John Libbey Eurotext

Selection criteria and preoperative investigation of patients with focal epilepsy who lack a localized structural lesion Volume 2, numéro 4, Décembre 2000

Disorders causing INLE

Cerebral microdysgenesis

Histopathologic anomalies of neocortical architecture are frequently observed in patients with chronic epilepsy. Meencke and Janz [1, 2] noted minor histological changes ("microdysgenesis" ) in the frontal cortex of patients with primary generalized epilepsy, including increased ectopic neurons in white matter. While similar abnormalities appear in the molecular layer of normal brains and dyslexic patients [3, 4], microdysgenesis is regarded as being intrinsically epileptogenic. In a study that compared fifty patients undergoing superficial temporal resections with 33 control autopsy specimens, severe neuronal ectopia was identified in 42% of epilepsy patients but not in controls [5].

The histological features of microdysgenesis result from disordered neuronal proliferation and migration. Nissl-stained tissue sections reveal abnormal dendritic morphology including pyramidal and non-pyramidal neurons with vertically oriented basal dendrites, and abnormally shaped apical dendrites [6]. Scattered neurons in subcortical white matter are common. In the absence of major malformation, these features suggest some form of neurochemical alterations as a basis for the lowering of seizure threshold.

The benefit of surgery for seizures due to microdysgenesis is controversial. While microscopic disorders of cortical development often have a poor response to therapy, [7], neuronal ectopia and neuronal clustering are linked to more favorable clinical outcome [5]. This diversity may reflect differences in case selection or surgical protocols.

Meningitis and encephalitis

Bacterial and viral infections of the central nervous system increase the risk of seizures and typically begin within five years. A general retrospective cohort study identified an 11-fold increase in the risk for seizures [8] that was non-uniform for different infections. The risk was sixteen-fold for encephalitis, four-fold for bacterial meningitis, and two-fold for aseptic meningitis. Viral encephalitis was the most significant risk factor for partial seizures.

Central nervous system infection before age 4 years is also a risk factor for hippocampal sclerosis, but does not account for all cases of post-infectious epilepsy, even seizures of temporal lobe origin. The Herpes simplex virus, for example, classically attacks insular and temporal neocortex leading to atrophy without hippocampal sclerosis [9].

Chronic focal encephalitis (Rasmussen syndrome)

The syndrome of partial seizures due to chronic focal encephalitis (CFE) is a rare but well recognized cause of medically refractory epilepsy [10, 11]. Rasmussen syndrome is associated with relentless inflammatory unilateral hemispheric destruction. Seizures are often the presenting feature, most commonly simple or complex partial events. Status epilepticus is the initial presentation in 20% of patients. Seizure patterns often vary as the disorder progresses, but remain frequent, severe and drug-resistant.

CFE patients exhibit progressive unilateral ventricular dilatation and cortical atrophy although anomalous courses are not rare. A patient reported by Zupanc et al. [12] had repeatedly normal MR imaging studies for several months after seizure onset, and a 12 year old boy manifested a focal area of increased signal in the left frontal lobe one year after seizure onset [13]. It appears likely that an unknown proportion of patients exhibit uncontrolled partial epilepsy prior to neuroimaging abnormalities.

Functional imaging may be more sensitive in the early detection of Rasmussen syndrome. MR spectroscopy reveals decreased NAA signal intensity throughout the affected hemisphere [14]. Both PET and SPECT demonstrate regions of abnormal metabolism and blood flow according to the severity of the involvement [15].

Metabolic and degenerative disorders

Seizures are the rule in most neurodegenerative and metabolic disorders, and serial seizures or status epilepticus are common. While generalized seizures such as myoclonic or tonic attacks occur commonly, localization-related features are also frequent, especially in younger patients. Their clinical features may be indistinguishable from other causes of partial epilepsy and must be excluded by virtue of other clinical signs. It must be emphasized that neuroimaging studies tend to be normal in the early stages of a degenerative illness, furthering diagnostic certainty.

Idiopathic partial epilepsy

The idiopathic partial epilepsies are common childhood-onset partial seizure disorders, accounting for approximately 25% of epilepsy in pre-adolescent patients. The benign evolution is well known, with seizure-freedom and disappearance of epileptiform EEG features before the end of the second decade. The favorable prognoses mandate restraint with regard to antiepileptic drug therapy [16].

A small proportion of children with idiopathic benign partial epilepsy experience a fulminant course that is medically resistant. Panayiotopoulos [17] reported on a late-onset variant of childhood onset occipital paroxysms with predominantly simple partial diurnal seizures consisting of visual symptoms (hallucinations) followed by hemiclonic movements, automatisms and migraine headaches. The EEG was indistinguishable from the benign early-onset group in revealing asymmetric repetitive occipital or posterior temporal spikes or sharp and slow waves that attenuated with eye opening. Complete seizure control is achieved in approximately 60% of patients [18, 19]. Patients with this variant are more likely to have an abnormal perinatal history, neurological examination and EEG background [20].

Benign partial epilepsy with centrotemporal spikes is the most common, and best-delineated form of benign partial epilepsy [21]. Prognosis for remission is excellent [22]. DNA analysis in some families implicates linkage to chromosome 15q14 associated with potential dysregulation of the alpha 7 AchR subunit gene or a closely linked region [23].

An atypical non-benign evolution occurs in children who have a younger age of onset and multiple seizure types, including partial or generalized atonic drop attacks, and generalized continuous spike-wave complexes during sleep [24]. Status epilepticus is rare [25]. One family with typical centro-temporal epileptiform discharges was reported to have a dominantly inherited pattern of oral and speech dyspraxia [26].

The presence of brain lesions in patients with BECTS suggests that some patients suffer from a symptomatic disorder mimicking an idiopathic syndrome. Malformations of cortical development, anaplastic tumors and hippocampal sclerosis occur in rare patients [27-30]. Patients with atypical evolution of BECTS must therefore be evaluated for the presence of a structural lesion indicative of symptomatic partial epilepsy.

Neurophysiology investigation

Non-invasive studies

The EEG is critical to the presurgical evaluation of children with INLE. EEG data help classify the epilepsy and exclude non-surgical syndromes. As discussed above, EEG data help to identify the idiopathic partial epilepsies or an underlying progressive disturbance such as Alpers' syndrome. After surgical candidacy is established, the EEG further assists in the localization of the seizure focus and predicts surgical outcome.

Newer techniques such as magnetoencephalography, 3-D dipole source localization algorithms and computerized processing add additional information to the raw EEG [31-33] but their impact remains to be determined.

The scalp EEG helps define the overall extent of epileptogenic involvement. It is usually adequate to lateralize or regionalize the seizure focus, but due to complexities in the generation and propagation of epileptic discharges, detects only 10 to 50% of interictal spikes at the scalp. Ictal onsets characterized by focal low amplitude fast activity are easily overlooked.

The yield of scalp EEG is enhanced through additional electrodes (supraorbital, anterior temporal, etc.). Unlike adults, sphenoidal electrodes have little utility in children who have a lower incidence of hippocampal sclerosis and mesial temporal seizure foci. However, children referred for monitoring are more likely to have frequent seizures, and their length of stay is often less than one week. The child with infrequent seizures should have antiepileptic drugs (AEDs) withdrawn and placed in activating situations (i.e. exercise).

EEG interpretation must allow for rapidly changing patterns that indicate dynamic postnatal processes including myelination, synaptic connectivity, dendritic pruning and neuronal dropout. The immature neonatal cortex cannot support sustained or widespread hypersynchronized cortical discharges, even with diffuse pathological changes. As a consequence, the EEG may de-emphasize the extent of epileptogenic region in the very young child. This limitation is compounded by spatio-temporal dissociation of clinical and electrographic seizures.

In contrast, infants are more prone to bilaterally synchronous or generalized discharges. Generalized discharges in patients with partial epileptic disorders occur during a narrow time window and are therefore a developmental phase. With advancing chronological age, the EEG and clinical semiology of partial seizures is more easily classified into standardized patterns.

The task of defining consistent focality is particularly challenging when focal epileptiform activity propagates rapidly. Seizures in children are generally easier to lateralize, but more difficult to localize compared to adults. Not uncommonly, focal epileptiform patterns are identified interictally prior to the phase of generalization [34]. Focality may be determined by alterations of EEG background including polymorphic slowing or attenuated fast frequencies. In some cases, focal intermittent fast activity is the only clue to primarily localized epileptogenic dysfunction [35].

Intracranial EEG monitoring (IEM)

With advances in non-invasive technology, fewer patients require chronic intracranial monitoring. Its contribution to the presurgical evaluation has been challenged, particularly in adults with lesional epilepsy in the temporal lobe. As surgical candidacy shifts towards younger age groups however, invasive recording has regained its utility, especially for children with normal imaging studies and subtle malformations of cortical development.

The main goal of IEM in the INLE patient is to generate a "surgical diagram" that accurately depicts the location and extent of the epileptogenic region (ER), its relationship to eloquent cortex and the planned resection. This approach is mandatory when the location and extent of the ER cannot be adequately determined through non-invasive means. However, IEM is costly and associated with increased morbidity. It should therefore be utilized on an individual basis. IEM is not an "exploratory procedure" if the non-invasive evaluation provides no information about side or approximate location of the ER. Due to limitations of sampling and interpretation, IEM may not always successfully define a discrete ictal onset or alter the ultimate surgical strategy and outcome [36].

