Statistical modeling of ICEEG features that determine resection planning

Accepted Manuscript Title: Statistical Modeling of ICEEG Features that Determine Resection Planning Author: Matthew C. Davis Devin R. Broadwater Winn H. Mathews A. Lebron Paige Jennifer L. DeWolfe Ro A. Elgavish Kristen O. Riley Lawrence W. Ver Hoef PII: DOI: Reference:

S0303-8467(16)30188-3 http://dx.doi.org/doi:10.1016/j.clineuro.2016.05.017 CLINEU 4418

To appear in:

Clinical Neurology and Neurosurgery

Received date: Revised date: Accepted date:

28-7-2015 18-2-2016 16-5-2016

Please cite this article as: Davis Matthew C, Broadwater Devin R, Mathews Winn H, Paige A Lebron, DeWolfe Jennifer L, Elgavish Ro A, Riley Kristen O, Ver Hoef Lawrence W.Statistical Modeling of ICEEG Features that Determine Resection Planning.Clinical Neurology and Neurosurgery http://dx.doi.org/10.1016/j.clineuro.2016.05.017 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Statistical Modeling of ICEEG Features that Determine Resection Planning

Matthew C. Davis, M.D.,1 Devin R. Broadwater, B.S.,2 Winn H. Mathews, M.D.,3 A. Lebron Paige, M.D.,4 Jennifer L. DeWolfe, M.D.,4 Ro A. Elgavish, M.D.,4 Kristen O. Riley, M.D.,1 Lawrence W. Ver Hoef, M.D.4 1. Department of Neurosurgery, University of Alabama at Birmingham, Birmingham, AL, United States. 2. University of Alabama at Birmingham School of Medical, Birmingham, Alabama, United States 3. School of Medicine, University of South Alabama, Mobile, AL, United States. 4. UAB Epilepsy Center, University of Alabama at Birmingham, Birmingham, AL, United States.

Correspondence: Devin R. Broadwater, B.S. Department of Neurosurgery University of Alabama at Birmingham 1720 Second Avenue South FOT 1062 Birmingham, AL 35294-3410 Phone: 334-477-8576 E-mail: [email protected] Keywords: epilepsy; electroencephalography; neurology; neurosurgery; outcome assessment

Abstract Object: The interpretation of intracranial EEG (ICEEG) recordings is a complex balance of the significance of specific rhythms and their relative timing to seizure onset. Ictal and interictal findings are evaluated in light of findings from cortical stimulation of eloquent cortex to determine the area of resection. Patients and Methods: Patients with ICEEG electrodes and subsequent surgical resection were retrospectively identified. Only the first 15 seconds of ictal activity, which was divided into five 3-second epochs, was considered. Every electrode in each patient was considered a separate observation in a logistic regression model to predict whether the cortex under a given electrode was included in the planned resection. Results: 19 included patients had a total of 37 unique seizures. Recordings from a total of 1306 electrodes were analyzed. The strongest predictors of resection of cortex underlying a given electrode was the presence of low-voltage fast activity in Epoch 1, rhythmic spikes in Epoch 1, interictal paroxysmal fast activity, and low-voltage fast activity in Epoch 2. High-amplitude beta spikes and rhythmic slow waves were also significant predictors in Epoch 1. Interictal spikes had a higher odds ratio of affecting the planned resection if described as “continuous” or “very frequent”. The presence of motor or language cortex were the strongest negative predictors of resecting underlying cortex. Conclusions: Here we describe a novel model of ictal and interictal patterns significantly associated with the inclusion of cortex underlying a given ICEEG electrode in the surgical resection plan.

1. Introduction: The interpretation of ictal intracranial EEG (ICEEG) recordings for resection planning is a complex balance of the significance of specific rhythms and their relative timing to seizure onset. Various interictal findings are also felt to be indicative of the epileptogenic zone. Ictal and interictal findings are then evaluated in light of findings from cortical stimulation of eloquent cortex to determine the area of resection.

Epilepsy surgery can produce the greatest likelihood of a cure in intractable epilepsy patients; however, it requires a focal onset in order to achieve seizure freedom. Further, greater accuracy localizing the lesion helps preserve eloquent cortex while improving seizure free outcomes. Thus, methods to localize lesions as precisely as possible is of upmost importance (Engel et al 1996, Rosenow et al 2001). Modalities used in the localization of epileptogenic foci and determining feasibility of resection include conventional and functional MRI, magnetoencephalography (MEG), single proton emission computed tomography (SPECT), positron emission tomography (PET), Wada testing, scalp video-EEG, and ICEEG (Van Paesschen et al 2007, Rathore et al 2013, Pittau et al 2014, Rheims et al 2013, Holmes et al 2010). When noninvasive tests are concordant and demonstrate an epileptogenic focus in a surgically favorable area, patients may proceed directly to resection. When the epileptogenic zone cannot be localized with confidence using non-invasive techniques, or is located within eloquent cortex, ICEEG is frequently employed prior to surgical resection (Van Gompell et al 2008).

