Eur Radiol (2008) 18: 119–129

DOI 10.1007/s00330-007-0710-2 NEURO

Karel Deblaere
Eric Achten
Structural magnetic resonance imaging
in epilepsy

Received: 23 November 2006

Abstract Because of its sensitivity hippocampal sclerosis and malforma-
Revised: 8 May 2007 and high tissue contrast, magnetic tions of cortical development, two
Accepted: 4 June 2007 resonance imaging (MRI) is the tech- major causes of medically intractable
Published online: 25 September 2007 nique of choice for structural imaging focal epilepsy.
# Springer-Verlag 2007 in epilepsy. In this review the effect of
K. Deblaere . E. Achten using optimised scanning protocols
Department of Neuroradiology, and the use of high field MR systems Keywords Brain . Imaging .
Ghent University Hospital, on detection sensitivity is discussed. Epilepsy
Ghent, Belgium
Also, the clinical relevance of ade-
K. Deblaere (*) quate imaging in patients with focal
MR Department—1K12, epilepsy is highlighted. The most
Ghent University Hospital, frequently encountered MRI findings
De Pintelaan 185, 9000 Ghent, Belgium
e-mail: [email protected] in epilepsy are reported and their
Tel.: +32-9-2406674 imaging characteristics depicted. Im-
Fax: +32-9-2404969 aging focus will be on the diagnosis of

Introduction imaging and good interpretation of the images. The

radiologist plays a crucial role in the detection of lesions
The prevalence of epilepsy is estimated to be between 0.5 that cause epilepsy. This role is even more pronounced
and 1% of the population. The incidence of new cases is when imaging is performed on patients with refractory
estimated to be approximately 30–50/100,000 person-years partial epilepsy who might be considered for epilepsy
[1]. Besides the disease related morbidity, epilepsy also has surgery, such as patients with mesial temporal sclerosis or
a high social, psychological and economical impact, both focal malformations of cortical development. It is not only
on patients and their environment. important to detect possible lesions but also to depict the
The diagnosis of epilepsy is merely clinical, but medical full extent of the lesions, as this may influence post-
imaging, especially magnetic resonance imaging (MRI) surgical outcome.
has had a high impact on the diagnostic work-up as well as
therapeutic planning. It seems that this role is still
becoming more and more important in patients with Who should be imaged?
refractory epilepsy, as newer techniques reveal more
lesions that are amenable to surgical therapy. According to the International League Against Epilepsy
This review does not aim at being complete from an (ILAE) [2], “everybody with epilepsy should have, in the
imaging findings point of view. The main goal is to stress ideal situation, a high quality MRI”. Considering the
the clinical importance and consequences of adequate therapeutic impact that MRI has on the management of
120

epilepsy patients [3] this may indeed be the optimal ictal discharges and not necessarily indicate the origin of
scenario, both for the benefit of the patients as well as from the discharges [14, 15].
a cost-benefit point of view.
The diagnostic yield of MRI depends on many factors,
one being the patient population it is applied to. The Epilepsy-optimised scanning protocols
diagnostic yield in detecting lesions possibly causing the
seizures ranges from 12.7% to 14% of patients presenting In a study by Von Oertzen et al. [16], it was demonstrated
with newly diagnosed epilepsy [4] or a first seizure [5]. The that a standard scanning protocol is insufficient to detect
yield goes up to 26% when optimal MRI is applied in an the majority of lesions in epilepsy patients with refractory
epilepsy patient population with localisation-related epi- epilepsy. In this study, a comparison was made between
lepsies [6]. The highest yield (65–83%) in detecting standard MRI and epilepsy-specific MRI findings in patients
epileptogenic lesions is observed in patient groups with with focal epilepsy. This study compared the results of a
intractable (temporal lobe) epilepsy who are considered standard MRI reported by “non-expert” radiologists, stan-
candidates for epilepsy surgery [7–9]. dard MRI evaluated by “epilepsy expert” radiologists, and
According to the existing evidence, MRI is not indicated epilepsy-specific MRI read by “expert” radiologists in
in patients with primary idiopathic generalised epilepsy 123 consecutive patients undergoing epilepsy surgery
and pediatric patients with febrile seizures and benign evaluation. The radiological findings were correlated with
rolandic epilepsy [5, 10]. postoperative histological examination. The sensitivity of
“non-expert” reports of standard MRI for focal lesions was
39%; of “expert” reports of standard MRI, 50%; and of
Imaging recommendations epilepsy-dedicated MRI, 91%. Hippocampal sclerosis was
missed in 86% of cases using standard MRI. Standard
Every patient with epilepsy, except for the patient MRI failed to detect 57% of focal epileptogenic lesions.
categories mentioned above, should have a high quality Although MRI is considered a very sensitive technique,
MRI. The term “high quality MRI” may be subject of this study illustrates the impact of optimising imaging
discussion, because MRI is already considered a very protocols on the diagnostic sensitivity.
sensitive technique. However, the imaging sequences Imaging protocols should be specifically tailored
used, slice positioning and experience of the radiologist according to the clinical information provided [17]. For
may play an important role in the sensitivity of our MRI instance, in a patient with complex partial epilepsy (CPE)
technique. the protocol should be aimed at detecting mesiotemporal
abnormalities, especially hippocampal sclerosis. In patients
with Jacksonian epilepsy, imaging emphasis should be on
Epilepsy screening protocol the frontal (rolandic) cortex, and in patients with gelastic
epilepsy the hypothalamic region should be carefully
An epilepsy screening protocol is recommended in those examined in search for a hamartoma [17]. Good clinical
patients presenting with a first seizure or in the general information provided by the referring colleague is, there-
work-up of a patient diagnosed with epilepsy. To rule out fore, essential for optimal imaging in every single case.
an evolving brain process, this protocol could include axial An optimised imaging protocol should, besides the
T2- and T1-weighted images, coronal FLAIR images and standard imaging protocol, at least comprise one coronal
diffusion-weighted images. It is recommended that gado- acquisition [T2-weighted and/or fluid-attenuated inversion
linium should not be administered in a routine fashion in recovery (FLAIR)] and a high resolution, T1-weighted,
the work-up of the epilepsy patient as it seldom gives gradient echo three-dimensional (3D) volume acquisition
additional information. Only when the clinical presentation (MPRAGE, SPGR). The first sequence allows to inspect
is suggestive (e.g. meningitis) or when initial imaging structures such as the hippocampal formation. The latter
findings are indicative (mass lesion, vascular lesion etc.) gives the opportunity to view the images in a multiplanar
should gadolinium be administered [11]. way for better anatomical localisation, helps to avoid
Approximately 50% of patients with primary and partial volume effects in the detection of malformations of
metastatic brain tumours are affected by seizures [12]. cortical development (MCDs) and can be used to correct
Tumours located in the frontal, frontoparietal and temporal for the head rotation in the coronal plane. With the current
regions are more likely to cause epilepsy than lesions techniques of parallel imaging and the upcoming high
located elsewhere in the brain and seizures are more field systems, high resolution 3D measurements can be
common in patients aged 30–50 years [13]. acquired, without compromising the signal-to-noise ratio
When using diffusion-weighted imaging in patients in (SNR). One of these 3D measurements can replace
the postictal phase, radiologists should be aware that the multiple measurements made with the same weighting
encountered signal changes may be the consequence of but in different scanning planes (see Fig. 7).
 121

