Among the advantages digital EEG has provided over the formerly. Peri-Ictal and Interictal, Intracranial Infraslow Activity INVITED REVIEW

Transcription

1 INVITED REVIEW Peri-Ictal and Interictal, Intracranial Infraslow Activity Tawnya Constantino* and Ernst Rodin Summary: It is widely assumed that the recording of EEG infraslow activity (ISA) requires direct current amplifiers. Yet, it has been shown during the past decade that conventional EEG systems can record activity between 0.01 and 0.1 Hz and that this frequency band contains additional information, especially in regard to seizure onset. To delineate the characteristics of background ISA, 24-hour scalp and intracranial recordings obtained from 5 patients during long-term monitoring were investigated. Scalp recordings had been sampled but intracranial tracings were continuous over periods of up to 3 days. Although scalp recordings were subject to artifact, intermittent genuine ISA increase could occur episodically in the interictal state not only distant from the ictal onset zone but also in the contralateral hemisphere. Intracranial recordings were limited, in this study, to portions of one hemisphere but likewise revealed that distant areas could show ISA increase lasting minutes or hours. This was at times not detectable when only the conventional frequency band was viewed. It is concluded that ISA information is contained in the clinical setting of intensive video-monitoring studies for the detection of the epileptogenic zone and can provide additional information above and beyond what is seen in the conventional frequencies. Key Words: EEG epilepsy monitoring, Infraslow preictal, Infraslow interictal, Infraslow sleep recording, Seizure prediction. (J Clin Neurophysiol 2012;29: ) Among the advantages digital EEG has provided over the formerly used analog EEG systems is the ability for post hoc changes of the acquired data. Furthermore, because digital systems no longer have a fixed time constant of 0.3 seconds, slower frequencies are routinely recorded and these can be recovered during data analysis by changes of filter settings. This has allowed the emergence of information, which previously could only be obtained with direct current amplifiers (DC) amplifiers. Ikeda et al. (1996, 1999) were the first investigators to demonstrate that, what is commonly referred to as, DC shifts can be recorded from the scalp and intracranially with a widely used commercial EEG system (Nihon Kohden), which has an input filter of Hz. The observation was subsequently confirmed by several authors (Bragin et al., 2005; Hughes et al., 2005; Mader et al., 2005; Fell et al., 2007; Rodin and Modur, 2008; Rodin et al., 2008, 2009; Shi et al., 2012). They also noted, in agreement with Ikeda et al. (1996, 1999), that these shifts have a smaller electrical field than conventional frequencies and may therefore have clinical utility in delineating the epileptogenic zone. There was only From the *Division of Neurology, Intermountain Medical Center, Neuroscience Institute, Murray, Utah, U.S.A.; and Department of Neurology, University of Utah, Salt Lake City, Utah, U.S.A. Presented at American Clinical Neurophysiology Meeting, February 2012, San Antonio, TX. Address correspondence and reprint requests to Tawnya Constantino, MD, Division of Neurology, Intermountain Medical Center, Neurosciences Institute, South Office Building, Suite 810, Murray, UT 84107, U.S.A.; tawnya.constantino@imail.org. Copyright Ó 2012 by the American Clinical Neurophysiology Society ISSN: /12/ one report (Gross et al., 1999) that raised a serious question about clinical usefulness because the shifts, as recorded from stainless steel depth electrodes, were encountered infrequently and at times unrelated to the seizure onset zone as identified by the conventional frequency band. Although ictal onset baseline shifts are being recorded by conventional EEG systems because a time span of seconds rather than minutes is involved, Vanhatalo et al. (2003, 2010) have emphasized that accurate recording of these shifts require DC amplifiers. These authors have also pointed that the recording of background sleep infraslow activity (ISA), defined as 0.02 to 0.2, require DC amplifiers (Vanhatalo et al., 2004), which this article will refute. Although the commonly used EEG systems can attenuate amplitudes and shorten the duration of wave forms below the specified input filter, when compared with DC systems, they do not alter the topography of events, which is the most important aspect for clinicians (Rodin et al., 2006). Review of the existing literature on ISA shows that the literature largely dealt