Despite the many efforts to characterize ictal onset in ICEEG recordings, a systematic approach to decision making in creation of a resection map is lacking. As a first step toward a comprehensive approach to decision making in resection planning, we describe here a novel model of ictal and interictal patterns significantly associated with the inclusion of cortex underlying a given ICEEG electrode in the surgical resection plan. These results will aid future studies and clinicians in defining a resection plan for epilepsy surgery. 2. Patients and Methods: 2.1 Patients Nineteen patients were retrospectively identified who underwent implantation of intracranial electrodes and subsequently surgical resection of epileptogenic tissue. Patients with only subtemporal electrode strips were excluded due to the relatively small contiguous area of cortex sampled. The study cohort included ten females and nine males, ranging in age from fifteen to fifty one years, with a mean age of twenty nine years. All patients were evaluated in a dedicated Epilepsy Monitoring Unit, and all surgical procedures were performed by a single neurosurgeon (KR). Intracranial electrodes were placed subdurally over suspected ictal foci based on findings from the noninvasive evaluation and consensus plan from surgical planning meetings. Noninvasive evaluation included MRI, scalp video-EEG monitoring, and neuropsychological testing in all patients. MEG, PET, and ictal/interictal SPECT were performed as deemed necessary. Cortical mapping of eloquent language, motor, and sensory cortex was performed with bedside electrical stimulation of intracranial electrodes. Five epileptologists reviewed ICEEG recordings from included patients. Each noted ictal and interictal findings specific to individual electrodes on a preset template to ensure consistent data collection.

Implanted electrode sets included both subdural surface grids and depth electrodes. Seventy six total electrode sets were implanted with an average of four per patient. Thirty eight large surface grids, twenty six strip grids, seven L-shaped “Spencer” grids, and five depth electrodes were implanted. Grids ranged in size from 64-electrode 8x8 grid to an 8-electrode 2x4 grid. 65.7% of strip electrodes were 4-electrode 1x4 strips. The majority of depth electrodes were 6-electrode depths.

2.2 Classification of epileptiform activity Ictal and interictal observations from the ICEEG recordings went into planning the resection of the epileptogenic zone. The parameters used in the design of the classification system were synthesized from the extant literature and clinical experience (Jung et al 1999, Williamson et al 1993, Lee et al 2000). Ictal findings were recorded for the first 15 seconds of ictal activity, split into five, three-second Epochs: Epoch 1 (1-3 seconds), Epoch 2 (4-6 seconds), Epoch 3 (7-9 seconds), Epoch 4 (10-12 seconds), and Epoch 5 (13-15 seconds).

Herald spikes and diffuse dysnchronization at ictal onset were the first parameters recorded. The presence or absence of ictal activity was noted in each Epoch by assessing spike morphology as well as the electrodes involved. Morphology was subdivided into low-voltage fast activity, high amplitude  spikes, rhythmic spikes (, , or ), or rhythmic sinusoidal waves (, , or ) with each electrode recorded as any, but only one of the above morphologies. The location of each electrode was recorded by lobe and sub-lobar location: Frontal (frontal polar, lateral frontal, dorsal frontal, mesial frontal, or orbital frontal), Temporal (mesial temporal, lateral temporal, basal temporal, or temporal polar), Parietal (mesial parietal, anterior parietal, posterior superior

parietal, or posterior inferior parietal), and Occipital (mesial occipital, lateral occipital, basal occipital, or occipital polar).

The first ictal morphology seen was used for each electrode observation. As such, if an electrode showed low-voltage fast activity in Epoch 1 and high amplitude  spikes in Epoch 4, only the observation from Epoch 1 would be recorded. Onset of clinical activity relative to electrographic ictal onset was recorded in seconds; negative if clinical seizure began before EEG onset, positive if after EEG onset, 0 for coincident, or “none” for clinically silent seizure activity.

Interictal findings for each electrode recorded included spikes, paroxysmal fast activity, and pathologic delta waves. Interictal spike activity was further subclassified by spike amplitude and frequency. Spike frequency was divided as follows: continuous (one spike every 1-3 seconds), very frequent (every 3-10 seconds), frequent (every 10-60 seconds), and infrequent (>60 seconds between spikes). Electrodes without ictal or interictal epileptiform activity were not included in the final analysis.