The authors also recommend adding a gradient echo T2 signal on T1-weighted images and volume loss [26]
image (T2*) or the more recent susceptibility weighted (Fig. 1). A recent study by Briellman et al. [28] demon-
image (SWI) to the scanning protocol to detect hemosid- strated that the high signal on T2 and FLAIR images is due
erin depositions, small cavernomas or vascular anomalies. to dentate gliosis. Loss of dentations can also be a sensitive
marker for HS [29].
Three different variants of HS were described by
MRI findings in intractable epilepsy Jackson et al. [30]. First there is the classical type of HS
(pure HS), described as abnormalities restricted to the
Hippocampal sclerosis (HS) hippocampus only. The second type is where there are
signal abnormalities in the tissues surrounding the hippo-
HS is the most frequently encountered cause of refractory campus of a similar type to the HS. The authors called this
temporal lobe epilepsy. In a study with 2,200 adult out- type HS+, the “plus” referring to areas involved additional
patients with a follow-up between 1 and 7 years, Semah et al. to the hippocampus. The third type of HS occurs in the
[18] reported that patients with HS have the highest degree of presence of a distinct other type of pathology such as a
intractability with anti-epileptic drugs (AEDs), together with tumour, post-traumatic injury, etc. The authors called this
temporal lobe epilepsy in general. These results suggest that type +HS, suggesting that the HS is secondary, possibly as
medial temporal lobe epilepsy (mTLE), most frequently a consequence of the effect of the seizures (which
caused by HS, is a condition that is difficult to treat from the originated in the other distinct lesion) on the ipsilateral
start. Surgical treatment for certain types of medically hippocampus. This type is recognised with increased
refractory epilepsy can be an alternative and recent advances frequency and might be present in 30% of the cases
in diagnostic procedures have greatly increased interest in where HS is found [30]. The implication of this is that
surgical therapy in the past two decades [19]. Surgical radiologists should be aware that in 30% of the patients
treatment to cure seizures has been particularly recom- with MR features of HS another lesion can be present in the
mended for certain “surgically remediable syndromes” [20], same hemisphere that may cause the epilepsy. The findings
especially mTLE, the most common form of human epilepsy of MRI features compatible with HS should not satisfy the
and the most refractory to pharmacotherapy [18]. Averaged physician in his/her quest for detecting lesions, but should
over all studies, about 60% of the patients with intractable alert the radiologist that other pathologies may be present
mTLE selected for surgery become seizure-free [19]. Wiebe that are the primary source of the epilepsy (+HS) or that are
et al. [21] conducted an intention-to-treat randomised, secondary to the HS (HS+).
controlled trial of surgery for mTLE and found that 58% of Finally, when no structural anomalies can be observed in
the patients in the surgical group were free of seizures a patient with suspected mTLE other dedicated MR
impairing awareness compared with 8% in the group techniques, such as hippocampal volume measurements
randomised to continued medical therapy (P<0.001). [31], MR spectroscopy [32], T2 relaxometry [33] and
The role of MRI in patients with mTLE is such that the diffusion tensor imaging [34, 35] of the hippocampus, can
diagnosis of HS on MRI has a positive predictive value on help lateralising mTLE. These techniques are not routinely
seizure outcome after surgery [22–24]. implicated and not available to every radiologist, and are
The facts that (1) patients with mTLE, and especially considered by the authors as beyond the scope of this
those with HS, are highly refractory to drug therapy, that review.
(2) epilepsy surgery may be the only alternative to become
seizure free and that (3) patients are less likely to be oper-
ated on when no hippocampal anomalies are being detected Malformations of cortical development (MCDs)
on MRI, underline the impact of MRI on patient manage-
ment and stress the importance of adequate imaging. MCDs are generally qualified according to the stage in
Optimal imaging of the hippocampus and mesial which cortical development was disrupted. These mal-
temporal lobe is performed by 2–3 mm paracoronal images formations also known as “cortical dysplasias” or “neuro-
orthogonal to the long axis of the hippocampus. So far the nal migration disorders” have been recognized increasingly
“gold standard” scan protocol included coronal T1 inver- in patients with drug refractory epilepsy. Interest in these
sion recovery (T1 IR) and coronal FLAIR sequences [25, malformations has raised because of the revolution in
26]. Nowadays, with better image resolution and higher molecular biology and also due to the in vivo diagnosis by
SNR on high field systems, more and more centres prefer high resolution MRI techniques [36, 37]. However,
using high resolution T2-weighted images, with or without uniformity in classification schemes has hampered further
inverting the contrast, to delineate the internal layered research as different terms and classifications were used in
structure of the hippocampus [27]. different centres, leading to confusion when communicat-
Typical imaging features of a sclerotic hippocampus are ing about this topic. So far, the most comprehensive class-
a high signal on FLAIR and T2-weighted images, a low ification has been proposed by Barkovich et al. [38] (Table 1).
122

