Semiology

The clinical semiology of a seizure provides the first clue to localizing its onset. Penfield described an assortment of mental, sensory, and physical aspects of seizures specific to various brain regions.⁷ These range from mental phenomena such as déjà vu and ill-defined epigastric rising to olfactory and gustatory auras to motor convulsions. The first manifestation of a given seizure most accurately reflects the area of onset, whereas seizure spread can lead to later involvement of other areas outside to true site of seizure onset.

Seizures of temporal lobe origin are the most likely to have auras. Those originating in the mesial temporal structures classically begin with an ill-defined epigastric sensation accompanied with panic or autonomic disturbance. Neocortical temporal seizures commonly have auditory, visual, or perceptual hallucinations. Head turning, posturing, automatisms, and behavioral arrest are common accompaniments of temporal seizures.

Frontal lobe seizures are often brief, occur in clusters, and may include complex motor posturing, versive eye movements, or vocalization. They can rapidly spread yielding early generalization, drop attacks, or early signs of temporal lobe activation.

Parietal lobe seizures often present with somatosensory symptoms but may have abdominal symptoms such as nausea or a gustatory sensation. Occipital seizures may be heralded by visual phenomena including scotoma.⁸

Neuropsychology

The classic neuropsychology workup of an epilepsy patient consists of a variety of standardized tests and questionnaires that establish a profile of the patient’s cognitive, emotional, and behavioral abilities. This testing quantifies the patient’s abilities, deficits, and coping strategies. Such information can localize areas of dysfunctional cortex, which may be the site of seizure onset. Neuropsychology also predicts the deficits likely to be incurred by resection of a given focus and the impact that such deficits will have on the individual’s life.⁹

In many emerging technologies such as functional magnetic resonance imaging (fMRI), the neuropsychologist plays a vital role in administering verbal and functional tasks and assessing the patient’s cooperation with such tests. As these tests are very dependent on cooperation from the child while in the MRI machine, the neuropsychologist plays a vital role in extracting and interpreting information from an otherwise uncooperative patient.^9,10

Electroencephalography

Electrophysiologic data are acquired by placing electrodes on the scalp and recording electrical differences between them. Interictal EEG can reveal epileptiform discharges such as spike and sharp waves, providing clues to the lateralization and localization of seizure onset, but prolonged video EEG provides an electrophysiologic picture of the seizures themselves. Video EEG data can confirm that spells are seizures (as opposed to breath-holding or other nonepileptic spells) and in many instances begin to localize the onset of the seizure. Most centers want to record several typical seizures and will continue monitoring until this is accomplished.¹¹

Ictal video EEG will precisely localize the seizure focus in about a third of patients with temporal lobe epilepsy.¹² Precise localization is even less common in those with extratemporal epilepsy where seizures tend to spread rapidly. Rapid spread from the orbitofrontal or posterior parietal cortex can falsely localize to the temporal lobe.¹³

Dense-array EEG is a newer technology that holds promise for more precise seizure localization. The dense array consists of 256 electrodes held to the head by an elastic mesh, giving substantially more information and spatial resolution than the standard 32-electrode array.¹⁴

Imaging

Surgical epilepsy cases are sometimes divided into lesional and nonlesional. Classically lesional cases from well-circumscribed pathologies such as cavernoma, dysembryoplastic neuroectodermal tumors (DNETs), and ganglioglioma have a significantly better prognosis than nonlesional cases. The identification of a structural abnormality on MRI is a strong predictor of localization.¹⁵ When the scalp EEG confirms seizure onset from the lesion, localization is strongly suggested, and resection often proceeds without invasive mapping.