with ictal baseline shifts, and information on interictal ISA is currently missing. Yet these frequencies may also be important in a thorough assessment of the epileptogenic zone. It is a well-known fact that the results of surgical removal of epileptogenic tissue tend to decay over time (de Tisi et al., 2011; Geller and Devinsky, 2006; Keogan et al., 1992; Moretti Ojemann and Jung, 2006; Tanriverdi et al., 2008; Wieser et al., 2003) and even patients who have enjoyed complete cessation of seizures may relapse when anticonvulsant medications are withdrawn (Schmidt et al., 2004). This indicates that additional epileptogenic tissue remains undetected with currently used methods. In addition, efforts are being made by using a variety of algorithms to predict when a seizure might occur. These results have so far not been very satisfactory because initial promising observations were frequently not reproducible (Carney et al., 2011; Hughes, 2008; Minasyan et al., 2010; Mormann et al., 2007). To address the possibility that ISA might provide additional information in regard to these questions, the current preliminary investigation was undertaken. METHODS The records of five consecutive patients who had undergone long-term intracranial video-monitoring to further define the epileptogenic zone were investigated. Previously archived data from scalp recordings were also available in four of them, and the EEG system was provided by XLTEK, which has an input filter 0.05 Hz. The sampling rate for intracranial data was 512 Hz and 256 Hz for scalp data. Intracranial strip and grid electrodes were of platinum (Ad-Tech Medical Instruments, Racine, WI), whereas the scalp electrodes were of Ag/AgCl. Electrode coverage for intracranial recordings ranged from 26 to 58 electrodes. Grids and strips had been placed according to seizure semiology, interictal scalp data, and in four instances magnetoencephalography recordings and radiologic information. Electrode placement was always unilateral and limited 298 Journal of Clinical Neurophysiology Volume 29, Number 4, August 2012

2 Journal of Clinical Neurophysiology Volume 29, Number 4, August 2012 Interictal, Intracranial Infraslow Activity to the side of the suspected seizure origin. Scalp recordings were obtained from 26 electrodes, which included T1/T2 and IO1/IO2 leads. For intracranial recordings, the de-identified video and EEG data for each day were placed separately on an external drive and subsequently analyzed with the BESA software (MEGIS, Gräfelfing, Germany). Seizures were investigated for area of onset, with special emphasis on the minute that preceded the occurrence of electrical changes in the conventional frequencies (1 to 70 Hz). Thereafter, the individual files were downsampled to 10 Hz to further assess preictal and interictal ISAs. The files were then viewed on the maximum window of 20 minutes. The high-pass filter of the software was left open and the low-pass filter was set to 0.1 Hz, which in all instances revealed pure ISA. To investigate whether consistent changes could be demonstrated that might herald a seizure 20 minutes before its occurrence, the data were also subjected to power spectral analysis. For these analyses, the raw data were used without software filters. The reference electrode for the intracranial electrodes was a needle electrode inserted into the temporalis muscle. This resulted frequently in contamination of the interictal recordings and they were, therefore, routinely transformed to an average common reference of all artifact-free channels. Interictal scalp recordings showed motor/muscle movement and eye movement artifact to such an extent that reliable interpretation of the waking state data was not possible and only sleep data that were artifact free were evaluated. Scalp recordings were also reformatted to an average common reference montage. Clinical or subclinical seizures occurred 21 times during intracranial recording periods ranging up to 3 days. Partial seizures were seen in 20 instances, and 1 patient had a tonic seizure characterized by extension of both arms. During scalp recordings, 14 partial and 1 tonic clonic seizures were recorded over a period of up to 7 days. During scalp recordings, 14 partial seizures were found and analyzed. In contrast to intracranial recordings, those from the scalp were not contiguous because they had been obtained at various times before the start of this investigation and archived data had to be used. Interictal epochs of only 1-minute duration had been saved except for epochs when a patient event had occurred. In these instances, 35 minutes of data were available. The