Intracranial EEG for seizure focus localization: evolving techniques, outcomes, complications, and utility of combining surface and depth electrodes

Transcription

1 CLINICAL ARTICLE Intracranial EEG for seizure focus localization: evolving techniques, outcomes, complications, and utility of combining surface and depth electrodes Yasunori Nagahama, MD, 1 Alan J. Schmitt, MD, 2 Daichi Nakagawa, MD, 1 Adam S. Vesole, BS, 3 Janina Kamm, PsyD, 2 Christopher K. Kovach, PhD, 1 David Hasan, MD, 1 Mark Granner, MD, 2 Brian J. Dlouhy, MD, 1,4 Matthew A. Howard III, MD, 1,4 and Hiroto Kawasaki, MD 1 Departments of 1 Neurosurgery and 2 Neurology, University of Iowa Hospitals and Clinics, Iowa City; 3 Carver College of Medicine, University of Iowa Hospitals and Clinics, Iowa City; and 4 Pappajohn Biomedical Institute, University of Iowa Carver College of Medicine, Iowa City, Iowa OBJECTIVE Intracranial electroencephalography (ieeg) provides valuable information that guides clinical decisionmaking in patients undergoing epilepsy surgery, but it carries technical challenges and risks. The technical approaches used and reported rates of complications vary across institutions and evolve over time with increasing experience. In this report, the authors describe the strategy at the University of Iowa using both surface and depth electrodes and analyze outcomes and complications. METHODS The authors performed a retrospective review and analysis of all patients who underwent craniotomy and electrode implantation from January 2006 through December 2015 at the University of Iowa Hospitals and Clinics. The basic demographic and clinical information was collected, including electrode coverage, monitoring results, outcomes, and complications. The correlations between clinically significant complications with various clinical variables were analyzed using multivariate analysis. The Fisher exact test was used to evaluate a change in the rate of complications over the study period. RESULTS Ninety-one patients (mean age 29 ± 14 years, range 3 62 years), including 22 pediatric patients, underwent ieeg. Subdural surface (grid and/or strip) electrodes were utilized in all patients, and depth electrodes were also placed in 89 (97.8%) patients. The total number of electrode contacts placed per patient averaged 151 ± 58. The duration of invasive monitoring averaged 12.0 ± 5.1 days. In 84 (92.3%) patients, a seizure focus was localized by ictal onset (82 cases) or inferred based on interictal discharges (2 patients). Localization was achieved based on data obtained from surface electrodes alone (29 patients), depth electrodes alone (13 patients), or a combination of both surface and depth electrodes (42 patients). Seventy-two (79.1%) patients ultimately underwent resective surgery. Forty-seven (65.3%) and 18 (25.0%) patients achieved modified Engel class I and II outcomes, respectively. The mean follow-up duration was 3.9 ± 2.9 (range ) years. Clinically significant complications occurred in 8 patients, including hematoma in 3 (3.3%) patients, infection/osteomyelitis in 3 (3.3%) patients, and edema/compression in 2 (2.2%) patients. One patient developed a permanent neurological deficit (1.1%), and there were no deaths. The hemorrhagic and edema/compression complications correlated significantly with the total number of electrode contacts (p = 0.01), but not with age, a history of prior cranial surgery, laterality, monitoring duration, and the number of each electrode type. The small number of infectious complications precluded multivariate analysis. The number of complications decreased from 5 of 36 cases (13.9%) to 3 of 55 cases (5.5%) during the first and last 5 years, respectively, but this change was not statistically significant (p = 0.26). CONCLUSIONS An ieeg implantation strategy that makes use of both surface and depth electrodes is safe and effective at identifying seizure foci in patients with medically refractory epilepsy. With experience and iterative refinement of technical surgical details, the risk of complications has decreased over time. KEYWORDS invasive electroencephalography; medically intractable epilepsy; subdural grid; subdural monitoring; subdural strip ABBREVIATIONS ATL = anterior temporal lobectomy; ieeg = intracranial electroencephalography; MCD = malformation of cortical development; MTS = mesial temporal sclerosis; VNS = vagal nerve stimulator. SUBMITTED July 24, ACCEPTED January 15, INCLUDE WHEN CITING Published online May 25, 2018; DOI: / JNS AANS 2018, except where prohibited by US copyright law 1

2 Intracranial electroencephalography (ieeg) provides more accurate information about patterns of epileptic discharges than noninvasive examinations, and it provides valuable diagnostic information that helps guide surgical decision-making. 7 There are, however, many challenges associated with invasive intracranial monitoring, and a range of strategies has been adopted across institutions to address these challenges. The overarching objective is to place sufficient numbers of electrodes in appropriately targeted brain regions in order to accurately define a patient s interictal and ictal seizure network, while at the same time avoiding surgical complications. The complexity of this field is reflected in the broad range of implantation strategies employed at different epilepsy surgery centers. At some institutions, surface electrode arrays are used exclusively while depth electrodes are used exclusively at others. The advantages and disadvantages of these two approaches have been debated. 8,21,22,24 Other groups have postulated that a combined approach using both surface and depth electrodes may lead to more accurate localization of seizure onsets and patterns of propagation. 8,21 This is the strategy that we have adopted at the University of Iowa, whereby almost all patients who are candidates for intracranial monitoring undergo implantation of both surface and depth electrodes. The diagnostic utility of ieeg methods must be balanced against the risks of surgical complications, including hematoma, infection, malignant cerebral edema, and brain compression from intracranial electrodes. 3,6, 10, 11, 13, 14,17,20, 23,25 Some studies have shown that higher complication rates may correlate with a greater number of implanted electrodes, longer monitoring duration, and a less experienced surgical team. 3,10 Other studies reported no association between complications and the number of electrodes placed or monitoring duration. 