Clinical Neurophysiology 127 (2016) 179–186

Contents lists available at ScienceDirect

Clinical Neurophysiology

journal homepage: www.elsevier .com/locate /c l inph

High frequency oscillations are less frequent but more specific toepileptogenicity during rapid eye movement sleep

http://dx.doi.org/10.1016/j.clinph.2015.05.0191388-2457/� 2015 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.

⇑ Corresponding author at: Department of Neurosurgery, Tohoku UniversitySchool of Medicine, 1-1 Seiryo-machi, Aoba-ku, Sendai 980-8574, Japan. Tel.: +8122 717 7230; fax: +81 22 717 7233.

E-mail address: iwa@nsg.med.tohoku.ac.jp (M. Iwasaki).

Rie Sakuraba a, Masaki Iwasaki b,⇑, Eiichi Okumura a, Kazutaka Jin a, Yosuke Kakisaka a, Kazuhiro Kato a,Teiji Tominaga b, Nobukazu Nakasato a

a Department of Epileptology, Tohoku University School of Medicine, Sendai, Japanb Department of Neurosurgery, Tohoku University School of Medicine, Sendai, Japan

a r t i c l e i n f o

Article history:Accepted 13 May 2015Available online 30 May 2015

Keywords:High-frequency oscillationsRapid eye movement sleepIntracranial electroencephalographySeizure outcomeEpilepsy surgery

h i g h l i g h t s

� Occurrence of high frequency oscillations (HFOs) is generally suppressed during rapid eye movementsleep (REM) compared with slow wave sleep (SWS).

� Relatively frequent HFOs during REM were associated with area of surgical resection in patient withseizure freedom.

� Weaker suppressive effect of REM on HFOs may provide a specific marker of epileptogenicity.

a b s t r a c t

Objective: We hypothesized that high frequency oscillations (HFOs) are differently suppressed duringrapid eye movement sleep (REM) between epileptogenic and less epileptogenic cortices, and that the sup-pressive effect can serve as a specific marker of epileptogenicity.Methods: Intracranial electroencephalography (EEG) was recorded in 13 patients with drug-resistant epi-lepsy. HFOs between 80 and 200 Hz were semi-automatically detected from total 15-min EEG epochseach for REM and slow wave sleep (SWS). z-Score of HFO occurrence rate was calculated from the base-line rate derived from non-epileptogenic cortex. Intracranial electrodes were labeled as REM dominantHFO (RdH) if REM z-score was greater than SWS z-score or as SWS dominant HFO (SdH) if SWSz-score was greater than REM z-score. Relationship of electrode location to the area of surgical resectionwas compared between RdH and SdH electrodes.Results: Out of 1070 electrodes, 101 were defined as RdH electrodes and 115 as SdH electrodes. RdH elec-trodes were associated with the area of resection in patients with postoperative seizure freedom(P < 0.001), but not in patients without seizure freedom.Conclusions: HFOs near the epileptogenic zone are less suppressed during REM.Significance: The less suppressive effect of REM may provide a specific marker of epileptogenicity.� 2015 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights

reserved.

1. Introduction

High-frequency oscillations (HFOs) are electroencephalography(EEG) markers of epileptogenicity, and can be categorized into rip-ples (80–200 Hz) and fast ripples (>200 Hz) according to their fre-quency. Interictal HFOs are associated with the seizure onset zone(Staba et al., 2002; Jirsch et al., 2006; Urrestarazu et al., 2007;

Jacobs et al., 2008, 2009; Bagshaw et al., 2009; Zijlmans et al.,2009; Crépon et al., 2010; Wang et al., 2013) and removal of thebrain region hosting high rates of interictal HFOs is related to goodseizure outcome after surgery (Jacobs et al., 2010; Akiyama et al.,2011; Zijlmans et al., 2012; Haegelen et al., 2013; Okanishi et al.,2014). In addition, a strong association is supported by theobserved increases in occurrence of both HFOs and seizures follow-ing reduction of medication (Zijlmans et al., 2009).

Interictal HFOs are generated both by epileptogenicity and byphysiological processes related to specific brain functions.Therefore, the diagnosis of epileptogenicity must distinguish

180 R. Sakuraba et al. / Clinical Neurophysiology 127 (2016) 179–186

pathological HFOs from physiological HFOs. For example, hip-pocampal ripples are important in memory consolidation(Girardeau and Zugaro, 2011). Memory processing is associatedwith HFOs distributed over the amygdala, hippocampus and speci-fic neocortical areas in humans (Kucewicz et al., 2014). HFOs canbe evoked by visual stimuli in the occipital cortex (Nagasawaet al., 2012) and by somatosensory stimuli in the somatosensorycortex (Hashimoto, 2000). Fast ripples are more strongly relatedto epileptogenicity than ripples (Bragin et al., 1999a,b; Stabaet al., 2002, 2004; Engel et al., 2003). However, pathologicalHFOs are still difficult to differentiate from physiological HFOs.

The present study investigated the effect of sleep stages onepileptic activities as an indicator for the differentiation of epilep-tic from physiological HFOs. Previous studies showed that bothspikes and HFOs appear less frequently in rapid eye movementsleep (REM) than in slow wave sleep (SWS). Interestingly, epilepticspikes are infrequent during REM, but the area of spikes is morespecific to the primary epileptogenic areas (Lieb et al., 1980;Sammaritano et al., 1991) or hemispheres (Ochi et al., 2011).However, no studies have demonstrated such a difference in HFOdistribution between REM and SWS (Bagshaw et al., 2009). Herewe hypothesized that REM suppresses HFO occurrence, but has dif-ferent effects in epileptogenic and less epileptogenic cortices, sothat evaluation of the sleep-related HFO suppression would be use-ful in the diagnosis of epileptogenicity.