clinical characteristics of the patients and the areas of intracranial electrode coverage are shown in Table 1. RESULTS Preictal Intracranial Recordings The determination of the precise moment of seizure onset can at times be difficult even for experts in EEG. In the current study, this was facilitated in four patients because their seizures were accompanied by high-frequency gamma activity and ictal onset baseline shifts in one or more channels when appropriate filter settings were used. In the remaining patients, all seizures were initiated by a series of spikes that terminated in typical rhythmic seizure discharges in that electrode. In all instances, electrical seizure onset preceded clinical phenomena by several seconds. When the viewing window was expanded to 1 minute from seizure onset and the preceding minutes studied with open highpass and a low-pass filter of 0.1 Hz, an increase in the usual background ISA was discernible in all but one seizure. An example for seizure onset (case 2 in the table), as seen by conventional frequencies and with open filters for a 1-minute segment, is shown in Figs. 1A and 1B as recorded on an average reference montage. The data consist of four 4-contact right-sided subtemporal strips arranged from most anterior to most posterior (channels 1 to 16), andchannels17to26reflect a lateral anterior temporal 5 2-strip. Electrode 1 was situated most anterior inferior (channel 17) at the temporal pole and channel 22 its anterior superior counterpart. Channels 21 and 26 were recorded from the most posterior electrodes of the strip. With conventional filter settings (1 to 70 Hz, Fig. 1A), seizure onset was characterized by initial attenuation of activity in most channels, as demonstrated by the marker, and the highest amplitude rhythmic discharges later during the seizure were in channel 7. When the high-pass filter was removed, several baseline shifts became apparent during the attenuation phase and of even higher amplitude in the lateral temporal strips, where phase reversals could also be observed at the beginning of rhythmic activity. The earliest highest negative shifts were in channels 13 and 14, which recorded from the most medial electrodes of the most posterior subtemporal strips. The amplitude at channel 13 was 887 mv and for the highest deflection during the onset of rhythmicactivityinchannel23was2.5mv. When the viewing window was expanded to 20 minutes, the conventional frequency band, as shown in Fig. 2A, provided no further information. But when ISA was evaluated with an open high-pass filter and a 0.1 Hz low-pass filter (Fig. 2B), it was apparent that a prolonged increase in amplitude was present in the most anterior subtemporal strip. It started 11 minutes before the conventional seizure onset, and in its most lateral contact (channel 4), it continued throughout the seizure into the postictal state. It is also noteworthy that the lateral temporal strip showed early phase reversals. These were initially in channels 19 and 20, subsequently in channels 18 and 19, and just before conventional seizure onset between channels 22 and 23. The amplitude at channel 22 was 2.6 mv and toward the end of the seizure, the amplitude in channel 21 was 4.9 mv. TABLE 1. Patient Clinical Profile of Patients Age (years)/gender Seizure Semiology Scalp EEG (ictal) MRI MEG IC EEG (ictal) 1 48/F Temporal L F-T Cerebellar atrophy L MT L MT 2 20/F Temporal R temporal Normal R temporal R MT 3 33/M Temporal L temporal Previous T lobectomy L inf. F, L temporal L temporal L ant. cingulate 4 32/F Temporal L C-T Left frontal encephalomalacia L frontal L temporal Multiple cav. Mal. (insula) 5 29/M Temporal L F-T Small L hem. L MT L MT R, right; L, left; F-T, fronto-temporal; C, central; T, temporal; MT, mesial temporal; MEG, magnetoencephalography; IC, intracranial; M, male; F, female; Multiple cav. Mal., multiple cavernous malformations; Small L hem., small left hemisphere. Copyright Ó 2012 by the American Clinical Neurophysiology Society 299

3 T. Constantino and E. Rodin Journal of Clinical Neurophysiology Volume 29, Number 4, August 2012 FIG. 1. A, Subdural strip recordings of the onset of a partial seizure in patient 2. Total display time is 1 minute, vertical stripes delineate 1 second, calibration bar ¼ 500 mv. Filter setting 1 to 70 Hz. Electrode array: first 16-channel 4 subtemporal strips from the anterior to the middle temporal lobe; channels 17 to 26, 5 2 lateral anterior temporal strip. The seizure starts with attenuation of activity at the marker. B, Same data but filter setting changed; high-pass filter open and low-pass filter set to 0.1 Hz. Several early baseline shifts are present during the attenuation phase and of higher amplitude at the time of rhythmic seizure onset. Amplitude at channel 13 was 887 mv and 2.5 mv at channel Copyright Ó 2012 by the American Clinical Neurophysiology Society

