651

Acute Intracranial Hemorrhage:

Intensity Changes on Sequential MR Scans
at 0.5 T

Robert D. Zimmerman1 Thirty-seven patients underwent

MR imaging at 0.5 T within 7 days of a CT-docu-
Linda A. Heier1 mented intracranial hemorrhage.
A total of 57 hematomas were evaluated. Twelve
Robert B. Snow patients underwent serial scanning and 12 patients had multiple hemorrhages into
different intracranial compartments. The appearances of the hematomas on spin-echo
David P. C. Liu1
American Journal of Roentgenology 1988.150:651-661.

(SE) images with a short repetition time (TR) of 500 msec and short echo time (TE) of
Anna B. Kelly1
32 msec (SE 500/32), long TR/intermediate TE (SE 2000/60), and long TR/long TE (SE
Michael D. F. Deck1 2000/120) were carefully evaluated with specific attention to the precise time after ictus.
Hematomas showed heterogeneous, complex, rapidly changing intensities. There was
a significant amount of variation among patients, especially between the third and
seventh days. Hematomas studied between 12 and 24 hr after hemorrhage were mildly
hyperintense on short TR scans and markedly hyperintense on long TR (intermediate
and long TE) scans (stage I). These findings in acute hemorrhage have received lfttle
prior attention. Over the next 1-2 days, hematomas became iso- to mildly hypointense
on short TR scans and markedly hypointense on long TR scans (stage II). Hypointensity
on long TR scans has previously been described at high field strengths; our communi-
cation demonstrates that this phenomenon is seen routinely at intermediate field
strengths as well. Hematomas became markedly hyperintense on short TR scans
beginning on approximately the fourth day postictus and redeveloped hyperintensity on
long TR scans approximately 5-6 days after ictus (stage Ill). By the end of the first week
they were hyperintense on all pulse sequences (stage IV).
MR findings on the first day after intracranial hemorrhage (in particular, subtle
hyperintensity on short TR scans) probably allow for a specific diagnosis, while the
variable, heterogeneous, and rapidly changing intensities noted between days 2 and 7
are often less specific.

The MR appearance of intracranial hemorrhage has been the subject of many

This article appears in the January/February communications. Findings in acute hemorrhage, particularly within 48 hr of ictus,
1988 issue of AJNR and the March 1988 issue of remain controversial and incompletely delineated [1 -9]. The intensity of hematomas
AJR
is known to be dependent on multiple factors including time from ictus and pulse
Received March 1 3, 1987; accepted after revi-
sion August 12, 1987.
sequence employed [1 , 6, 7]. Recently, it has been suggested that the field strength
of the MR imager may also affect intensity such that “acute” (under 7 days)
Presented at the annual meeting of the American
Society of Neuroradiology. San Diego, January hemorrhage may be detected more easily at ultralow (0.02 1) [4] or high (1 .5 1)
1 986, and presented in part at the Symposium field strengths [5] than at intermediate field strengths (0.1 -0.6 T). Unfortunately, all
Neuroradiologicum, Stockholm, June 1986.
prior studies of acute hemorrhage are limited by the small number of patients
I Department of Radiology, New York Hospital-
evaluated in each series. Extrapolations from these limited data have led to
Cornell University Medical Center, 525 E. 68th St.,
New York, NY 10021 . Address reprint requests to confusing and often contradictory conclusions. To clarify the findings in acute
R. D. Zimmerman. hemorrhage on MA, we studied 37 patients within 1 week of an acute hemorrhagic
2 Department of Neurosurgery. New York Hos- ictus with specific attention to the precise relationship between the intensity of the
pital-Comell University Medical Center, New York, lesion and the time from ictus. The implications of these findings on previously
NY 10021.
presented theories of hematoma intensity will be discussed, as will the efficacy of
AJR 150:651-661, March 1988
0361 -803X/88/1 503-0651
MR and CT in the clinical evaluation of patients with suspected acute intracranial
0 American Roentgen Ray Society hemorrhage (AIH).
 652 ZIMMERMAN ET AL. AJR:150, March 1988

TABLE 1: Relationship Between Time After Ictus and Cause and Location of Hemorrhage

No. by Day
Cause/Location Total
1 2 3 4 5 6 7
Cause:
Trauma 0 2 3 6 2 3 5 21
Hypertension 1 2 2 1 1 0 0 7
Vascular 1 1 0 0 0 2 1 5
Neoplasm 0 1 0 1 0 0 0 2
ldiopathic/miscellaneousb 0 1 0 1 1 1 4 8
Location of hematoma:
Epidural 0 0 1 1 0 2 0 4
Subdural 1 1 1 5 2 1 5 16
Parenchymal 2 6 3 5 2 4 6 28
Subarachnoid 0 2 0 1 0 1 1 5
Intraventricular 1 2 1 0 0 0 0 4
Total 4 11 6 12 4 8 12 57
a Inciucies aneurysms, arteriovenous malformations, and venous angiomas.
b Includes hemorrhage without known cause, venous thrombosis, and amyloid angiopathy.

Subjects and Methods Short TR/Short TE Images

American Journal of Roentgenology 1988.150:651-661.