2.3 Other parameters affecting resection Epoch of ictal spread, specific electrodes overlying eloquent cortex (language, motor, or sensory), electrodes overlying a known MRI lesion, and electrodes selected for underlying cortical resection were also recorded.

2.4 Predictors of seizure-free outcomes

The following demographic factors were collected: Gender, race, number of antiepileptics at the time of surgery, number of seizure types, age at seizure onset, age at surgery, and underlying neurologic condition. Operative factors collected included: procedure performed, tissue pathology, intra-operative electrocorticography (ECOG) findings, intra-operative complication, and post-operative complication. Long-term outcomes data included: Engle classification at 3, 6, 12, and 24, and 36 months, number of antiepileptics at each clinic visit, decrease or increase in antiepileptic doses, and pre- and post-operative driving status.

2.5 Statistical Analysis In total 1,331 individual electrodes were included. As each different ictal pattern was classified separately, the total number of unique electrode recordings was 2,589. Each electrode was counted as a separate observation.

A total of 131 seizures with 37 distinct seizure types were identified. When a given patient had multiple seizures with consistent ictal EEG patterns, a single representative ICEEG was analyzed. Nine patients had two or more distinct ictal patterns recorded. In each of these cases a representative ICEEG sample for each ictal pattern was analyzed and weighted according to the fraction of total seizures for that patient. Each patient was given equal weighting in the final analysis regardless of the number of seizures or seizure types observed. Finally, we fit a multivariate logistic regression model of surgical resection risk by electrographic activity adjusted for all covariates under study.

Sample case: Patient A had three electrode sets implanted, consisting of an 8-electrode 2x4 grid, a 6-electrode 1x6 strip, and a 35-electrode “Spencer” grid, for a total of 49 electrodes. This patient had two distinct seizure types recorded, bringing the total number of observations to 98, each of which was weighted by 0.5 in the final analysis. This method of separating and recording data on an electrode-by-electrode and seizure-by-seizure basis provided greater statistical power and allowed detection of multiple significant independent variables.

Lastly, we used bivariate 2 tests to identify associations between covariates of interest and categorical postoperative outcomes. Only associations with P < .05 were taken as significant. SAS 9.2 (SAS Institute, Cary, North Carolina) was used to perform all statistical analyses. The institutional review board of the University of Alabama at Birmingham reviewed this study.

3. Results: Nineteen included patients underwent intracranial EEG monitoring with subdural strips and grids prior to resection. Two patients additionally had depth electrodes implanted. A total of thirteen cortical resections and six temporal lobectomy procedures were included. One patient had a history of traumatic brain injury prior to the onset of seizures, one had seizures that began following an ischemic stroke, and two patients had a history of encephalitis prior to the onset of seizures. Fifteen patients had cryptogenic epilepsy. A minimum twelve-month follow-up was achieved for each patient, with a mean twenty-two month follow-up time. No intra-operative complications were observed for any patients. Six patients had minor postoperative complications, including three transient neurologic deficits. No permanent neurologic deficits occurred.

58% of patients were Engel Class 1 more than twelve months postoperatively. Four patients had returned to driving; all of these patients had Engel 1a seizure control, no residual intra-operative spikes, and no post-operative complications. Relative to cortical resection, temporal lobectomy was not associated with either incidence of residual intra-operative spikes (, p=0.2585) or rate of seizure freedom at greater than twelve months (, p=0.2937). Additionally, presence of residual spikes was not associated with postoperative seizure freedom (, p=0.622). 11 of the 19 included patients had no focal lesion on MRI prior to resection. Of those patients, 7 had Engel Class I after resection, 2 Engel Class 2, 1 Engel Class 3, and 1 Engel Class 4. Of the remaining 8 patients who had abnormal MRIs, 6 had focal imaging abnormalities in the suspected epileptogenic region. 4 of these patients were Engel Class I postoperatively and the remaining 2 were Engel Class 4. The two patients with nonfocal MRI changes were both Engel Class I after resection. We found no significant difference in postoperative seizure freedom rates between patients with and without focal lesions on MRI (p=0.9).