Thus, although removal of the entire region of significant electroencephalographic abnormalities is generally required for seizure-freedom in non-lesional epilepsy [37, 38], not all patients do require invasive studies. One- stage resections based on interictal PET data in conjunction with intraoperative ECoG have been performed on a limited number of patients with a single hypometabolic region at one center. It is probably too early to know whether this approach will ultimately prove successful [39]. One-stage procedures are indicated in children with partial seizures arising from the anterior temporal lobe when scalp EEG and PET/SPECT data are congruent [40].

IEM may be necessary in otherwise straightforward cases of temporal lobe epilepsy if non-invasive studies suggest seizure origin in the posterior temporo-occipital base/convexity or if the ER is encroaching upon language cortex in the dominant hemisphere [41]. IEM is also indicated if tailored procedures such as selective amygdalo-hippocampectomy or lateral neocorticectomy rather than lobectomy are contemplated. In children with non-lesional extra-temporal epilepsy where the role of functional imaging is still uncertain [40, 42-44], invasive recording may be the only means to determine the true extent of the ER.

The electrographic patterns that define the extent of optimal resection are not universally agreed upon. Regions of active spiking on the pre-excision ECoG are used to define the epileptogenic zone and resection plane in patients undergoing one-stage excisional procedures [45]. Prominent interictal spiking is considered significant if it shows consistent focality or rhythmic features, occurs in trains of focal "fast" activity, or is associated with focal attenuation of background. Focal burst suppression is always regarded as significant. In contrast, infrequent spikes, spikes without consistent focality or "rim" spikes recorded on the post-resection ECoG are ignored.

In subdurally implanted patients, the ictal onset zone is the single most critical factor in defining the ER [36, 46, 47]. Secondary foci that consistently activate intra-ictally [48] (during a seizure) are included in the resection if they appear in proximity to the primary ictal focus. Similar to the ECoG data, regions of prominent interictal spiking and background abnormalities are also considered significant.

Talairach and colleagues [49] pioneered depth recording of the EEG to localize seizure origin in surgical candidates. The use of depth electrodes is particularly valuable in temporal lobe epilepsy [50], but may also contribute useful information about deeply seated foci in other lobes as well [51, 52]. When depth electrodes are placed accurately and targeted strategically, delineation of seizure onset is well defined and improves seizure outcome. The combined use of subdural and depth electrodes in the preoperative evaluation of epilepsy has been reported, and MR imaging of both depth and subdural electrodes is a safe procedure [53].

Neuroimaging

MRI

By definition, patients with "non-lesional" epilepsy have a normal routine CT/MRI exam. However, more detailed analysis using novel MRI sequencing techniques can often identify subtle abnormalities of gyral architecture or the gray-white matter interface. Additional surface phased-array coils improve signal-to-noise ratio and spatial resolution in superficial cortex and the hippocampus [54]. In a study employing three-dimensional reconstruction, gyral abnormalities were detected in 6/16 patients with extratemporal seizures and unremarkable routine MRI exams [55]. High field strength magnets enhance the yield in INLE [56], and volumetric acquired sequences on T1 weighted MR imaging permit co-registration with abnormalities on functional imaging studies.

Over-diagnosis may result from inadequate normative data in the first few years of life. For example, incomplete myelination decreases gray-white matter contrast on T2 weighted images and compromises the identification of cortical abnormalities. Special MRI sequences are effort-intensive and impractical. We generally utilize special sequences only in regions of interest identified on other non-invasive studies.

MRS

Magnetic resonance spectroscopy (MRS) assesses cerebral metabolites and neurotransmitters in epileptogenesis. Anatomical specificity is lacking as MRS is based on the chemical properties of protons in magnetic fields. Concentrations of N-acetyl aspartate (NAA), creatinine (Cr), and choline are estimated through proton H MRS. Epileptogenic foci show a decreased ratio of NAA/Cr and choline due to neuronal loss and gliosis [57]. High-energy phosphate compounds, inorganic phosphate, and pH are assessed using phosphorus MRS; in general, epileptic foci are associated with increased pH and inorganic phosphate, and decreased phosphate monoesters [58, 59].

Positron emission tomography (PET)

PET scans define focal metabolic abnormalities at resolutions as low as 5 mm [60], with the type of abnormality varying according to the ligand used. Multiple metabolic tracers are available including fluoro 2-deoxyglucose (FDG), tracers specific for cerebral blood flow such as oxygen and dihydrogen oxide (O H2O), and receptor density ligands such as in opiates or flumazenil. In children, INLE, subtle malformations of cortical development are more sensitive to flumazenil PET than FDG PET.