2. Methods

This study was approved by Tohoku University InstitutionalReview Board.

2.1. Patients

This study included 13 consecutive patients, 5 males and 8females aged 12–41 years (mean age 25.4 years), withdrug-resistant epilepsy who underwent extra-operative intracra-nial EEG monitoring for surgical treatment of epilepsy betweenMay 2012 and January 2014 in the Tohoku University HospitalComprehensive Epilepsy Program. All patients were qualified forintracranial electrode implantation after comprehensivepre-surgical evaluation and patient management conference.Consequently, 12 patients underwent resection surgery and onepatient was excluded from surgical treatment. Six patientsachieved seizure freedom at 1 year postoperative follow-up exam-ination (Engel’s classification class I). The clinical characteristics ofthe patients are summarized in Table 1. Preoperative epilepsy diag-nosis included temporal lobe epilepsy in five patients, frontal lobeepilepsy in four, fronto-temporal lobe epilepsy in two, parietal lobeepilepsy in one, and occipital lobe epilepsy in one patient. Sevenpatients underwent surgery on the left. Brain magnetic resonanceimaging (MRI) found no abnormalities in four patients.

2.2. Implantation of intracranial electrodes

Implantation of intracranial electrodes was designed to coverthe probable epileptogenic area and, if necessary, to study the rela-tionship between the epileptogenic area and functional cortex.Electrodes were implanted under general anesthesia, guided by aneuronavigation system (Brainlab�, Brainlab AG, Feldkirchen,Germany). Depth electrodes were inserted using a framelessstereotactic system (VarioGuide™, Brainlab AG).

Depth electrodes consisted of six cylindrical platinum contactsof 1.3-mm length and 1.1-mm diameter (Ad-Tech MedicalInstrument Corporation, Racine, WI). All six contacts were alignedat 10-mm intervals in one type, and the first four contacts were

aligned at 5-mm intervals, the next contact at a 15-mm interval,and the last contact at a 10-mm interval in the other type. All con-tacts were mounted on a 1.1-mm wide flexible plastic probe.Subdural or strip electrodes consisted of platinum discs with3-mm diameter, embedded on a silicone sheet at inter-electrodedistances of 10 mm (Ad-Tech Medical Instrument Corporation).The exposed area of the disc was 2.3 mm in diameter. A previousstudy showed that intracerebral electrodes with contacts between0.2 and 5 mm2 possessed similar HFO detection abilities (Châtillonet al., 2013). Depth and subdural electrodes were treated as havingsimilar recording properties.

The location and number of electrodes are summarized inTable 2. Subdural electrodes were combined with depth electrodesin 11 patients. Two patients with temporal lobe epilepsy under-went bilateral electrode implantation. Medians of 18 depth elec-trodes (range 0–30) and 80 subdural electrodes (range 8–130)were used per patient. A total of 1192 contacts, including 198depth and 994 subdural electrodes, were implanted across allpatients. A total of 114 depth contacts were located in thehippocampus.

A total of 122 contacts (10.2%) were excluded as ‘‘bad elec-trodes’’ from later analysis, so the EEGs were obtained from theremaining 1070 contacts. The bad electrodes included 105 contactsclearly located in the white matter and 17 contacts with physicallybad condition, such as disconnection. Disconnection was judged onraw signal waveforms as no visible EEG or contamination withhigh intensity noise.

2.3. Anatomical location of electrodes and surgical resection

Three-dimensional magnetization prepared rapid gradient echo(3D-MPRAGE) imaging and three-dimensional computed tomogra-phy (3D-CT) were performed before electrode implantation and onday 7 after implantation. These anatomical images were coregis-tered by linear affine transformation. The transformation was com-pleted by maximization of mutual information by changing thecoordinates and angles of the target images using Amira version5 (Visualization Sciences Group, Inc., Burlington, MA). Locationsof electrodes were visualized as metallic artifacts in thepost-implantation CT scans coregistered on the preoperative3D-MRIs. Postoperative 3D-MPRAGE imaging was coregistered tothe preoperative 3D-MPRAGE image and CT scan. The area of sur-gical resection and the relationship to the electrode location weredefined visually on the fusion image. Pre- and postoperativeimages were obtained with a 3-T MRI scanner (MAGNETOM Trio3T, Siemens AG, Munich, Germany) and post-implantation imageswith a 1.5-T MRI scanner (Intera Achiva 1.5T NOVA DUAL, RoyalPhilips, Amsterdam, the Netherlands).

2.4. EEG recording

Intracranial EEG signals were sampled and recorded at 1000 Hz(Neurofax EEG-1200, Nihon-Kohden Co., Tokyo, Japan), simultane-ously with scalp EEG and electromyography (EMG)/electro-oculography for sleep staging. Scalp electrodes wereattached to 21 locations according to the international 10–20 sys-tem together with bilateral surface anterior temporal electrodes(Silverman, 1960). Two electrodes were attached to the mentalismuscle for EMG. Two electrodes were attached to the upper leftto the left eye and to the lower right to the right eye, respectively,for electro-oculography.