A total of 131 seizures with 37 distinct seizure types were recorded. Nine patients had two or more distinct ictal patterns recorded. Recordings from a total of 1306 electrodes were analyzed. The strongest predictors of resection of cortex underlying a given electrode was the presence of low-voltage fast activity in Epoch 1, rhythmic spikes in Epoch 1, interictal paroxysmal fast activity, and low-voltage fast activity in Epoch 2 (Table 1). High-amplitude beta spikes and rhythmic slow waves were also significant predictors in Epoch 1, but were not significant in later epochs. Interictal spikes had a higher odds ratio of affecting the planned resection if described as “continuous” or “very frequent”, but less frequent spikes were also significant predictors. The

presence of motor or language cortex was the strongest negative predictor of resection of underlying cortex, however eloquent sensory cortex was not found to significantly predict subsequent resection. The presence of a focal MRI lesion under an electrode location was positively associated with resection (i.e. odds ratio >1), but not significantly so. This is presumably due to the fact that most of the patients in our cohort were MRI-negative. Figure 1 demonstrates examples of ictal abnormalities including herald spikes/spike burst, diffuse desynchronization, low voltage fast activity, and rhythmic spike activity. Figure 2 demonstrates interictal frequent high amplitude spikes and paroxysmal fast activity.

4. Discussion: When non-invasive studies are inadequate to proceed directly to surgical resection, ICEEG is frequently employed. ICEEG has the highest spatial and temporal resolution for identification of the ictal onset zone, and remains the gold standard for presurgical workup of epilepsy (Holmes et al 2010). ICEEG also identifies interictal abnormalities, and can be used for functional mapping of underlying cortical function (Van Gompell et al 2008, Spencer et al 2008, Rodin et al 2008, De Curtis et al 2009, Placantonakis et al 2010). Subdural electrode strips or grids are most frequently employed, but epidural, depth, and foramen ovale electrodes are also used, often in combination (Nair et al 2008, Wellmer et al 2012, Bancaud et al 1969, Spencer et al 1990, Sheth et al 2014).

Larger grids provide greater neocortical coverage, but require a larger craniotomy and have higher associated complication rates (MacDougall et al 2009b). While a smaller craniotomy is preferred when the seizure focus is believed to localize within a small area, failure to place

electrodes over the epileptogenic area results in a nondiagnostic study. Repeat, larger craniotomy carries increased complication risk, and is associated with lower long-term seizure freedom (MacDougall et al 2009b). Insertion of strip electrodes through burr holes carries lower complication risk than craniotomy, but allows for less cortical coverage and increases risk for a nondiagnostic study (MacDougall et al 2009).

Both ICEEG seizure onset patterns and extent of resection have been associated with surgical outcomes (Wetjen et al 2009, Kim et al 2010). Resection including all areas with pathologic delta waves and frequent interictal spikes is associated with improved postoperative seizure freedom in extratemporal lobe epilepsy (Kim et al 2010). A strong correlation between localized low-amplitude fast activity (20-60Hz “rapid discharges”) and ictal onset has been identified (Alarcon et al 1995, Allen et al 1992, Fisher et al 1992, Gotman et al 1995, Guggisberg et al 2008, Spencer et al 1992, Wendling et al 2003, Mierlo et al 2013). Intracranial beta-gamma activity also characterizes seizure onset in neocortical ictal onset zones (Bartolomei et al 2001, Francione et al 2003, Jung et al 1999, Lee et al 2000, Salanova et al 1992, Tassi et al 2002, Wendling et al 2003). Surgical resection of underlying cortex correlates with postoperative Engel classification in these patients (Alarcon et al 1995, Gompel et al 2008, Placantonakis et al 2010, Wetjen et al 2009, Palm Desert International Conference 1993). Recently, emerging evidence about high frequency oscillations and infraslow activity are also associated with ictal localization, and may predict surgical outcomes (Dolezalova et al 2013, Modur 2014, Modur et al 2012, Perucca et al 2014). Focal high frequency oscillations (>20 Hz) at seizure onset have been shown to be predictive of long-term seizure freedom in patients with extratemporal lobe epilepsy (Wetjen et al 2009).

