Approximately 20 to 30% of patients with epilepsy are considered drug refractory,[1] defined by the International League Against Epilepsy (ILAE) as the persistence of
seizures despite adequate trials of two well-tolerated and appropriately chosen anti-epileptic
drugs (AEDs) under the supervision of an experienced neurologist over the course of
at least 1 year.[2] This population represents the potential pool of candidates for surgical treatment,
but not all of these patients are amenable to resective surgery. Successful surgical
intervention for epilepsy requires a thorough investigation to identify those patients
with clearly localizable epileptic foci, with the goal being freedom from seizure
with no permanent neurologic deficits.
The goal of any presurgical evaluation in a patient with drug-resistant epilepsy is
to identify a localized region of epileptic cortex and subsequently resect or disconnect
that abnormal cortex. The initial investigation relies on a detailed clinical history
and neurologic exam with consultation of both the patient and family members who have
witnessed the seizures of the patient. The presence of auras, semiology, and timing
of seizures are vital pieces of knowledge that inform seizure onset localization.
Signs and symptoms, suggestive of a focal seizure onset, are further investigated
through long-term video electroencephalogram (EEG) monitoring. Additional testing
may include behavioral evaluation and neuropsychological testing. Various neuroimaging
techniques are also employed to identify structural and/or metabolic abnormalities
that may be the source of the seizures.[3] In the perfect scenario, all data streams are concordant with ictal EEG, confirming
that the clinical seizures do indeed arise from a clearly identified cortical region.
However, in many cases, there remains some level of discordance in the functional
and anatomic seizure localizations, and in these situations, invasive intracranial
investigation with implanted electrodes may also be necessary to determine the precise
zone of seizure onset.
Most focal epilepsy arises in the temporal lobe, notably the mesial temporal lobe,
followed by the frontal, parietal, and occipital lobes.[4] In some cases, the epileptic zone is multilobar or even panhemispheric. In cases
of clear cut mesial temporal sclerosis (MTS), between 65 and 85% of patients are cured
of epilepsy following resection.[5]
[6] Surgical success is much less predictable when bilateral MTS is present or when
no identifiable temporal lobe lesion exists. This is even true in the extratemporal
epilepsies. The most important predictor of a favorable surgical outcome is the identification
of a clear cut structural lesion that is consistent with the clinical semiology of
the seizures and the ictal EEG onset.[7] Therefore, advanced neuroimaging techniques have become increasingly important in
identifying and defining epileptogenic lesions that might previously have gone undetected.
In this article, we discuss the advances in neuroimaging techniques that have allowed
surgeons to more precisely identify the epileptic zone and subtle lesions that may
not have been detectable in the past.
Magnetic Resonance Imaging
Magnetic resonance imaging (MRI) is the most useful imaging modality employed in the
investigation of epilepsy. The most common sequences to obtain are thin cut T1- and
T2-weighted images to evaluate white and gray matter distinction as well as evaluate
the fluid spaces. Fluid attenuated inversion recovery (FLAIR) sequences allow for
the evaluation of edema, subcortical sclerosis, and focal atrophy. Hemosiderin and
calcium-sensitive sequences, such as gradient echo (GRE) and susceptibility-weighted
imaging (SWI), can also be useful to detect small vascular abnormalities or evidence
of remote hemorrhage. Most images are acquired at 1.5T in the standard axial, coronal,
and sagittal planes, though oblique coronal slices perpendicular to the long axis
of the hippocampus are useful in suspected cases of MTS. This particular slice orientation
can allow for better comparison of hippocampi to determine the degree of asymmetry
between the two sides as well as any loss of internal architecture ([Fig. 1]).
Fig. 1 Oblique coronal T2 (a) and T2-FLAIR (b) MR images in a patient with intractable complex partial seizures and a history of
febrile convulsions as an infant. Note the slightly decreased volume and loss of internal
architecture on T2-weighted images as well as the increased T2 signal intensity in
the right hippocampus (arrows). FLAIR, fluid attenuation inversion recovery; MR, magnetic
resonance.
Careful examination of the images is the responsibility of the epilepsy surgery team,
since every presurgical test informs another, and all investigations are complementary.
Seizure semiology and scalp EEG often suggest where the zone of seizure onset is most
likely and focus attention to that area or lobe of interest. If a lesion is suspected,
then additional images can be obtained in coronal or oblique slices through the area
of interest. This can often convert the idiopathic nonlesional case into a lesional
case with a more favorable surgical outcome.