2.5. EEG samples for HFO analysis

Video-EEG monitoring was performed for 14 days in 11 patientsand for 21 and 26 days in the other two patients (Cases 2 and 13).

Table 1Clinical characteristics of 13 patients.

Case Sex/age Epilepsydiagnosis

Scalp EEGlocalization

MRI FDG-PEThypometabolism

IntracranialEEG seizureonset zone

Surgery Pathology Seizure outcomeat 1 year(Engel’sclassification)

Interictalspikes

Ictalonset

Patients with seizure freedom1 F/15 Lt TLE Lt T Lt hemi Lt HA, Lt ITG FCD,

Lt STG FCDLt T Lt mT Lt ATL and

removal of LtITG FCD

HS, FCD type I I

2 M/35 Lt TLE Blt aT Rt T Normal Normal Lt Hip Lt ATL HS, FCD type I I3 F/27 Lt TLE Lt T Lt T Lt HA, Lt T-P scar Lt T-P Lt mT Lt extended T

lobectomyHS, gliosis I

4 M/41 Rt FLE Rt F-C Rt C Rt F FCD None Lesion Lesionectomy FCD type II I5 M/20 Lt FLE Lt aT Lt hemi Normal None Lt orbF Resection of

Lt orbF cortexMicrodysgenesis I

6 F/14 Rt OLE Rt hemi Rt O Multiple tubers Rt O Lesion Lesionectomyof Rt O tuber

Cortical tuber I

Patients without seizure freedom7 F/26 Lt TLE Lt F-T Lt T Normal Lt mT Lt mT Lt ATL Microdysgenesis IV8 F/32 Lt TLE Blt T Blt T Lt mT tumor Lt mT Lt Hip Lt ATL Ganglioglioma III9 F/35 Rt FTLE Blt F-T Rt T Normal None Rt Hip Rt ATL,

resection ofRt orbF cortex

Gliosis III

10 F/12 Lt FLE Lt aT, Lt F Lt hemi Lt F atrophy None Lt F-P Lt Flobectomy,resection ofpost-centralgyrus andinsula

Normal III

11 M/16 Rt PLE Rt T, Vertex Rt hemi Rt P atrophy Rt P Rt mP Resection ofRt P cortex

FCD type I III

12 M/22 Rt FTLE Rt F, Rt T Rt hemi Rt T and orbF scar Rt T-P Diffuse Rt ATL Gliosis IV13 F/35 Lt FLE Lt F F Vertex Lt F scar Lt F / No surgery /

TLE = temporal lobe epilepsy; FLE = frontal lobe epilepsy; OLE = occipital lobe epilepsy; FTLE = fronto-temporal lobe epilepsy; PLE = parietal lobe epilepsy; hemi = hemi-sphere; T = temporal; Blt = bilateral; a = anterior; F = frontal; C = central; O = occipital; HA = hippocampal atrophy; ITG = inferior temporal gyrus; FCD = focal cortical dys-plasia; STG = superior temporal gyrus; P = parietal; m = medial; orb = orbito; Hip = hippocampus; ATL = anterior temporal lobectomy with amygdalohippocampectomy;HS = hippocampal sclerosis.

Table 2Number of depth and subdural electrodes and relationship to surgical resection.

Case Depth electrodes Subdural electrodes Total Electrodes inside resection Electrodes outside resection

Analyzed Bad Location Analyzed Bad Location

Patients with seizure freedom1 13 11 Hip, FCD 79 1 blT, latPO 104 31 612 15 9 Bil Hip 55 3 Bil blT, Rt latF 82 20 503 4 8 Hip 66 0 blT, latP 78 46 244 9 3 FCD 60 0 mFP, latFP 72 12 575 10 8 OrbF 102 0 mF, bF, latF 120 22 906 0 0 / 129 1 bO, mFPO, latTPO 130 18 111

Patients without seizure freedom7 5 7 Hip 16 6 blT 34 15 68 8 10 Lt Hip, Tumor 52 0 Bil blT 70 19 419 6 12 Hip, OrbF 120 2 blT, latFP, bF, mFP 140 24 102

10 8 4 Hip 122 0 blT, latFP, bF, mFP 134 47 8311 4 14 Cingulate gyrus 84 0 latTPO, mFP 102 20 6812 12 18 OrbF, mF, Hip 8 0 latFT 38 11 913 0 0 / 83 5 lat FP, mFP, bF 88 / /

Hip = hippocampus; FCD = focal cortical dysplasia; Bil = bilateral; OrbF = orbito-frontal; m = medial; F = frontal; blT = basal lateral temporal; lat = lateral; P = parietal;O = occipital; b = basal; T = temporal.

R. Sakuraba et al. / Clinical Neurophysiology 127 (2016) 179–186 181

Simultaneous scalp EEG was recorded between days 7 and 9 (threenights) after electrode implantation in all except one patient (Case13) in whom the EEG was recorded between days 16 and 18 (threenights). Five-minute EEG segments were clipped and stored foreach sleep stage from each night for offline HFO analysis(Zelmann et al., 2009). Data recorded within 2 h of seizures were

excluded from sampling. Sleep staging was performed using thescalp EEG recordings based on the 2007 American Academy ofSleep Medicine Manual for the Scoring of Sleep and AssociatedEvents (Iber et al., 2007). The segments from REM and non-rapideye movement stage 3 sleep were used for later analysis.Non-rapid eye movement stage 3 sleep, or SWS, was defined on

182 R. Sakuraba et al. / Clinical Neurophysiology 127 (2016) 179–186

the scalp EEG recordings as epochs containing at least 20% deltawaves. REM was defined by epochs containing low-amplitudemixed-frequency EEG, low chin EMG tone, and rapid eyemovement.