Once intracranial electrodes are in place, ICEEG influences the final resection map far more than the previous non-invasive studies. However, the interpretation of ICEEG findings and drawing of the resection map is a complex process that involves assigning a relative weight to the character, timing, and location of the evolving ICEEG ictal pattern in light of interictal recordings, imaging findings, and cortical mapping results. In this study we describe which features among the myriad information provided in the course of intracranial monitoring most consistently guided the epileptologist to include or exclude a patch of cortex in the resection map. This study is purely descriptive, not prescriptive, with the goal of describing how a group of experienced epileptologists interpret a complex data set, but not to prescribe one approach to this task over another. It is also important to note that the choice of where to place intracranial electrodes is an extremely important and complex decision making process, but once the electrodes are implanted, it is the icEEG results including mapping that guide actual location and extent of resection more than the preoperative evaluation, though certainly not to the exclusion of the preoperative data. Due to the inherently limited nature of icEEG sampling, any such study is subject to discovery bias in that we cannot know what icEEG would show in areas where there were no electrodes. However, in each of the cases described in this study there was sufficient sampling to progress to resection, that is none were non-localized after icEEG, and seizure free outcomes were comparable to other top tier epilepsy surgery programs, suggesting that the putative ictal focus was covered in most if not all cases. Nevertheless, a prospective study of resection based solely on ICEEG output compared to resection by clinical analysis would be instrumental in determining the effectiveness of this method.

If the effectiveness of epilepsy surgery is to be studied rigorously, there needs to be a significant degree of uniformity in the process of deciding how to draw the resection map. As a first step toward establishing criteria for translating ICEEG finding into extent of resection, we statistically modeled the practice at an active epilepsy surgery center to assess which findings at a given electrode site were predictive of including the underlying cortex in the resection map. By considering each electrode as a separate observation, the logistic regression model had sufficient power to evaluate numerous factors. In this environment, the presence of low-voltage fast activity in the first 6 seconds, rhythmic spike activity in the first 3 seconds, and interictal paroxysmal fast activity were the strongest predictors that the underlying cortex would be resected. Language and motor findings on cortical stimulation were negative predictors, but sensory cortex was not significantly associated with likelihood of resection. If a multicenter trial studying the efficacy of neocortical epilepsy surgery is undertaken, an analysis such as this at each participating institution could help determine if final resection planning is done in a similar fashion.

One limitation of this study is the investigation of cases using primarily subdural grid and rare depth electrodes, rather than the cases consisting solely of stereotactic implantation of depth electrodes (SEEG). While SEEG has been the principle ICEEG approach in many European centers for decades and has been adopted in recent years by several centers in North America, the subdural grid based approach remains the primary approach employed in the majority of Level 4 Epilepsy Centers in the United States. A grid-based approach carries a distinct advantage for this study, in that the geographic coverage provided by the contiguous grid of electrodes allows for direct assessment of the spatial boundaries of specific ICEEG patterns, while the

targeted sampling of an SEEG approach often leaves several centimeters between electrodes (though not between contacts on a given electrode). Therefore, the precise border of the ictogenic zone often cannot be directly observed with SEEG and can only be inferred to lie somewhere between two adjacent electrodes. While SEEG has distinct advantages in sampling cortex that is not easily accessed by grids, such as the orbital and mesial frontal areas or the insular/opercular cortices, grid based approaches are still appropriate for many patients. Lastly, each of the ICEEG patterns described in this study are also seen in SEEG recordings, though the network-based approach commonly used to interpret SEEG may not rely as heavily on ascertaining distinct geographic boundaries of the ictal onset zone.

5. Conclusions: Standardized methods for creating a resection map are essential for assessing the effectiveness of neocortical epilepsy surgery. In our practice, ictal early low voltage fast activity, ictal rhythmic spikes, and interictal paroxysmal fast activity on ICEEG were strongly associated with ultimate cortical resection. Motor and language cortex were the strongest negative predictors for inclusion in the planned resection, but sensory cortex was not a significant predictor of resection. Further studies may determine which resection map creation strategy results in the greatest rate of postoperative seizure freedom.

Acknowledgements: This manuscript has not been previously published, in whole or in part, and is not under consideration by any other publisher. All authors were involved in the research and writing of this paper. This paper was supported by a grant from the Kaul Foundation. Ethical approval was obtained from the Institutional Review Board at the University of Alabama at Birmingham. Thank you for your consideration of this manuscript for publication in Clinical Neurology and Neurosurgery.