6,11,25 In one study, the authors suggested that initial extensive electrode coverage may reduce the risk that a second surgery will be required for placement of additional electrodes. 11 In the current study, we performed a retrospective review and analysis of all the patients who underwent craniotomy and electrode placement for ieeg over the recent 10-year period from January 2006 through December 2015 at our institution. The goals of the study were to describe our evolving ieeg techniques and evaluate our outcomes and complications over the course of this time epoch. Because we used a combination of subdural surface electrodes and stereotactically placed depth electrodes in most patients, we also evaluated how effective and clinically useful each category of electrode was in defining a patient s seizure network. Methods Patients All adult and pediatric patients who underwent placement of intracranial electrodes for long-term seizure monitoring between January 2006 and December 2015 at the University of Iowa Hospitals and Clinics were identified from a prospectively maintained epilepsy surgery database. The database included 91 patients, all of whom underwent either unilateral or bilateral craniotomy for electrode placement. The current study was approved by the University of Iowa institutional review board. Presurgical Workup All patients in this study underwent presurgical epilepsy evaluation, including video-eeg, high-resolution structural MRI, interictal PET scanning with FDG, and neuropsychological testing. The clinical history and results of the presurgical workups in individual cases were reviewed and discussed at a multidisciplinary epilepsy surgery conference, which was composed of epileptologists, neuroradiologists, neuropsychologists, EEG technologists, epilepsy service nurses, and neurosurgeons. Furthermore, Wada testing and/or functional MRI were performed to determine the language-dominant side; to delineate functionally eloquent areas, such as speech, motor, and sensory cortices; or to evaluate memory reserve in patients in whom medial temporal resection was a possibility. Patients were determined to be candidates for ieeg if the results of the workups suggested that they were likely to have a resectable seizure onset zone. Patients with clear-cut mesial temporal seizures in whom all presurgical workup results concordantly pointed to unilateral medial temporal lobe seizure onset underwent anterior temporal lobectomy (ATL) without ieeg. Typically, patients had predominantly unilateral neocortical seizures with ill-defined onset, unilateral temporal lobe seizures with ambiguous mesial versus lateral onset, predominantly unilateral mesial temporal seizures with less frequent contralateral involvement, or unilateral temporal lobe seizures with early involvement of other areas. Surgical Technique and Invasive Monitoring Craniotomy and Exposure In all cases, intracranial electrodes were placed during a craniotomy procedure performed under general anesthesia. Aspects of the surgical technique evolved during the study period, and methods currently in use are described. After removal of the bone flap, hemostatic agents (e.g., Oxycel rolls) were placed between the bone and the dura along the craniotomy margin. Dural tack-up sutures were placed with nonabsorbable stitches to prevent bleeding from the epidural space. The dura was opened and excised circumferentially, leaving a 5-mm to 10-mm dural cuff all around the craniotomy margin (Fig. 1A). Bleeding from the edge of the remaining dura was coagulated with bipolar cautery to minimize the risk of postoperative blood accumulation. Short radial cuts were made to the dural cuff to allow for postoperative expansion of the brain. The dural cuff was tented up and reflected toward the bony edge circumferentially around the craniotomy margin with dural sutures either to the pericranium or to the temporalis muscle for further hemostasis (Fig. 1A). Placement of Intracranial Electrodes We use a combination of grid, strip, and depth electrodes of different sizes and shapes in order to achieve optimal electrode coverage for purposes of characterizing the seizure network and performing functional mapping 2

3 Y. Nagahama et al. FIG. 1. Placement of intracranial electrodes. A: Intraoperative photograph showing a left-sided craniotomy and exposure for a pediatric patient who underwent ieeg with bilateral but predominantly left-sided coverage. Note that the peripheral dural cuff is tented up with black dural sutures to the pericranium or temporalis muscle. B: Intraoperative photograph obtained in the same patient, showing placement of surface and depth electrodes secured to the dural cuff. C: Postoperative axial CT scan showing the elevated placement of the bone flap on the left side. D: The postoperative reconstruction image showing the locations of surface electrode contacts and entry points for depth electrodes. Figure is available in color online only. of eloquent brain regions. Strip electrodes consisted of a single row of a variable number of electrode contacts, whereas grid electrodes were designed with variable numbers of electrode contact rows and columns. Strip electrodes have less implant volume and are more flexible than grid electrodes and cause less brain compression. We preferentially use strip electrodes to gain access to cortical surfaces that abut the dura but cannot be directly visualized without retracting the brain (e.g., ventral temporal and frontal lobes, cortex lining the interhemispheric fissure). We also use strip electrodes to obtain coverage in the brain regions with marked surface curvature (temporal, frontal, and occipital poles), and when the overall area of surface coverage required is very large. Once appropriate exposure and hemostasis were obtained, we placed electrodes in the following order: surface electrodes (grids and/or strips) over ventral areas (ventral frontal and/or temporal areas), surface electrodes (grids and strips) over the convexity, and then depth electrodes (Fig. 1B). The location of each ventral surface electrode was adjusted depending on the anatomy of bridging veins. Every effort was made to preserve bridging veins. Each ventral surface strip electrode was carefully inserted, and if we encountered resistance while inserting an electrode or if we visualized bridging veins, we either changed the location of or aborted placement of that particular electrode. Use of Grid Surface Electrodes When grids are used, they are typically placed over the relatively flat portion of the lateral convexity. Multiple cuts are made in each grid to partially separate each into multiple rows or columns so that the implant is more flexible and produces less mass effect on the brain. 3