2.6. Semi-automated detection of HFO

Detection of HFOs was performed offline using a custom-madeapplication operating under MATLAB version 2013a (Math Works,Natick, MA). EEG signals were referenced to the averaged signal.Recorded signals were filtered between 80 and 200 Hz (in bothcases, a finite impulse response filter of order 32 was used). Thesignal envelope was calculated using a Hilbert transformation(Fig. 1). Then, the local maxima were automatically detected usinga threshold set to 5 standard deviations of the baseline (Staba et al.,2002). The baseline was manually defined from the selected EEGsegments of at least 1 s/min with no outstanding oscillatory activ-ities. HFOs were defined by events containing at least six consecu-tive oscillations (Staba et al., 2002). Events less than 15 ms apartwere considered as a single event. Detected events were markedon the EEG waveforms (Fig. 2). False positive events were assessedvisually in retrospect and rejected if necessary.

2.7. REM/SWS dominant HFO

Relative occurrence rate of HFOs was calculated and comparedbetween REM and SWS. For this calculation, the baseline occur-rence rate of HFOs was first obtained from electrodes located inthe non-epileptogenic zone for each sleep stage. In this study,the non-epileptogenic zone was defined as areas outside the resec-tion, anatomical epileptogenic lesion, eloquent cortex, hippocam-pus, and interictal spike zone in patients with seizure freedomafter surgery. Physiological HFOs are more detectable in eloquentcortex and hippocampus (Hashimoto, 2000; Girardeau andZugaro, 2011; Kucewicz et al., 2014). A total of 196 electrodes werelocated in the non-epileptogenic zone in four patients with postop-erative seizure freedom (40–63 electrodes per patient). The othertwo seizure-free patients had no electrodes located within thenon-epileptogenic zone. Then, z score of the HFO occurrence ratewas calculated from the baseline for each electrode for each sleepstage. Electrodes were categorized into two groups according tothe z scores during REM and SWS. ‘‘REM dominant HFO’’ (RdH)electrode was defined as an electrode with REM z score of higherthan 2 as well as higher than the SWS z score. Similarly, ‘‘SWS

Fig. 1. Automated detection of HFOs. Band-passed EEG waveform at 80–200 Hz (A)and its envelope after Hilbert transformation (B). Local maxima (red circles) withamplitude more than 5 times standard deviation of the baseline (red line) wereautomatically detected. Only events containing at least six consecutive oscillationsin the band-passed EEG were defined as HFOs (oscillatory event between greenlines). (For interpretation of the references to color in this figure legend, the readeris referred to the web version of this article.)

dominant HFO’’ (SdH) electrode was defined as an electrode withSWS z score of higher than 2 as well as higher than the REM zscore.

2.8. Statistical analysis

The occurrence rate of HFOs per minute was compared betweenREM and SWS with Wilcoxon signed-rank test. Repeated measuresANOVA was used to examine the main effects of sleep stages (REMor SWS) and patients, and the interaction between the two factorson the occurrence rate. The level of significance was set at 0.001.

The relationship to the area of surgical resection was comparedbetween RdH and SdH electrodes in patients with postoperativeseizure freedom (n = 6), with Fisher’s exact test. The level of signif-icance was set at 0.001. This relationship was also examined inpatients who did not achieve postoperative seizure freedom(n = 6).

3. Results

A total of 22,158 HFOs were identified from 769 electrodes dur-ing SWS (125 to 9746 HFOs per patient), and 5396 HFOs were iden-tified from 453 electrodes during REM (15 to 1286 HFOs perpatient). A total of 285 electrodes were located within the resec-tion, of which 240 and 195 electrodes detected at least one HFOevent during SWS and REM, respectively. One patient (Case 13)was excluded from surgical treatment because no EEG changeswere associated with her seizures. This patient was excluded fromthe later analysis on the relationship of HFOs to surgical resection.

3.1. Occurrence rate of HFOs during REM and SWS

The occurrence rate of HFOs per electrode was significantlylower during REM (mean 0.3/min, range 0.0–11.7/min) than duringSWS (mean 1.4/min, range 0.0–47.4/min, P < 0.0001, Wilcoxon test,see Fig. 2). The occurrence rate per patient was lower during REMin all patients (Supplementary Table S1). Two-way repeatedANOVA showed the main effect of sleep stages, in which the occur-rence rate was lower during REM than SWS, and the interaction, inwhich the occurrence rates were different between patients. Thebaseline occurrence rate from non-epileptogenic cortex was0.04 ± 0.22/min during REM and 0.26 ± 0.78/min during SWS(P < 0.0001, Wilcoxon test). Two-way repeated ANOVA showedthe main effect of sleep stages and interaction by patients on theoccurrence rate.