Thirty-seven patients were included in this study. In each case, a Hematomas evaluated 12-24 hr after an acute ictus (Fig.
CT-documented hemorrhage had occurred within 1 week of the initial 2) were subtly and uniformly hyperintense relative to white
MR. The time from clinical ictus to MR imaging was identified as matter in three examples of AIH (Figs. 2C and 2D) and
precisely as possible, as were the cause and location of the hemor-
isointense in one case. The average intensity of AIH dimin-
rhage (Table 1). All patients were evaluated on a O.5-T MR imaging
ished over the next 2 days (Fig. 1). Appropriately timed serial
device.” When hematomas in multiple compartments (1 2 cases) were
studies were not performed in any of these patients and,
present, these were evaluated separately, since they often had
different intensities (see Results). Serial studies (1 2 cases) were also therefore, the diminishing intensity could not be demonstrated
evaluated individually and included within the overall tabulation. Thus, directly. Lesions studied at 24-48 hr had variable intensity,
the total number of hematomas evaluated (57) exceeded the number ranging from mild hypo- to mild hyperintensity (Fig. 3). The
of patients (37). Ti -weighted spin-echo (SE) images with a short two hematomas that were hyperintense were evaluated early
repetition time (TA) of 500 msec and short echo time (TE) of 32 msec on the second day (30 and 32 hr, respectively). Hematomas
(SE 500/32) were obtained in all cases. Moderately T2-weighted long studied on the third to fourth postictal day (Figs. 3-5) dem-
TR/intermediate TE (SE 1 500/90) single-echo images were obtained onstrated the greatest variation; mild hyperintensity was the
in the initial six patients, while the remaining 31 patients had multiecho most commonly encountered pattern (Figs. 4C and 5B). Fea-
scans with a TA of 2000 msec and two echoes (TE 60-i 20 msec)
=
tures were similar to those seen on day 1 , except that small
or four echoes (TE 30-i 20 msec). These scans were subdivided
=
foci of more marked hyperintensity were often seen within
into long TA/intermediate TE images (for example, SE 2000/60), in
which spinal fluid was gray, and long TA/long TE images (for example,
these older hematomas (Fig. 3G). Thereafter, there was a
SE 2000/i 20), in which spinal fluid was white. The intensity of the dramatic and progressive increase in intensity (Fig. 6B), which
hemorrhages was qualitatively compared with that of white matter was most prominent near the periphery of the hematoma.
on all images. More precise quantitative measurements were not This phenomenon was directly observed in five patients stud-
attempted, because hematomas showed quite heterogeneous inten- ied serially within the first week of ictus (Figs. 4E and 5E) and
sities. The MA findings on each pulse sequence were compared with in another four patients in whom follow-up examinations were
each other and with CT findings to determine the relative clinical performed after 1 week (1 0-28 days).
efficacy of each pulse sequence and of MA, in general, when com-
pared with CT in the evaluation of AIH.
Long TA/Intermediate to Long TE Images

Results Hematomas evaluated within 24 hr of ictus were typically

hyperintense (Figs. 2E-2H); hypointensity was present in only
Hematomas undergo rapid, reversible changes in intensity
one case studied at 24 hr. Thereafter, there was a dramatic
that vary with the pulse sequences used. Therefore, relation-
decrease in intensity. Hypointensity was present in seven of
ships between time and intensity for each pulse sequence
1 1 hemorrhages evaluated during the second postictal day
have been tabulated and are presented in graphic form (Fig.
(Fig. 1). The four hematomas without hypointensity were in
1). The effects of two additional factors, recurrent hemorrhage
two patients with multiple hemorrhages evaluated before 36
and site of hemorrhage, were also evaluated.
hr. On the third postictal day, hypointensity was seen in five
of six hematomas, and the hypointense components were
“ Technicare. larger and/or darker than those seen on day 2 (Fig. 4D). The
 AJR:150, March 1988 MR OF ACUTE INTRACRANIAL HEMORRHAGE 653

o....o TR 500 I TE 32
‘-S TR 2000 I TE 60 Markedly
MarkedLy Hyperintense
Hyperintense J
Mildly
MildLy
Hyperintense
Hyperintense /‘........TR 500 I TE 30

ISOINTENSE -
ISOINTENSE I-

Mildly L
MIldly
Hypointense r Hypointerise
/ TR2Q00/TE6Q
Markedly Markedly / ---TR2000/TEL2Q
\----,
Hypointense Hypointense

1 2 3 4
---
5 6
I

7 1 2 3 4 5 6 7
Day Day
A B
Fig. 1.-Time/intensity curves for acute intracranial hemorrhage.
A, All hematomas are plotted. Intensity on each day is given relative to white matter. Range of intensities is shown by vertical bars and mean intensities
for all hematomas on each day are open circles (TR = 500 msec, TE = 32 msec) or solid circles (TR = 2000 msec, TE = 60 msec).
B, Mean intensities only are plotted for SE 500/32 and SE 2000/60-120 scans. Cases In which multiple episodes of hemorrhage were present have
been eliminated to simplify curves.
American Journal of Roentgenology 1988.150:651-661.

