For patients with intractable epilepsy, the role of intracranial surgery has been progressively recognized as a well-tolerated and clearly beneficial treatment option.^1–3 Improvements in quality of life are now frequently documented alongside the success rates with surgical approaches, and early surgical consideration for medical failures has gained acceptance as a routine part of the comprehensive approach to refractory seizure management. Although some patients can proceed to an operative intervention based on clinical presentation and noninvasive tests,^3,4 others require more investigation before surgical resection can be considered. This latter group may have insufficient scalp electroencephalography (EEG) documentation of their seizure disorder or an unknown correlation between seizures with a suspected lesion or substrate; they also may lack localization of an ictal onset that is amenable to surgical resection. In other cases, surgical resection for seizure control may pose significant risk if adjacent regions harbor the eloquent cortex. When language, motor, or sensory function may be involved, intracranial electrodes can be used for functional mapping.

In these more challenging cases, subdural strips and grid electrodes have become more important and increasingly useful.^5,6 Intracranial EEG and stimulation can provide the final necessary information that permits consideration of a surgical approach. This chapter will discuss factors influencing the strategies chosen for subdural electrode use when evaluating a patient with partial epilepsy.

Noninvasive Monitoring: Limits and Complementary Roles

EEG provides a noninvasive physiological examination of the brain that allows characterization of neuronal discharge patterns. It provides a graphic representation of frequency and amplitude, and also demonstrates the temporal relationship and spatial distribution of electrical activity from the cerebral cortex.⁷ EEG identifies both normal and pathologic states, with the latter role being essential in the diagnosis and treatment of seizure disorders.^8,9 However, the role of a standard EEG is limited in that the typical events of interest (seizures) do not reliably occur while the test is performed.^10,11 To address this dilemma, long-term, continuous, noninvasive monitoring can be performed in an epilepsy monitoring unit (EMU). The EMU also provides a safe environment where supervision in an inpatient environment minimizes the risk when medications are tapered to help capture a seizure.

Despite the ability of an EMU to greatly improve the likelihood of capturing an event, there are still significant limitations with scalp EEG.¹² The use of scalp contacts requires that the electrical activity of the brain must be detectable through several layers of tissue, each of which causes various impediments to the process of recording an EEG. When using superficial electrodes, the scalp reduces the amplitude of EEG signals. Even trying to abrade the scalp to improve the conductivity does not provide a significant improvement in the EEG.¹³ Studies of bone indicate that it is a poor conductor of electrical energy.^14,15 The skull, dura, and cerebrospinal fluid (CSF) have been demonstrated to attenuate the signal strength of the brain’s activity, affecting amplitude,¹⁶ and with attenuation, the ability to discern true signal from ambient noise is reduced. Without the effects of the skull and scalp, subdural electrode amplitudes are more than an order of magnitude greater than signals reaching the scalp. Also, although long-term scalp electrodes require attention and the application of electrolyte, subdural electrodes have reliably low impedances that are maintained during long-term monitoring.

Artifacts, or extraneous noise relative to true cortical electrical activity, are another limitation of scalp EEG, even in an EMU. So-called internal artifacts are related to concomitant patient physiology, such as electrocardiogram activity, eye movement, swallowing, and other voluntary or involuntary muscle movement. External artifact may arise from any source of electrical interference, including 50 or 60 cycle noise from standard alternating current or the use of adjacent electrical equipment. This latter effect can be more prevalent in the inpatient setting of the EMU, given the amount of technology in the hospital environment.¹⁷ Additional artifact results from irregularities in the skull, such as after craniotomy, and produces a “breach rhythm.”¹⁸ Subdural electrodes provide a stronger signal for recording (better signal-to-noise ratio) and are not subject to most of the other artifacts noted above.

Upon passing through the calvarium and scalp, electrical currents experience some diffusion, or “smearing,” and can appear over a greater distribution than the focal nature of their onset.^19,20 This limits the accurate localization of an epileptic focus. The frequency of scalp EEG recordings is also distinct from what is obtained from subdural electrodes. Scalp EEG demonstrates only a portion of the high frequencies seen on subdural recordings. However, this is not felt to be due to the skull and scalp as much as the result of summation of polyphasic cortical activity.²¹

Another aspect that limits scalp EEG is the irregular geometric relationships between the scalp (determined by the skull) and much of the brain’s cortical surfaces. Although the scalp approximates the surface of a sphere, the underlying cortex has a much more complex shape with infolded gyri and deep clefts at the sylvian and interhemispheric fissures. Because scalp EEG is most sensitive to radial (perpendicular) oriented discharges or dipoles,²² electrical activity coming from cortical discharges not in this orientation may be poorly discerned. Another problem related to spatial relationships is the longer distance between the basal regions of the brain and external electrodes. This distance attenuates the electrical signal such that it is poorly detected at the scalp.²³ With regard to all issues mentioned, placement of subdural electrodes overcomes many of these obstacles.