The z score of the HFO occurrence rate was 1.37 ± 4.56 duringREM (�0.19 to 54.40) and 1.43 ± 4.87 during SWS (-0.34 to60.32). The z score was higher than 2 in 159 electrodes duringREM (1 to 43 electrodes per patient) (Supplementary Table S2). Atotal of 101 electrodes (9.4%) were defined as RdH electrodes (0to 21 electrodes per patient), and 62 were present in six patientswith postoperative seizure freedom. The z score was higher than2 in 174 electrodes during SWS (0 to 57 electrodes per patient)(Supplementary Table S2). A total of 115 electrodes (10.7%) weredefined as SdH electrodes (0 to 49 electrodes per patient), and 87were present in six patients with postoperative seizure freedom.The remaining 854 electrodes (79.8%) were not defined as RdH orSdH electrodes. The z score was 2 or smaller during both REMand SWS in those electrodes.

3.2. Association of RdH/SdH electrodes with surgical resection

Table 3 summarizes the relationships between RdH and SdHelectrodes and the area of surgical resection. In six patients withpostoperative seizure freedom, 44 and 18 RdH electrodes, and 36

Fig. 2. Detection of HFOs in a patient with postoperative seizure freedom (Case 2). Preoperative MRI found no abnormalities, and combined subdural and depth electrodeswere implanted in the bilateral temporal lobes and hippocampi. Ictal EEG revealed seizure onset from the left hippocampus. Left anterior temporal lobectomy andamygdalohippocampectomy were performed, and the patient became seizure-free for 2 years. Upper figure shows anatomical fusion images of 3D-MPRAGE and post-implantation CT. Intracranial electrodes are shown in green, and RdH and SdH electrodes are labeled with red stars and blue circles, respectively. RdH electrodes aredistributed more specifically in the area of surgical resection (shaded with orange) compared to SdH electrodes. Lower figure shows semi-automated detection of HFOs with acustom application. Band-passed filtered EEG waveforms at 80–200 Hz are shown for SWS (left) and REM (right). Detected HFO events are marked with pink boxes. Expandedimage of the event marking is also shown in the boxed figure. EEGs included in the surgical resection are marked with dotted orange lines. HFO events are more frequent anddistributed in SWS than in REM. Note that relative frequency, not absolute frequency, of HFOs to the baseline (z score) was compared between REM and SWS in this study.

R. Sakuraba et al. / Clinical Neurophysiology 127 (2016) 179–186 183

and 51 SdH electrodes were located inside and outside the resec-tion, respectively. RdH electrodes were associated with the areaof resection (P < 0.001, Fisher’s exact test, Figs. 2 and 3). In thesix patients without postoperative seizure freedom, 19 and 20

RdH electrodes, and 15 and 13 SdH electrodes were located insideand outside resection, respectively. No significant association wasfound between RdH electrodes and surgical resection (Figs. 4and 5).

Table 3Number of RdH and SdH electrodes and relationship to the area of resection.

Case RdH electrodes (n = 101) SdH electrodes (n = 115)

Insideresection

Outsideresection

Insideresection

Outsideresection

Patients with seizure freedom1 13 1 6 12 10 3 4 63 18 3 7 04 0 0 7 65 2 0 0 16 1 11 12 37

Patients without seizure freedom7 0 0 1 08 4 2 4 19 2 7 0 3

10 5 8 10 611 0 0 0 312 8 3 0 013 No surgery

184 R. Sakuraba et al. / Clinical Neurophysiology 127 (2016) 179–186

4. Discussion

The present study revealed that REM dominant HFOs may be avaluable marker of epileptogenicity. The location of electrodesdetecting relatively frequent HFOs during REM was associatedwith the area of surgical resection in patients with postoperativeseizure freedom. In this study, the HFO occurrence rate ofeach sleep stage was normalized to the baseline of thenon-epileptogenic cortex, and relative changes in the occurrencerate were evaluated between REM and SWS. The present resultsshow that the influence of sleep stages on HFO occurrence was dif-ferent for the epileptogenic and less epileptogenic cortices. HFOsare generally suppressed during REM, but the suppressive effectis weaker near the epileptogenic zone (Fig. 6). Therefore, the sleepeffect must be considered in distinguishing pathological fromphysiological HFOs.

4.1. Influence of sleep on occurrence of HFOs

This study found that the occurrence rate of HFOs was lowerduring REM than during SWS, in accordance with previous studiesin human using depth microelectrodes and macroelectrodes (Stabaet al., 2004; Bagshaw et al., 2009). SWS is associated with the max-imum occurrence rate of HFOs. Alternating periods of neuronalquiescence and intense discharge within the cortico-thalamic

Fig. 3. Association of RdH and SdH electrodes to the area of surgical resection inpatients with postoperative seizure freedom (electrode n = 149, patient n = 6). RdHelectrodes are associated with the area of resection compared with SdH electrodes(P < 0.001).

networks generate cyclic hyperpolarizing-depolarizing fluctua-tions in membrane potential, associated with slow sleep oscillationduring SWS (Steriade et al., 1993, 2001). This pattern of activitydecreases the temporal variability between neuronal dischargesand synchronizes the timing of postsynaptic inputs, increasingthe probability of activating the hippocampal neurons that gener-ate ripple activity (Staba et al., 2004). In contrast, rhythmic pat-terns of neuronal discharge are replaced by tonic firing patternsduring REM (Steriade et al., 1993), which would decrease the syn-chrony and increased variability of inputs to the hippocampus, soreducing ripple activity (Staba et al., 2004).