4 Stereotactic Placement of Depth Electrodes Each depth electrode was placed using a frameless stereotactic navigation system (StealthStation Navigation, Medtronic), as previously described. 16,18 Briefly, a target point and a proposed entry point for each depth electrode were planned on the navigation workstation prior to surgery. We placed depth electrodes after all the ventral and convexity surface electrodes were placed. A customdesigned guide tube was attached to a standard flexible arm of a Greenberg retractor system (Symmetry Surgical), which was in turn secured to the Mayfield head clamp (Integra). On the StealthStation, the patient s 3D T1-weighted MRI scans had been coregistered with the patient s head. The final entry point for each depth electrode was determined based on presurgical planning but with some adjustments to avoid surface blood vessels, electrodes, sulci, and the sylvian fissure. A 2-mm-long small corticectomy was made at the entry point. A StealthStation Navigus probe was inserted into the guiding tube and manually positioned as indicated by the navigation so that the trajectory of the electrode path went through the entry point and the target. We marked the depth of the target from the entry point, which was calculated by the navigation device, on a slotted cannula (Ad-Tech Medical Instrument Corp.) and an electrode. The Navigus probe was removed from the guide tube, and a cannula sleeve guide (Ad-Tech Medical Instrument Corp.) was slowly inserted into the guide tube. The slotted cannula was inserted into the cannula sleeve guide and advanced into the brain parenchyma to the premarked depth. A stylet of the two-piece slotted cannula was removed, and a depth electrode was advanced through the slotted cannula to the precalculated depth. Once the electrode was placed, the slotted cannula and guide tube were removed. Sampling of the Medial Cortex Cortical areas forming the medial surface of the hemisphere were sampled using multiple approaches. In some instances, strip or small grid electrodes were advanced into the interhemispheric fissure. In other cases, depth electrodes were placed. Depth electrodes were particularly useful in cases in which the anatomy of the interhemispheric fissure was abnormal, such as in patients with a history of trauma, meningitis, or encephalitis. The depth electrodes were placed within the midline cortex using oblique trajectories from the dorsal cortical entry points or lateral trajectories with entry points with the lateral cortex. Securing of Intracranial Electrodes Each electrode was secured in place by stitching a corner or a cable to the nearby dura or another grid electrode at multiple locations with 4-0 Nurolon sutures to prevent unintended movements of the electrode (Fig. 1B). Each ventral surface electrode was typically secured to the dura at the distal end of the electrode. Each electrode over the convexity was usually secured at two locations, one at the leading edge of the electrode and the other at the other end of the electrode or at an exiting cable. Each depth electrode was gently bent to conform to the curvature of the cortical surface and typically secured to a neighboring surface grid electrode or the nearby dura. Duraplasty and Closure The cables of all the intracranial electrodes were tunneled out approximately 3 cm away from the scalp incision line. This procedure was performed prior to dural closure. If the cables were passed through the scalp after dural closure, electrodes might be inadvertently pulled and unknowingly displaced from their original positions while their cables were being tunneled through the scalp. After all the electrode cables were tunneled, an expansive duraplasty was performed using a commercially available dural substitute in a watertight fashion. The dural substitute was sutured to the dural cuff while the dural cuff was kept tented up to the surrounding craniotomy edge. Keeping the dural cuff tented up maintained hemostasis and also minimized the risk of unintended movements of the electrodes sutured to the dural cuff. More recently, we used fibrin sealant over the dural sutures to improve the watertight closure. In all of the electrode implantation cases, the bone flap was replaced. In recent cases, we replaced the bone flap in a slightly elevated position using rectangular titanium plates bent in a stair shape to mitigate the mass effect exerted by the electrodes (Fig. 1C). 15 Before scalp closure, we placed a 4-contact strip electrode in the subgaleal space, which was used to provide a reference and ground contacts. In the latter half of this series, vancomycin powder was sprinkled over the subcutaneous space after replacement of the bone flap in order to reduce postoperative infection. 1,12 A cutaneous drain was placed, which was typically removed on postoperative day 1, and the scalp and temporalis muscle were reapproximated in the standard fashion. Each electrode extension cable was individually secured to the scalp with a 3-0 nylon skin suture. Postimplantation Management Postoperatively, the patients underwent imaging, typically high-resolution brain MRI, brain CT, and skull radiography to evaluate for mass effect exerted by electrodes, brain edema, and hemorrhagic complications and to localize electrode contacts (Fig. 1D). Prophylactic intravenous antibiotics (usually cefazolin or either vancomycin or clindamycin for patients allergic to cefazolin) were given intraoperatively during the electrode implantation surgery, continued throughout the monitoring period, and typically continued for 3 5 days after the second surgery until the patients were discharged from the hospital. One dose of dexamethasone (10 mg) was given intravenously during the first and second surgeries, but no additional doses were routinely administered due to a concern that steroids may decrease the frequency of habitual seizures, as previously shown 2 and to a theoretical increased risk of infection and impaired wound healing. Antiepileptic drugs were gradually tapered postoperatively using a protocol tailored to each individual patient. During monitoring, the surgical wound was checked and cleaned daily. If a CSF leak was noted from the cable exit sites, additional sutures were placed at the bedside for tight sealing. Monitoring was typically scheduled for 2 weeks in adult patients and for 7 10 days in pediatric patients. When no sufficient seizures were recorded by the end of the scheduled monitoring period, the monitoring was ex- 4