development of hypointensity from days 1 to 3 was observed Discussion

directly in four hematomas studied serially during this period
The MR features of AIH are complex, fascinating, and
(Figs. 3E and 3H). From days 4-7, there was a trend toward
incompletely understood. Different investigators have re-
increasing intensity, but AlHs studied during this period
ported a variety of findings and have come to often contradic-
showed the greatest variation in intensity on long TR images
tory conclusions about the physical basis of these phenomena
(Figs. 4 and 6).
and their clinical implications [1-9]. In our judgment, this
confusion and contradiction arise from attempts to extrapo-
late general principles from very limited clinical and/or exper-
Comparison of Pulse Sequences
imental data. This type of analysis worked well with CT, in
There was poor correspondence between short and long which the densities of hematomas are relatively uniform and
TA scans. The intensities encountered within individual he- change slowly [1 0, 1 1 ]. However, the intensity patterns and
matomas were different, as was the degree of heterogeneity shifts that occur with MR are much more complex. Hematoma
(Fig. 4). The timing and rate of change of intensity reversals intensity is typically heterogeneous, and changes rapidly and
were also different (note the differences in the slopes of the continuously during the first postictal week. Findings in our
time/intensity curves in Fig. 1 B). By contrast, hematomas had series indicate that these shifts are bidirectional (AIH decreas-
similar appearances on long TR/intermediate TE and long ing and then increasing in intensity) and subject to significant
TA/long TE scans; the only discernible difference was accen- variation among patients. Thus, general conclusions drawn
tuation of hypointensity on the long TE sequence (Figs. 5C from limited clinical experience have proved to be incorrect
and SD). or, at least, incomplete. Previous attempts to reconcile con-
tradictory findings have centered on the effects of extrinsic,
equipment-dependent factors, such as variations in pulse
Additional Factors That Affect A/H Intensity sequences and, more recently, the field strength of the MR
imager [4, 5]. Although these factors are undoubtedly impor-
The AIHs in 1 2 patients with CT and/or clinical evidence of tant, their true contributions remain unclear. Data from our
multiple episodes of hemorrhages were more heterogeneous series indicate that at least some of the apparent discrepan-
in appearance and more likely to vary from “typical’ intensities cies disappear when a large number of cases are analyzed
than those encountered in patients with single hemorrhagic with careful attention to both the precise time after ictus and
events (Fig. 4). The effect of hematoma location was analyzed the exact intensities encountered. These data also make it
by evaluating the patients with hemorrhages into multiple possible to gain a firmer understanding of the underlying
intracranial compartments. In nine of 12 cases, intensity dif- biochemical changes that produce these alterations in inten-
ferences were observed. The greatest differences occurred sity and to delineate four stages in AIH evolution, each with
in patients in whom extraaxial or parenchymal hematomas a typical pattern of intensity on multiple pulse sequences.
were combined with hemorrhages into spinal fluid compart- On short TA/short TE scans, AIH is hyperintense relative
ments (Fig. 3). In patients with simultaneous extraaxial and to brain when studied between 1 5 and 24 hr (Fig. 2). The
parenchymal hematomas, the intensities tended to be similar initial mild hyperintensity is unexpected, since most authors
(Fig. 2). report that hematomas are iso- to hypointense relative to
 654 ZIMMERMAN ET AL. AJR:150, March 1988

Fig. 2.-Dural arteriovenous malformation,

with parenchymal and subdural hemorrhage. CT
was performed 9 hr after and MR 15 hr after
hemorrhage.
A and B, Unenhanced CT scans show right
temporal parenchymal hematoma and right con-
vexity and interhemispheric subdural hemato-
mas.
C and D, SE 500/32 images reveal both pa-
renchymal (C) and subdural (D) hematomas to
be mildly hyperintense to white matter.
E and F, SE 2000/60 images. Hematomas are
moderately hyperintense to white matter.
G and H, SE 2000/120. Hematomas are mark-
edly hyperintense. Hematomas are of equal in-
tensity but subdural hematoma is better seen on
MR than on CT because of superior anatomic
display and absence of bone artifacts. Paren-
chymal hematoma is seen on both studies but
perhaps more easily characterized as hemor-
rhaglc on CT. Note subtle hypointense rims at
margins of hematomas (arrows).
American Journal of Roentgenology 1988.150:651-661.

- ‘ ‘

t’-1

‘:“
‘4 .-..

“v_
 AJR:150, March 1988 MR OF ACUTE INTRACRANIAL HEMORRHAGE 655

Fig. 3.-Traumatic epidural and subarachnoid

hematomas.
A and B, CT scans at 21 hr after trauma show
hyperdense frontal epidural hematoma and focal
left suprasellar subarachnoid hematoma (A).
C, SE 500/32 scan at 36 hr. Epidural hema-
toma is isointense to brain but easily detected;
subarachnold hematoma cannot be seen (nor
was it seen on long TR Images at this time).
D and E, SE 2000/30 (proton-densIty) (0) and
SE 2000/90 (heavIly T2-weighted) (E) scans at
36 hr. Epidural hematoma is hyperintense to
brain. Nodules of hypointensity are seen withIn
epidural and subgaleal hematomas.
F and G, SE 500/32 scans at 72 hr. Subarach-
noid hematoma Is now hyperintense to white
matter (F); epidural hematoma (G) Is now mildly
hypointense to white mafter- medial hypointense
rim (G) represents displaced dura (arrow). Epi- ; ,
dural hematoma Is well seen despite lack of
.4 ..,
contrast due to improved anatomic display. a
H, SE 2000/60 scan at 72 hr. Epidural and
subgaleal hematomas have become markedly
hypointense to white matter (compare with E).
American Journal of Roentgenology 1988.150:651-661.