High-Magnetic Field Strength Imaging
One of the most touted advances in MRI has been the development of higher magnetic
field strength. Initially, small surface coils were placed over the area of interest
to improve spatial resolution, but these could only improve visualization under the
coil, and therefore placement was extremely important ([Fig. 2]). The advantages of 3T MRI and higher magnetic field imaging are improved spatial
and contrast resolution over the entire brain, with the detractions being increased
susceptibility to motion artifact and increased imaging times. The MR image clarity
at 3T can certainly improve the level of confidence that a lesion has been identified,
but in most cases, all these lesions can also be seen at 1.5T, at least in retrospect.[8] The difference is in the certainty of the diagnosis, which is very important in
making a surgical decision ([Fig. 3]).
Fig. 2 Standard T2 fast spin echo axial MR image (a) and image obtained with a surface coil over the temporal regions (b) in a patient with intractable complex partial seizures and a very active interictal
spike discharge on scalp EEG in the right temporal leads. The abnormality was clearly
identified as a small cavernous angioma, and in retrospect, there was a suspicious
area on standard MRI, but this was not clear enough on its own to make a firm diagnosis.
EEG, electroencephalogram; MR, magnetic resonance.
Fig. 3 Coronal T2-weighted MR images at 1.5T (a) and 3.0T (b) in a young boy with intractable seizures. The initial MRI at 1.5T was interpreted
as normal, but the dysplastic cortex along the collateral sulcus and the periventricular
gray matter in the right temporal lobe that is so evident on the 3T images can also
be appreciated in retrospect on the 1.5T images (arrows). MR, magnetic resonance;
MRI, magnetic resonance imaging.
While 3T MRI machines are now widely available, an increasing number of 7T machines
are currently being deployed. In one prospective study, 21 consecutive patients, with
evidence of focal onset seizures but lacking an identifiable lesion on conventional
MRI, underwent further imaging using 7T MRI. In 29% of these patients, GRE and FLAIR,
performed at 7T strength, revealed a distinct lesion, with histopathological diagnosis
of focal cortical dysplasia in all patients who underwent surgical resection.[9] The same group also reported that 7T imaging allowed for more effective visualization
of foci of polymicrogyria.[10]
Volumetric Analysis
High-resolution thin cut MRI allows for the construction of advanced volumetric and
morphological models. Volumetric analysis of the hippocampus has been a tool in the
evaluation of temporal lobe epilepsy (TLE), with seizure lateralization occurring
primarily on the side with reduced hippocampal volume.[11] These quantitative methods are useful in detecting subtle mesial temporal atrophy.
Methods are also available to render these three-dimensional (3D) volumetric MRI datasets
into accurate cortical surface renderings of the individual patient's brain. Cortical
anatomy can be examined for subtle abnormalities of gyral and sulcal patterns. These
3D images can also be used to visualize the location and extent of subcortical abnormalities
because the underlying gray matter and white matter may be examined using readily
available computerized image processing tools[12] ([Fig. 4]).
Fig. 4 Axial T2-weighted MRI (a) of a child with intractable generalized seizures due to right hemispheric schizencephaly
and a 3D surface rendering (b) of the same child that demonstrates the extent of the perisylvian gyral abnormalities
more clearly. Note the deepened sylvian fissure that extends into the parietal lobe;
the wide flattened gyri; and the lack of normal tertiary sulcal patterns. MRI, magnetic
resonance imaging.
Advanced image processing techniques can also create cortical thickness maps with
automated color-coded measurements of the cortical mantle and the gray–white junction.
The human cortical gray matter is typically 2 to 4 mm thick, but varies between Brodmann's
areas. Cortical dysplasias are often characterized by thickened and disorganized gray
matter, and subtler cases, too difficult to see on standard imaging, can sometimes
be detected with these techniques.[13]
[14] At the very least, areas that appear thickened on cortical thickness maps can be
examined much more carefully and attentively on the original MR image sequences. Quantitative
imaging can also define the total volumes of gray and white matter in each lobe, which
may also inform the trained observer that a subtle lobar abnormality exists that could
not be appreciated on direct inspection.[15] Qualitative evaluation is currently in clinical use, while quantitative automated
methods remain under investigation.[16]
[17]
Myelin Maps
An even newer technique for measuring myelin concentration using multicomponent-driven
equilibrium single-pulse observation of T1 and T2 (McDESPOT) may also play an important
role in evaluating the white matter of patients with epilepsy, especially pediatric
patients.[18] This technique compares the myelin concentration of the subject with an age-matched
control database and highlights differences. These differences are then presented
visually for examination and can be used to focus attention on the cortical area of
interest ([Fig. 5]).