5 tended up to approximately 3 weeks as necessary. At the end of the monitoring, video-eeg data were reviewed, and it was determined if the patients had a seizure focus or foci that could be safely resected. If the patients were good candidates for surgical procedures, after informed consent was obtained they underwent electrode removal and seizure focus resection or disconnection on the same day. When planned seizure focus resection involved areas near the eloquent cortex (language and primary motor areas), extraoperative functional cortical stimulation mapping was performed at the end of the ieeg monitoring period. The patients who were not candidates for resective surgery underwent electrode removal only. Data Collection The basic demographic and clinical information was obtained from the prospectively maintained epilepsy surgery database. Additional detailed information was further collected through retrospective review of the medical records (by Y.N., A.J.S., and A.S.V.). The gathered information included the following variables: age, sex, a history of prior surgical interventions, electrode coverage (laterality, areas, and number of electrodes), monitoring duration, monitoring results, extent of seizure focus resection, pathological findings, outcomes, and complications. Assessment of Outcome and Complications Seizure outcome was assessed based on information contained in the last known clinic visit note at the University of Iowa Hospitals and Clinics as of November 2016, using the modified Engel class. 9 Adverse events were considered to be clinically significant complications if they resulted from surgeries (either the first surgery for electrode implantation or the second surgery for electrode removal with/without seizure focus resection) and resulted in additional unplanned surgical interventions. Assessment of the Utility of Surface Versus Depth Electrodes in Seizure Focus Localization We evaluated the utility of surface versus depth electrodes in seizure focus localization. The seizure onset zone was defined as the area involved in seizure activity within the first second of seizure onset. For each case with recorded seizures, the seizure onset zone was categorized based on its detection by surface electrodes, depth electrodes, or a combination of both. Data Analysis Data were analyzed using the JMP program (version for Windows, SAS Institute). We evaluated correlations between clinical variables that we gathered and the clinically significant complications (i.e., surgically treated mass effects due to intracranial hemorrhage, edema, and compression by electrodes). Univariate and multivariate analyses were used to determine the relationships between the clinical variables and these complications. The variables with p < 0.3 in the univariate analysis were used in the forward and backward stepwise multivariate analysis model. Significance was set as p < TABLE 1. Demographic data for ieeg cases Characteristics Value No. of patients 91 Male 49 Female 42 Mean age in yrs ± SD (range) 29 ± 14 (3 62) Adult ( 18 yrs) 69 (75.8) Pediatric (<18 yrs) 22 (24.2) History of prior surgeries 13 (14.3) VNS placement 7 (7.7) ATL 2 (2.2) VNS & ATL 1 (1.1) Tumor resection/biopsy 3 (3.3) Values are presented as the number of patients (%) unless stated otherwise. Furthermore, the Fisher exact test was used to evaluate whether there was a significant difference in the rate of complications during the first and second half of the study period (from 2006 to 2010 and from 2011 to 2015, respectively). Results Demographics and Presurgical Workup During the 10-year period from January 2006 through December 2015, 91 patients (49 males and 42 females) underwent craniotomy and placement of intracranial electrodes for extraoperative invasive ieeg (Table 1). The age of the patients averaged 29 ± 14 years and ranged from 3 to 62 years. This group included 22 pediatric patients younger than 18 years (24.2%). Thirteen patients (14.3%) had undergone prior procedures, such as vagus nerve stimulator (VNS) placement (7 patients), ATL (2 patients), both VNS placement and ATL (1 patient), and tumor resection/biopsy (3 patients). At least one lesion was noted on preoperative structural MRI in 57 patients (62.6%, lesional). Twenty patients (22.0%) were diagnosed as having mesial temporal sclerosis (MTS) on MRI. The other 34 cases (37.4%) were nonlesional. Implantation of Intracranial Electrodes Forty-five patients (49.5%) underwent unilateral craniotomy (24 right-sided and 21 left-sided cases), whereas 46 patients (50.5%) underwent bilateral craniotomy (20 predominantly right-sided, 25 predominantly left-sided, and 1 symmetric craniotomies; Table 2). In predominantly one-sided cases, a limited number of electrodes (typically a few depth electrodes with or without a few strip electrodes) were placed through a small craniotomy contralateral to a large craniotomy side in order to rule out contralateral ictal onset. The electrode coverage included most frequently the temporal lobe (86 cases, 94.5%) and the frontal lobe (74 cases, 81.3%), followed by the parietal lobe (25 cases, 27.5%), the insula (11 cases, 12.1%), and the occipital lobe (10 cases, 11.0%). Depth electrodes were used in almost all cases (89 patients, 97.8%), and mesial 5