4.2. Sleep and differentiation of pathological from physiological HFOs

Discrimination of pathological from physiological HFOs is nec-essary for reliable identification of the epileptogenic zone and sur-gical decision making. This study focused on the effect of sleep onHFO generation. Most previous studies performed HFO analysisduring SWS (Jacobs et al., 2008, 2009, 2010; Haegelen et al.,2013) or non-rapid eye movement sleep (Bragin et al., 2002;Akiyama et al., 2011; Okanishi et al., 2014), because HFOs are fre-quent and easy to detect. Currently, there is no doubt about therelationship between HFO generation and epileptogenicity(Jacobs et al., 2008, 2010; Haegelen et al., 2013; Okanishi et al.,2014). However, HFOs are generated not only by epileptogenicmechanisms, but also by physiological processes.

This study showed that the effect of sleep on the occurrence ofpathological HFOs was different from that of physiological HFOs.Previous studies on non-epileptic animals have shown that rippleswere present in the neocortex during the depolarizing phase ofslow oscillation under ketamine-xylazine anesthesia as well asduring natural SWS. Ripples are also present, but less pronounced,during the activated states of waking and REM (Grenier et al.,2001). On the other hand, a significant number of epileptic HFOsco-occur with epileptic spikes and fast waves (Wang et al., 2013).Interictal epileptic spikes become infrequent (Sammaritano et al.,1991; Malow et al., 1997, 1998), but the spiking area is localizedin the primary epileptogenic areas during REM in patients withpartial epilepsy (Lieb et al., 1980; Ochi et al., 2011). Although theeffect of sleep stage on epileptic HFOs is not clearly understood,it is possible that physiological HFOs are well suppressed butepileptic HFOs are relatively maintained together with epilepticspikes and localized to the epileptogenic zone during REM.

4.3. Association of RdHs and epileptogenic area

RdHs can serve as a valuable marker of the epileptogenic zone.In this study, RdH electrodes were associated with the area ofresection in patients with seizure freedom, compared with SdHelectrodes. It is important to note that the true extent of the epilep-togenic zone was not known. On the contrary, the area of resectionwas known to include the epileptogenic zone and seizure onsetzone, as indicated by the postoperative seizure freedom. The speci-ficity of RdHs to the epileptogenic zone should be furtherinvestigated.

Many previous studies have compared the occurrence of HFOswith the seizure onset zone (Staba et al., 2002; Jirsch et al., 2006;Urrestarazu et al., 2007; Jacobs et al., 2008, 2009; Bagshaw et al.,2009; Zijlmans et al., 2009; Crépon et al., 2010; Wang et al.,2013). However, the reliability of the seizure onset zone as a mar-ker of epileptogenic zone largely depends on electrode coverageand postoperative seizure outcome. In practice, considerable num-bers of HFOs are also found outside the area of resection as well asthe seizure onset zone. Quantitative aspects of HFOs, such asamplitude, frequency, and duration, should be investigated to

Fig. 4. The distribution of RdH and SdH electrodes in a patient without postoperative seizure freedom (Case 10). Compared with Fig. 2, majority of RdH electrodes werelocated outside the resection.

Fig. 5. Association of RdH and SdH electrodes to the area of surgical resection inpatients without postoperative seizure freedom (electrode n = 67, patient n = 6).There is no association of RdH electrodes with the surgical resection.

Fig. 6. Schematic explanation of the influence of sleep stages on the occurrence ofHFOs. Detected HFOs comprise of pathological and physiological HFOs, althoughone-by-one differentiation is currently impossible and was not attempted in thisstudy. In normal cortex, physiological HFOs are suppressed during REM.Physiological HFOs are usually infrequent (rightmost), but can be relativelyfrequent in the functional cortex such as the primary sensori-motor cortex andhippocampus (second right). The irritative cortex presents with frequent epilepticHFOs during SWS, which are well suppressed during REM. The suppressive effect isweaker in the epileptogenic cortex, where relatively frequent HFOs are observedduring REM.

R. Sakuraba et al. / Clinical Neurophysiology 127 (2016) 179–186 185

measure the degree of epileptogenicity, before HFOs can be appliedas a diagnostic tool in epilepsy surgery.

A previous study of the effect of sleep stages on HFO rates foundthat HFO rates were higher in the seizure onset zone than in thenon-seizure onset zone, but the relationship was similar in bothsleep stages (Bagshaw et al., 2009). The EEG was recorded withfour to nine depth electrode leads, implanted mostly into the hip-pocampus and amygdala. HFO occurrence rates were compared atthe group level between the inside and outside of the seizure onsetzone, although the seizure onset zone was not necessarily sup-ported by postoperative seizure freedom. The present studyincluded patients with various types of epilepsies and wide corticalcoverage was achieved with subdural electrodes. The occurrencerate of HFOs was decreased at most electrodes regardless of theepileptogenic zone during REM as shown previously (Bagshawet al., 2009). However, the present study compared HFO rates atthe group level, and investigated the suppressive effect of REMon HFO generation at the electrode level. A minority of electrodeswere characterized as RdH or SdH electrodes, and these electrodeswere associated with the area of surgical resection in patients withseizure freedom.

Both RdH and SdH are epileptogenic, but RdH is more epilepto-genic than SdH. It is of note that RdH does not pinpoint the epilep-togenic zone. As seen in our results and previous studies, HFOsoccur not only in epileptogenic or seizure-onset zone, but also inthe surrounding ‘‘less-epileptogenic’’ area. However, the degreeof epileptogenicity may not be linear to the distance from theepileptogenic zone, and subdural electrodes do not cover corticalsurface equally. Therefore, RdH and SdH electrodes were mixedin the area of resection, and some were even located outside theresection in this study.