Fig. 5 T2-weighted coronal images of a 6-year-old girl with severe intractable seizures
with a left supplementary motor area focus. All MRIs were reportedly normal, but a
small transmantle dysplasia was identified in the left superior frontal gyrus (a, arrow) and McDESPOT imaging demonstrated a focal area of decreased myelin concentration
in the same area (b, arrow) along with more diffuse myelin loss in the deeper white matter tracts. McDESPOT,
multicomponent-driven equilibrium single-pulse observation of T1 and T2; MRI, magnetic
resonance imaging.
Diffusion Tensor Imaging
Diffusion tensor imaging (DTI) is a relatively new technique that can better visualize
white matter. The directionality of the white matter fiber tracts is color coded based
upon the differential mobility of water molecules along versus perpendicular to the
fiber bundles, a term known as anisotropy. Where axonal bundles are intact and parallel,
anisotropy is high; where axonal tracts are spatially disorganized or intermingled,
anisotropy is low.[13]
[19]
[20]
[21] These color-coded anisotropy maps can be analyzed to see if the coefficient of anisotropy
is maintained or normal in the various lobes of the brain as an indirect measure of
dysfunctional connections. In cases of medial temporal epilepsy, it has been found
that associated limbic white matter tracts, such as the fornix and cingulum, can show
evidence of injury.[22] The use of this technique in epilepsy surgery remains investigational.
Diffusion tensor imaging has become a unique tool in surgical planning by identifying
eloquent tracts, such as the corticospinal tract for motor function or the arcuate
fasciculus for language.[23] For deep epileptic foci near eloquent pathways, tractography may be used to generate
a surgical plan that avoids injury.[24] In MTS, tractography is being used to evaluate Meyer's loop, which carries contralateral
superior visual quadrant information. Following either anterior temporal lobectomy
or selective amygdalohippocampectomy, greater than 40% of patients experience visual
field deficits due to disruption in Meyer's loop.[25] In one series of 12 patients, selective amygdalohippocampectomy was performed either
through the subtemporal approach or a transcortical approach, depending on which method
allowed for avoidance of Meyer's loop as visualized by DTI tractography. Postoperatively,
only 25% of patients experienced a visual field deficit.[26]
A future direction for DTI application is the development of functional connectome
maps for the elucidation of functional seizure tracts. In these methods, DTI maps
are used to develop a topographical map of seizure network nodes. Through computational
simulations of seizure spread, the most important nodes or foci for initiation and
propagation of seizures may be identified, and targeted resection may be performed.
This remains a purely research endeavor at the moment and not ready for clinical application.[27]
From a mathematical viewpoint, DTI is limited by the fact that each 2-mm voxel is
treated as a single diffusion compartment (white matter, gray matter, or CSF only),
whereas anatomically, such a voxel would actually comprise some combination of white
matter, gray matter, and CSF.[28] Two other analyses being investigated include Composite Hindered and Restricted
Model of Diffusion (CHARMED) and Neurite Orientation Dispersion and Density Imaging
(NODDI). In CHARMED, during reconstruction, each voxel is modeled with an intracellular
and an extracellular component. By removing the artifact from the extracellular component,
one anticipates higher resolution of axonal integrity and direction.[29]
NODDI models each voxel as a combination of three compartments: intracellular, extracellular,
and CSF. Measures that may be obtained through this analysis include the orientation
dispersion index (ODI), which reflects the degree to which neurites are dispersed.