r’
s

, .i

L:-#{149}’”
L.. ,
- .

rfr
1’Ad1
“.: i

‘S ‘ :‘ i
j
 656 ZIMMERMAN ET AL. AJR:150, March 1988

Fig. 4.-Recurrent hypertensive thalamic

hemorrhage.
A, Initial CT scan 24 hr after initial ictus shows
left thalamic hematoma.
B, Follow-up study I day later because of
Increasing drowsiness shows recurrent hemor-
rhage with enlargement of hematoma.
C, SE 500/32 scan 72 hr after initial hemor-
rhage. Lesion Is predominantly hypointense with
small central focus of hyperintensity.
D, SE 2000/60 scan at the same time. Eccen-
tric hypointense focus is noted in lateral aspect
of hematoma. Curvilinear hypointensity is also
noted within and at margin of hematoma (short
arrows). More laterally placed linear hypointen-
shy corresponds to internal capsule (long ar-
row).
E, SE 500/32 scan at 6 days. Hematoma is
diffusely hyperintense.
F, SE 2000/60 scan at 6 days. Lesion is pro-
dominantly hyperintense with small focus of hy-
pointensity near center of hematoma. Hypoln-
tense rim marks edge of hematoma (arrow).
American Journal of Roentgenology 1988.150:651-661.
 AJR:i50, March 1988 MR OF ACUTE INTRACRANIAL HEMORRHAGE 657

Fig. 5.-Hypertensive lenticular nucleus he-

matoma.
A, CT scan at 24 hr.
B, SE 500/32 scan at 72 hr. Lesion is isoin-
tense to white matter.
C and D, SE 2000/60(C) and SE 2000/120(D)
images at 72 hr. Nodular hypointense focus at
center of hematoma is surrounded by hyperin-
tensity. Since hypointense focus is smaller than
hematoma seen on CT, peripheral hyperintensity
represents combination of both hemorrhage and
edema.
E-G, Follow-up MR scans at 7 days. Hema-
toma is diffusely hypointense on all sequences.
Thin peripheral rim Is seen at margin of lesion
on SE 2000/60 (F) and SE 2000/120 (G) scans.
American Journal of Roentgenology 1988.150:651-661.

..,

rJ ,I

!
 658 ZIMMERMAN ET AL. AJR:150, March 1988

Fig. 6.-Birth trauma with retro-third ventric-

ular subarachnoid and subdural hematoma.
A, CT scan 3 days after birth. Subarachnoid
hematoma is noted behind third ventricle.
B, SE 500/32 image 6 days after birth. Retro-
third ventricular hematoma is hyperintense.
Small-convexity hematoma (arrows) could not
be seen on CT.
C and 0, SE 2000/60(C) and SE 2000/120 (D)
images 6 days after birth. Hematoma becomes
progressively less intense and mimics intensity
of subcutaneous and intraorbltal fat. Thus, retro-
ventricular lesion could be confused with con-
genital midline fat-containing lesion.
American Journal of Roentgenology 1988.150:651-661.

brain throLghout this period [2, 5-7]. Careful review of pub- Over the next day, the mean intensity diminishes, and by
lished images, however, reveals subtle diffuse or focal hyper- the end ofthe second day, hematomas are iso- to hypointense
intensity within hematomas that are generally characterized relative to brain (Fig. 3). This decrease in intensity (compare
as isointense (see Fig. 1 B of Dooms et al. [7]). The underlying Figs. 2 and 3) may be from Ti prolongation. An alternate
basis of this finding is unclear. Hyperintensity on short TR/ explanation, however, is that the T2 shortening that occurs
short TE scans usually results from Ti shortening (relative to during this period (see below) may be of sufficient magnitude
adjacent brain), and yet in vivo [1 , 7, 1 2] and in vitro [1 2, 13] to diminish intensity on these so-called Ti -weighted short TR
calculations indicate that Ti is initially prolonged relative to scans in which the TE is 32 msec.
white matter. It seems likely, therefore, that the subtle hyper- Between the third and fifth days after hemorrhage, the
intensity initially encountered on short TR/short TE scans is intensity on short TR/short TE scans dramatically and ab-
related to changes in other MR parameters. Since T2 prolon- ruptly increases so that hematomas become markedly hyper-
gation does not generally produce hyperintensity on short intense (more prominently at the periphery than at the center
TR/short TE scans [14, 1 5], high proton density must be of the AIH) in as little as 24 hr (Figs. 4-6). This phenomenon
considered the probable cause of this phenomenon. Hyper- has been demonstrated in previous reports [1 -8] and is
intensity caused by high proton density may be directly ob- believed to be secondary to a paramagnetic effect of the
served on an SE 2000/30 (“proton-density”) scan (Fig. 3D). oxidative breakdown product methemoglobin, which begins
The high proton density of acute extravasated blood is prob- to appear on about the third posthemorrhage day [1 6, 17].
ably a reflection of the high water content of liquid blood By the end of the first week, all hematomas show marked
before clot formation and retraction. hyperintensity on short TR/short TE scans.
 AJR: 1 50, March 1988 MR OF ACUTE INTRACRANIAL HEMORRHAGE 659