4.4. Limitations

This study has several limitations. The analysis was limited tothe ripple band between 80 and 200 Hz, and fast ripples werenot investigated because of 1000 Hz sampling, whereas fast ripplesare more relevant to epileptogenicity than ripples (Bragin et al.,1999a,b; Staba et al., 2002; Engel et al., 2003). Analysis on theeffect of sleep on HFOs of higher frequency range is required.

False positive or false negative detection of HFOs might occurwith the semi-automated detection procedure used in this study.However, semi-automated detection is essential for analyzinglarge amounts of data. In this study, 1192 contacts are analyzedfor 5 min during the REM and SWS in 13 patients. Several differentautomated detection systems have been proposed (Zelmann et al.,2012). The sensitivity and specificity of HFO detection largelydepend on the parameters and types of detector. This study usedrelatively strict parameters, i.e. thresholds for amplitude, number

186 R. Sakuraba et al. / Clinical Neurophysiology 127 (2016) 179–186

of oscillations, and inter-event interval, to minimize false positivedetection.

The influence of sleep on HFO generation might differ betweenanatomical regions, for example, hippocampus versus neocortex.These areas have special properties of HFOs (Hashimoto, 2000;Girardeau and Zugaro, 2011; Nagasawa et al., 2012; Kucewiczet al., 2014). Anatomical position of electrode should also be con-sidered to improve diagnostic precision. Investigation on theanatomical difference in sleep effect is currently underway.

The occurrence of HFOs was not compared to epileptic spikes inthis study. This investigation is important since spikes are also themarker of epileptogenicity, and HFOs can co-occur with spikes.HFOs co-occurring with epileptic spikes may be more relevant tothe epileptogenic zone than only HFOs (Wang et al., 2013).

5. Conclusions

REM dominant HFOs may be a valuable marker of epilepto-genicity. HFOs are generally suppressed during REM, but the influ-ence of sleep stages on HFO occurrence was different for theepileptogenic and less epileptogenic cortices. The location of elec-trodes hosting relatively frequent HFOs during REM was associatedwith the area of surgical resection in patients with postoperativeseizure freedom. The suppressive effect of REM on HFOs is weakernear the epileptogenic zone.

Acknowledgements

This research was partially supported by a Grant-in-Aid forScientific Research (No. 25462240) from the Japan Society for thePromotion of Science. The author MI was supported by theTakeda Science Foundation. We wish to thank Dr. YoshiyukiNishio for his comments and advice on the research approach.

Conflict of interest: Authors have no conflict of interest todisclose.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.clinph.2015.05.019.

References

Akiyama T, McCoy B, Go CY, Ochi A, Elliott IM, Akiyama M, et al. Focal resection offast ripples on extraoperative intracranial EEG improves seizure outcome inpediatric epilepsy. Epilepsia 2011;52:1802–11.

Bagshaw AP, Jacobs J, Levan P, Dubeau F, Gotman J. Effect of sleep stage on interictalhigh-frequency oscillations recorded from depth macroelectrodes in patientswith focal epilepsy. Epilepsia 2009;50:617–28.

Bragin A, Engel J, Wilson C, Fried I, Buzsáki G. High-frequency oscillations in humanbrain. Hippocampus 1999a;9:137–42.

Bragin A, Engel J, Wilson CL, Fried I, Mathern GW. Hippocampal and entorhinalcortex high-frequency oscillations (100–500 Hz) in human epileptic brain andin kainic acid-treated rats with chronic seizures. Epilepsia 1999b;40:127–37.

Bragin A, Wilson CL, Staba RJ, Reddick M, Fried I, Engel J. Interictal high-frequencyoscillations (80–500 Hz) in the human epileptic brain: entorhinal cortex. AnnNeurol 2002;52:407–15.

Châtillon CE, Zelmann R, Hall JA, Olivier A, Dubeau F, Gotman J. Influence of contactsize on the detection of HFOs in human intracerebral EEG recordings. ClinNeurophysiol 2013;124:1541–6.

Crépon B, Navarro V, Hasboun D, Clemenceau S, Martinerie J, Baulac M, et al.Mapping interictal oscillations greater than 200 Hz recorded with intracranialmacroelectrodes in human epilepsy. Brain 2010;133:33–45.

Engel Jr J, Wilson C, Bragin A. Advances in understanding the process ofepileptogenesis based on patient material: what can the patient tell us?Epilepsia 2003;44(Suppl. 12):60–71.

Girardeau G, Zugaro M. Hippocampal ripples and memory consolidation. Curr OpinNeurobiol 2011;21:452–9.

Grenier F, Timofeev I, Steriade M. Focal synchronization of ripples (80–200 Hz) inneocortex and their neuronal correlates. J Neurophysiol 2001;86:1884–98.

Haegelen C, Perucca P, Châtillon CE, Andrade-Valença L, Zelmann R, Jacobs J, et al.High-frequency oscillations, extent of surgical resection, and surgical outcomein drug-resistant focal epilepsy. Epilepsia 2013;54:848–57.

Hashimoto I. High-frequency oscillations of somatosensory evoked potentials andfields. J Clin Neurophysiol 2000;17:309–20.