It also allows for the calculation of the intracellular volume fraction, which is
a reflection of neurite density.[30] In one preliminary study, it has been shown that NODDI can be used to identify occult
focal cortical dysplasia.[31]
Perfusion Imaging
Although not in widespread use, a variety of MRI sequences have been developed with
application for imaging in epilepsy. Arterial spin labeling (ASL) is a contrast-free
perfusion measuring sequence and involves radiofrequency bursts in the neck to cause
inverted magnetization of arterial water. Following a delay to allow the inverted
water to perfuse the cortex, the desired regions are imaged. Control labeling is also
performed in which the magnetic inversion step is not performed. By comparing the
differences in magnetic inversion between the ASL and control sequences, one can obtain
an estimate of cerebral perfusion at various locations. Limitations of ASL stem from
a relatively low signal to noise ratio, which is susceptible to artifacts.[32] In one study of 16 patients with epilepsy, comparing ASL to PET, despite having
slightly decreased resolution, ASL was able to identify the same abnormalities in
perfusion as PET imaging.[33] In a more recent study of 164 patients with seizures, ASL imaging was able to detect
alterations in perfusion in 39% of patients, with highest sensitivity if the imaging
was performed within 5 hours of seizure onset.[34] Another perfusion-weighted sequence is dynamic susceptibility contrast (DSC) MR
perfusion. In this modality, alterations in T2 signal are induced by passage of paramagnetic
contrast agents through the cerebral microvasculature, which can then be used to determine
the perfusion of tissue with measures including cerebral blood volume (CBV) and cerebral
blood flow (CBF).[35] DSC imaging may be distorted or inaccurate in regions of calcification or hemorrhage
due to its variable effects on the contrast agent signal. Despite the differences
in techniques, both ASL and DSC have been used to lateralize temporal lobe epilepsy
without the need for radioactive agents.[36]
Functional MRI
Functional MRI (fMRI) is based on the principle that metabolically active regions
of brain have increased blood flow, known as activation flow coupling. The T2*-weighted
blood oxygenation level dependent (BOLD) signal is viewed as a reflection of neural
activity.[37] During an fMRI study, the patient is asked to perform or imagine a series of tasks
while actively being imaged. Extremity movement tasks allow for the detection of sensorimotor
regions of brain, while speech and comprehension tasks allow for identification of
regions responsible for language.
The utility of fMRI in epilepsy surgery relates to the identification of eloquent
regions of brain; notably motor and sensory function when lesions are adjacent to
these cortical areas. Pending the location of the lesion, surgical approaches as well
as extent of resection may be decided. Memory function may also be assessed via fMRI,
but this method remains in development and does not obviate the need for invasive
WADA testing when indicated. The more common use of fMRI is for lateralization and
localization of language ([Fig. 6]). In cases of medial temporal epilepsy, the side of language dominance determines
the extent of posterior resection. On the nondominant hemisphere, up to 6 cm of temporal
lobe may be resected, while on the dominant side, up to 4 cm may be safely resected.
Fig. 6 Functional MRI of a 26-year-old man with new onset seizures characterized by ictal
coughing and speech arrest. MRI demonstrated a small enhancing lesion in the left
subcentral gyrus (a, yellow). The patient underwent fMRI for tongue sensorimotor mapping (mauve) and
language mapping (green), and the images were rendered with superimposed cortical
venous structures (a, white). Surgery was performed awake with right tongue contraction and sensory responses
marked (mauve) and speech arrest (green) marked with tags on the cortical surface
(b). Lesionectomy while testing for speech was performed successfully without neurological
deficit. Pathology demonstrated ganglioglioma. fMRI, functional MRI.
Some investigation has been done into simultaneous scalp EEG/fMRI as a method for
coupling the temporal resolution of EEG with the spatial resolution of fMRI. The promise
of such imaging is the ability to follow the propagation of the seizure in real time
across the brain and thereby identify the entire epileptic network.[38] Although actively capturing an ictal event is quite difficult, it is hoped that
interictal spikes may also be used to mark subthreshold events.
Magnetoencephalography
As a technology, magnetoencephalography (MEG) is most similar in functionality to
EEG. The premise of MEG is that synchronous neural activity leads to the generation
of magnetic dipoles, which are recorded by a multidetector array ([Fig. 7]). In epilepsy, MEG has been investigated to isolate the onset and propagation of
interictal spikes as a proxy for epileptogenic activity; however, full clinical utility
has not been achieved.[39] Benefits of MEG are that magnetic dipoles do not experience artifacts from skull
and muscle artifacts and so allow for noninvasive recording simultaneously over the
entire brain. Limitations stem from the low signal to noise ratio that exists if not
enough concurrent dipoles are present. Also, given the expense of imaging, long-term
MEG is often not possible.
Fig. 7 A 19-year-old male patient had mainly nocturnal seizures beginning with tingling
in the left thigh followed by tonic extension of the left arm and flexion of the right
arm. Diurnal events were associated with frequent falls. Interictal scalp EEG was
unhelpful and ictal EEG appeared bilateral from the onset. MEG was performed and revealed
a single dipole source in the right frontal lobe (a, arrows). Reinterpretation of the MRI demonstrated a small area of high-intensity
signal in the right parasagittal cortex (b, arrows), and coronal MRI revealed a small transmantle dysplasia (c, arrows). Focal cortical resection of the dysplasia cured the patient of his seizures.
Pathology demonstrated a cortical dysplasia Type IIb. EEG, electroencephalogram; MRI,
magnetic resonance imaging.