On long TR scans, hematomas are more heterogeneous hematoma and between hematomas) as well as the rapid
and show greater variation in intensity than on short TR changes in intensity that occur over time. Unfortunately, the
scans. The lesions are generally hyperintense relative to brain intensity patterns encountered at 0.5 T do not match those
when evaluated i2-24 hr after ictus (Fig. 2). Small nodules predicted by in vitro data. Hypoxic, intact RBCs are located
of hypointensity may be identified within the AIH, and a at the center of hematomas, and thus central hypointensity
peripheral rim of hypointensity is commonly seen (Fig. 2). should be the commonly encountered pattern [5]. In our
More generalized hypointensity, of variable degree and oc- experience, however, hypointense foci are often located ec-
cupying a variable amount of the hematoma (as identified by centrically (Fig. 4), and, as previously stated, a thin rim of
CT), develops late in the first or early in the second postictal hypointensity is seen at the margin of many clots (Figs. 2, 4,
day and becomes most marked by the third to fourth postictal and 5). The eccentric location of the hypointense focus may
day (Figs. 3-5). reflect areas of recurrent hemorrhage (Fig. 4); one of the
The inital hyperintensity and early conversion to hypoin- striking features in our series was the evidence of clinically
tensity were seen directly in serial studies in two of our silent rehemorrhage (as documented by serial CT and/or MR
patients (Fig. 3) and in one published example [1 8]. These studies demonstrating increase in hematoma size) in six
serial changes were documented in an experimental hema- cases. Given the apparent sensitivity of clot intensity to
toma studied by Di Chiro et al. at 0.6 T [i 2]. The early deoxyhemoglobin concentration, even minimal rehemorrhage
hyperintensity on long TR scans is indicative of T2 prolonga- might significantly alter the intensity of the hematoma.
tion relative to normal brain. Whether this is a property of The hypointense peripheral rim is more puzzling. This phe-
normal, nonfiowing blood or represents some early postextra- nomenon has been described in chronic hemorrhage, where
vasation biochemical change is not known, but it is probably it is said to reflect a magnetic susceptibility effect of hemosid-
related to the fact that, as a fluid, blood has a higher water enn within macrophages in the capsule of aging hematomas
American Journal of Roentgenology 1988.150:651-661.

content and thus a longer T2 than normal brain tissue does. [5]. The hypointense rim in acute hematomas is less promi-
The subsequent development of hypointensity is indicative of nent but otherwise similar in appearance. At this stage of
T2 shortening. This phenomenon was initially described by hematoma evolution, hemosidenn-laden macrophages are not
Gomori et al. [5], who postulated that it was a paramagnetic present [23-25]; thus, another explanation for this phenom-
effect of the deoxyhemoglobin within intact hypoxic RBCs. enon must be sought. Edelman et al. [2i ] noted this hypoin-
Paramagnetic Ti shortening does not occur with deoxyhe- tense rim and suggested that it might represent a border-
moglobin despite the presence of four unpaired electrons, zone phase-shift phenomenon. In our experience, similar rims
because its molecular structure does not allow the proximity have also been seen at the margin of all abscesses and some
(3 A) of unpaired electrons necessary to produce Ti short- metastatic foci [26, 27]. Pathologic studies in these cases
ening [i 9]. Before cell lysis, however, the deoxyhemoglobin have demonstrated the presence of macrophages, but with-
may produce T2 shortening because of its effect on magnetic out hemosiderin. We believe that this represents a paramag-
susceptibility (the ratio between external applied magnetic netic effect of free radicals produced by macrophages during
field and the internal field generated by the sample). The active phagocytosis [27] and believe it may account for the
intracellular deoxyhemoglobin produces stronger magnetic rim hypointensity encountered in acute hematomas as well.
fields than those generated in the deoxyhemoglobin-free ex- A far more striking and troubling discrepancy between
tracellular space. This heterogeneity of magnetic field strength predictions based on in vitro MR spectroscopy and clinical
within the sample leads to the presence of local magnetic data is the dependence of magnetic susceptibility on the
gradients between the intra- and extracellular spaces, result- square of the field strength of the MR unit [20, 22]. This led
ing in rapid dephasing (short T2) of protons that freely diffuse Gomori et al. [5] to state that hypointensity would be seen
across the cell membrane [5, 20-22]. only at high field strengths (i .5 T). This is clearly not the case
This hypothesis receives strong support from two inde- in clinical practice. Hypointensity is routinely seen on inter-
pendent sources. First, it has recently been demonstrated mediate-field-strength systems (0.35-0.6 T), as demonstrated
that the hypointensity within hematomas is seen more com- by us (Figs. 3-5) and in other publications [8, i 2, 18, 21],
monly and extensively with gradient-echo pulse sequences even though magnetic susceptibility is theoretically one-ninth
than with long TR/long TE SE pulse sequences [2i]. Gradient- that seen at 1 .5 T. More surprisingly, acute hematomas
echo scans are more sensitive to magnetic susceptibility studied at 0.i 5 T (magnetic susceptibility i/i 00 that seen at
effects because they are more dependent on T2* than on T2. i .5 T) and even at 0.02 T (magnetic susceptibility i/5625 that
Second, magnetic susceptibility effects of deoxyhemoglobin seen at 1 .5 T) also demonstrate mild hypointensity (see Fig.
have been demonstrated in vitro with MR spectroscopy, even 2E of Sipponen et al. [i ], Fig. 2C of DeLaPaz et al. [3], and
before the development of MR imaging [16, 20]. Figs. 1 B and 6C of Sipponen et al. [4]). Based on a compari-
With MR spectroscopy, the degree of T2 shortening is son of published images, it appears that the degree of hy-
proportional to the square of both the hemoglobin concentra- pointensity does indeed vary with field strength, but to a
tion and the magnetic field strength [20, 22]. The in vitro data lesser extent than that predicted by in vitro studies. The
help to explain some of the phenomena encountered in our discrepancy between experimental prediction and clinical ob-
clinical cases, but in other important ways are at variance servation should not come as a complete surprise, given the
with clinical experience. The dependence of intensity on the complexity of hemorrhagic brain lesions. Reactive changes
square of the deoxyhemoglobin concentration is probably the within and/or adjacent to hematomas could modify the inten-
cause of the variability of intensity (both within an individual sities encountered on clinical images. Additional factors, un-
 660 ZIMMERMAN ET AL. AJR:150, March 1988