The distinction between lesional and nonlesional epilepsies has blurred a bit as improved imaging has revealed cortical dysplasias, focal sclerosis, or other pathologies previously noted only by histology. Higher-field MRI, diffusion tensor imaging, and other MRI methods (e.g., MR spectroscopy and fMRI) may also show abnormalities. Voxel-based MRI postprocessing has been recently shown to help visualize blurred gray–white matter junctions or abnormal extension of cortical bands otherwise not recognizable on MRI.¹⁶ A negative MRI may still allow for further surgical planning. In a recent series of pediatric temporal lobectomies, half of those eventually shown to have histologically confirmed mesial temporal sclerosis had normal hippocampus on MRI.¹⁷ In such cases, other imaging such as positron emission tomography (PET), single-photon emission computed tomography (SPECT), and magnetoencephalography (MEG) can help to further define the extent of the epileptic-onset zone and the functionality within and surrounding the malformation.

PET provides a tomographic image of brain glucose utilization, while SPECT images blood flow. Both can reveal areas of metabolic derangement (hypometabolism interictally and hypermetabolism during a seizure) that can represent a seizure focus.^18,19 MEG collects minute magnetic potentials generated by bands of synchronized neural currents, a technique that can reveal epileptic foci as well as functional circuits (e.g., motor units).²⁰ The specific role for these various modalities is not determined and considerable variability exists across programs.²¹

Electrocorticography

A significant amount of information about the epileptogenicity and function of an area of tissue can be obtained from intraoperative electrical monitoring. ECoG data are obtained by placing electrodes on the exposed brain and observing the EEG record for epileptiform discharges (spike and sharp waves).^23,24 Anesthetic considerations are important to avoid suppressing or altering the EEG record. Benzodiazepenes should be avoided except at induction or after ECoG. Inhalational agents above 0.5 MAC will dampen the record beyond interpretation. A combination of sevoflourane and dexmetomidate is preferred at our (BO and JGO) institution, but a total intravenous anesthetic is also possible.²⁵

Some functional information can be obtained in the anesthetized patient. A distinct pattern of cortical activity (phase reversal) of evoked-somatosensory responses can localize the central sulcus. Direct stimulation of motor cortex can elicit motor responses that are visualized or recorded on EMG. Awake craniotomy in cooperative older children (sometimes as young as 10) can be used to map language cortex.²⁵ Current is passed through a hand-held bipolar stimulator (Ojemann stimulator) to inhibit an area of cortex while the patient is performing speech tasks. Speech arrest induced during this procedure reflects an area of vital speech function.²⁴

ECoG guidance adds little risk to an epilepsy resection, and can avoid the second operation and likely some morbidity of implanted electrode monitoring. The information gained, however, is limited to the anesthetized, interictal state. Time constraints also limit electrophysiologic data.

Implanted Electrodes (Grids, Strips, Depths)

Implanted electrodes allow for EEG monitoring and stimulation mapping outside of the operating room. Recordings include the awake, drowsy, and sleep states and ideally will include several of the patient’s seizures. Monitoring typically continues for 5 to 7 days but can be extended if additional information is needed. Very brief intracranial monitoring (24 or 48 hours) can be a useful technique in toddlers who would not tolerate prolonged implantation.

Intracranial monitoring electrodes are typically placed in the subdural space but epidural placement is employed at times, and yields comparable information.^26,27 Placement below the skull dramatically increases the amplitude of cortical electrical signals collected and diminishes muscle artifact, the main source of noise. Implanted electrodes also greatly increase the spatial resolution of the EEG. A typical grid consists of 64 electrodes on an 8×8 cm array, whereas scalp EEG typically employs 32 electrodes to cover the entire head.

Effective definition of an epileptic focus relies on placement of electrodes as close to the region of interest as possible while electrodes covering distant sites, including the opposite hemisphere can exclude multifocal onset of seizures.

Implanting electrodes additionally allows for prolonged functional mapping outside the operating room by stimulating through the same electrodes.² By applying current across two neighboring electrodes, the intervening area of cortex (about a centimeter of tissue) is inhibited. Deficits observed can identify the function of the cortex and the deficits likely to be incurred by resection. Lower stimulation of motor cortex may produce muscle contraction rather than inhibition. Many children who cannot tolerate with awake intraoperative mapping will cooperate with extraoperative mapping.³⁰ Additionally, the mapping can take place during complex activities such as drawing, writing, or playing a musical instrument potentially identifying important integrative functions.