References: [1] Alarcon G, Binnie CD, Elwes RD, Polkey CE. Power spectrum and intracranial EEG patterns at seizure onset in partial epilepsy. Electroencephalogr Clin Neurophysiol. 1995;94(5):326-337. [2] Allen PJ, Fish DR, Smith SJ. Very high-frequency rhythmic activity during SEEG suppression in frontal lobe epilepsy. Electroencephalogr Clin Neurophysiol. 1992;82(2):155-159. [3] Bancaud J, Angelergues R, Bernouilli C, et al. Functional stereotaxic exploration (stereoelectroencephalography) in epilepsies. Rev Neurol (Paris). 1969;120(6):448. [4] Bartolomei F, Wendling F, Bellanger JJ, Régis J, Chauvel P. Neural networks involving the medial temporal structures in temporal lobe epilepsy. Clin Neurophysiol. 2001;112(9):17461760. [5] De Curtis M, Gnatkovsky V. Reevaluating the mechanisms of focal ictogenesis: The role of low-voltage fast activity. Epilepsia. 2009;50(12):2514-2525. doi:10.1111/j.15281167.2009.02249.x. [6] Doležalová I, Brázdil M, Hermanová M, Horáková I, Rektor I, Kuba R. Intracranial EEG seizure onset patterns in unilateral temporal lobe epilepsy and their relationship to other variables. Clin Neurophysiol. 2013;124(6):1079-1088. doi:10.1016/j.clinph.2012.12.046. [7] Engel J., Jr . Principles of epilepsy surgery. In: Shorvon S, Dreifuss F, Fish D, editors. The Treatment of Epilepsy. Oxford: Blackwell; 1996. pp. 519–529. [8] Fisher RS, Webber WR, Lesser RP, Arroyo S, Uematsu S. High-frequency EEG activity at the start of seizures. J Clin Neurophysiol. 1992;9(3):441-448.

[9] Francione S, Vigliano P, Tassi L, et al. Surgery for drug resistant partial epilepsy in children with focal cortical dysplasia: anatomical-clinical correlations and neurophysiological data in 10 patients. J Neurol Neurosurg Psychiatr. 2003;74(11):1493-1501. [10] Gotman J, Levtova V, Olivier A. Frequency of the electroencephalographic discharge in seizures of focal and widespread onset in intracerebral recordings. Epilepsia. 1995;36(7):697-703. [11] Guggisberg AG, Kirsch HE, Mantle MM, Barbaro NM, Nagarajan SS. Fast oscillations associated with interictal spikes localize the epileptogenic zone in patients with partial epilepsy. Neuroimage. 2008;39(2):661-668. doi:10.1016/j.neuroimage.2007.09.036. [12] Holmes MD, Tucker DM, Quiring JM, Hakimian S, Miller JW, Ojemann JG. Comparing noninvasive dense array and intracranial electroencephalography for localization of seizures. Neurosurgery. 2010;66(2):354-362. doi:10.1227/01.NEU.0000363721.06177.07. [13] Jung WY, Pacia SV, Devinsky R. Neocortical temporal lobe epilepsy: intracranial EEG features and surgical outcome. J Clin Neurophysiol. 1999;16(5):419-425. [14] Kim DW, Kim HK, Lee SK, Chu K, Chung CK. Extent of neocortical resection and surgical outcome of epilepsy: intracranial EEG analysis. Epilepsia. 2010;51(6):1010-1017. doi:10.1111/j.1528-1167.2010.02567.x. [15] Lee SA, Spencer DD, Spencer SS. Intracranial EEG seizure-onset patterns in neocortical epilepsy. Epilepsia. 2000;41(3):297-307. [16] Rosenow F, Lüders H. Presurgical evaluation of epilepsy. Brain. 2001;1683-1700. Doi: 10.1093/brain/124.9.1683 [17] MacDougall KW, Burneo JG, McLachlan RS, Steven DA. Outcome of epilepsy surgery in patients investigated with subdural electrodes. Epilepsy Res. 2009;85(2-3):235-242. doi:10.1016/j.eplepsyres.2009.03.014.

[18] MacDougall KW, Steven DA, Parrent AG, Burneo JG. Supplementary implantation of intracranial electrodes in the evaluation for epilepsy surgery. Epilepsy Res. 2009;87(1):95-101. doi:10.1016/j.eplepsyres.2009.07.003. [19] Modur PN, Vitaz TW, Zhang S. Seizure localization using broadband EEG: comparison of conventional frequency activity, high-frequency oscillations, and infraslow activity. J Clin Neurophysiol. 2012;29(4):309-319. doi:10.1097/WNP.0b013e318262435d. [20] Modur PN. High frequency oscillations and infraslow activity in epilepsy. Ann Indian Acad Neurol. 2014;17(Suppl 1):S99-S106. doi:10.4103/0972-2327.128674. [21] Nair DR, Burgess R, McIntyre CC, Lüders H. Chronic subdural electrodes in the management of epilepsy. Clin Neurophysiol. 2008;119(1):11-28. doi:10.1016/j.clinph.2007.09.117. [22] Palm Desert International Conference on the Surgical Treatment of the Epilepsies. Surgical Treatment of the Epilepsies. 2nd ed. (Engel J, ed.). New York: Raven Press; 1993. [23] Perucca P, Dubeau F, Gotman J. Intracranial electroencephalographic seizure-onset patterns: effect of underlying pathology. Brain. 2014;137(Pt 1):183-196. doi:10.1093/brain/awt299. [24] Pittau F, Grouiller F, Spinelli L, Seeck M, Michel CM, Vulliemoz S. The role of functional neuroimaging in pre-surgical epilepsy evaluation. Front Neurol. 2014;5:31. doi:10.3389/fneur.2014.00031. [25] Placantonakis DG, Shariff S, Lafaille F, et al. Bilateral intracranial electrodes for lateralizing intractable epilepsy: efficacy, risk, and outcome. Neurosurgery. 2010;66(2):274-283. doi:10.1227/01.NEU.0000363184.43723.94.