6 TABLE 2. Electrode coverage Characteristic Value Laterality Unilateral Rt-sided 24 (26.4) Lt-sided 21 (23.1) Bilateral 46 (50.5) Rt > lt 20 (22.0) Rt < rt 25 (27.5) Rt = lt 1 (1.1) Coverage of lobes Frontal 74 (81.3) Temporal 86 (94.5) Parietal 25 (27.5) Occipital 10 (11.0) Insula 11 (12.1) Patients w/ depth electrodes 89 (97.8) Mesial temporal w/ depth electrodes 81 (89.0) Insula w/ depth electrodes 11 (12.1) Mean no. of implants & electrodes ± SD (range) Grids 2.9 ± 1.8 (0 10) Grid contacts 96.3 ± 64.4 (0 199) Strips 5.7 ± 3.5 (0 19) Strip contacts 31.3 ± 24.5 (0 148) Depths 4.3 ± 2.1 (0 11) Depth contacts 23.5 ± 15.9 (0 78) Total contacts 151 ± 58 (24 255) Values are presented as the number of patients (%) unless stated otherwise. temporal structures were the most frequently targeted sites (81 patients, 89.0%). A combination of grid, strip, and depth electrodes was utilized in essentially all cases in order to achieve optimum electrode coverage. The number of grid, strip, and depth electrodes used in individual cases averaged 2.9 ± 1.8 (range 0 10), 5.7 ± 3.5 (0 19), and 4.3 ± 2.1 (0 11), respectively (Table 2). The total number of electrode contacts used in individual cases averaged 151 ± 58 (range ). The total number of contacts for each electrode type, i.e., grid, strip, and depth electrodes, averaged 96.3 ± 64.4 (range 0 199), 31.3 ± 24.5 (0 148), and 23.5 ± 15.9 (0 78), respectively. Invasive Monitoring The monitoring duration averaged 12.0 ± 5 days and ranged from 1 to 30 days (Table 3). In total, a seizure focus was localized in 84 (92.3%) cases, and seizure focus resection was performed in 72 (79.1%) cases (Tables 3 and 4). In 82 of 84 patients (90.1%) seizure foci were localized based on ictal onset (Table 3). Among these 82 cases, a seizure focus was detected with surface electrodes only, depth electrodes only, and both in 28 cases (34.1%), 13 cases (15.9%), and 41 cases (50.0%), respectively. In this TABLE 3. Invasive monitoring results Characteristic Value Mean monitoring duration in days ± SD (range) 12.0 ± 5.1 (1 30) Seizure focus identified 82 (90.1) Surface vs depth electrodes Surface electrodes 28/82 (34.1) Surface & depth electrodes 41/82 (50.0) Depth electrodes 13/82 (15.9) Resection/disconnection 70 Resection offered but not performed 2 Resection not indicated/offered 10 Bilateral onsets 7 Multiple independent foci 3 Seizure focus suspected based on interictal 2 (2.2) discharges Surface vs depth electrodes Surface electrodes 1/2 (50) Surface & depth electrodes 1/2 (50) Depth electrodes 0/2 (0) Resection/disconnection 2 Seizure focus not identified/suspected 7 (7.7) Ambiguous/diffuse/bilateral/multifocal 4 No seizure despite monitoring extension 1 Early electrode removal/complications 2 Seizure focus resection, total cases 72 (79.1) Complications Hematoma 3 (3.3) Infection/osteomyelitis 3 (3.3) Edema/brain compression 2 (2.2) Permanent deficit 1 (1.1) Death 0 (0) Values are presented as the number of patients (%) unless stated otherwise. cohort of patients, seizure focus resection was performed in 70 cases, was offered but not performed in 2 cases, and was not offered in 10 cases due to bilateral and/or multiple independent seizure foci. In the remaining 2 cases (2.2%), the seizure foci were inferred based on the locations of interictal epileptic discharges (e.g., bursts of gamma activity, spike-wave discharges) because the patients did not have seizures during the period of monitoring (Table 3). The interictal discharges were recorded from subdural surface electrodes in one case and by both surface and depth electrodes in the other case. Nineteen patients did not undergo resective surgery (Table 5). Bilateral independent medial temporal seizure foci were found in 8 patients. In 4 patients multiple diffuse seizure onsets were recorded. A seizure focus could not be localized in 3 patients, including 1 patient in whom no seizures occurred despite extension of the monitoring period to 29 days. Intracranial monitoring was aborted due to significant complications in the other 2 cases (epidural 6

7 TABLE 4. Preoperative MRI findings and types of surgical procedure MRI Finding Surgical Approach ATL Extended ATL ATL or Extended ATL & T Lesion ATL or Extended ATL & Extratemporal T or T lesion Extratemporal Total MTS +, lesion MTS +, lesion MTS, lesion MTS, lesion Total Extended ATL = ATL with extended temporal lobe resection; Extratemporal = resection of ablation outside of temporal lobe; T = resection of temporal lobe without medial temporal resection; T lesion = resection of lesion in the temporal lobe without medial temporal resection; + = positive; = negative. Values are presented as the number of patients. hematoma in one case and intraparenchymal hemorrhage in the other). In 1 case, resective surgery was declined. In another case resective surgery was not performed due to concern for memory function because the seizure focus was localized to the medial temporal lobe of the languagedominant side in the patient who had an atrophic contralateral hippocampus. Resection Seizure focus resection was ultimately performed in 72 patients (79.1% of all cases). Relationships between preoperative MRI findings and types of surgery performed are summarized in Table 4. Among the 72 resection cases, MTS was found in 15 patients (20.8%) on preoperative MRI (MTS + in Table 4). Thirty-four (47.2%) patients had lesions other than MTS (lesion +), including 2 patients with both (MTS + lesion +). Twenty-five (34.7%) cases were nonlesional (MTS - lesion -). It is worth noting that in 37 of 57 patients without MTS on preoperative MRI (MTS- and ATL or extended ATL in Table 4) seizure foci were localized in the medial temporal lobe with or without additional foci, and therefore anterior and medial temporal lobe resection was performed in these patients. Overall, seizure foci were localized to the temporal lobe in 56 patients (77.7%), solely in the temporal lobe in 44 patients (61.1%), solely outside of temporal lobe in 16 patients (22.2%), and both in 12 patients (16.7%). Anterior and medial temporal lobe resection (ATL) was performed in 52 patients (72.2%), among whom 28 patients (38.9%) underwent standard ATL alone, whereas the remaining 24 patients underwent more extensive resection in addition to ATL in the temporal lobe to remove a wider temporal neocortical seizure onset zone (5 cases, 6.9%), lesion in the temporal lobe (7 cases, 9.7%) or outside of the temporal lobe (12 cases, 16.7%) (Table 4). Four patients (5.6%) underwent resection of seizure foci in the temporal lobe without resection of the amygdala and the hippocampus. Sixteen cases (22.2%) had seizure foci only outside of the temporal lobe. Temporal plus epilepsy, in which seizure foci were localized in both the medial temporal lobe and outside of the temporal lobe, was noted in 12 cases (16.7%), most commonly involving the frontal lobe (10 cases, 13.9%). If we define lesionectomy as resection of an MRIpositive lesion, we performed 47 lesionectomies (65.3%), of which 15 (20.8%) were ATL for MRI-positive MTS, 32 (44.4%) were resection of lesions other than MTS, and 2 cases (2.8%) had both MTS and non-mts lesions (Table 4). Type of Electrode and Resection Among 13 patients whose seizure foci were identified only by depth electrodes, all patients