Positron Emission Tomography
Fludeoxyglucose F18 positron emission tomography (18F-FDG PET) imaging is based on the principle that metabolic activity of cortex is
reflected in the rate of uptake of glucose. 18F-FDG is a glucose analogue with an attached positron emitting fluoride atom. Imaging
may be performed in either the ictal or interictal period. During ictal imaging, there
should be evidence of increased metabolic activity at the seizure onset zone, while
in interictal PET, there is often evidence of hypometabolism at the seizure onset
zone[40] ([Fig. 8]). In cases of temporal lobe epilepsy, FDG-PET imaging has been found to have greater
than 80% sensitivity,[41]
[42] while in cases of extratemporal epilepsy, sensitivity was less.[42] An issue with metabolic imaging is spatial resolution, as the region of abnormal
activity is oftentimes larger than the true seizure focus. As such, it remains a poor
modality for identifying an exact surgical plan of resection.
Fig. 8 Interictal PET images in the axial (a), coronal (b), and sagittal (c) planes in an adult with intractable temporal lobe seizures. Note the decreased metabolism
in the left temporal lobe affecting primarily the inferior and mesial temporal neocortex
(a and b [arrows] and c). PET, positron emission tomography.
Single Photon Emission Computed Tomography
Single photon emission computed tomography (SPECT) is a nuclear medicine imaging modality
that relies on the presence of a radioactive tracer to determine both quantitatively
and qualitatively the rates of regional cerebral perfusion. Interictal SPECT is less
sensitive in detecting alterations in perfusion, with 44% sensitivity, and as such
has fallen out of favor as a primary imaging modality in epilepsy. One method to increase
the sensitivity of SPECT is known as subtraction ictal and interictal SPECT coregistered
to MRI (SISCOM) and involves the challenge of obtaining SPECT imaging at the time
of seizure onset ([Fig. 9]). To properly perform an ictal SPECT, the tracer must be injected and imaging performed
within the first minute of seizure onset to most effectively identify the seizure
onset zone. Benefits of this method are that changes in perfusion are directly superimposed
on MRI imaging to provide more exact anatomic localization of perfusion abnormalities.
However, this is a very difficult technique to perform in any institution, and false
localizations can occur if there is a delay in injection. A meta-analysis of this
modality has not demonstrated significant clinical utility.[43]
Fig. 9 Ictal and interictal axial SPECT images in a patient with focal and generalized seizures
with a suspected frontal or temporal onset. Peri-ictal injection of the radioisotope
reveals increased uptake in the left inferior frontal region (a, arrow), confirming a suspicion of focal interictal hypometabolism in the same area
(b, arrow). SPECT, single photon emission computed tomography.
Intraoperative Imaging
In addition to utilizing multimodal imaging for surgical preplanning, the widespread
deployment of neuronavigation systems from multiple companies has allowed surgeons
to visualize and confirm abnormal lesions in the intraoperative setting. As software
sophistication increases, information from both anatomical and functional imaging
studies and tractography can be integrated to compare the real-time operative location
with alternative functional and anatomic analyses. Hence, following a good registration,
the surgeon can confidently maximize resection of lesional tissue while avoiding injury
to eloquent tracts. A technology that is also gaining widespread attention is intraoperative
MR imaging (iMRI). Using this technology, the patient may be imaged intraoperatively
while still under general anesthesia. The availability of this most updated information
allows the surgeon the freedom to perform further resection in a targeted fashion
at border zones, which may not be readily apparent by visual inspection. In one institutional
study of surgery for patients with medically refractory epilepsy, it was found that
12% of patients underwent second-look resection following iMRI verification of incomplete
resection.[44]
Conclusion
Successful epilepsy surgery relies on the precise localization and subsequent resection
of a well-defined epileptogenic focus. A high degree of concordance between multiple
data streams yields the highest rates of seizure freedom. By combining information
from multiple imaging modalities, the surgeon is more certain of the precise location
and extent of the resection target. Advanced imaging techniques are proving valuable
at all stages of treatment. At the presurgical stage, novel methods are allowing for
the visualization of lesions that previously were undetectable. In cases where standard
video EEG may yield equivocal localization, more sensitive imaging modalities allow
for the possibility of new targets for invasive recording. Functional imaging and
tractography allow for the elucidation of eloquent structures and their avoidance
during surgery. With the ability to coregister structural and functional imaging in
modern neuronavigation systems, the efficacy and safety of epilepsy surgery are improved.
