121.4
103.5
106.9
aI.,
Normal G/W
1974, 1981; Fujimoto et 1976; Mrsulja et aI., 1983). Proton NMR is re
edema fluid (Kogure et aI.,
H2O
Tl
markably sensitive and able to detect cerebral
G/T
115.5
105.9
104.7
edema associated with ischemia, as well as to sepa
H/W
124.5
104.7
107.0
HIT
118.4
107.1
104.9
rat e normal structures based on their different water contents. Therefore, proton NMR imaging is
Ischemia
able to pictorialize the early changes of cerebral
107.0
G/W
123.8
105.5
G/T
117.1
104.1
I03.S
H/W
124.S
IOS.7
107.1
HIT
IIS.O
107.3
103.9
G
109.2
111.0
102.5
W
107.1
IOS.8
102.4
H
107.4
113.0
102.5
ischemia. The mechanisms of this early detection
T
107.S
112.S
103.4
are discussed later in this section.
ischemia in vivo. Lesions were detected as early as
90 min (Spetzler et aI., 1983), 2 h (Buonanno et aI., 1983), and 3 h (Mano et aI., 1983) after ischemia. In the present study, we detected, with excellent spa
Ischemia-normal
tial resolution, abnormality as early as 30 min after
Although most of the proton NMR signals re
Values of relaxation times and water content are tabulated in
flect water in t he tissue, a considerable amount of
Table 2. G, W, H, and T denote gray matter. white matter. hip pocampus, and thalamus. respectively. and combined abbrevia tions denote ratios, e.g .. G/W
them from lipids may contribute. Recent studies on
means gray matter- to- white
proton chemical shift imaging revealed that signals
matter ratio.
from tissue s u c h as subcutaneous fat, bone marrow, and muscle are contaminated with a con siderable amount of signals from lipids (Pykett and Rosen,
firmed in every case by ATP and K + pictures, re spectively; an example is shown in Fig.
5. Water
water; those from lipids are negligible (Pykett and
contents of the cortex, hippocampus, and thalamus
Rosen,
are plotted against calculated T 1 and T2 values im mediately before death (Fig. tween them was excellent (r =
1983) because of a short T2 of membrane
lipids in the brain.
6). Correlation be 0.892 and 0.744 for
In the present NMR spect roscopic experiment on excised rat brains, in which only signals from water
Tl and T2, respectively). Correlations between TI
are concerned, TI and T2 relaxation times were
and T2 relaxation times and water and electrolyte
strongly correlated wit h tissue water content (Fig.
content were similar to those of the spectroscopic study (Fig.
1983; Sepponen et aI., 1984). However, in
the case of the brain, signals are entirely from
I), as has been previously reported (Go and Edzes, 1975; Naruse et aI., 1982). In a normal rat brain, the
7).
difference of Tl between gray matter or the hippo
DISCUSSION
campus and white matter or the thalamus (15.524.5%) was three to four times greater than that of water content (4.7-7.0%). In the case of T2, this
E nergy failure of the brain caused by cerebral ischemia facilitates the development of brain edema
T? (ms)
" (5)
y = 0.456)( r = 0.987
P < 0.001
+
l"' :t
FIG. 1. Spectroscopic T1 and T2 relaxation times were plotted against water content of t�e tissue. T1 was linearly dependent on water con tent. However, T2 values of each region were prolonged independently with each other, showing the mechanism of T2 prolongation to be not due solely to the increase in water content. G, W, H, and T denote gray matter, white matter, hippocampus, and thalamus, respectively. Primes denote ischemia.
y c 1.45x p < 0.05
J
Cereb Blood Flow Metab, Vol. 6. No.2, 1986
216
H. KATO ET AL.
Na
c:
(m�.q/kg)
y
_
W: y
�
H"
Y
T: y "lOa
= �
= _
_
40.4x
-
c:
2999.3
( P
0.925
10.9:.: -
I
38.lx
0.971
500
\\ •
0
0
•
•
0
0
o
o
o o
200
00 o
• •
•
0
•
o
•
300
74
76
78
80
FIG. 2. Correlations between spectroscopic T1 and T2 relaxation times and water, Na , and K+ con tent of normal and ischemic (60 min ischemia plus 60 min reperfu sion) rat brain. (Open circles), gray matter; (filled squares), white matter; (filled c ircles), hippo campus; (open squares), thalamus. Primes denote ischemia. Na con tent was nonlinearly correlated with water content of the tissue, which resulted in a similar correla tion between Na' content and the T1 or T2 relaxation time. K' con tent showed poor correlation with water content as well as with T1 and T2 relaxation times.
•
0
0
0
o
0
0
0
0
�, 74
76
80
78
11
11 (.
0
400
)
(s)
1.3
1. 3
c'
�
1.2
1.2
1.1
1.1
t
y
r
1.0
I
=
0.0026x
co
0.933
p