[26] Rathore C, Kesavadas C, Sarma SP, Radhakrishnan K. Usefulness of Wada test in predicting seizure outcome following anterior temporal lobectomy. Epilepsy Res. 2013;107(3):279-285. doi:10.1016/j.eplepsyres.2013.09.003. [27] Rheims S, Jung J, Ryvlin P. Combination of PET and Magnetoencephalography in the Presurgical Assessment of MRI-Negative Epilepsy. Front Neurol. 2013;4:188. doi:10.3389/fneur.2013.00188. [28] Rodin E, Modur P. Ictal intracranial infraslow EEG activity. Clin Neurophysiol. 2008;119(10):2188-2200. doi:10.1016/j.clinph.2008.07.222. [29] Salanova V, Andermann F, Olivier A, Rasmussen T, Quesney LF. Occipital lobe epilepsy: electroclinical manifestations, electrocorticography, cortical stimulation and outcome in 42 patients treated between 1930 and 1991. Surgery of occipital lobe epilepsy. Brain. 1992;115 ( Pt 6):1655-1680. [30] Sheth SA, Aronson JP, Shafi MM, et al. Utility of foramen ovale electrodes in mesial temporal lobe epilepsy. Epilepsia. 2014;55(5):713-724. doi:10.1111/epi.12571. [31] Spencer SS, Guimaraes P, Katz A, Kim J, Spencer D. Morphological patterns of seizures recorded intracranially. Epilepsia. 1992;33(3):537-545. [32] Spencer SS, Spencer DD, Williamson PD, Mattson R. Combined depth and subdural electrode investigation in uncontrolled epilepsy. Neurology. 1990;40(1):74-79. [33] Spencer SS, Goncharova II, Duckrow RB, Novotny EJ, Zaveri HP. Interictal spikes on intracranial recording: behavior, physiology, and implications. Epilepsia. 2008;49(11):18811892. doi:10.1111/j.1528-1167.2008.01641.x. [34] Tassi L, Colombo N, Garbelli R, et al. Focal cortical dysplasia: neuropathological subtypes, EEG, neuroimaging and surgical outcome. Brain. 2002;125(Pt 8):1719-1732.

[35] Van Mierlo P, Carrette E, Hallez H, et al. Ictal-onset localization through connectivity analysis of intracranial EEG signals in patients with refractory epilepsy. Epilepsia. 2013;54(8):1409-1418. doi:10.1111/epi.12206. [36] Van Gompel JJ, Worrell GA, Bell ML, et al. Intracranial electroencephalography with subdural grid electrodes: techniques, complications, and outcomes. Neurosurgery. 2008;63(3):498-505; discussion 505-506. doi:10.1227/01.NEU.0000324996.37228.F8. [37] Van Paesschen W, Dupont P, Sunaert S, Goffin K, Van Laere K. The use of SPECT and PET in routine clinical practice in epilepsy. Curr Opin Neurol. 2007;20(2):194-202. doi:10.1097/WCO.0b013e328042baf6. [38] Wellmer J, von der Groeben F, Klarmann U, et al. Risks and benefits of invasive epilepsy surgery workup with implanted subdural and depth electrodes. Epilepsia. 2012;53(8):1322-1332. doi:10.1111/j.1528-1167.2012.03545.x. [39] Wendling F, Bartolomei F, Bellanger JJ, Bourien J, Chauvel P. Epileptic fast intracerebral EEG activity: evidence for spatial decorrelation at seizure onset. Brain. 2003;126(Pt 6):14491459. [40] Wetjen NM, Marsh WR, Meyer FB, et al. Intracranial electroencephalography seizure onset patterns and surgical outcomes in nonlesional extratemporal epilepsy. J Neurosurg. 2009;110(6):1147-1152. doi:10.3171/2008.8.JNS17643. [41] Williamson PD, French JA, Thadani VM, et al. Characteristics of medial temporal lobe epilepsy: II. Interictal and ictal scalp electroencephalography, neuropsychological testing, neuroimaging, surgical results, and pathology. Ann Neurol. 1993;34(6):781-787. [42] Yuan J, Chen Y, Hirsch E. Intracranial electrodes in the presurgical evaluation. Neurol Sci. 2012; 33:723-729. Doi: 10.1007/s10072-012-1020-2.