4 Journal of Clinical Neurophysiology Volume 29, Number 4, August 2012 Interictal, Intracranial Infraslow Activity FIG. 2. A, Same patient, same seizure but on a 20-minute window. Vertical stripes delineate 10 seconds. Filter settings 1 to 3.3 Hz. The seizure is at the end of the display window. Calibration bar ¼ 300 mv. B, Same data but high-pass filter open and lowpass filter set to 0.1 Hz. Calibration bar ¼ 1 mv. Infraslow activity buildup most marked in the first (most anterior) subtemporal strip that starts 11 minutes before conventional seizure onset. At higher amplifications, still earlier involvement of the inferior electrodes of the lateral temporal strip was seen. Amplitudes of early activity in channel 20 (note the phase reversal with channel 19) 1.2 mv, channel 4 at onset 1.3 mv, channel 22 immediately before conventional electrical seizure onset 2.6 mv, and channel 21 before the end of the seizure 4.9 mv. Copyright Ó 2012 by the American Clinical Neurophysiology Society 301

5 T. Constantino and E. Rodin Journal of Clinical Neurophysiology Volume 29, Number 4, August 2012 Preictal Scalp Recordings For this aspect of the study, reliable data were severely limited because, in all but one instance, seizures arose from the waking state and the onset was contaminated by artifacts. In the seizure, which arose from sleep, an increase in ISA background amplitudes was observed 4 minutes before conventional electrical seizure onset. Because of the limited case material and inasmuch as other manuscripts in this issue deal with this topic, it was not further investigated. Interictal Intracranial Raw Data When the high-pass filter was left open and the low-pass filter was set to 0.1 Hz, interictal background ISA amplitudes varied considerably between channels and between patients. Furthermore, there was a difference in regard to the waking and sleeping states. A typical example from the patient whose ictal data have been shown in the previous figures is presented in Figs. 3A and 3B for a 20-minute epoch. The data were obtained on a day when no clinical or subclinical seizures had occurred. As Fig. 3A demonstrates, background activity ranged from 105 mv (channel 3) to 315 mv (channel 15) in the waking state and 46 to 151 mv for the same channels during sleep, Fig. 3B. From this background emerged intermittently higher amplitude activity ranging from100 mv up to 3 mv in a waxing and waning manner when the raw data were viewed on screens showing 20 minutes. These increases in activity could be limited to one channel in various portions of the file but frequently involved two or more neighboring electrodes and at times even all of the available electrodes. When only one channel was involved, electrode artifact had to be considered. But the classic electrode pops were clearly discernible as a sudden high-amplitude, 14 to 30 mv, monophasic or biphasic sequence that returned to the baseline within approximately 80 seconds. In these five cases, ISA amplitude increase was not limited, in all but one patient, to a discrete area that could be regarded as the epileptogenic zone but also occurred independently in distant strips or grids. It could last several minutes up to 2.5 hours. A 20-minute example is shown in Fig. 4A with conventional filter settings and Fig. 4B for ISA. The patient was asleep at the time, and the conventional filter settings did not reveal diagnostic changes. The corresponding ISA in Fig. 4B, however, showed marked discharges in the lateral temporal 5 2-strip (channels 17 26), which appeared to progress from the anterior inferior tip to the superior row of electrodes. The highest discharge in channel 24 had amplitude of 2.9 mv. This local increase in amplitudes was usually unrelated to clinical or subclinical seizures, although in two instances the event button was pressed by the patient during these discharges because of the subjective awareness of a possible seizure. But since inspection of the conventional frequency band had failed to reveal obvious changes in the recording, it was regarded as a false alarm. Figures 5 and 6 show examples of interictal ISA increase in patient 5 and demonstrate that the previous examples were not an isolated finding. The recording electrode array was similar to case 2 but an additional 4-contact strip had been placed on the inferior frontal gyrus (channels 27 to 30). Although Fig. 5 shows highest activity in