The greater sensitivity when EEG contacts are on the brain versus the scalp is well illustrated by the work of Tao and colleagues.²⁴ They recorded simultaneous scalp EEG while monitoring intracranial EEG during evaluation for temporal lobe epilepsy surgery. The group compared cerebral discharges with and without scalp EEG activity and estimated the volume of brain involved by the number of subdural contacts having concurrent depolarization. Monitoring interictal spikes, they found that the source area had to reach a critical size before reliably producing recordable scalp potentials: 90% of cortical spikes with a source area >10 cm² produced scalp EEG spikes, whereas only 10% of cortical spikes having <10 cm² of source area produced scalp potentials. They also found a minimal threshold where source areas <6 cm produced no scalp potentials. This last finding is consistent with the much earlier work of Cooper et al.²⁵

The limitations of sensitivity and spatial resolution of scalp EEG diminish its utility in trying to localize the focus of partial events when considering surgical options. In distinction, subdural electrodes can provide the precise localization necessary for both complete resection and preservation of the functional cortex because each electrode is quite specific for the underlying cortex.

Finally, noninvasive EEG may not demonstrate any pathology in certain abnormal states, particularly partial seizures. Seizure types that have been shown to occur in the absence of EEG abnormalities include seizures of extratemporal origin (particularly frontal lobe), simple partial seizures, and complex partial seizures that do not generalize.^26,27 Spencer et al. estimated that noninvasive EEG is unable to sufficiently identify the epileptogenic zone in ~25% of patients with medically uncontrolled epilepsy.²⁸

This first case study demonstrates how subdural electrodes can obtain a much cleaner EEG relative to scalp recordings. The study also demonstrates two limitations of subdural recordings. Even with invasive monitoring, precise localization of a seizure onset may still be difficult (the intracranial EEG in Videos 32-1a and 32-1b revealed only regional suppression). Furthermore, subdural electrode data may be limited if the onset appears at the edge of the grid (edge effect). Moving or adding grids to mitigate these issues is discussed in the section Electrode Selection and Placement.

Noninvasive EEG, while providing the majority of information for the diagnosis and characterization of epilepsy, may prove insufficient for more complex cases. Invasive EEG can overcome many of the technical limitations of scalp recordings, and when considering surgical resection in complex cases, intracranial EEG becomes a significant adjunct if not mandatory part of the evaluation.

Imaging (Anatomical and Functional): Limits and Complementary Roles

Three-dimensional neuroimaging techniques have advanced greatly from initial anatomical detail provided by computed tomography (CT) and MRI.²⁹ The detailed examination of intracranial anatomy is now part of the routine evaluation for patients with epilepsy,^30,31 and the appearance on MRI of typical lesions associated with epilepsy have been well defined.³² Despite the strides in neuroimaging, many patients with epilepsy will still have negative imaging studies.³³

Imaging of the normal and abnormal physiological epiphenomena in the brain falls within the category of functional imaging. These tests include positron emission tomography (PET), magnetoencephalography (MEG), and SPECT.³⁴ Ictal SPECT has been noted to be helpful as a guide to optimize placement of intracranial electrodes. In successful cases of epilepsy surgery, there appears a high rate (70%) of colocalization between the resected area and the area of maximal increased perfusion identified by ictal SPECT.³⁵ More recently, differential SPECT imaging of ictal activity has been combined with anatomical detail to help localize seizure onset. SISCOM has been shown to have increased utility over SPECT for imaging ictal onset.³⁶ Similarly, MEG data can give localizing information about interictal activity.³⁷ However, while providing additional information, SISCOM or MEG findings alone are not usually sufficient to determine either resectability or the extent of resection, particularly in extratemporal cases. Although noninvasive imaging data may not be conclusive, using functional imaging superimposed on MR anatomy can be helpful to plan subsequent subdural monitoring³⁸ (Figure 32-2). The results of seizure control are much improved when resection is guided by SISCOM with subsequent intracranial monitoring. The results of postsurgical seizure control are 60% excellent when the SISCOM abnormality is resected versus 20% excellent when resection does not include the SISCOM focus.³⁶ PET imaging typically identifies regions of hyper- or hypometabolism. However, the practical use and technical needs of PET limit capturing ictal events reliably, similar to MEG. Interictal PET is commonly used to identify regional hypometabolism, such as in medial temporal sclerosis (MTS). In cases where it correlates well with temporal lobe epilepsy, it may be sufficient to guide resection. Additionally, in identifying extratemporal abnormalities, current PET studies may provide information that could support other results, but by themselves they will not likely identify focal epileptogenic regions or guide resection.