had deep seizure foci: 7 foci in the unilateral medial temporal lobe, 1 focus in the bilateral medial temporal lobe, 1 focus in the unilateral medial temporal lobe and concomitant lateral temporal cortex next to a neoplastic lesion, and 4 foci in extratemporal areas (3 lesional cases and 1 nonlesional case). We performed ATL, extended ATL and resection of a temporal lesion, and resection of an extratemporal lesion in 6, 1, and 4 cases, respectively. Two patients did not undergo resective surgery. As for the patients whose seizure foci were identified by both depth and surface electrodes, both types of electrodes played a crucial role in identifying seizure foci and determining the extent of resection in more than one-third of TABLE 5. Preoperative MRI findings and reason for no resection MRI Finding Bilateral Onset Multifocal Not Localized Declined Dominant Medial Temporal Lobe Aborted Total MTS +, lesion MTS +, lesion MTS, lesion MTS, lesion Total Values are presented as the number of patients. 7

8 TABLE 6. Preoperative MRI findings and pathology of 72 resection cases MRI Finding Pathology MTS Gliosis MCD Neoplasm Total MTS +, lesion MTS +, lesion MTS, lesion MTS, lesion Total Values are presented as the number of patients. TABLE 7. Location and pathology of 72 resection cases MTS Gliosis MCD Neoplasm No Abnormality Hippocampus 18 (25) 17 (23.6) 9 (12.5) 2 (2.8) 4 (5.6) Amygdala NA 14 (19.4) 14 (19.4) 0 4 (5.6) Cortex NA 33 (45.8) 24 (33.3) 0 6 (8.3) MRI lesion 13 (18.1) 11 (15.3) 27 (37.5) 9 (12.5) 0 NA = not applicable. Values are presented as the number of patients (%). these patients. Among 42 patients whose seizure focus was detected or inferred by both depth and surface electrodes, in 16 patients depth electrodes and surface electrodes picked up seizure onset zones that were geographically remote from each other. Among them, 6 underwent resection of anterior medial temporal lobe and extratemporal seizure foci, 6 underwent resection of extratemporal seizure foci, 1 had resection of a lesion in the temporal lobe, and 3 did not undergo resective surgery. In the remaining 26 patients, depth electrodes and surface electrodes seemed to pick up seizure onset of the same structures, and therefore we could have found seizure foci by either type of electrode. Most of such overlapping detection of seizure foci by depth and surface electrodes occurred in the medial temporal lobe. Fourteen of these patients underwent standard ATL, 2 had ATL plus resection of a temporal lesion, 3 had ATL plus resection of an extra-temporal lobe, 2 had resection of an extratemporal lobe, and 5 did not have resection. Pathology Pathology results in relation to preoperative MRI findings are summarized in Table 6. Because multiple samples were obtained from multiple brain sites in each patient, the total number of pathologies was 114. All patients had at least one abnormal pathology finding. MTS was found in 18 patients, including 6 MRI-MTS negative patients whose MRI findings were negative for MTS. Gliosis, malformation of cortical development (MCD), and neoplasm were found in 47, 40, and 9 brain sites, respectively. Of note, we found 31 abnormalities in nonlesional cases (MTS, lesion ). Pathology findings in relation to brain sites are summarized in Table 7. MTS and gliosis were found almost equally in the hippocampus in 18 (25%) and 17 (23.6%) patients, respectively, followed by MCD in 9 patients (12.5%). In the amygdala, gliosis and MCD were found in 14 patients each. We found gliosis in the cortex in almost half of the cases (45.8%), followed by 24 cases (33.3%) of MCD. MRI-positive lesions were diagnosed as MTS, gliosis, MCD, and neoplasm in 13 (18.1%), 11 (15.3%), 27 (37.5%), and 9 (12.5%), respectively. Outcome Among the 72 patients who underwent seizure focus resection, 47 (65.3%) patients achieved seizure freedom (modified Engel class I) (Table 8). Engel classes II, III, and IV were obtained in 18 (25.0%), 6 (8.3%), and 1 (1.4%) patients, respectively. The follow-up duration averaged 3.9 ± 2.9 years and ranged from 0.1 to 10.5 years. The outcomes based on pathological findings are summarized in Table 9. MCD alone was most frequently encountered (25 patients), followed by gliosis in 19 patients, MTS in 10 patients, dual findings of MCD and MTS in 8 patients, tumors in 8 patients, and a cavernoma in 1 patient. In the 2 patients in whom resective surgery was performed based on interictal discharges, one patient achieved modified Engel class IA outcome at 3.1 years after right ATL and the other achieved class IB outcome at 3.5 years after left posterior temporal lesionectomy, confirming appropriate localization of the epileptogenic zones. Table 10 summarizes outcome in relation to the types of surgical procedures. ATL achieved the best outcome in terms of the percentage of Engel class I outcome, followed by nonmedial temporal lobe resection. Outcome of extended ATL, extended ATL with resection of a temporal or extratemporal lesion achieved about 60%, although less than 50% of resection of an extratemporal lesion without ATL achieved Engel class I. Among the 19 patients who did not undergo resective surgery, 6 patients subsequently underwent VNS placement (their outcomes are not included in the tables). Among the cohort of 72 patients who underwent resective surgery, 4 patients subsequently underwent VNS placement (their outcomes were Engel classes IIIA at 7.6 years, IVA at 6.9 years, IC at 7.1 years, and IIIA at 2.1 years, and included in the tables), 1 patient underwent repeat ieeg and additional resection after the study period (Engel class IIIA outcome at 5.7 years before the repeat resection was included in this study, although the patient again became seizure free after the second resective surgery), and 1 patient underwent additional resection at an outside hospital for tuberous sclerosis followed by VNS placement during the study period (the Engel class IC outcome at 7.9 years was included in the tables). Complications Hemorrhagic Complications Clinically significant hemorrhagic complications occurred in 3 (3.3%) cases during the monitoring period after electrode implantation surgery (Table 3). One of the patients required emergency evacuation of an epidural hematoma during the monitoring period, continued the 8