Fig. 1 Coronal FLAIR (a) and

T2-weighted images (b) in a
patient with left temporal lobe
epilepsy. The left hippocampus
is smaller, has lost internal
structure and has a high signal
on both T2-weighted and
FLAIR images (arrows). Con-
firmed left HS on pathology

MRI for MCDs should be performed using high- use in assessing MCDs. Barkovich et al. [36] suggest that
resolution T1-weighted gradient echo volume acquisitions if a study is obtained before the age of 8 months, the
(MPRAGE, SPGR). The inversion recovery-like contrast physician should rely on the T2-weighted images, and if
of the 3D T1-weighted images provides a strong contrast the study is obtained at the age of 15 months but before
between grey and white matter. the age of 30 months, the physician should rely more
Additional FLAIR/T2-weighted images are necessary, on the appearance of the T1-weighted images. If any
preferably 3D, or when not available, at least two doubt exists, a repeat examination should be performed
orthogonal directions for cross-referencing [39]. at a later stage [36].
In the neonate and infant, FLAIR images have little The imaging characteristics and clinical correlations of
contrast between grey and white matter and have little or no some of the most frequently encountered MCDs are
reported below.

Table 1 Simplification of the classification scheme according to

Barkovich et al. [38] Heterotopia
Classification scheme Heterotopia is the presence of clusters of normal grey
matter (neurons) in abnormal locations along the migration
1. Malformations due to abnormal neuronal and glial proliferation or path from the ventricular wall to the normal peripheral
apoptosis location of the cortex. Classically, heterotopia is divided in
a. Decreased proliferation/increased apoptosis: microcephalies three groups based on clinical and imaging characteristics:
b. Increased proliferation/decreased apoptosis: megalencepalies subependymal or periventricular heterotopia, subcortical
c. Abnormal proliferation (abnormal cell types): heterotopia and band or laminar heterotopia.
– non-neoplastic (tuberous sclerosis, cortical dysplasia with Subependymal or periventricular heterotopia consists of
balloon cells, hemimegalencepalie) round or ovoid nodules of grey matter that reside in the
– neoplastic (association with disordered cortex): DNET, wall of the lateral ventricles or that lie directly lateral to the
ganglioglioma, gangliocytoma ventricular wall (Fig. 5). These nodules remain isointense
to grey matter on all sequences, which helps us to
2. Malformations due abnormal neuronal migration
differentiate these lesions from the periventricular lesions
a. Lissencephaly/subcortical band heterotopia spectrum found in tuberous sclerosis. Subependymal or periventric-
b. Cobblestone complex ular heterotopias do not calcify and do not enhance after
c. Heterotopia (subependymal, subcortical, marginal glioneuronal) contrast administration. These heteropias may be found in
3. Malformations due to abnormal cortical organisation isolation or combined with other developmental anomalies,
a. Polymicrogyria and schizencepahly most frequently callosal agenesis and Chiari II malforma-
b. Cortical dysplasia without balloon cells tions. In these latter cases, presentation depends merely on
c. Microdysgenesis
the associated anomalies [36]. Patients with the isolated
form may present with epilepsy but also with milder
4. Malformations of cortical development, not otherwise classified
symptoms of developmental delay. Sometimes subepen-
 123

normal brains. The condition in which sulcation is

completely absent is referred to as complete lissencephaly
or also called agyria, whereas in patients with incomplete
lissencephaly (also called pachygyria) (Fig. 3) only a few
broad gyri are present, bordered by shallow sulci. Several
authors suggested that the terms “lissencephaly”, “agyria”
and “pachygyria” should only be used when the cortex is
abnormally thick [36].
The severity of the lissencephaly, together with the
presence of associated anomalies, determines the clinical
presentation of the disorder.
Several genetic mutations are reported as the cause of
lissencephaly and imaging characteristics differ according
to the causative mutation. Please refer to the extensive
review of Barchovich et al. [36] for a detailed description
of the several subtypes. As a radiologist it is important to
describe in detail the most and least affected portions of
the brain, the presence and absence of a cell sparse zone in
the (thickened) cortex and possible abnormalities to the
cerebellum and corpus callosum. These findings help to
identify the possible genetic mutations and give proper
genetic counselling to the parents.

Polymicrogyria (PMG) and schizencephaly

Fig. 2 A 6-year-old girl with severe developmental delay and
epilepsy. Coronal T1 IR shows incomplete lissencephaly in the PMG is the most commonly encountered MCD in imaging
occipital region with presence of some shallow sulci. Note the series of patients with medically refractory epilepsy.
typical three layered cortex (see text): outer dense cortical layer
(short thick arrow), followed by a cell sparse layer (long dotted Schizencephaly is less frequently encountered and often
arrow) and a thick inner layer of grey matter (long arrow). This associated with PMG. Both anomalies can be found in an
layered aspect is lost in the parietal cortex which shows one thick isolated form or can be associated with other MCDs.
layer of grey matter with very shallow sulci (pachygyria, thick Clinical presentation varies and ranges from developmental
dotted arrow) delay to epilepsy and focal or diffuse neurological deficits.

dymal heterotopia may be an incidental finding with no

clinical impact.
Subcortical heterotopia is a very heterogeneous group
and the affected patients most frequently present with
seizures. Large bilateral subcortical heterotopia may be
associated with impaired cognitive development [40].
Imaging characteristics of these heterotopias are the same
as those of subependymal heterotopia and distribution of
these nodules may vary from localised groups to long
subcortical curvilinear bands of grey matter nodules.
Some authors consider band heterotopia (Fig. 3) as a
mild form of lissencephaly, and this is discussed in the next
paragraph.

Lissencephaly and pachygyria

The most typical disorder resulting from arrests in neuronal Fig. 3 Sagittally acquired 3D T1-weighted image (MPRAGE) of a
13-year-old girl with learning difficulties and epilepsy. In the frontal
migration is called lissencephaly or “smooth brain” region there is pachyria (short thick arrows), whereas more posterior
(Fig. 2). In this condition the sulcation is reduced and the band heteropia (long arrows) is demonstrated in these images with
depth of the sulcation is more shallow than observed in high grey/white matter contrast
124

Imaging characteristics of PMG vary according the

myelinisation of the brain. In the first stage (neonatal), the
cortex appears to be thin and irregularly ondulating
(cortical ribbon) (Fig. 4). At a later stage, when myelinisa-
tion progresses, the cortex becomes thicker (5–8 mm) than
normal cortex with an irregular surface, both at the inner
and outer surface (Fig. 5). As opposed to pachygyria,
normal sulci are not observed in the affected part of the
cortex. The distribution can be very variable and
theoretically all parts of the brain can be affected, but the
most frequently affected is the (posterior) perisylvian
region. Typical distribution patterns can be observed, one
of the best known bilateral, symmetrical PMG syndromes
is bilateral perisylvian polymicrogyria [41] (Fig. 5).
Schizencephaly is characterised by a cleft extending
from the pial surface of the brain to the ventricle. The cleft
is lined by dysmorphic appearing grey matter. Usually
there is a subdivision in open lip schizencephaly (large
open cleft) and closed lip schizencephaly (smaller defect,
CSF filling not always visible). Other diagnostic clues,
especially in the closed lip variant, are a deformity of the
ventricle at the site of the cleft pointing towards the cleft
and the cortex radiating into the cleft (Fig. 6).

Fig. 5 Axial T1 IR image of a 5-year-old boy with epilepsy and

learning difficulties. The image shows clearly bilateral subepen-
dymal and periventricular heterotopic grey matter (short arrows).
This patient has also bilateral perisylvian polymicrogyria (long
arrows) visible as an irregular thickening of the cortex and loss of
normal sulci

Focal cortical dysplasias

In the classification scheme of Barkovich et al. [38], the

existence of focal cortical dysplasias (FCDs) was not
emphasised. These FCDs are localised malformations
increasingly associated with refractory epilepsy and more
frequently being operated on in epilepsy centres. Lack of a
uniform classification scheme for these FCDs made
organisation of imaging, EEG and neuropathology data
between centres difficult. Therefore, a panel of specialists
[42] on this matter proposed a classification scheme based
on histopathological findings. Recently, imaging findings,
clinical and electrophysiological data were correlated with
the classification scheme.

Mild MCDs

These are intracortical lesions characterised by dyslamina-

tion and columnar disorganisation. They represent the
mildest form of the histopathological spectrum of FCD and
may represent the earlier term “microdysgenesis”, a term
Fig. 4 Axial T2-weighted image of a 6-month-old girl with status
epilepticus shows a region with right frontal polymicrogyria that should be abandoned according to the panel.
(arrows) that can be seen as a thin ondulating cortical ribbon with Mild MCDs may be related to epilepsy and possibly
abnormal gyration other behavioural and cognitive abnormalities. Mild MCD,
 125

according to this recent classification, is probably not

detectable by current MRI techniques and diagnosis is
usually retrospective [42].

FCD

Type I
no dysmorphic neurons or balloon cells
Type I FCDs have been shown to lack highly specific
EEG patterns and to be most frequently localised in the
temporal lobes. On MRI they present as focal hypo-
plasia with shrinkage and moderate signal intensity
alterations in the white matter core. Ipsilateral hippo-
campal sclerosis was often present [43, 44].
Type II
Taylor-type FCD [dysmorphic neurons without (type
IIA) or with balloon cells (type IIB)]
Type II or Taylor-type FCDs are generally located
extratemporal and are frequently associated with ictal-
like patterns on direct cortical recordings [42–44].
These lesions are thought to have a high degree of
intrinsic epileptogenity [45, 46]. This type is the most
frequently identified FCD on MRI; however, MRI still
Fig. 6 Coronal reconstruction of a sagittally acquired 3D T1-
weighted image (MPRAGE) of a 5-year-old girl. A closed lip can be can be normal [47]. MRI characteristics of Taylor-type
seen ranging from the left posterior ventricular surface to the pial FCDs are [48–50]: (1) focal areas of increased cortical
surface and lined with heterotopic grey matter (long arrows). Note thickness, (2) blurring of the grey/white matter junc-
also that the cortex is radiating into the cleft (thick arrow) of this tion, (3) increased signal on T2-weighted or FLAIR
closed lip schizencephaly images (more likely to occur in balloon cell containing
lesions (type IIB) and (4) extension of cortical tissue
with increased signal from the cortical surface tapering
to the ventricle [51] (transmantle dysplasia) (Fig. 7).
Although some overlap exists, type I and type II FCDs
can usually be distinguished with MRI. Because type II

Fig. 7 Sagittally acquired 3D

FLAIR image (a) with axial
multiplanar reconstruction (b) at
3 T. Previous imaging at 1.5 T
was considered normal. These
images show a high signal trail
(arrow in b), ranging from the
ventricular surface to the right
posterior insular cortex. The
cortex itself shows a focal
thickening and high signal on
FLAIR images (double arrow).
T1-weighted images (not
shown) demonstrated a minimal
blurring of the edge between
grey and white matter in this
Type II focal cortical dysplasia
126

FCDs have a better prognosis after surgery [43], a table and calcifications may be present in up to 30% of
provisional MR diagnosis may be important for cases. Scalloping of the inner table is not specific to DNET
surgical planning and prognosis. Palmini et al. [42] but is suggestive of a slowly evolving cortical lesion.
stated that if these initial results are confirmed, these Gangliogliomas can occur at any age but most
findings may have a significant impact on surgical commonly in children and young adults, with a prediliction
planning and will emphasise even more the need for in for the temporal lobe. The lesion is usually a well-
vivo MRI identification of subtypes of FCD as post- circumscribed cortically based mass, often partially cystic
surgical prognosis differ according FCD subtypes. and with no or little edema. Contrast enhancement is
variable, ranging form no enhancement to dense homoge-
neous enhancement [54]. The tumour volume of gan-
gliogliomas in early childhood (<10 years) is significantly
larger than that of gangiogliomas encountered in adult
Dysplastic tumours patients, especially the cystic part [55].

Dysembryoplastic neuroepitheliomas (DNETs) and gan-

gliogliomas are tumoural lesions frequently associated with Mass lesions
refractory epilepsy. These lesions may be surrounded with
cortical regions displaying abnormal cytoarchitecture and Besides the patient group with a primary or metastatic
large (dysmorphic) neurons. Both lesions have a good tumour presenting with a seizure, there is also an entity
prognosis after surgery. known as classic long-term epilepsy-associated tumours.
MRI in DNETs shows a cortical and well-delineated These generally younger patients have long-standing drug
lesion exceeding the thickness of the normal cortex and refractory epilepsy associated with classic epilepsy asso-
involving the white matter. There is no surrounding edema ciated tumours such as ganglioglioma, DNET (cfr supra),
or mass effect [52]. DNETs can be found anywhere in the pleomorphic xanthoastrocytoma and pilocytic astrocyto-
brain but most frequently involving the temporal lobe mas. It is recommended that these tumours should be
(Fig. 8). These lesions are typically identified before the recognised and removed early because of the high chance
age of 20 and long-standing drug resistant epilepsy is the of seizure freedom and the rare but possible risk of
most common symptom. The presence of a triangular- malignant degeneration [56].
shaped or wedge-shaped (width maximal at the cortex
tapering towards the ventricle) cortically-based mass, the
presence of “septations” isointense to grey matter may be
helpful to support the diagnosis of DNET [53]. Contrast
enhancement is rare; there may be scalloping of the inner

Fig. 8 Axial T2-weighted

image (a) of a 22-year-old
woman with chronic drug re-
fractory epilepsy shows a lesion
in the right anterior mesial tem-
poral lobe with high signal
intensity and slightly lobulated
edges. There is no surrounding
edema. T1-weighted image after
administration of gadolinium
shows no enhancement (b).
Dysembryoplastic neuroepithe-
lioma confirmed on pathology
 127

Vascular lesions

The two most frequently encountered vascular lesions

associated with epilepsy are arteriovenous malformations
(AVMs) and cavernomas. The latter are more frequently
associated with seizures than AVMs.
AVMs are generally easily recognised as a “bag of black
worms” due to the flow voids of the tightly packed vessels.
There is little or no mass effect and there may be
surrounding gliosis (Fig. 9).
Seizures and epilepsy are the most common symptom-
atic presentation of supratentorial cavernomas. Besides the
seizures and epilepsy, potential risk of bleeding and/or
focal neurological deficit play a role in the management of
these lesions [57]. These “popcorn-like” lesions have a
mixed signal core, due to the locules filled with blood in
different stages of degradation. There is a hypointense rim
surrounding the lesions on T2* weighted images due to
hemosiderin. As previously mentioned, gradient echo T2-
weighted images (T2*) are extremely useful in identifying
small cavernomas (Fig. 10).

Acquired lesions (infection, insult or trauma)

Imaging characteristics of acquired lesions, such as

Fig. 9 Axial T2-weighted image in a 34-year-old man presenting infection, infarction or trauma causing epilepsy, will not
with first seizure shows an AVM in the right posterior insular region. be discussed further, as these are well known to those
Note the multiple flow voids in the lesions (“bag of black worms”, reading brain MRIs on a regular basis. Gradient echo
long arrow) and the large draining vein (thick arrow)
T2*-weighted images may render additional information
on the presence of hemosiderin in the case of post-
traumatic epilepsy, with a preferential involvement of the
frontobasal and temporal lobes.

Fig. 10 Coronal T2-weighted

and axial T2* image of a 20-
year-old woman suffering from
epilepsy and auditory hallucina-
tions. The T2-weighted image
(a) shows a small lesion with
mixed signal intensities in the
cortex of the posterior part of
the superior temporal sulcus at
the left side (arrow). The axial
T2*-image (b) shows a clear
susceptibility artifact (arrow)
caused by the hemosiderin
present in this small cavernoma
128

Benefits of high field MRI and high resolution surface coil vated a change in clinical management. At 3.0 T, PA MRI
imaging resulted in the detection of a new lesion in 65% of the
patients in the subgroup (23/40) of patients with prior 1.5-
Although MRI in epilepsy patients with optimal scanning T MRIs interpreted as normal. This diagnostic improve-
protocols at 1.0 and 1.5 Tesla (T) machines yields a large ment combined with magnetoencephalographic (MEG)
number of positive studies, some small or subtle lesions guidance may lead to more lesions to be detected [59] as
remain undetected [27]. The recent clinical implementation well as better understanding of the relationship between
of clinical high field (3.0 T) MR scanners and the use of epileptiform activity and the presence or absence of
high resolution surface coils has expanded our imaging structural lesions on MRI [60]. As MRI techniques
armory in detecting epileptogenic lesions in the human continue to evolve, more and more lesions will be
brain. Recently Knake et al. [58] conducted an imaging detected. Whether these lesions are the cause of the
study in 40 patients with medical intractable focal epilepsy will become one of the most important questions.
epilepsies using phased array (PA) coils at 3.0 T and Therefore, the multidisciplinary approach with clinicians,
compared the findings with standard imaging at 1.5 T. other imaging modalities and electrophysiologists, as it is
Additional diagnostic information was obtained in 48% already current practice in the epilepsy centres, will
using 3.0-T PA MRI compared with prior 1.5-T MRIs. In become a conditio sine qua non in the work-up of the
37.5% of the patients this additional information moti- epilepsy patient.

References
1. Sander JW (2003) The epidemiology of 9. Lefkopoulos A, Haritanti A, 19. Engel J Jr, Wiebe S, French J et al
epilepsy revisited. Curr Opin Neurol Papadopoulou E, Karanikolas D, (2003) Practice parameter: temporal
16:165–170 Fotiadis N, Dimitriadis AS (2005) lobe and localized neocortical resec-
2. Commission on Neuroimaging of the Magnetic resonance imaging in 120 tions for epilepsy: report of the Quality
International League Against Epilepsy patients with intractable partial sei- Standards Subcommittee of the Amer-
(1997) Recommendations for neuro- zures: a preoperative assessment. Neu- ican Academy of Neurology, in asso-
imaging of patients with epilepsy. roradiology 47:352–361 ciation with the American Epilepsy
Epilepsia 38:1255–1256 10. Bernal B, Altman NR (2003) Evidence- Society and the American Association
3. Kuzniecky RI (2005) Neuroimaging of based medicine: neuroimaging of of Neurological Surgeons. Neurology
epilepsy: therapeutic implications. seizures. Neuroimaging Clin N Am 60:538–547
NeuroRx 2:384–393 13:211–224 20. Langfitt JT (1997) Cost-effectiveness
4. Berg AT, Testa FM, Levy SR, Shinnar 11. Elster AD, Mirza W (1991) MR imag- of anterotemporal lobectomy in medi-
S (2000) Neuroimaging in children ing in chronic partial epilepsy: role of cally intractable complex partial epi-
with newly diagnosed epilepsy: a contrast enhancement. AJNR Am J lepsy. Epilepsia 38:154–163
community-based study. Pediatrics Neuroradiol 12:165–170 21. Wiebe S, Blume WT, Girvin JP,
106:527–532 12. Schaller B, Ruegg SJ (2003) Brain Eliasziw M (2001) A randomized,
5. King MA, Newton MR, Jackson GD tumor and seizures: pathophysiology controlled trial of surgery for temporal-
et al (1998) Epileptology of the first- and its implications for treatment lobe epilepsy. N Engl J Med 345:311
seizure presentation: a clinical, electro- revisited. Epilepsia 44:1223–1232 22. Kuzniecky R, Burgard S, Faught E,
encephalographic, and magnetic 13. Liigant A, Haldre S, Oun A et al (2001) Morawetz R, Bartolucci A (1993)
resonance imaging study of 300 Seizure disorders in patients with brain Predictive value of magnetic resonance
consecutive patients. Lancet 352: tumors. Eur Neurol 45:46–51 imaging in temporal lobe epilepsy
1007–1011 14. Hufnagel A, Weber J, Marks S et al surgery. Arch Neurol 50:65–69
6. Griffiths PD, Coley SC, Connolly DJA (2003) Brain diffusion after single 23. Berkovic SF, McIntosh AM, Kalnins
et al (2005) MR imaging of patients seizures. Epilepsia 44:54–63 RM et al (1995) Preoperative MRI
with localisation-related seizures: initial 15. Szabo K, Poepel A, Pohlmann-Eden B predicts outcome of temporal lobecto-
experience at 3.0 T and relevance to the et al (2005) Diffusion-weighted and my: an actuarial analysis. Neurology
NICE guidelines. Clin Radiol 60:1090– perfusion MRI demonstrates parenchy- 45:1358–1363
1099 mal changes in complex partial status 24. Gilliam F, Faught E, Martin R et al
7. Urbach H, Hattingen J, Von Oertzen J epilepticus. Brain 128:1369–1376 (2000) Predictive value of MRI-
et al (2004) MR imaging in the 16. von Oertzen J, Urbach H, Jungbluth identified mesial temporal sclerosis for
presurgical workup of patients with S et al (2002) Standard magnetic surgical outcome in temporal lobe
drug-resistant epilepsy. AJNR Am J resonance imaging is inadequate for epilepsy: an intent-to-treat analysis.
Neuroradiol 25:919–926 patients with refractory focal epilepsy. Epilepsia 41:963–966
8. Cakirer S, Basak M, Mutlu A, Galip J Neurol Neurosurg Psychiatry 25. Jack CR Jr, Rydberg CH, Krecke KN
GM (2002) MR imaging in epilepsy 73:643 et al (1996) Mesial temporal sclerosis:
that is refractory to medical therapy. 17. Urbach H (2005) Imaging of the diagnosis with fluid-attenuated inver-
Eur Radiol 12:549–558 epilepsies. Eur Radiol 15:494–500 sion-recovery versus spin-echo MR
18. Semah F, Picot MC, Adam C et al imaging. Radiology 199:367–373
(1998) Is the underlying cause of
epilepsy a major prognostic factor for
recurrence? Neurology 51:1256
 129

26. Jackson GD, Berkovic SF, Duncan JS, 38. Barkovich AJ, Kuzniecky RI, Jackson 50. Lee BC, Schmidt RE, Hatfield GA,
Connelly A (1993) Optimizing the GD, Guerrini R, Dobyns WB (2001) Bourgeois B, Park TS (1998) MRI of
diagnosis of hippocampal sclerosis Classification system for malformations focal cortical dysplasia. Neuroradiolo-
using MR imaging. AJNR Am J of cortical development: update 2001. gy 40:675–683
Neuroradiol 14:753–762 Neurology 57:2168–2178 51. Barkovich AJ, Kuzniecky RI, Bollen
27. Sawaishi Y, Sasaki M, Yano T, 39. Ruggieri PM, Najm I, Bronen R et al AW, Grant PE (1997) Focal transmantle
Hirayama A, Akabane J, Takada G (2004) Neuroimaging of the cortical dysplasia: a specific malformation of
(2005) A hippocampal lesion detected dysplasias. Neurology 62:S27–S29 cortical development. Neurology
by high-field 3 Tesla magnetic reso- 40. Barkovich AJ, Kjos BO (1992) Gray 49:1148–1152
nance imaging in a patient with tem- matter heterotopias: MR characteristics 52. Daumas-Duport C, Scheithauer BW,
poral lobe epilepsy. Tohoku J Exp Med and correlation with developmental and Chodkiewicz JP, Laws ER Jr, Vedrenne
205:287–291 neurologic manifestations. Radiology C (1988) Dysembryoplastic neuroepi-
28. Briellmann RS, Kalnins RM, Berkovic 182:493–499 thelial tumor: a surgically curable
SF, Jackson GD (2002) Hippocampal 41. Kuzniecky R, Andermann F, Guerrini tumor of young patients with intracta-
pathology in refractory temporal lobe R (1993) Congenital bilateral perisyl- ble partial seizures. Report of thirty-
epilepsy: T2-weighted signal change vian syndrome: study of 31 patients. nine cases. Neurosurgery 23:545–556
reflects dentate gliosis. Neurology The CBPS Multicenter Collaborative 53. Fernandez C, Girard N, Paz Paredes A,
58:265–271 Study. Lancet 341:608–612 Bouvier-Labit C, Lena G, Figarella-
29. Oppenheim C, Dormont D, Biondi A 42. Palmini A, Najm I, Avanzini G et al Branger D (2003) The usefulness of
et al (1998) Loss of digitations of the (2004) Terminology and classification MR imaging in the diagnosis of
hippocampal head on high-resolution of the cortical dysplasias. Neurology dysembryoplastic neuroepithelial tumor
fast spin-echo MR: a sign of mesial 62:S2–S8 in children: a study of 14 cases. AJNR
temporal sclerosis. AJNR Am J 43. Tassi L, Colombo N, Garbelli R et al Am J Neuroradiol 24:829–834
Neuroradiol 19:457–463 (2002) Focal cortical dysplasia: neuro- 54. Tampieri D, Moumdjian R, Melanson
30. Jackson GD, Briellmann RS, pathological subtypes, EEG, neuro- D, Ethier R (1991) Intracerebral gan-
Kuzniecky RI (2005) Temporal lobe imaging and surgical outcome. Brain gliogliomas in patients with partial
epilepsy. In: Kuzniecky RI, Jackson 125:1719–1732 complex seizures: CT and MR imaging
GD (eds) Magnetic resonance in epi- 44. Colombo N, Tassi L, Galli C et al findings. AJNR Am J Neuroradiol
lepsy, 2nd edn. Elsevier, Amsterdam (2003) Focal cortical dysplasias: MR 12:749–755
31. Achten E, Deblaere K, De Wagter C imaging, histopathologic, and clinical 55. Provenzale JM, Ali U, Barboriak DP,
et al (1998) Intra- and interobserver correlations in surgically treated pa- Kallmes DF, Delong DM, McLendon
variability of MRI-based volume mea- tients with epilepsy. AJNR Am J RE (2000) Comparison of patient age
surements of the hippocampus and Neuroradiol 24:724–733 with MR imaging features of gan-
amygdala using the manual ray-tracing 45. Palmini A, Gambardella A, Andermann gliogliomas. AJR Am J Roentgenol
method. Neuroradiology 40:558–566 F et al (1995) Intrinsic epileptogenicity 174:859–862
32. Kuzniecky R (2004) Clinical applica- of human dysplastic cortex as sug- 56. Luyken C, Blumcke I, Fimmers R et al
tions of MR spectroscopy in epilepsy. gested by corticography and surgical (2003) The spectrum of long-term
Neuroimaging Clin N Am 14:507–516 results. Ann Neurol 37:476–487 epilepsy-associated tumors: long-term
33. Bernasconi A, Bernasconi N, 46. Gambardella A, Palmini A, Andermann seizure and tumor outcome and neuro-
Caramanos Z et al (2000) T2 relaxo- F et al (1996) Usefulness of focal surgical aspects. Epilepsia 44:822–830
metry can lateralize mesial temporal rhythmic discharges on scalp EEG of 57. Awad I, Jabbour P (2006) Cerebral
lobe epilepsy in patients with normal patients with focal cortical dysplasia cavernous malformations and epilepsy.
MRI. Neuroimage 12:739–746 and intractable epilepsy. Electroence- Neurosurg Focus 21:e7
34. Salmenpera TM, Simister RJ, Bartlett P phalogr Clin Neurophysiol 98:243–249 58. Knake S, Triantafyllou C, Wald LL
et al (2006) High-resolution diffusion 47. Bastos AC, Comeau RM, Andermann F et al (2005) 3T phased array MRI
tensor imaging of the hippocampus in et al (1999) Diagnosis of subtle focal improves the presurgical evaluation in
temporal lobe epilepsy. Epilepsy Res dysplastic lesions: curvilinear refor- focal epilepsies: a prospective study.
71:102–106 matting from three-dimensional mag- Neurology 65:1026–1031
35. Dumas de la Roque A, Oppenheim C, netic resonance imaging. Ann Neurol 59. Moore KR, Funke ME, Constantino T,
Chassoux F et al (2005) Diffusion 46:88–94 Katzman GL, Lewine JD (2002)
tensor imaging of partial intractable 48. Chan S, Chin SS, Nordli DR, Goodman Magnetoencephalographically directed
epilepsy. Eur Radiol 15:279–285 RR, DeLaPaz RL, Pedley TA (1998) review of high-spatial-resolution
36. Barkovich AJ, Raybaud CA (2004) Prospective magnetic resonance imag- surface-coil MR images improves lesion
Malformations of cortical development. ing identification of focal cortical dys- detection in patients with extratemporal
Neuroimaging Clin N Am 14:401–423 plasia, including the non-balloon cell epilepsy. Radiology 225:880–887
37. Barkovich AJ, Raybaud CA (2004) subtype. Ann Neurol 44:749–757 60. Lee DH (2006) High-resolution mag-
Neuroimaging in disorders of cortical 49. Bronen RA, Vives KP, Kim JH, netic resonance imaging (MRI), surface
development. Neuroimaging Clin N Fulbright RK, Spencer SS, Spencer DD coil MRI, and magnetoencephalogra-
Am 14:231–254, viii (1997) Focal cortical dysplasia of Tay- phy. Adv Neurol 97:279–283
lor, balloon cell subtype: MR differen-
tiation from low-grade tumors. AJNR
Am J Neuroradiol 18:1141–1151