Most patients with temporal lobe epilepsy, ipsilateral MTS on MRI, and concordant hypometabolism on PET can move directly to resective surgery without additional studies. However, there is a small population of patients with MTS with an indication for intracranial monitoring. Approximately 10% of patients with MTS have a lack of concordance between the side of hippocampal atrophy on MRI and the EEG abnormality.^39,40 Thus, subdural (or depth) monitoring may be able to resolve discordance or “rule in” a temporal lobectomy on the side of radiographic MTS. In concert with this is evidence that one out of seven patients with MTS has another lesion that might be a cause for seizures, such as a vascular malformation, low-grade tumor, or malformation of cortical development.⁴¹ Properly identifying the cause of seizures in these dual-pathology patients is important, as the results of removing the extratemporal lesion alone are not associated with a high rate of seizure freedom.⁴² Technologies that increase the sensitivity and specificity of MRI and other modalities will likely advance with time to better detect epileptogenic lesions.⁴³ However, there will continue to be a role for subdural monitoring whenever there is suspicion of a cortical focus and ambiguity between EEG and the anatomical findings or when there are dual or multiple pathologies.

When there is concordance between EEG, anatomical, and functional imaging data, there is likely no need for intracranial monitoring.⁴⁴ Yet with disparities in data, the role of both anatomical and functional imaging can be employed to guide intracranial electrode implantation. With an ability to confine implantation to a specific region of interest, the extent of implant surgery may be reduced yet would still cover the most suspected areas. Because the risk of complications has been shown to increase 40% with every additional 20 subdural electrodes⁴⁵ and >80% with bilateral implantation, the modalities of imaging and subdural implantation become complementary to minimize patient risk and increase the yield for ictal zone determination and successful surgery.

The precursor to implanted subdural electrodes was their use intraoperatively as pioneered by Berger in 1929. These initial investigations were performed with the electrodes placed outside the dura in patients with skull defects.⁴⁶ Others have also championed the use of epidural electrodes in various formats,^47–49 typically because of the ease of insertion and the lower potential for infection. However, the desired result of increased sensitivity and greater spatial resolution with epidural placement is still inferior to subdural electrodes that lie directly in contact with the cortex.⁵⁰ Penfield was the first to document the use of subdural electrodes with a surgical procedure.⁵¹ This was presented around 1940, but it was more than 40 years before frequent reports began to appear in the literature describing the use of subdural grid electrodes.⁵ Although the use of subdural electrodes has become more common, no standardized approach has been identified. Variability exists in indications, implant techniques or paradigms, and the number and type of electrodes.⁵ Without agreement on a uniform approach, it becomes difficult to make fair comparisons between institutions or any assessment of methods, and a multitude of preferences exist. However, there are clearly advantages to specific approaches. Monitoring regions of the cortex such as the subtemporal and interhemispheric areas cannot be reached with epidural electrodes, yet they are accessible to subdural electrodes, typically via craniotomy. In frontal lobe epilepsy, medial frontal discharges are not detected with scalp EEG⁵² and thus are an excellent indication for subdural electrodes. Regardless of any preference or approach, any time scalp recordings cannot confirm correlation between an imaging abnormality and the patient’s seizures, invasive monitoring is indicated. Even in lesional cases, the best surgical outcomes have been found to occur when an MRI visible focus is consistent with an EEG focus and can be completely removed.⁵³