Jacobs J, LeVan P, Chander R, Hall J, Dubeau F, Gotman J. Interictal high-frequencyoscillations (80–500 Hz) are an indicator of seizure onset areas independent ofspikes in the human epileptic brain. Epilepsia 2008;49:1893–907.

Jacobs J, Levan P, Chtillon CD, Olivier A, Dubeau F, Gotman J. High frequencyoscillations in intracranial EEGs mark epileptogenicity rather than lesion type.Brain 2009;132:1022–37.

Jacobs J, Zijlmans M, Zelmann R, Chatillon CÉ, Hall J, Olivier A, et al. High-frequencyelectroencephalographic oscillations correlate with outcome of epilepsysurgery. Ann Neurol 2010;67:209–20.

Jirsch JD, Urrestarazu E, LeVan P, Olivier A, Dubeau F, Gotman J. High-frequencyoscillations during human focal seizures. Brain 2006;129:1593–608.

Kucewicz MT, Cimbalnik J, Matsumoto JY, Brinkmann BH, Bower MR, Vasoli V, et al.High frequency oscillations are associated with cognitive processing in humanrecognition memory. Brain 2014;137:2231–44.

Lieb JP, Joseph JP, Engel J, Walker J, Crandall PH. Sleep state and seizure foci relatedto depth spike activity in patients with temporal lobe epilepsy.Electroencephalogr Clin Neurophysiol 1980;49:538–57.

Malow BA, Kushwaha R, Lin X, Morton KJ, Aldrich MS. Relationship of interictalepileptiform discharges to sleep depth in partial epilepsy. ElectroencephalogrClin Neurophysiol 1997;102:20–6.

Malow BA, Lin X, Kushwaha R, Aldrich MS. Interictal spiking increases with sleepdepth in temporal lobe epilepsy. Epilepsia 1998;39:1309–16.

Nagasawa T, Juhász C, Rothermel R, Hoechstetter K, Sood S, Asano E. Spontaneousand visually driven high-frequency oscillations in the occipital cortex:Intracranial recording in epileptic patients. Hum Brain Mapp 2012;33:569–83.

Ochi A, Hung R, Weiss S, Widjaja E, To T, Nawa Y, et al. Lateralized interictalepileptiform discharges during rapid eye movement sleep correlate withepileptogenic hemisphere in children with intractable epilepsy secondary totuberous sclerosis complex. Epilepsia 2011;52:1986–94.

Okanishi T, Akiyama T, Tanaka S, Mayo E, Mitsutake A, Boelman C, et al. Interictalhigh frequency oscillations correlating with seizure outcome in patients withwidespread epileptic networks in tuberous sclerosis complex. Epilepsia2014;55:1602–10.

Sammaritano M, Gigli GL, Gotman J. Interictal spiking during wakefulness and sleepand the localization of foci in temporal lobe epilepsy. Neurology1991;41:290–7.

Silverman D. The anterior temporal electrode and the ten-twenty system.Electroencephalogr Clin Neurophysiol 1960;12:735–7.

Staba RJ, Wilson CL, Bragin A, Fried I, Kucewicz MT, Cimbalnik J, et al. Quantitativeanalysis of high-frequency oscillations (80–500 Hz) recorded in humanepileptic hippocampus and entorhinal cortex. J Neurophysiol 2002;88:1743–52.

Staba RJ, Wilson CL, Bragin A, Jhung D, Fried I, Engel Jr J. High-frequency oscillationsrecorded in human medial temporal lobe during sleep. Ann Neurol2004;56:108–15.

Steriade M, McCormick DA, Sejnowski TJ. Thalamocortical oscillations in thesleeping and aroused brain. Science 1993;262:679–85.

Steriade M, Timofeev I, Grenier F. Natural waking and sleep states: a view frominside neocortical neurons. J Neurophysiol 2001;85:1969–85.

Urrestarazu E, Chander R, Dubeau F, Gotman J. Interictal high-frequency oscillations(10–500 Hz) in the intracerebral EEG of epileptic patients. Brain2007;130:2354–66.

Wang S, Wang IZ, Bulacio JC, Mosher JC, Gonzalez-Martinez J, Alexopoulos AV, et al.Ripple classification helps to localize the seizure-onset zone in neocorticalepilepsy. Epilepsia 2013;54:370–6.

Iber C, Israel S, Chesson AL, Quan SF. The AASM Manual for the Scoring of Sleep andAssociated Events. Rules, Terminology and Technical Specifications.Westchester, IL: American Association of Sleep Medicine; 2007.

Zelmann R, Zijlmans M, Jacobs J, Châtillon CE, Gotman J. Improving theidentification of High Frequency Oscillations. Clin Neurophysiol2009;120:1457–64.

Zelmann R, Mari F, Jacobs J, Zijlmans M, Dubeau F, Gotman J. A comparison betweendetectors of high frequency oscillations. Clin Neurophysiol 2012;123:106–16.

Zijlmans M, Jacobs J, Zelmann R, Dubeau F, Gotman J. High-frequency oscillationsmirror disease activity in patients with epilepsy. Neurology 2009;72:979–86.

Zijlmans M, Jiruska P, Zelmann R, Leijten FSS, Jefferys JGR, Gotman J. High-frequencyoscillations as a new biomarker in epilepsy. Ann Neurol 2012;71:169–78.