PET scanning is usually performed in the interictal state. Ictal studies are more difficult due to short tracer half-life and the need for steady state mathematical modeling. Postictal metabolic changes may be evident up to 48 hours. Scalp EEG is obtained during and after the tracer injection to look for interictal discharging and subclinical seizures. While scalp electrodes do not affect PET scanning, intracranial electrodes can produce erroneous regional hypometabolism on FDG PET.

The focal areas of hypometabolism commonly seen interictally are believed to represent reduced blood flow and inhibition or deafferentation of neurons surrounding the epileptogenic focus. The degree of hypometabolism usually varies across regions with more severe hypometabolism in the ictal onset zone (figure 1a) [61]. The more profound the hypometabolism, the better the prognosis, especially in temporal lobe epilepsy. In children with cortical dysplasia, areas of very active spiking may reveal hypermetabolic interictal states.

PET data can identify focal dysfunction in INLE due to infantile spasms and anterior temporal foci [39]. Hypometabolism is most often observed in temporo-parieto- occipital regions; partial seizures and focal EEG findings in these regions may antedate the onset of spasms. More recent studies recognize the limitations of PET imaging. Hypometabolic regions may be non-specific or reflect secondary sites that revert to normal once the primary seizure focus is removed. Furthermore, the size and location of observed focalities may vary during interictal, ictal and postictal states. These considerations become especially critical with frequent seizures or near-continuous epileptiform EEG activity. Thus, PET data could easily mischaracterize the extent of the ER, resulting in failure or unnecessarily large resections. These concerns suggest caution in relying too heavily on PET data for surgical planning in INLE. Newer PET ligands such as flumazenil or serotonin may, pending validation, be more specific than FDG [62].

Single photon emission tomography (SPECT)

The major strength of SPECT lies in its ability to document peri-ictally increased blood flow. SPECT images are formed from injected radiotracer photons in the cortex. Ligands include Tc hexamethylpropleneamine oxime (HMPAO) and Tc ethylcysteinate dimer (ECD) that are fixed in brain for up to 6 hours. SPECT also measures neurotransmitter systems including benzodiazepines and muscarine cholinergic receptors [63].

Variable SPECT results in the same patient are unrelated to interictal spike or seizure frequency [64]. Similar to PET, interictal SPECT demonstrates regional hypoperfusion over a region of cortex larger then regions of MRI or EEG abnormality [65]. Cortical dysplasia may reveal increased perfusion in the interictal state reflecting the heightened metabolic demand of almost continuous discharges.

Ictal SPECT scans demonstrate regional hyperperfusion (figure 1b), higher sensitivity and greater specificity than interictal SPECT [66]. Improved yield is possible with seizures of 90 seconds or more; injection must occur within 30 seconds for best results. The yield is lower for simple partial seizures or secondarily generalized seizures whereas simple partial seizures rarely lateralize. The origin of secondary generalized seizures is often the most hyperperfused region [67].

Ictal SPECT is particularly useful for characterizing syndromes such as hypothalamic hamartomas or tuberous sclerosis complex where positive findings may obviate the need for invasive monitoring. There are limitations however, as hyperperfusion often extends well beyond the ictal onset zone to regions involved in propagation. While secondary propagation sites usually demonstrate hyperperfused "blushes", the primary epileptogenic focus often has a configuration that does not conform morphologically to an anatomic gyrus.

Subtraction imaging of ictal and interictal SPECT may be used if the ictal images are inconclusive. However, ictal hyperperfusion may persist for variable periods of time, and compromise the interictal study if obtained too soon after the ictal scan. There is relatively poor spatial resolution with the subtraction SPECT scans but co-registration to volumetric MRI scan using a surface matching technique improves anatomical localization [68].

Comparison between modalities

There is no definitive comparison of the relative sensitivities of MRI, peri-ictal SPECT and interictal PET [64]. MRI assessment is solely visual and subjective as it must account for the normal variation of gyral morphology. PET and ictal SPECT assist the localizing process in INLE, or lesional cases with discordant ictal EEG and MRI data. Interictal SPECT has a high rate of hypoperfusion contralateral to the ictal onset zone (10%), and poor correlation with the epileptogenic zone [64]. PET and ictal SPECT show low sensitivity and high specificity for extratemporal foci and higher sensitivity and moderate specificity for temporal lobe epilepsy [64].

Ictal SPECT is less sensitive than PET for nonlocalized extratemporal lobe seizures. PET is more sensitive than ictal SPECT for temporal lobe epilepsy as judged by subdural EEG ictal localization, and correlation with surgical outcome [64]. PET's main problem is its cost and availability. Although SPECT is a less sensitive technique for determining the epileptogenic zone compared to PET, it is the only modality that has the capacity to image during a seizure and therefore, complements PET data [64].

Received September 15, 2000 / Accepted November 10, 2000