affected by field strength, could thus contribute to the devel- evaluated in our series (Fig. 4F).
opment of hypointensity at low fields or militate against. full The other source of intensity variation is the location of the
expression of hypointensity at high field strengths, lessening hematoma. When time/intensity relationships for AIH in differ-
the dependence of intensity on field strength. ent intracranial compartments were studied, no clear trends
The next major change in hematoma intensity on long TR were discerned, but, when scans demonstrating hemorrhage
scans is the redevelopment of hyperintensity (Figs. 4 and 5), into multiple compartments were evaluated, it was noted that
which begins on approximately the fourth postictal day. (This the intensities were different in nine of i 2 cases. This differ-
reversal of intensity may occur quickly; that is, in 24-48 hr.) ence was most marked when hemorrhage into a spinal fluid
Thus, hematomas pass through a second isointense phase compartment (intraventricular or subarachnoid) was seen in
(Fig. 6) to become hyperintense relative to brain by the end conjunction with either a parenchymal or extraaxial hematoma
of the first week. The redevelopment of hyperintensity is a (Fig. 3). Subarachnoid hemorrhage was particularly difficult to
result of RBC lysis. The overall deoxyhemoglobin concentra- detect on MR. Acute (under 3 days) and/or diffuse subarach-
tion decreases and deoxyhemoglobin becomes homogene- noid hemorrhage could not be detected (Fig. 3C). This poor
ously distributed in the extracellular space. This eliminates visualization of diffuse subarachnoid hemorrhage has already
the heterogeneous magnetic field, and thus local magnetic been noted [3, i 3]. Some investigators have ascribed it to
gradients within the sample disappear. The increased inten- the higher oxygen tension of CSF when compared with brain,
sity, therefore, results from a simple subtraction of the T2 which leads to a low deoxyhemoglobin concentration [28,
shortening effect of the intracellular deoxyhemoglobin; the 29]. In the absence of deoxyhemoglobin, T2 shortening and
intrinsically prolonged T2 of blood is unmasked and the hypointensity do not develop during the acute posthemor-
intensity returns to its initial high level. Additional prolongation rhage period. Another factor that may contribute to the poor
of T2 during this period may result from an increase in water visualization of diffuse subarachnoid hemorrhage is the effect
American Journal of Roentgenology 1988.150:651-661.

within the hematoma as cells lyse and the clots age [23, 24]. of spinal fluid pulsation on observed intensity. Recent studies
It is important to remember that the increase in intensity that indicate phase-shift effects of spinal fluid pulsation may con-
occurs on long TR scans during this period is not caused by tribute significantly to the observed intensity within spinal fluid
the presence of methemoglobin, which produces the nearly spaces [30]; therefore, it is possible that the effects of hem-
simultaneous increase in intensity seen on short TR scans. orrhage on spinal fluid intensity might be masked.
The paramagnetic effect of methemoglobin produces marked In summary, the complexity of the intensities encountered
Ti and, to a lesser extent, T2 shortening. On long TR scans, in AIH is a true reflection of the complex biochemical changes
all intracranial soft tissues have undergone complete inter- that occur during the first posthemorrhagic week. These
pulse recovery; thus, a decrease in Ti does not increase the changes are clearly interdependent, but the extent to which
intensity of the hemorrhage relative to brain. The T2 short- each occurs, the timing of the occurrence, and the location
ening effect of methemoglobin should actually cause dimin- within an individual hematoma vary. The overall intensity and/
ished signal intensity, but this effect is obviously overcome or specific intensity pattern encountered within any individual
by more powerful determinants of T2. High proton density hematoma will, therefore, reflect the nearly unique interplay
may also contribute to the increased intensity encountered of these factors at one particular moment. Under these cir-
on long TR scans at this stage, especially when used in cumstances, the variable, heterogeneous, and somewhat un-
conjunction with a short or intermediate TE. predictable appearance of these lesions is to be expected.
In addition to these already complex phenomena, two Since a wide range of intensities are encountered in AIH, it
additional factors significantly affect hematoma intensity. The is obvious that no single sequence and no single intensity can
first factor is recurrent hemorrhage, which is extremely difficult characterize acute hemorrhage. It is, however, possible to
to diagnose in any individual case because rapid intensity define characteristic (although not necessarily specific) com-
changes occur with (Fig. 4) or without (Fig. 5) rehemorrhage. binations of intensities on multiple-pulse sequences at four
In general, rehemorrhage appears to be a major source of the different stages of AIH evolution:
heterogeneity and variability of hematoma intensity. The ad- Stage I (under 24 hr)-subtle hyperintensity on short TR/
mixture of extravasated blood of various ages probably ac- short TE scans and moderate to marked hyperintensity on
counts for the almost random pattern of heterogeneous inten- long TR/intermediate TE and long TR/long TE scans, respec-
sity encountered within recurrent hemorrhages (Fig. 4). The tively (Fig. 2). This combination is relatively specific (especially
presence of unsuspected preexisting hemorrhage and the when a thin, hypointense rim is also present), since most
development of postictal rehemorrhage explain, at least in lesions hyperintense on long TR scans are hypo- to isointense
part, the overall variations in intensity encountered among on short TR scans.
different patients. For instance, MR performed in an elderly Stage II (1 -3 days)-iso- to mild hypointensity on short
patient 60 hr after the onset of mild hemiparesis and obtun- TR/short TE scans and marked hypointensity on long TR/
dation demonstrated a large subdural hematoma that was intermediate to long TE scans (Figs. 3-5). Although charac-
“prematurely” hyperintense on short TR/short TE scans. It is teristic, this pattern is not as specific as previously suggested
most likely, given the patient’s clinical and CT findings, that [5]. Marked hypointensity may also be encountered in densely
this was a rehemorrhage into a clinically silent chronic sub- calcified lesions [3i ] and noncalcified nonhemorrhagic gran-
dural hematoma. Postictal rehemorrhage may account for the ulomatous masses [26, 32, 33].
persistence of hypointensity on long TR scans into the sixth Stage III (4-6 days)-hyperintensity on short TR/short TE
and seventh days in five of the examples of rehemorrhage images and markedly variable intensity on long TR/interme-
 AJR:150, March 1988 MR OF ACUTE INTRACRANIAL HEMORRHAGE 661