Cortical stimulation for mapping can induce seizure activity, causing inhibition beyond the intended stimulation both temporally and spatially. The EEG must be monitored for after-discharges reflective of stimulation induced seizure.

Grid electrodes consist of a broad array of contacts that can cover a large area of cortex. They can be particularly helpful in defining an epileptic zone within a lobe or defining the relationship of a lesion to seizure onset. Convexity sites are particularly amenable to grid placement.³¹ Extraoperative language and motor mapping are typically performed through grids as they give ample coverage of the hand and face motor area as well as the typical language sites.

Strip electrodes consist of a single or double row of electrodes spaced along a flexible strip that can be safely passed along the subdural space. Coverage of the interhemispheric cortex, orbitofrontal cortex, and mesial temporal lobe can be achieved by passing strip electrodes, all locations poorly monitored by scalp recordings.^28,32 A strip passed along the undersurface of the temporal lobe will typically place the distal electrode just at the parahippocampal gyrus, providing good monitoring of the mesial temporal structures.³³ Placement of strip electrodes requires only a burr hole, allowing for limited monitoring of distant sites, including the contralateral hemisphere without requiring a craniotomy.

Distant and mesial temporal recordings can also be obtained with depth electrodes. These electrodes line a narrow probe that passes directly into the region of interest. Some centers also place depth electrodes directly within or along the deep margins of a lesion infiltrating the white matter (e.g., cortical dysplasia) to evaluate the depth of resection necessary.^28,34 These various electrodes are often used in combination to monitor all areas of interest.

Complications

Electrode implantation procedures require two separate operations (occasionally more) with a prolonged hospitalization between. The most common complications are cerebrospinal fluid (CSF) leakage, fever, and infection. Minor CSF leak is common while electrodes are in place, but rarely problematic. While low-grade fevers are common and wound infection much less so, suspicion should remain high in order to recognize and treat infections promptly.³⁵

Cerebral edema combined with the mass effect of the implants can cause elevated intracranial pressure. This occurs more commonly with more electrodes implanted. Mannitol and steroids typically will control intracranial pressure, but on rare occasions, grids may need to be removed.

Strip electrodes have fewer complications than grids, 1% overall compared to 3% to 4%. Strips also appear to be safer than depths. Isolated cases of permanent neurologic deficit and death have been reported.^35–39 The risk of complications, as well as the additional hospitalization time, need to be balanced against the information gained by invasive monitoring.

Indication

These procedures are primarily indicated in patients with pre-existing unihemispherical damage and typical neurologic deficits such as hemianopia and hemiparesis combined with drug resistant epilepsy. If these cases occur in early infancy, are combined with holohemispheric dysplasias or extensive damage and severe drug resistant seizures they are frequently called “catastrophic epilepsy.” Other typical diagnoses seen in these patients include hemimegalencephaly, multilobar cortical dysplasia, and various disorders of gyration such as polymicrogyria or lisencephaly. Hemimegalencephaly (HME) is a quite rare malformation of the cortical development arising from an abnormal proliferation of anomalous neuronal and glial cells that leads to the hypertrophy of the whole affected cerebral hemisphere. Epilepsy typically presents in infancy with severe, drug-resistant seizures. Many infants with HME undergo hemispherotomy before 2 years of age. HME patients have more operative complications and worse seizure outcomes than other hemispherotomy patients.^45,46

Frequently hemispheric damage is due to perinatally acquired brain defects or intrauterine hemorrhagic damage. Perinatal stroke is in many ways an ideal pathology for hemispherotomy as patients typically have little function stored in the affected hemisphere and rarely incur new deficits from the operation. Development and cognition often improve when seizures are controlled and epilepsy medicines are decreased or discontinued. Many of these children have cystic encephalomalacia in continuity with the ventricular system or separated by a thin membrane. This expanded space gives additional surgical access and easier visualization of the deep anatomy that needs to be disconnected, but this anatomy can be distorted. Care must be taken to maintain orientation.