Figure Legends Figure 1. Examples of several ictal abnormalities, including herald spikes/spike bursts, diffuse desynchronization, low voltage fast activity, and rhythmic spike activity Figure 2. Examples of interictal abnormalities, including frequent high amplitude spikes and paroxysmal fast activity

Figure 1

Low Voltage Fast activity

Rhythmic Spike Activity

Herald Spikes Diffuse Desynchronization

Figure 2

Paroxysmal Fast Activity

Interictal High Amplitude Spikes

Table 1: Likelihood that a given electrographic finding at a given electrode will associate with resection of underlying cortex Parameter

Beta

Odds

95% Confidence Limits

p-value

0.945

2.747

0.08

Ratio ICTAL FINDINGS Herald Spikes

0.4767

1.611

Diffuse

(-)1.2601* 0.284*

0.2

0.401

<.0001

2.0309*

7.621*

2.861

20.296

<.0001

2.3911*

10.926*

4.965

24.043

<.0001

1.1891*

3.284*

1.283

8.409

0.0132

1.1039*

3.016*

1.21

7.515

0.0178

1.1133*

3.044*

1.667

5.561

0.0003

1.9803*

7.245*

1.772

29.619

0.0058

Desynchronization Seen at Ictal Onset Epoch 1 (sec 1-3) Rhythmic Spike Activity Low-Voltage Fast Activity High Amplitude Beta Spikes Rhythmic Sinusoidal Wave Activity Epoch 2 (sec 4-7) Rhythmic Spike Activity Low-Voltage Fast

Activity High Amplitude

1.3326*

3.791*

1.047

13.725

0.0424

(-)1.3708

0.254

0.029

2.261

0.2192

1.327*

3.77*

1.856

7.659

0.0002

0.7177

2.05

0.855

4.913

0.1076

14.3028

>999.999

<0.001

>999.999

0.9804

Beta Spikes Rhythmic Sinusoidal Wave Activity Epoch 3 (sec 8-10) Rhythmic Spike Activity Low-Voltage Fast Activity High Amplitude Beta Spikes Rhythmic

No occurrence of this activity

Sinusoidal Wave Activity Epoch 4 (sec 11-12) Rhythmic Spike

0.4813

1.618

0.451

5.808

0.4605

0.6296

1.877

0.756

4.657

0.1745

<0.001

>999.999

0.9945

Activity Low-Voltage Fast Activity High Amplitude Beta Spikes

(-)12.2422 <0.001

Rhythmic

0.6966

2.007

0.144

28.062

0.6047

1.1035*

3.015*

1.633

5.564

0.0004

(-)0.3265

0.721

0.008

66.905

0.8876

3.3324

28.006

0.352

>999.999

0.1355

(-)0.4796

0.619

0.098

3.915

0.6103

1.4734*

4.364*

1.774

10.737

0.0013

1.4266*

4.164*

1.879

9.228

0.0004

Frequent (every 10- 0.5825*

1.791*

1.128

2.843

0.0135

Sinusoidal Wave Activity Epoch 5 (sec 13-15) Rhythmic Spike Activity Low-Voltage Fast Activity High Amplitude Beta Spikes Rhythmic Sinusoidal Wave Activity

INTERICTAL FINDINGS Spike Frequency Continuous (every 3 sec) Very Frequent (every 3-10 sec)

60 sec)

Infrequent (>60

0.7761*

2.173*

1.195

3.953

0.011

1.65038

5.208*

2.308

11.752

<.0001

0.4604

1.585

0.93

2.702

0.0907

sec) Paroxysmal Fast Activity Pathologic Delta Waves

CORTICAL MAPPING Electrodes overlying Eloquent Cortex Language

(-)1.6547* 0.191*

0.07

0.519

0.0012

Motor

(-)2.1212* 0.12*

0.04

0.362

0.0002

Sensory

(-)0.3863

0.68

0.266

1.736

0.4195

0.4515

1.571

0.63

3.913

0.3323

IMAGING FINDINGS Electrode Overlying Known Lesion