diate to long TE scans. When marked hypo- or hyperintensity 10. New PF, Aronow S. Attenuation measurements of whole blood and blood
fractions in computed tomography. Radiology 1976;121 :635-640
are present on long TA scans, hemorrhage may be diagnosed
ii. Scotti G, Terbrugge K, Melancon D, et al. Evaluation ofthe age of subdural
with confidence; however, if scans are obtained just as inten- hematomas by computerized tomography. J Neurosurg 1977;47:31 1-315
sity is beginning to increase on long TA scans, findings may 12. Di Chiro G, Brooks RA, Girton ME, et al. Sequential MR studies of
be indistinguishable from those seen in fat-containing lesions intracerebral hematomas in monkeys. AJNR 1986;7: 193-1 99

(Fig. 6). 13. Chakeres DW, Bryan RN. Acute subarachnoid hemorrhage: in vitro corn-
panson of magnetic resonance and computed tomography. AJNR
Stage IV (more than 6-7 days)-hyperintensity on short 1986;7:223-228
TR/short TE and long TA/intermediate-to-long TE scans. This 1 4. Bydder GM, Magnetic resonance imaging of the brain. Radiol Clin North
combination is considered specific, but it may be encountered Am 1984;22:779-794
rarely in intracranial dermoids [34]. 1 5. Wehrli FW, MacFall JR. Shutts P. Breger R, Herfkens RJ. Mechanisms of
contrast and NMR imaging. J Comput Assist Tomogr 1984;8:369-380
The ability to detect and characterize hematomas is de-
1 6. Brooks RA, Battocletti JH, Sances A, et al. Nuclear magnetic relaxation in
pendent on anatomy as well as intensity; therefore, the degree blood. IEEE Trans Biomed Eng 1975;22:12-18
to which a lesion distorts normal anatomy also has a signifi- 17. Bradley WG, Schmidt PG. EffeCt of methernoglobin formation on the MR
cant effect on hematoma detection. Subarachnoid hemor- appearance of subarachnoid hemorrhage. Radiology 1985;156:99-i03
18. Harms SE, Siemers PT, Hildenbrand T, Plum G. Multiple spin echo mag-
rhage is difficult to visualize on MA, not only because it fails
netic resonance imaging of the brain. Radiographics 1986;6:117-134
to alter intensity, but also because it fails to produce anatomic 19. Wolf GL, Bumett KR, Goldstein EJ, Joseph PM. Contrast agents for
distortion (Fig. 3). Parenchymal hematomas (Fig. 2) are always magnetic resonance imaging. In: Kressel HY, ed. Magnetic resonance
detected on MA, but they may be difficult to specifically annual 1985. New York: Raven, 1985:231-266
characterize as hemorrhagic lesions; conversely, extraaxial 20. Brittenham GM, Farrell DE, Harris JW, et al. Magnetic susceptibility meas-
urernents of human iron stores. N EngI J Med 1982;307 :1671-1675
hematomas (subdural and epidural, Figs. 2 and 3) have char-
21 . Edelman RR, Johnson K, Buxton R, et al. MR of hemorrhage: a new
acteristic and relatively specific anatomic configurations that approach. AJNR 1986;7:751-756
American Journal of Roentgenology 1988.150:651-661.

make the diagnosis of hematoma possible regardless of in- 22. Thulborn KR, Waterton JC, Matthews PM, Radda GK. Oxygenation de-
tensity. In fact, because of the absence of bone artifacts and pendance of the transverse relaxation time of water protons in whole
blood at high fields. Biochim Biophys Acta 1982;174:265-270
the excellent contrast between brain and adjacent spinal fluid,
23. Stehbens WE. Intracerebral and intraventncular hemorrhage. In: Pathology
small, extraaxial hematomas are detected much more easily
of cerebral blood vessel. Saint Louis: Mosby, 1972:284-350
on MR than on CT (Fig. 2) [6, 35]. 24. Adams RD. Sidman AL. Introduction of neuropathology. New York: Mc-
In conclusion, the complex nature of the intensity changes Graw Hill, 1968:177-178
encountered in acute intracranial hemorrhage and the diffi- 25. Undenberg R. Pathology of craniocerebral injuries. In: Newton TH, Potts
DE, eds. Radiology of the skull and brain, vol 3. Anatomy and pathology.
culty in distinguishing some of these changes from those seen
Saint Louis: Mosby, 1977:3049-3087
in other lesions suggest that CT will remain the procedure of 26. Zimmerman AD, Becker AD, Devinsky 0, et al. MRI features of cerebral
choice as the screening test in the evaluation of any patient abscesses and other intracranial inflammatory lesions (exduding AIDS
suspected of acute intracranial hemorrhage. Once the hem- patients). Presented at the Symposium Neuroradiologicum, Stockholm,
orrhagic nature of the lesion is identified on CT, MR can be June 1986
27. Fleming CA, Zimmerman AD, Becker AD, Deck MDF. The diagnostic
used to increase diagnostic information by more accurately significance of rim intensity and edema patterns in the differentiation of
demonstrating the extent of the abnormality, assessing the intracranial mass lesions on MRI. Presented at the annual meeting of the
time from hemorrhage, and detecting the presence of under- American SOciety of Neuroradiology, New York City, May 1987
lying lesions. 28. Kemp SS, Grossman Al, lp CV, et al. The importance of oxygenation in
the appearance of acute subarachnoid hemorrhage on high field magnetic
resonance imaging. Presented at Symposium Neuroradiologicum, Stock-
helm, June 1986
REFERENCES
29. Gornori JM, Grossman RI. Mechanisms responsiblefor the MR appearance
1 . Sipponen JT, Sepponen RE, Sivula A. Nuclear magnetic resonance (NMR) and evolution of intracranial hemorrhage. Radiology 1986;161(P):364
imaging of intracerebral hemorrhage in the acute and resolving phases. J 30. Rubin JB, Enzmann DR. Imaging spinal CSF pulsation by 2DFT magnetic
Comput Assist Tomogr 1983;7:954-959 resonance: significance during clinical imaging. Harmonic modulation of
2. Han JS, Kaufman B, Alfidi RJ, et al. Head trauma evaluated by magnetic proton MA precessional phase by pulsatile motion: origin of spinal CSF
resonance and computed tomography: a comparison. Radiology 1984; flow phenomenon. Presented at the annual meeting of the American
150:71-77 Society of Neuroradiology, San Diego, January 1986
3. DeLaPaz RL, New PF, Buonanno J, et al. NMR imaging of intracranial 31 . Dot RF, New PF, Pile-Speilman J, et al. Detection of intracranial calcifica-
hemorrhage. J Comput Assist Tomogr 1984;8:599-607 tions by MA. AJNR 1986:7:801-810
4. Sipponen JT, Sepponen RE, Tanttu JI, Sivula A. Intracranial hematomas 32. Becker RD, Zimmerman AD, Sze G, Haimes AB, Deck MDF. MR of
studied by MR imaging at 0.17 and 0.02T. J Comput Assist Tomogr subdural ernpyema and other extra-axial inflammatory lesions. Presented
1985;9:698-704 at the annual meeting of the American Society of Neuroradiology, New
5. Garnet JM, Grossman RI, Goldberg HI, Zimmerman RA, Bilaniuk LT. York City, May 1987
Intracranial hematomas: imaging by high field MR. Radiology 1985; 33. Sherman JL, Hayes WS, Stem BJ, Citrin CM, Pulaski PD. MR evaluation
157: 87-93 of intracranial sarcoidosis: comparison with CT. Presented at the annual
6. Snow RB, Zimmerman RD. Gandy SE, Deck MDF. Comparison of magnetic meeting of the American Society of Neuroradiology, New York City, May
resonance imaging and computed tomography in the evaluation of head 1987
injury. Neurosurgery 1986;18:45-52 34. Newton DA, Larson TC, Dillon WP, Newton TH. Magnetic resonance
7. Dooms GC, Uske A, Brant-Zawadzki M, et al. Spin echo MR imaging of characteristics of cranial epidermoid, dermoid and teratornatous tumors.
intracranial hemorrhage. Neuroradiology 1986;28: 132-138 Presented at the annual meeting of the American SOciety of Neuroradiol-
8. Zimmerman RD. Deck MDF. Intracranial hematomas: imaging by high field ogy, New York City, May 1987
MR. Radiology 1986;159:565 35. Zimmerman AD, Snow RB, Gandy SE, et al. Comparison of MRI and CT
9. Zimmerman RA, Bilaniuk LT, Grossman RI, et al. Resistive NMR of in the evaluation of head trauma: review of initial 100 cases. Presented at
intracranial hematomas. Neuroradiology 1985;27: 16-20 the Symposium Neuroradiologicum, Stockholm, June 1986