9 TABLE 8. Outcome in 72 resection cases Modified Engel Epilepsy Surgery Outcome Class No. of Patients (%) Class I 47 (65.3) IA 27 IB 10 IC 7 ID 3 Class II 18 (25.0) IIA 8 IIB 7 IIC 3 IID 0 Class III 6 (8.3) IIIA 6 IIIB 0 Class IV 1 (1.4) IVA 1 IVB 0 IVC 0 The mean follow-up duration ± SD was 3.9 ± 2.9 (range ) years. TABLE 9. Outcome in 72 resection cases based on pathology results Pathology No. of Patients Modified Engel Outcome Score* Class I Class II Class III Class IV Gliosis (58) 7 (37) 1 (5) 0 (0) MTS 10 6 (60) 3 (30) 1 (10) 0 (0) MTS & MCD 8 6 (75) 1 (13) 1 (13) 0 (0) MCD (68) 5 (20) 2 (8) 1 (4) Neoplasm 8 6 (75) 2 (25) 0 (0) 0 (0) Ganglioglioma 5 4 (80) 1 (20) 0 (0) 0 (0) DNET 1 0 (0) 1 (100) 0 (0) 0 (0) Grade II oligodendroglioma 1 1 (100) 0 (0) 0 (0) 0 (0) Grade III oligodendroglioma 1 1 (100) 0 (0) 0 (0) 0 (0) Cavernoma 1 1 (100) 0 (0) 0 (0) 0 (0) NA (MST) 1 0 (0) 0 (0) 1 (100) 0 (0) Total DNET = dysembryoplastic neuroepithelial tumor; MST = multiple subpial transection. * Values are presented as the number of patients (%). monitoring, and ultimately underwent left ATL (pathology: MCD). This patient achieved a modified Engel class IB outcome at 5.7 years (Fig. 2A). The other 2 patients developed large hematomas with significant neurological decline (right-sided epidural hematoma in one case and left frontal intraparenchymal hemorrhage around one of depth electrodes in the other), requiring emergency removal of the electrodes and termination of the monitoring without seizure focus resection (Fig. 2B and C). No patients suffered hemorrhagic complications after the surgery for electrode removal with or without seizure focus resection. Infectious Complications Three cases were complicated by surgical site infections (Table 3). The infections in 2 of these cases were both noted within the few weeks after electrode removal and seizure focus resection (Enterobacter aerogenes in one case and coagulase-negative Staphylococcus in the other case), requiring debridement and removal of a bone flap. One of these patients had undergone left ATL (pathology: MTS), resulting in an Engel class IIIA outcome at 7.6 years, whereas the other patient had undergone right ATL (pathology: gliosis) and ultimately achieved Engel class IA at 3.1 years. The third patient who had undergone left ATL and left frontal lesionectomy (pathology: gliosis) presented with delayed surgical site infection about 19 months later, requiring debridement and removal of the bone flap. Although clearly purulent fluid was noted intraoperatively, no organisms grew on cultures. The patient was empirically treated with a 4-week course of intravenous cefepime. Seizure focus resection in this patient resulted in Engel class IB outcome at 1.9 years. Cerebral Edema/Brain Compression Complications Clinically significant cerebral edema and brain compression requiring additional surgical interventions occurred in 2 (2.2%) cases (Table 3). Significant aphasia and confusion developed in one of these patients, requiring an additional surgery during the monitoring for temporary removal of the bone flap (Fig. 3A). The patient ultimately underwent right ATL (pathology: gliosis) and achieved Engel class IA outcome at 1.9 years. The other patient developed aphasia after electrode implantation, which persisted throughout the monitoring period, as we reported previously. 15 His seizure focus was localized to the left temporal lobe, and the aphasia precluded functional mapping of the language area prior to tailored left ATL (Fig. 3B). This patient underwent electrode removal without ATL and recovered from aphasia soon after the electrodes were removed. He subsequently underwent tailored left ATL during an awake craniotomy and intraoperative language mapping (pathology: MCD) several months later. The patient achieved Engel class IA outcome at 3 years. Permanent Deficits One patient (1.1%) in the series developed a permanent neurological deficit related to a surgical complication (Table 3). This patient developed a left frontal intraparenchymal hemorrhage associated with depth electrode placement, as described above (Fig. 2C), and has a moderate aphasia. No patients experienced other clinically significant complications and permanent neurological deficits specifically due to the surgery for electrode removal with or without seizure focus resection. No deaths occurred during this series. Statistical Analysis of Complications The multivariate analysis for the hematoma and edema/ compression complications (5 cases) showed that these complications were significantly associated with a greater 9

10 TABLE 10. Types of surgical procedure and outcome Engel Class Surgery ATL Extended ATL ATL or Extended ATL & T Lesion ATL or Extended ATL & Extratemporal T or T Lesion Extratemporal Total I * II III IV Total % of class I % of class I & II Values are presented as the number of patients unless stated otherwise. Extended ATL = ATL with extended temporal lobe resection; Extratemporal = resection of ablation outside of temporal lobe; T = resection of temporal lobe without medial temporal resection; T lesion = resection of lesion in the temporal lobe without medial temporal resection. * One patient who had resection of an extratemporal lesion underwent ATL at an outside hospital and achieved class IA. This patient was moved from the extratemporal category in Table 4 to the extended ATL and extratemporal category in Table 10. number of the implanted electrode contacts (p = 0.01), but not with age, history of prior cranial surgery (craniotomy or burr hole placement), laterality of the procedure (unilateral left sided or predominantly left sided), the duration of monitoring, or the number of grid/strip/depth electrodes (Table 11). Due to the small number of the infectious complications (3 cases), separate multivariate analysis could not be performed. Of note, the duration of the monitoring and the total number of electrode contacts for each of these 3 patients were 14 days and 188 contacts, 22 days and 224 contacts (the case of delayed surgical site infection), and 14 days and 167 contacts, respectively. There were a total of 5 complications (2 cases of hematoma, 1 case of edema/brain compression, and 2 cases of infection) among the 36 (13.9%) cases performed during the first 5 years of this series. There were a total of 3 complications (1 case each of hematoma, edema/brain compression, and infection) among the 55 cases (5.5%) performed during the latter 5 years. The Fisher exact test comparing the rates of all the types of the complications during the first and second 5 years of this series showed no significant difference (OR 0.60, 90% CI ; 5 of 36) vs (3 of 55); p = 0.26). One patient required an additional operation to reposition some of the electrodes and place additional electrodes because the original electrodes did not cover the seizure onset zone. The seizure focus was subsequently identified, and the patient underwent resective surgery. Discussion Invasive extraoperative monitoring with ieeg provides essential diagnostic information to guide epilepsy surgery when the results of noninvasive studies are inconclusive. 