The choice of electrode strips versus grids is based on the location, accessibility, and degree of coverage desired of suspected epileptogenic regions. Subdural strips (a single line of electrodes) have a lower risk than grids by virtue of their smaller size and may be inserted via a burr hole, thus avoiding a full craniotomy. Subdural electrodes are usually manufactured from soft, flexible, thin transparent Silastic sheets that have small metal contacts and wires embedded within them. Although both stainless steel and platinum contacts are available, the MRI compatibility of the latter allows useful postimplant imaging. Some centers make their own strips and grids.⁵ Commercial providers offer an array of choices in varying dimensions (e.g., Ad-Tech Medical Instrument Corp., Racine, Wisconsin, and PMT/Permark Corp., Chanhassen, Minnesota). A common configuration spaces the 5 mm electrode centers 1 cm apart. The length and shape of the strip/grid are chosen after a review of the noninvasive EEG recordings, imaging studies, and other clinical information, such as seizure semiology and language localization. Various sizes are available; we have found 4 × 5 cm grids to be useful for subfrontal (orbitofrontal) and subtemporal coverage, whereas larger 6 × 8 and 8 × 8 cm grids are useful when covering large areas of the cerebral convexity. Subdural strips have also been used to supplement grids, add regional coverage, or monitor medial temporal lobe regions.⁵⁴ Dual-sided strips and grids are also available for placement in the sylvian fissure, within schizencephalic abnormalities, and in the interhemispheric fissure. Both strips and grids have electrode leads that exit the Silastic sheet and must be brought through the dura and bone upon closure of the surgical wound. These leads are tunneled under the galea and exit through small sites some distance from the scalp incision. Case Study 2 demonstrates an implant methodology on a nonlesional case.

There are a few distinct groups of patients with whom common or uniform strategies of implanting electrodes may apply. Examples of populations where this may be seen include clinically suspected temporal lobe epilepsy with very rapid spread/generalization. When this occurs in the presence of bilateral MTS or bilateral temporal hypometabolism,^57,58 equivocal PET data, or in nonlesional cases, bilateral temporal subdural strips may be a standard implant routine. Bilateral hippocampal depths may also be chosen for these situations. However, as noted previously, there is no standardized methodology regarding choice of subdural strips, subdural grids, depth electrodes, or some combination thereof. Although various authors have expressed their preferences, the common theme in choosing an implant methodology is that each strategy is tailored to the specific clinical needs of each patient. The basic goals are to define as best as possible the focal onset of a seizure and delineate any regions of eloquent brain (usually language or motor function). In lesional cases where the correlation between the lesion and the seizure are highly suspected by noninvasive data, no implant may be necessary. However when there are disparities, the implant plan would clearly include coverage of the lesion, as well as coverage of whatever region of the brain is suspected by EEG or imaging studies (MRI, PET, MEG, and SISCOM). In some cases confirmation can be achieved by use of several strips placed via burr holes. This is the least invasive manner of using subdural electrodes. However, the caution with subdural electrodes is that placement is done without direct visualization. Although delicate subdural structures such as veins may be encountered while sliding strips into a desired location, the relative safety of subdural strip placement is well documented.⁵⁹ Because subdural strip placement can be done relatively easily, strips are sometimes employed to determine regional onset and define or “rule in” the possibility of a more extensive subdural grid implantation. Such would be the case when there is rapid spread of EEG and laterality of onset is not clear, when extensive or multiple lesions are present in different lobes of the brain, or when no lesions are present, and a uniform, stereotyped, but nonlateralized onset is noted. Sometimes strips are used as part of a staged surgical plan that guides subsequent implantation of a subdural grid strategy. In these cases, strips can help define laterality and perhaps regionally, while follow-up implantation of grids can define discrete focality. Although extensive grid placement could be an alternative approach in some of these circumstances, a staged methodology may help avoid an extensive craniotomy using a large number of grid electrodes, and the increased risk proportional to the number of grid electrodes has been documented. Difficulties with subdural strips through burr holes include a limited ability to guide placement over focal (vs regional) areas of interest and problems with adhesions between the dura and the cortex. This can occur when there has been prior surgery for tumors or trauma or in cases with a past history of infection. When placement is not critical, the surgeon can attempt to modify the location of a strip by sliding it to one side of an obstacle. However, burr hole insertion of strips is routinely accomplished over the cerebral convexities and has even been performed to place strips covering the medial and basal temporal lobes and in the interhemispheric fissure.