Hemispherotomy may also be considered in diseases such as Sturge-Weber syndrome, a sporadically occurring phakomatosis consisting of unilateral facial port-wine stain in the V1 distribution, glaucoma, and unilateral leptomeningeal angioma. The intracranial angioma causes progressive cortical atrophy and calcification. Epilepsy occurs in 75% to 90% of Sturge-Weber patients, usually presenting in infancy. In some patients, the angioma is focal enough to be resected, but most are candidates for hemispherotomy.^47,48

Other pathologies potentially amenable to hemispherotomy include Rasmussen’s encephalitis, and postencephalitic and post-traumatic hemispheric damage. Rasmussen’s encephalitis or chronic focal encephalitis is a chronic T-lymphocyte inflammation that remains localized to one hemisphere. Presentation is usually between 5 and 10 years of age with acute, drastic onset of focal seizures and progressive loss of hemispheric function. Anticonvulsants typically have little success in controlling seizures. Medical treatments to control inflammation, including steroids, intravenous immunoglobulin (IVIG), tacrolimus, and plasmapheresis are the first line of therapy, although their efficacy is limited.^49,50

Timing of hemispherotomy has been debated in Rasmussen’s patients particularly. Permanent loss of hemispheric function eventually occurs in a large majority from the disease, prompting recent recommendations for early hemispherotomy. Some evidence supports improved cognitive outcomes with early surgery,^46,51 but this must be balanced against the acute loss of hemispheric function from surgery performed before hemispheric dysfunction is fully actualized.

As hemispherotomy disconnects the motor and visual cortices, hemiparesis and visual field defects are inevitable consequences of the operation. The procedure is more appealing if these deficits are already present or if they are an inevitable outcome of the underlying pathology. Postoperative hemiparesis affects the upper extremity more than the lower. Ability to walk is almost always maintained.^52,53 Improved development and cognition commonly follow hemispherotomy in cases where seizures stop and anticonvulsants can be weaned.^52,54–56

Diagnostics

The most important presurgical diagnostics include MRI and EEG recording of ictal events. MRI may show larger defects, frequently in the distribution of the middle cerebral artery, and malformations of cortical development. MRI findings include gross malformations of the hemisphere, ectopic gray matter, disorders of gyration, extensive postencephalitic damage, or holohemispheric atrophy. In the ideal case all ictal activity can be localized to the affected hemisphere. However, hemispherectomy may be indicated even in the setting of contralateral EEG or even MRI abnormalities. MRI abnormalities on the opposite hemisphere are particularly concerning for predicting poor success.^57,58

If the disease has started after the fourth year it is important to demonstrate transfer of language to the other hemisphere for which the intracarotid amobarbital test, or possibly functional MRI,¹⁰ is useful. However, the association between speech outcome and hemispherectomy is multifactoral with congenital etiologies having a more favorable outcome, even with age considered.⁵⁹

Choice of Approach

It has become clear that the disconnective procedures carry the same success rate, may be even higher success rates as those procedures that involve larger resective steps.^60,61 In the 10 to 15 years of available follow-up for these disconnective procedures the complications known from anatomic hemispherectomy have not been reported thus far, and in particular no cases of cerebral hemosiderosis, which was reported for earlier procedures. The availability of CT and MRI-scanning has also minimized the potential problem of shunt malfunctions. However, postoperative hydrocephalus is also much rarer in disconnective procedures such as trans-sylvian keyhole deafferentation or trans-sylvian keyhole disconnection, or the alternatives described by others.^40,43,44 Other advantages of the trans-sylvian keyhole disconnection procedure favored by us have been confirmed by other series.^43,60,