the second subtemporal strip, it is highest in the first one in Fig. 6. In both instances, there is spread to other strips, including the frontal one. Seizure onset, as determined by conventional frequencies and baseline shifts, was in the most medial contacts of the first and second subtemporal strip. These findings are typical for what can be expected when long-term recordings are available and the software filters are set to record pure ISA. Power Spectral Analysis for Preictal and Interictal Intracranial Data When only the first seizure of the day, which had lasted at least 50 seconds, was considered, the power for a 1-minute segment, starting with seizure onset, ranged from 404 mv 2 to a maximum of 110,414 mv 2. The data are, however, better appreciated in the context of background activity when the seizure was placed in the center of a 20-minute window. Tables 2 and 3 present the results of 20-minute epochs for the waking state, the seizure epoch, and midnight sleep. Table 2 deals with patients 1, 2, and 4, whereas table 3 shows the results from patients 3 and 5. The data were separated because, as will be seen, the values for the patients in Table 3 differed for unknown reasons, by a factor of 10, from those in Table 2 and only the ictal days are shown. For the other three patients additional data for a day when no seizure had occurred are also presented. Although the absolute values for patients 3 and 5, as shown in Table 3, are not meaningful, it will be seen that the relative distributions are similar to the ones in Table 2. The first column of the tables shows in the top row the state of the patient and the time of the start of the 20-minute epoch, the next three columns show the power in square micovolts, and in the fourth column, the range of frequencies is observed. The bottom row shows the number of channels involved in each of these power bands and provides information on spread of ISA increase. Although the seizures were partial, it is apparent that power was markedly increased in the ictal state not only in absolute values but also in the number of channels involved. Contrary to expectations, it was decreased during sleep as compared with the waking state. Furthermore, Table 2 shows that midnight sleep power was even lower during the interictal than on the ictal night. The data were also investigated for the 20-minute preictal and subsequent postictal period and for possible differences between midnight and early morning sleep. There was, however, only one patient (2) who had a tonic seizure in this group, whereas the others had only features of automatic behavior. In this patient, a preictal power buildup was noted, which persisted into the postictal epoch. This did not occur in the other patients. There were suggestive differences between midnight and early morning sleep, with the latter showing more power, at times approaching the waking levels, but the current data were insufficient to make a conclusive statement. Interictal Scalp Data Because these data were not from contiguous EEG epochs and waking portions were in part contaminated by artifact, only the sleep data will be presented. With windows of 10 to 20 minutes, background activity had amplitudes ranging from 5 to 20 mv in different channels, and it was usually highest in the anterior quadrant bilaterally. From this background arose intermittently higher amplitude discharges (50 to 150 mv) lasting several minutes that usually were also maximal in the frontotemporal areas. Although grid and strip implantation was on the left in these four patients, the scalp amplitude increases were always bilateral and could be higher on the side contralateral to the intracranial data. DISCUSSION Although this was a preliminary study and as such had limitations, which will be discussed later, it showed what the clinical electroencephalographer can expect to see in archived data, when the high-pass filter is left open and the low-pass filter is set to 0.1 Hz. In 302 Copyright Ó 2012 by the American Clinical Neurophysiology Society

6 Journal of Clinical Neurophysiology Volume 29, Number 4, August 2012 Interictal, Intracranial Infraslow Activity FIG. 3. A, Same patient interictal background activity in the waking state. Calibration bar ¼ 200 mv; 20-minute epoch, high-pass filter open and low-pass filter set to 0.1 Hz. Amplitudes range from 105 to 315 mv. B, Same patient interictal background during midnight sleep. Same amplifications and filter settings as in (A). The amplitude of background activity is considerably reduced and ranges from 36 to 151 mv. Copyright Ó 2012 by the American Clinical Neurophysiology Society 303

7 T. