Eur. Radiol.

(2001) 11: 17701783

DOI 10.1007/s003300000800

P. M. Parizel
S. Makkat
E. Van Miert
J. W. Van Goethem
L. van den Hauwe
A. M. De Schepper

Received: 19 September 2000

Revised: 28 November 2000
Accepted: 4 December 2000
Published online: 3 May 2001
 Springer-Verlag 2001

P. M. Parizel ( ) S. Makkat E. Van

Miert J. W. Van Goethem L. van den
Hauwe A. M. De Schepper
Department of Radiology,
University of Antwerp, Wilrijkstraat 10,
2650 Edegem, Belgium
E-mail: parizelp@uia.ua.ac.be
Phone: +32-3-8 21 37 32
Fax: +32-3-8 25 20 26

NE UR O

Intracranial hemorrhage:
principles of CT and MRI interpretation

Abstract Accurate diagnosis of intracranial hemorrhage represents a

frequent challenge for the practicing
radiologist. The purpose of this article is to provide the reader with a
synoptic overview of the imaging
characteristics of intracranial hemorrhage, using text, tables, and figures to illustrate time-dependent
changes. We examine the underlying
physical, biological, and biochemical factors of evolving hematoma
and correlate them with the aspect
on cross-sectional imaging techniques. On CT scanning, the appearance of intracranial blood is determined by density changes which
occur over time, reflecting clot formation, clot retraction, clot lysis
and, eventually, tissue loss. However, MRI has become the technique
of choice for assessing the age of an

Introduction
Intracranial hemorrhage is a potentially life-threatening
neurological condition. It results in a significant burden
on health care resources. Patient outcome is influenced by
an early and accurate diagnosis. Radiological assessment
of intracranial blood is challenging, due to the highly
variable appearance of intracranial blood, depending on
its age and location. It is therefore important to understand the underlying physical, biological, and biochemical factors of an evolving hematoma and to correlate them
with the aspect on cross-sectional imaging techniques.
Intracranial hemorrhages can be subdivided into intracerebral and extracerebral types. Intracerebral hemorrhages (ICH) can be classified according to their site

intracranial hemorrhage. On MRI

the signal intensity of intracranial
hemorrhage is much more complex
and is influenced by multiple variables including: (a) age, location,
and size of the lesion; (b) technical
factors (e.g., sequence type and parameters, field strength); and (c) biological factors (e.g., pO2, arterial
vs venous origin, tissue pH, protein
concentration, presence of a bloodbrain barrier, condition of the patient). We discuss the intrinsic magnetic properties of sequential hemoglobin degradation products. The
differences in evolution between
extra- and intracerebral hemorrhages are addressed and illustrated.
Keywords Brain Hemorrhage
CT MRI Hematoma

of origin and the mechanisms responsible for their occurrence. They are also known as intraparenchymal
hemorrhages. The incidence of ICH has a slight male
preponderance, steadily increases with age, and peaks in
the eighth decade. According to one prospective study,
the incidence of ICH is 13.9 per 100,000 per year in men
compared with 12.3 per 100,000 per year in women [1].
It is estimated that ICH accounts for 15 % of all strokes
[2, 3]. The mechanisms responsible for ICH include:
hypertension; hemorrhagic infarction; cerebral amyloid
angiopathy; rupture of vascular malformations; bleeding into primary or metastatic brain tumors; coagulopathies (due to the use of anticoagulants and thrombolytic
agents); sympathomimetic drug effect (amphetamines,
phenylpropanolamine, and cocaine); and vasculitis [2].

1771

a
Fig. 1 a, b Hyperacute epidural hematoma in a 41-year-old man.
This case illustrates clot formation as evidenced by the increasing
CT density of the hematoma. a Axial non-contrast CT scan upon
admission shows a right parietal epidural hematoma (EDH) in the
hyperacute stage. The scan was obtained less than 30 min after
head trauma. Average density of the hematoma is 49 HU. b Follow-up scan, obtained 66 min later, shows that the density of the
EDH has increased. Average density is now 67 HU. The density
increase reflects clot formation

Extracerebral hemorrhages comprise epidural hemorrhage (EDH), subdural hemorrhage (SDH), subarachnoid hemorrhage (SAH), and intraventricular
hemorrhage (IVH) [4]. Both EDH and SDH are most
commonly caused by craniocerebral trauma. The origin
of SAH can be traumatic (superficial contusion of the
cerebral gyri) or non-traumatic (aneurysm rupture, arteriovenous malformation, secondary extension from an
intracerebral hemorrhage). The same holds true for
IVH.
In this article, we review the basics of CT and MRI
interpretation of intracranial hemorrhage. We also analyze the evolution of the hematoma over time and its
effect on the imaging findings.

Computed tomography
In reviewing the principles of CT interpretation of ICH,
the basic physics of X-ray imaging must be taken into
account. Attenuation, defined as the removal of X-ray
photons from the beam, occurs in biologic tissues. The
attenuation properties of a tissue are linked to their
atomic number and physical density. In other words, attenuation of the X-ray beam is determined by the density of the electron clouds in the tissues it traverses [5].
The attenuation properties of intracranial blood are
determined by the aggregation of globin molecules in
the hematoma [6]. There is a linear relationship between CT attenuation, protein content (mainly hemo-

globin), and hematocrit [7]; however, artifacts located

close to the skull base can easily mimic hemorrhage on
spiral-CT scans [8].
Immediately after the hemorrhage, freshly extravasated blood exhibits a markedly heterogeneous appearance with mixed density values in the range of 4060
Hounsfield units (HU; Fig. 1a) [9, 10]. This is due to the
formation of a complex, inhomogeneous mass that contains red blood cells (RBCs), white blood cells (WBCs),
and small platelet clumps interspersed with protein-rich
serum. Parts of the hematoma exhibit density values,
which are only moderately higher than the density value
of the adjacent brain parenchyma; therefore, a hyperacute EDH may be difficult to distinguish from the adjacent cortical gray matter.
During the early hours of hemorrhage, the CT density values within the hematoma rapidly increase up to
6080 HU (Fig. 1b). This is due to the formation of a
meshwork of fibrin fibrils and globin molecules. The
globin (protein) component of the hemoglobin has a
high density [11]. Moreover, due to incipient clot retraction, the hematocrit may increase to 90 %, thereby
further augmenting the density. These phenomena explain why, during the first week, intracranial hematomas
appear on non-contrast CT scans as well-demarcated
hyperdense lesions [12].
In large hematomas, a horizontal fluidfluid level is
observed in the hyperacute and acute phase. This effect
is called the hematocrit effect [13]. The dependent
area, which has higher CT attenuation values, is believed to represent sedimented cellular elements of
blood. The supernatant portion, which has a lower CT
density, presumably represents blood serum (Fig. 2).
As the hematoma matures, clot retraction ensues.
This increases the attenuation coefficient to 80100 HU
in the center of the hematoma. A hypointense halo appears around the central nidus due to serum extrusion
and reactive vasogenic edema (Fig. 3).
This stage is short-lived because soon proteolysis begins, and the hematoma protein is degraded and ab-

1772

Fig. 2 Hemorrhagic sedimentation level on CT (the hematocrit

effect). The patient is an 81-year-old woman with an acute hypertensive hemorrhage. Axial non-contrast CT scan shows a large intraparenchymal hematoma in the left hemisphere. There is a fluidfluid level within the hematoma, which is characteristic of the
so-called hematocrit effect. The hypodense supernatant represents
blood serum, whereas the denser sediment corresponds to settled
blood cells. Smaller hemorrhagic foci of high attenuation (due to
clot formation) are observed near the anterior margin of the large
hematoma. Mass effect causes a subfalcial herniation. The left lateral ventricle is displaced and compressed

sorbed. During the weeks after the acute event, the

density of the hematoma decreases by an average of
0.71.5 HU per day, due to chemical breakdown of globin molecules [14, 15]. This process begins in the periphery and proceeds toward the center of the hematoma. Usually, the decrease in density does not correspond to the decrease in mass effect; thus, even when
the density change is no longer present, persistent mass
effect on CT can give a clue to the previous episode of
hemorrhage in the brain.
After a few weeks or even months, macrophages digest the blood breakdown products, ultimately resulting
in resolution of the clot. The typical residuum of an old
hematoma is a slit-like cavity lined with hemosiderin
[16]. The hemosiderin is easily recognizable on gradient-echo MR images but is almost impossible to detect
on a CT examination. Other end-stage appearances of
an old hematoma include: an area of hypodensity (due
to tissue loss); focal atrophy; calcification; or ventricular
enlargement [14, 17].
After contrast administration in the acute stage,
there is no enhancement of the hematoma; however,

Fig. 3 An 81-year-old man with a well-circumscribed right parietal

intraparenchymal hematoma. This non-contrast CT scan was obtained 12 days after acute bleeding. The maturing hematoma is
surrounded by a hypointense halo. This is due to clot retraction
and/or vasogenic edema. The central part of the hematoma has
high attenuation values, ranging between 80 and 100 HU

with time, there develops neovascularization in the surrounding brain tissue. This will enhance on contrast administration due to breakdown of blood-brain barrier
(BBB) in the vascularized capsule [18]. The resulting
ring-like enhancement can lead the radiologist or clinician to diagnose this condition erroneously as a brain
tumor or abscess.
Extra-axial hemorrhages, such as EDH and SDH,
also evolve through the same stages as intracerebral
hemorrhage. Additionally, the extra-axial blood collections cause mass effect that may lead to brain herniation. The CT appearance of acute SDH depends on the
interval between the last major episode of bleeding and
the time of examination. Subdural hemorrhage appears
as a peripheral crescent-shaped collection of blood
density clot lying between the inner table and the cerebral hemisphere. The majority of SDHs are due to the
rupture of bridging veins. Subacute hematomas can be
nearly isodense to the adjacent cerebral cortex and may
be difficult to differentiate from normal brain tissue
[19]. Chronic SDHs are encapsulated, crescent-shaped,
low-attenuation collections. The capsule of a chronic
SDH is a capillary-rich membrane that enhances on
contrast administration. A mixed density pattern, found
in approximately 5 % of chronic SDHs, can be due to
recurrent hemorrhage [20]. Alternatively, a heteroge-

1773

Fig. 4 Laminated appearance of an acute atypical subdural hematoma (SDH) in a 76-year-old man. The non-contrast CT scan has
been photographed using a wide window to improve visualization
of the extra-axial space. The heterogeneous aspect of the hematoma is due to a so-called atypical subdural hematoma. It can be very
difficult to differentiate from re-bleeding into a pre-existing
chronic SDH [13]. There is significant mass effect upon the brain,
which appears hypodense due to edema. The right lateral ventricle
is displaced across the midline but not compressed. There is a
midline shift with subfalcial herniation. The patient also had a
downward transtentorial herniation, and a Duret hematoma in the
brainstem (not shown). The patient died shortly after this CT scan

neous SDH with mass effect can indicate a so-called

atypical SDH (Fig. 4). It can be very difficult to differentiate from re-bleeding into a pre-existing chronic
subdural hematoma [13]. Calcification or ossification is
seen in 0.32.7 % of chronic SDHs, usually after a few
months or years [21].
Epidural hemorrhage assumes a focal biconvex configuration that may cross dural folds, such as falx and
tentorium, but not sutures (Fig. 5). On CT, acute EDH
shows high blood density and often a swirl-like lucency
in the clot [4]. In the subacute stage, it becomes more
homogeneous. In the chronic state, the clot is reabsorbed by the perivascular elements derived from the
dural vessels, resulting in the formation of a membrane
similar to that of SDH. On contrast administration the
displaced dura gets enhanced due to neovascularization.
The characteristic properties of EDH and SDH are
summarized in Table 1.
Most cases of SAH are due to rupture of an intracranial aneurysm, or of vascular malformation. Traumatic SAH results from injury to surface vessels on the

Fig. 5 Mixed density values in an acute right parietal arterial EDH

in a 1-day-old baby boy. A non-contrast CT scan was performed.
The hematoma is biconvex in appearance, and is limited by the
frontal and lambdoid sutures (EDH does not cross suture lines). It
contains inhomogeneous hypodense areas, representing blood that
is not yet clotted. This is indicative of acute bleeding. The mass
effect causes a subfalcial herniation, with compression of the right
lateral ventricle, and dilatation of the left lateral ventricle

pia or the arachnoidal meninges. It may be focal or diffuse. Computed tomography remains the investigation
of choice in the diagnosis of acute SAH. Serpentine or
linear areas of high attenuation in the subarachnoid
spaces or basal cisterns are the hallmark of SAH on CT
scans. The sensitivity of CT for detecting SAH depends
on the volume of the extravasated blood, the hematocrit, and the time elapsed after the acute event. The
density of the hematoma decreases rapidly over time,
due to dilution of the blood by cerebrospinal fluid
(CSF); thus, after only a few days, it may be impossible
to detect SAH on CT.
Intraventricular hemorrhage is identified in 15 % of
all patients with closed head injury [22]. Disruption of
subependymal veins due to shearing injuries [23], basal
ganglionic hemorrhage with subsequent extension to
adjacent ventricle [24], and reflux of SAH through the
foramina of Luschka and Magendie are thought to be
the most common causes of IVH. On CT, acute IVH
appears as high-density collections with or without a
fluidfluid level.

1774

Table 1 Epidural vs subdural

hematoma

Epidural hematoma (EDH)

Subdural hematoma (SDH)

Coup side

Contre-coup side

Associated with skull fracture in  90 % of cases No consistent relationship with skull fractures
Does not cross suture lines
Does cross suture lines
Not limited by falx or tentorium (may extend
from supra- to infratentorial or across midline)

Limited by falx and tentorium (confined to supraor infratentorial compartment, does not cross
midline)

Origin
Origin
Arterial (majority, due to tearing of one or
Venous, due to laceration of superficial bridging
more branches of the meningeal arteries,
cortical veins
most commonly the middle meningeal artery)
Venous (minority, due to laceration of a dural Arterial (minority, due to lacerations of superfivenous sinus, e.g., along the sphenoparietal
cial arteries, especially in combination with sesinus)
vere cerebral contusions)
Medical emergency
Magnitude of the mass effect caused by EDH is
directly related to the size of the extracerebral
collection

May be chronic
Magnitude of the mass effect caused by SDH is
more often associated with underlying parenchymal injury

CT is preferred imaging technique because:

Rapid accessibility

MRI is preferred imaging technique because:

MRI is more sensitive than CT, especially in the
detection of so-called isodense SDHs which may
be difficult to see on CT
Multiplanar imaging capability
Better definition of multi-compartmental nature
of SDH

Shows both the hemorrhage and the skull

fracture
MR can be useful for:
Detection of parenchymal repercussions
(edema, mass effect, herniations)

MR imaging
On MRI, the appearance of intracranial hemorrhage
undergoes complex signal intensity changes, which are
determined first and foremost by the age of the bleed.
The multiple variables, which influence the MRI findings, are categorized as follows [13, 25, 26]:
1. Intrinsic factors: age of the hematoma (formation of
degradation products, RBC integrity, clot retraction); size of the lesion; intra- or extra-axial location;
episodes of recurrent hemorrhage; and degree of dilution with CSF (for SAH and IVH)
2. Technical factors: e.g., type of MRI pulse sequence,
sequence parameters, magnetic field strength, receiver bandwidth
3. Biological factors: pO2; arterial vs venous origin; tissue pH; protein concentration; presence of a BBB;
condition of the patient.
Depending on the characteristic intensity patterns observed in their evolution, intracerebral hemorrhages can
be staged as hyperacute (first few hours), acute
(13 days), early subacute (37 days), late subacute
(47 days to 1 month), or chronic (1 month to years)
[26]. Although we can classify the evolution of a hem-

orrhage as per the above staging, the duration of the

stages can vary and different stages may co-exist
(Fig. 6).
The MR interpretation of intra-axial hemorrhages
requires an understanding of the evolving pathophysiology of the hematoma and also the complex chemistry
of hemoglobin denaturation.

Hemoglobin and iron

Hemoglobin is the oxygen-carrying pigment found in
RBCs. It is a large molecule (molecular weight 68,000)
that consists of four subunits. Each subunit contains a
heme moiety conjugated to a polypeptide. The four
polypeptides are collectively referred to as the globin
portion of hemoglobin. In the hemoglobin of normal
human adults (hemoglobin A), there are two a chains
(141 amino acid residues) and two b chains (146 amino
acid residues). Embedded in a non-polar gap in the
polypeptide chain is the heme ring. Heme is a porphyrin
derivative, which contains an iron molecule in its center:
This is the oxygen-binding site. The sixth coordination
site of the heme iron is occupied by molecular oxygen in
oxy-hemoglobin (oxy-Hb) and is vacant in deoxy-hemoglobin (deoxy-Hb) [27].

1775

Fig. 6 ai Evolution of a right occipital hematoma in a 27-year-old woman.

a Contrast-enhanced CT scan obtained 4 h after an acute clinical event (at
another hospital). There is a dense, oval-shaped hematoma in the right occipital lobe. Already on this hyperacute CT scan, the hematoma appears
surrounded by a thin hypodense halo, presumably reflecting edema. b Axial
T1-weighted spin-echo (SE) MR image after 48 h. c Axial T2-weighted turbo spin-echo (TSE) MR image after 48 h. On T1-weighted images, the hematoma is hypointense relative to white matter and isointense or slightly
hypointense relative to gray matter. On T2-weighted images, the hematoma
is markedly hypointense. This indicates the presence of intracellular deoxyhemoglobin. The rim of the hematoma is starting to become slightly hyperintense on T1-weighted images, suggesting the formation of intracellular
methemoglobin peripherally. On the T2-weighted images, the lesion is surrounded by a halo of vasogenic edema, which is of high signal intensity.
d Axial T1-weighted SE MR image after 7 days. e Axial TSE T2-weighted
MR image after 7 days. The hematoma now appears to have two compartments. The anterior portion is hyperintense on T1-weighted images, and of
low to intermediate signal intensity on T2-weighted images. This signal behavior indicates the presence of intracellular methemoglobin. The posteriormedial portion of the hematoma presumably represents a cavernous

malformation, which has bled and which still contains deoxyhemoglobin.

There is slightly more vasogenic edema than on the first MR examination
(compare e with c). f Axial SE T1-weighted MR image after 3 weeks. g Axial TSE T2-weighted MR image after 3 weeks. The anterior portion of the
hematoma is strongly hyperintense on T1-weighted images and T2-weighted
images. This is typical of extracellular methemoglobin. The posteriormedial portion of the hematoma now shows a mixed, reticulated appearance
with high-signal-intensity central components surrounded by a hypointense
rim on T2-weighted images, which represents hemosiderin deposition. The
vasogenic edema has completely disappeared (compare g with c and e).
h Axial SE T1-weighted MR image after 2 months. i Axial T2-weighted TSE
MR image after 2 months. The anterior portion of the hematoma has become isointense to gray matter on T1-weighted images and remains hyperintense T2-weighted images. This reflects an area of encephalomalacia containing proteinaceous fluid. The posteriormedial portion of the hematoma
still displays a mixed signal intensity pattern. Both components are surrounded by a hypointense rim on T2-weighted images, which is due to hemosiderin deposition. The volume of the lesion has markedly decreased as
compared with previous scans

1776

Fig. 6 gi

In the blood vessels, hemoglobin alternates freely

between the oxy-Hb form (as it leaves the pulmonary
circulation) and the deoxy-Hb form (as it leaves the
capillary bed). In both oxy- and deoxy-Hb, iron is in the
ferrous state (Fe2+). In oxy-Hb, iron is in the low-spin
ferrous form without any unpaired electron, whereas in
deoxy-Hb the heme iron is in the high-spin ferrous state
with four unpaired electrons. This is due to the partial
transfer of the outer orbital unpaired electrons to the
oxygen molecule during oxygenation [28].
The electrons in the outer orbital determine the
magnetic properties of the substance in a magnetic field.
In the case of hemorrhage, there are three types of
magnetic behavior: diamagnetic; paramagnetic; and superparamagnetic [29]. Substances without unpaired
electrons are diamagnetic (e.g., oxy-Hb) and those with
unpaired electrons are paramagnetic [e.g., deoxy- and
methemoglobin (met-Hb)] (Fig. 7). Substances with
very large numbers of unpaired electrons are superparamagnetic (e.g., hemosiderin).
Unpaired electrons cause field fluctuations owing to
the larger magnetic moment of the electrons. These
field inhomogeneities are due to an interaction between the electron dipole of the unpaired electrons in
the outer orbital and the nuclear dipole of adjacent
protons. The interaction is inversely proportional to
the sixth power of the inter-dipolar distance [30]. The
mechanism is known as proton-electron dipole-dipole
(PEDD) interaction or simply dipoledipole interaction. It enhances both T1 and T2 proton relaxation, al-

though the effect is pronounced on T1-weighted images. Since the effect drops off rapidly with distance, it
is of importance only when water molecules approach
the paramagnetic center closely as in met-Hb [31, 32].
In deoxy-Hb the iron atom does not fit in the center of
the ring and moves slightly out of the plane of the ring
thereby preventing access to water; thus, no PEDD
occurs [33].
When magnetically susceptible substances with unpaired electrons are brought in an external magnetic
field, this results in magnetic inhomogeneity, leading to
irreversible loss of phase coherence. This magnetic susceptibility effect, called T2 proton relaxation enhancement (T2 PRE), selectively shortens T2 without affecting T1 [32]. The local field gradients disappear when the
RBCs lyse (loss of compartmentalization). The T2 relaxation process is lost and the signal brightens on T2weighted images.

Chronological changes
In the hyperacute stage blood leaves the vascular system
(extravasation). The hematoma now consists mainly of
intact RBCs containing oxy-Hb. As discussed previously, oxy-Hb is diamagnetic since it does not have unpaired electrons [31]. On T1-weighted images, hyperacute hematomas are iso- to hypointense to brain because of their longer T1 times, due to the high water
content. On T2-weighted images there is hyperintensity
again due to the high water content; thus, hemorrhage
less than 12 h old behaves as a fluid collection and may
be indistinguishable from any other edematous mass.

1777

Fig. 7 Electron distribution in the

orbitals of iron (energy levels), in
oxidation states +2 and +3, in
various hemoglobin degradation
products. Small arrows indicate
magnetic dipole moments of single electrons. In each orbital,
magnetic moments pair with
those with opposite magnetic dipole moment. In unfilled orbitals,
electrons remain unpaired until
the shell is half full. In oxy-hemoglobin, there are no unpaired
electrons in the third orbital. The
substance is therefore diamagnetic. In deoxy-hemoglobin there are
four unpaired electrons in the
third orbital. This substances is
paramagnetic. The 4-s orbital is
empty in both cases, because iron
is in the ferrous (+2) state. In metHb, the third orbital contains five
unpaired electrons. Due to the
loss of one additional electron,
iron is now in the ferric (+3) state

This implies the importance of CT in the early hours of

head injury (Fig. 6a) [26, 34].
In the acute stage the clotted intact RBCs contain
deoxy-Hb [32]. Lack of access of water molecules to the
paramagnetic center of the protein prevents PEDD interaction; thus, there is no T1-shortening and hematomas appear iso- to hypointense to brain on T1-weighted
images. On the other hand, the presence of magnetically
susceptible deoxy-Hb within the intact RBC creates a
field gradient across the cell membrane. The diffusion
of water molecules in and out of the RBC in this magnetic field gradient results in dephasing of protons and
hence T2-shortening. T2-shortening also occurs due to
clot retraction and increase in hematocrit; therefore,
T2-weighted images demonstrate marked hypointensity
(Fig. 6b, c) [35].
In the early subacute stage clot retraction occurs and
deoxy-Hb is oxidized to met-Hb [27]. The RBCs are still
intact and iron in met-Hb is in the ferric state (Fe3+) with
five unpaired electrons. These unpaired electrons render met-Hb paramagnetic. T1 is markedly shortened due
to PEDD, resulting in increased intensity on T1weighted images [27]. The selective T2 PRE induced by
the intracellular met-Hb causes marked hypointensity
on T2-weighted images. Early subacute hemorrhage is,
therefore, bright on T1-weighted images and dark on
T2-weighted images (Fig. 6d, e).
In the late subacute stage the severely hypoxic RBCs
undergo cell lysis, met-Hb becomes extracellular, and
there is no more magnetic non-uniformity. Even though
there is decompartmentalization, iron is still in the ferric
state. On T1-weighted images late subacute hemorrhage
appears bright due to PEDD. Loss of the compartmen-

talization of met-Hb and the increased water content of

the lysed RBCs result in T2-lengthening (Fig. 6f, g) [36].
T1-weighted images are of use in the distinction of acute
and subacute hemorrhages, whereas T2-weighted images play an important role in the distinction of early
and late subacute hemorrhages.
In the chronic stage macrophages digest the clot and
lysed RBCs are transformed into proteinaceous fluid.
With continued oxidative denaturation, met-Hb is converted to high molecular complexes such as ferritin and
hemosiderin [37]. Both are crystalline storage forms of
iron [37]. In both cases, iron is in the ferric state. Ferritin
represents the main storage form of iron in the human
body. It is water soluble and shortens both T1 and T2
relaxation times, which results in a signal intensity
change on MRI. Hemosiderin, which is a degradation
product of ferritin, is water-insoluble and has a much
stronger T2-shortening effect than ferritin [38]. In hemosiderin, an alteration in the tertiary structure of the
globin molecule occurs such that the sixth coordination
site of the heme iron is occupied by a ligand from within
the globin molecule. This causes a very large number of
unpaired electrons making hemosiderin superparamagnetic [29, 39]. Hemosiderin does not exhibit a T1-shortening effect because of its larger cluster size and its water insolubility; therefore hemosiderin is isointense on
T1-weighted images. Conversely, T2-PRE occurs due to
lysosomal compartmentalization of hemosiderin, resulting in a very hypointense appearance on T2-weighted scans (Fig. 6h, i) [37]. Heme degradation products
are also phagocytosed by glial cells that surround the
clot. This brain stain can persist almost indefinitely if
the hemorrhage occurs in areas with intact BBB [40].

1778

Fig. 8 ac Hemosiderin deposition in the subcortical white matter

is the hallmark of hemorrhagic diffuse axonal injury. The patient is
a 16-year-old man who had suffered a closed head injury (MVA
with deceleration trauma) 5 months prior to this MR examination.
a Axial T2-weighted TSE MR image reveals two questionable hypointense lesions in the right frontal subcortical white matter.
b Axial fluid-attenuated inversion recovery (FLAIR) T2-weighted
MR image does not show areas of gliosis. c Axial gradient echo fast
low-angle shot T2*-weighted MR image displays multiple hypointense foci in both frontal lobes at the gray matterwhite matter
junction. The markedly hypointense appearance of the lesions,
their multiplicity, and topographical distribution are typical of hemosiderin deposits following hemorrhagic shearing injuries. This
example underscores the importance of obtaining heavily T2*weighted images whenever there is a clinical suspicion of old
hemorrhagic lesions

The accumulation of hemosiderin-laden macrophages

does not occur in the pituitary gland, which lacks a BBB
[41]. Hemosiderin deposits are extremely hypointense
on gradient-echo images due to their strong magnetic
susceptibility (Fig. 8). The T2-shortening due to magnetic susceptibility effects is enhanced on higher-fieldstrength systems and on gradient-echo images, whereas
it is reduced with fast spin-echo MR techniques [26].
The different stages of an evolving hematoma are
summarized in Table 2, and are correlated with MRI
signal intensities on T1- and T2-weighted images.
Like parenchymal hemorrhage, SDH undergoes five
stages of evolution and thus presents five different appearances on MRI [26]. Because of the higher oxygen
tension in the vascularized dura, the progression from one
stage to the next is slower in the extra-axial compartment
[42]. The first four stages are the same as for a parenchy-

mal hematoma with the same T1 and T2 characteristics.

Rebleeding into a pre-existing SDH is demonstrated
clearly on MRI (Fig. 9). In the chronic stage, there is continued oxidative denaturation of met-Hb that results in
the formation of nonparamagnetic hemichromes [28];
thus, the intensity of chronic SDH is less than that of subacute subdural hematomas, particularly on T1-weighted
images. On T2-weighted studies, most SDH are hyperintense. In the extra-axial compartment during the chronic
phase there is no hemosiderin rim per se since it is located
outside the BBB [42]; however, with repeated episodes of
subdural bleeding, hemosiderin is deposited due to the
poor clearance mechanism [42]. The SDH is then outlined
by a hypointense rim due to hemosiderin deposition.
Epidural hemorrhage also evolves through the same
stages as SDH. It is distinguishable from the latter on the
basis of its typical morphology and topography, as well as
by the low intensity of the fibrous dura mater between
the hematoma and the brain. This is most prominent in
the late subacute stage due to the hyperintensity of metHb on both T1- and T2-weighted images.
Both SAH and IVH differ from other extra-axial
hemorrhages in that they are mixed with CSF. Because
of the high ambient oxygen level (pO2) in this environment, the progression from one stage to the next is much
slower. Moreover, since IVH and SAH are mostly arterial in origin, the predominant form of hemoglobin is
oxy-Hb. Immediately after the extravasation of blood in
SAH or IVH, there is a shortening in T1 due to increase
in hydration-layer water owing to the higher protein
content of bloody CSF [43]. This results in areas of increased signal intensity on T1-weighted images and
proton-density-weighted images [44, 45]. A fluid-atten-

1779

Table 2 Sequential signal intensity (SI) changes of intracranial

hemorrhage on MRI (1.5 T). Hb hemoglogbin; e- electrons; PEDD
proton-electron dipole-dipole interaction; T2-PRE T2-proton relaxation enhancement; FeOOH ferric oxyhydroxide; isointense

relative to normal gray matter; increased SI relative to normal

gray matter; decreased SI relative to normal gray matter;
markedly decreased SI relative to normal gray matter

Hyperacute
hemorrhage

Acute hemorrhage

What happens

Blood leaves the

vascular system
(extravasation)

Deoxygenation with Clot retraction and

formation of deoxy- deoxy-Hb is oxidizHb
ed to met-Hb

Cell lysis (membrane Macrophages digest

disruption)
the clot

Time frame

< 12 h

Hours to days
(weeks in center
of hematoma)

A few days

47 days to 1 month

Weeks to years

Red blood cells

Intact erythrocytes

Intact, but hypoxic

erythrocytes

Still intact, severely

hypoxic

Lysis (solution of
lysed cells)

Gone; encephalomalacia with proteinaceous fluid

State of Hb

Intracellular oxy-Hb Intracellular deoxyHb

Intracellular met-Hb Extracellular met-Hb Hemosiderin (inso(first at periphery of

luble) and ferritin
clot)
(water soluble)

Oxidation state

Ferrous (Fe2+),
no unpaired e-

Ferrous (Fe2+),
four unpaired e-

Ferric (Fe3+), five

unpaired e-

Magnetic properties

Diamagnetic (c)

Paramagnetic (c > 0) Paramagnetic (c > 0) Paramagnetic (c > 0) FeOOH is superparamagnetic

SI on T1-weighted
images

or

(or ) (No PEDD

interaction)

(PEDD interaction)

SI on T2-weighted
images

(High water
content)

T2 PRE (susceptibility effect)

T2 PRE (suscep- No T2 PRE

tibility effect)
(loss of compartmentalization)

uation inversion recovery (FLAIR) sequence nulls or

greatly reduces the signal from CSF and produces heavy
T2-weighting, attained with long TE values; thus, SAH
and IVH appear hyperintense compared with the CSF
and surrounding gray matter on FLAIR images
Fig. 9 a, b Chronic SDH with recurrent hemorrhage in a 31-year-old
woman. a Coronal T1-weighted SE
image (TR/TE = 520/15 ms). b Axial T2-weighted TSE image (TR/
TE = 5900/90 ms). Both images
show a crescent-shaped SDH over
the right cerebral hemisphere. The
subdural hematoma is of mixed signal intensity and presents a laminated, two-layered appearance. The
outer layer is of high signal intensity
on both T1- and T2-weighted images. This signal intensity behavior
is typical of extracellular methemoglobin. The inner layer is of intermediate signal intensity on T1weighted images, and is markedly
hypointense on T2-weighted images. This indicates the presence of
intracellular deoxyhemoglobin. The
inner layer is the most recent site of
bleeding

Early subacute
hemorrhage

Late subacute
hemorrhage

Ferric (Fe3+), five

unpaired e-

(PEDD interaction)

Chronic hemorrhage

Ferric (Fe3+) 2000 5

unpaired e-

or (no PEDD
interaction)
T2 PRE (susceptibility effect)

(Fig. 10) [44]. In the chronic stage, after repeated episodes of SAH, hemosiderin may stain the leptomeninges, leading to superficial siderosis. This causes a hypointense lining of the brain surface on T2-weighted images (Fig. 11) [46, 47].

1780

Fig. 10 Subacute intraventricular and subarachnoid hemorrhage

in a 47-year-old woman typically appears as high signal foci on a
FLAIR sequence. Hemorrhagic sedimentation levels are seen in
the third ventricle, and in the occipital horns of the lateral ventricles. The increased signal intensity in the dependent areas represents methemoglobin. The FLAIR is the most sensitive sequence
for the detection of intraventricular and subarachnoid blood, because the signal from cerebrospinal fluid is suppressed

Biological factors
Apart from the mechanisms mentioned previously, certain biological factors profoundly influence the MR imaging appearance of hemorrhage [48]. These factors include: oxygenation of the affected tissue; protein concentration (which depends on hematocrit, clot matrix
retraction and RBC hydration); and integrity of the
BBB [35, 49].
Oxygenation (pO2)
In experimental studies using a mixture of RBC and
CSF, the T1 relaxation time is not affected by variations
in pO2 [40]. On the contrary, the T2 relaxation time of
water protons in blood depends on the state of O2 saturation of hemoglobin. The T2 relaxation rate (1/T2)
varies as the square of the concentration of deoxyhemoglobin [40, 50].
The more hypoxic the environment, the greater the
hypointensity of a hematoma on T2-weighted scans [48];
thus, the center of an intracranial hematoma, which is
more hypoxic, remains markedly hypointense for a

Fig. 11 Superficial siderosis in a 68-year-old man. Axial T2weighted TSE image shows a low-signal intensity rimming of the
brain surface, particularly the brainstem and upper vermis. This
leptomeningeal hemosiderin deposition is due to repeated episodes of subarachnoid (or intraventricular) hemorrhage

longer time than the periphery. However, in an environment that is less hypoxic, the transformation of oxyHb to deoxy-Hb is slowed. This is, for example, the case
in SAH or IVH, as already discussed. The pO2 of CSF is
approximately 43 mm Hg, almost the same as the pO2
of venous blood, which is 40 mm Hg.
In this context, it should also be remembered that
there are differences in maturation between hemorrhages of venous or arterial source. Venous blood already has a higher deoxy-Hb concentration (40 %) as
compared with arterial blood (5 % deoxy-Hb); therefore, venous blood will experience a more pronounced
and earlier loss of signal intensity on T2-weighted images than arterial blood. On the other hand, the higher
pO2 in an arterial hemorrhage will cause the conversion
of deoxy-Hb to met-Hb to occur more rapidly in an arterial than in a venous hemorrhage [51].
Protein concentration
Changes in protein concentration alter the MR characteristics of aqueous solutions. In experimental circumstances, T1 and T2 relaxation times in concentrated solutions decrease as the protein concentration increases
[52]. An increase in hematocrit produces shortening of
T1 and T2. A fluidfluid level, due to sedimentation of

1781

Fig. 12 a, b Hemorrhagic sedimentation level on MRI in a 77year-old woman with an acute

right putaminal hemorrhage.
a Sagittal T1-weighted SE image (TR/TE = 525/15 ms).
b Axial T2-weighted echo-planar image (TR/TE = 5600/
159 ms). Both images show a
large mass containing a horizontal fluidfluid level. The supernatant is hypointense on T1weighted images and hyperintense on T2-weighted images.
This signal behavior is typical
of a fluid collection and corresponds to blood plasma. The
dependent portion of the hematoma is of intermediate signal intensity on T1- and T2weighted images, and corresponds to sedimented blood
cells

cellular elements, can be observed in large hematomas in

the (hyper)acute phase. The dependent portion of the
hematoma is of intermediate to low signal intensity and
corresponds to sedimented blood (Fig. 12; compare with
Fig. 2). The increase in protein (hemoglobin) concentration caused by clot retraction results in hyperintense appearance on T1-weighted images and hypointense appearance on T2-weighted images [49, 53]. The shortening
in T1 relaxation time is due to a decrease in water content
and an increase in cross-linking between protein molecules that occur during clot-matrix formation [54].
Changes in RBC hydration have an effect on both T1and T2-relaxation times of blood [43]. Dehydrated RBCs
in an area of hemorrhage would cause a substantial decrease in signal intensity on T2-weighted images [55].
Effect of BBB integrity
Hemosiderin deposition, which occurs in the late subacute and chronic stages of ICH, is due to a temporary
defect in the BBB. Hemorrhages occurring in regions of
the brain which do not possess a BBB (e.g. SDH, EDH,

and pituitary hematomas) do not usually display persistence of low signal intensity on T2-weighted images [40,
41]. An exception is the so-called superficial siderosis, a
deposition of hemosiderin pigment on the surface of the
brain, observed after repeated episodes of SAH
(Fig. 11).

Conclusion
Diagnosis of intracranial hemorrhages is based on CT,
which identifies hemorrhage as a high-attenuation mass
within the brain substance, and MRI, which in addition
can provide a more accurate estimate of the stage of the
hemorrhage by identifying sequential patterns of transformation of the hemoglobin molecule within the hematoma.
Acknowledgement This work was supported by a Clinical Research Associate Grant of the Fund for Scientific Research, Flanders (F. W.O. Vlaanderen), Belgium (grant no. G.3C06.96). We
thank G. Van Hoorde for photographic assistance.

References
1. Giroud M, Gras P, Chadan N et al.
(1991) Cerebral haemorrhage in a
French prospective population study.
J Neurol Neurosurg Psychiatry 54:
595598
2. Kase CS (1995) Intracerebral haemorrhage. Baillieres Clin Neurol 4: 247278

3. Herman B, Leyten AC, van Luijk JH,

Frenken CW, Op de Coul AA, Schulte
BP (1982) Epidemiology of stroke in
Tilburg, the Netherlands. The population-based stroke incidence register.
Part 2. Incidence, initial clinical picture
and medical care, and 3-week case fatality. Stroke 13: 629634

4. Zimmerman RA, Bilaniuk LT (1982)

Computed tomographic staging of
traumatic epidural bleeding. Radiology
144: 809812

1782

5. Phelps ME, Gado MH, Hoffman EJ

(1975) Correlation of effective atomic
number and electron density with attenuation coefficients measured with
polychromatic X rays. Radiology 117:
585588
6. Dolinskas CA, Bilaniuk LT, Zimmerman RA, Kuhl DE, Alavi A (1977)
Computed tomography of intracerebral
hematomas. II. Radionuclide and
transmission CT studies of the perihematoma region. AJR 129: 689692
7. New PJF, Aronow S (1976) Attenuation
measurements of whole blood and
blood fractions in computed tomography. Radiology 121: 635640
8. Bahner ML, Reith W, Zuna I, Engenhart-Cabillic R, van Kaick G (1998)
Spiral CT vs incremental CT: Is spiral
CT superior in imaging of the brain?
Eur Radiol 8: 416420
9. Bergstrom M, Ericson K, Levander B et
al (1977) Variation with time of the attenuation values of intracranial hematomas. J Comput Assist Tomogr 1:
5763
10. Scotti G, Terbrugge K, Melancon D
et al. (1977) Evaluation of the age of
subdural hematomas by computerized
tomography. J Neurosurg 47: 311315
11. Brooks RA, Chiro G di, Patronas N
(1989) MR imaging of cerebral hematomas at different field strengths: theory and applications. J Comput Assist
Tomogr 13: 194206
12. Tans JT (1977) Computed tomography
of intracerebral hematoma. Clin Neurol
Neurosurg 79: 285295
13. Gean AD (1994) Imaging of head trauma. Raven Press, New York
14. Dolinskas CA, Bilaniuk LT, Zimmerman RA, Kuhl DE (1977) Computed
tomography of intracerebral hematomas. I. Transmission CT observations
on hematoma resolution. AJR 129:
681688
15. Cohen WA, Wayman LA (1992) Computed tomography of intracranial hemorrhage. Neuroimaging Clin North Am
2: 7587
16. Lennington BR, Laster DW, Moody
DM, Ball MR (1979) Pre-enhancement
ring density in resolving intracerebral
hematomas. Comput Tomogr 3:
105109
17. Kreel L, Kay R, Woo J, Wong HY,
Nicholls MG (1991) The radiological
(CT) and clinical sequelae of primary
intracerebral haemorrhage. Br J Radiol
64: 10961100
18. Laster DW, Moody DM, Ball MR
(1978) Resolving intracerebral hematoma: alteration of the ring sign with
steroids. AJR 130: 935939

19. Wilms G, Marchal G, Geusens E,

Raaijmakers C, Van Calenbergh F,
Goffin J, Plets C (1992) Isodense subdural haematomas on CT: MRI findings. Neuroradiology 34: 497499
20. Hashimoto N, Sakakibara T, Yamamoto K, Fujimoto M, Yamaki T (1992)
Two fluid-blood density levels in
chronic subdural hematoma. Case report. J Neurosurg 77: 310321
21. Iplikcioglu AC, Akkas O, Sungur R
(1991) Ossified chronic subdural hematoma: case report. J Trauma 31: 272295
22. LeRoux PD, Haglund MM, Newell
DW, Grady MS, Winn HR (1992) IVH
in blunt head trauma: an analysis of 43
cases. Neurosurgery 31: 678685
23. Parizel PM, zsarlak O, Van Goethem
JW et al. (1998) Imaging findings in diffuse axonal injury after closed head
trauma. Eur Radiol 8: 960965
24. Weisberg LA (1979) Computerized tomography in intracranial hemorrhage.
Arch Neurol 36: 422426
25. Gomori JM, Grossman RI, Yu-Ip C,
Asakura T (1987) NMR relaxation
times of blood: dependence on field
strength, oxidation state, and cell integrity. J Comput Assist Tomogr 11:
684690
26. Bradley WG Jr (1993) MR appearance
of hemorrhage in the brain. Radiology
189: 1526
27. Bradley WG, Schmidt PG (1985) Effect
of methemoglobin formation on the
MR appearance of subarachnoid hemorrhage. Radiology 156: 99103
28. Wintrobe MM, Lee GR, Boggs DR
et al. (1981) Clinical hematology. Lea
and Febiger, Philadelphia, pp 88102
29. Saini S, Frankel RB, Stark DD, Ferrucci JT (1988) Magnetism: a primer
and review. AJR 150: 735743
30. Bush CH (2000) The magnetic resonance imaging of musculoskeletal hemorrhage. Skeletal Radiol 29: 19
31. Gomori JM, Grossman RI (1988)
Mechanisms responsible for the MR
appearance and evolution of intracranial hemorrhage. Radiographics 8:
427440
32. Gomori JM, Grossman RI, Goldberg
HI, Zimmerman RA, Bilaniuk LT
(1985) Intracranial hematomas: imaging by high-field MR. Radiology 157:
8793
33. Fermi G, Perutz MF, Shulman RG
(1987) Iron distances in hemoglobin:
comparison of X-ray crystallographic
and extended X-ray absorption fine
structure studies. Proc Natl Acad Sci
USA 84: 61676168

34. Snow RB, Zimmerman RD, Gandy SE,

Deck MD (1986) Comparison of magnetic resonance imaging and computed
tomography in the evaluation of head
injury. Neurosurgery 18: 4552
35. Clark RA, Watanabe AT, Bradley WG,
Roberts JD (1990) Acute hematomas:
effects of deoxygenation, hematocrit,
and fibrin-clot formation and retraction
on T2-shortening. Radiology 175:
201206
36. Hackney DB, Atlas SW, Grossman RI
et al. (1987) Subacute intracranial hemorrhage: contribution of spin density to
appearance on spin-echo MR images.
Radiology 165: 199202
37. Thulborn KR, Sorensen AG, Kowall
NW et al. (1987) The role of ferritin and
hemosiderin in the MR appearance of
cerebral hemorrhage: a histopathologic
biochemical study in rats. AJR 154:
10531059
38. Vymazal J, Urgosik D, Bulte JW (2000)
Differentiation between hemosiderinand ferritin-bound iron using nuclear
magnetic resonance and magnetic resonance imaging. Cell Mol Biol 46:
835842
39. Bean CP, Livingstone JD (1987) Superparamagnetism. J Appl Physiol 30:
12051295
40. Grossman RI, Gomori JM, Goldberg
HI et al. (1986) Magnetic resonance
imaging of hemorrhagic conditions.
Acta Radiol Suppl 369: 5355
41. Piotin M, Tampieri D, Rfinacht DA,
Mohr G, Garant M, Del Carpio R,
Robert F, Delavelle J, Melanson D
(1999) The various MRI patterns of pituitary apoplexy. Eur Radiol 9: 918923
42. Fobben ES, Grossman RI (1986) MR
characteristics of subdural hematomas
and hygromas at 1.5 T. AJNR 10:
687696
43. Fabry ME, Eisenstadt M (1978) Water
exchange across red cell membranes. II.
Measurements by nuclear magnetic
resonance T1, T2, and T12 hybrid relaxation. The effects of osmolarity, cell
volume, and medium. J Membr Biol 42:
375398
44. Noguchi K, Ogawa T, Inugami A et al.
(1995) Acute subarachnoid hemorrhage: MR imaging with fluid attenuated inversion recovery pulse sequences.
Radiology 196: 773777
45. Noguchi K, Ogawa T, Inugami A,
Toyoshima H, Okudera T, Uemura K
(1995) MR of acute subarachnoid hemorrhage: a preliminary report of fluidattenuated inversion-recovery pulse sequences. AJNR 15: 19401943

1783

46. Grossman RI, Gomori JM, Bilaniuk LT,

Zimmerman RA, Goldberg HI (1985)
High-field MR imaging of superficial
siderosis of the central nervous system.
J Comput Assist Tomogr 9: 972975
47. Janss AJ, Galetta SL, Freese A et al.
(1993) Superficial siderosis of the central nervous system: magnetic resonance
imaging and pathological correlation.
Case report. J Neurosurg 79: 756760
48. Grossman RI, Gomori JM, Goldberg
HI et al. (1988) MR imaging of hemorrhagic conditions of the head and neck.
Radiographics 8: 441454
49. Hayman LA, Taber KH, Ford JJ, Bryan
RN (1991) Mechanisms of MR signal
alteration by acute intracerebral blood:
old concepts and new theories. AJNR
12: 899907

50. Thulborn KR, Waterton JC, Matthews

PM, Radda GK (1982) Oxygenation
dependence of the transverse relaxation
time of water protons in whole blood at
high field. Biochim Biophys Acta 714:
265270
51. Weingarten K, Zimmerman RD, DeoNarine V et al. (1991) MR imaging of
acute intracranial hemorrhage: findings
on sequential spin-echo and gradientecho images in a dog model. AJNR 12:
457467
52. Fullerton GD, Potter JL, Dornbluth
NC (1982) NMR relaxation of protons
in tissues and other macromolecular
water solutions. Magn Reson Imaging
1: 209226

53. Hayman LA, Ford JJ, Taber KH, Saleem A, Round ME, Bryan RN (1988)
T2 effect of hemoglobin concentration:
assessment with in vitro MR spectroscopy. Radiology 168: 489491
54. Hayman LA, Taber KH, Ford JJ et al.
(1989) Effect of clot formation and retraction on spin-echo MR images of
blood: an in vitro study. AJNR 10:
11551158
55. Taber KH, Ford JJ, Jensen RS et al.
(1992) Change in red blood cell relaxation with hydration: application to MR
imaging of hemorrhage. J Magn Reson
Imaging 2: 203208