7 However, there are risks associated with invasive ieeg, including hematoma formation, infections, malignant cerebral edema, and brain compression from intracranial electrodes. 3,6,10,11,13,14,17,20,23,25 A prior study demonstrated that the rate of complications improved over time with increasing experience and more refined surgical techniques and postoperative care. 10 Other authors performed a systematic review and meta-analysis and found an increase in the rate of complications, thought to be possibly attributable in part to increased use of larger grids. 23 In the current report, we aimed to examine our institutional experience over a recent 10-year epoch to study more updated outcomes and complications associated with use of a combined surface and depth electrode implantation strategy. The current study involved a total of 91 consecutive FIG. 2. Axial CT images obtained in 3 patients with postimplantation hemorrhagic complications. A: Left-sided acute epidural hematoma. B: Right-sided acute subdural hematoma, as well as subgaleal hematoma. C: Left frontal intraparenchymal hemorrhage (arrow) associated with a depth electrode (the high density dot is due to the artifact). 10

11 FIG. 3. Axial CT images obtained in 2 patients with postimplantation edema/compression complications. Right-sided (A) and left-sided (B) electrode placement with associated brain edema and compression, causing significant mass effect and midline shift. patients (both adult and pediatric) who underwent craniotomy and placement of intracranial electrodes for ieeg from January 2006 through December 2015 at the University of Iowa. The most common areas of coverage were in frontotemporal brain regions. Surface electrodes (grids and/or strips) were used in all patients in this series; depth electrodes were also placed in most of these patients (97.8%; 89 of 91 patients), most commonly targeting the medial temporal lobe structures (89.0%; 81 of 91 patients). Epilepsy centers that mainly use subdural surface grid and strip electrodes may infrequently utilize depth electrodes, but the approach at our center has been to use a combination of surface electrodes and depth electrodes with the intent of sampling both cortical and subcortical structures in order to provide effective coverage of epileptogenic networks. The total number of contacts used in individual cases averaged 151 ± 58 (range ), which is higher than the numbers reported in earlier series. 3,6,10,11,13,17,25 Our strategy of using extensive recording arrays is influenced by the fact that the epileptogenic zone in some patients may consist of multiple spatially distributed areas of high epileptogenicity. 5 For instance, the epileptogenic zone in some patients with temporal lobe epilepsy may extend beyond the medial and/or anterior lateral temporal neocortex to involve additional areas, such as the orbitofrontal area, insula, suprasylvian area, and temporoparieto-occipital junction (i.e., temporal plus epilepsy). 4,5,19 These spatially distributed regions of the epileptogenic network, if present but not recognized and resected, may contribute to surgical treatment failures. In fact, we extended resection from standard ATL to the neighborhood area in the temporal lobe or in the outside of temporal lobe in 24 cases (33.3%) in this series. In addition, invasive recording with comprehensive electrode coverage was often helpful in tailoring seizure focus resection not only in nonlesional cases but in lesional cases where peri-lesional areas may be part of the epileptogenic zone, and such epileptogenic zones were often not visible on MRI. It is also noteworthy that many of the abnormal pathology findings, especially MCD and gliosis, were not detected on preoperative MRI. In 50% of cases in which a seizure focus was localized, the focus was identified from recordings obtained from both surface and depth electrodes. In the remaining 50% of cases, the seizure focus was identified from recordings obtained from only one type of electrode: in 29 cases by surface electrodes and 13 cases by depth electrodes. These 13 patients had seizure foci in deep structures. In 16 of 42 patients whose seizure foci were identified with both depth and surface electrodes, both types of electrodes played a crucial role for the determination of surgical approach, and either one was indispensable. These results are consistent with those of earlier studies reporting that a combined approach using both surface subdural and depth electrodes provides more accurate information about the locations of seizure onsets and propagation patterns. 8,21 Each type of electrode has its advantages and disadvantages in terms of clinical utility. For example, lateral and dorsal neocortices are optimally covered using surface electrodes. Especially around the motor, sensory, and language cortices, surface grid electrodes have a clear advantage for performing functional mapping. Deep-seated structures, such as the insula, cingulate cortex, and cortical dysplasia located deep in the sulci, are better accessed with depth electrodes. The hippocampus and amygdala can be accessed using either subdural electrodes or depth electrodes. In particular, a long strip electrode along the parahippocampal gyrus is a very useful tool. However, accessing a deep area with surface electrodes is a fundamentally blind technique and it is not uncommon to be affected by adhesion of the cortical surface to the dura mater, location of bridging veins, and undulation of skull base bone. We believe that our TABLE 11. Univariate and multivariate analyses of hematoma and edema/brain compression complications (5 cases) in 91 cases Variables Univariate Analysis Multivariate Analysis OR 95% CI p Value OR 95% CI p Value Age Prior surgery* < Lt sided Monitoring duration No. of grids No. of strips No. of depths No. of total contacts * Includes prior craniotomy or burr hole placement (e.g., previous ATL, tumor resection, and tumor biopsy). 11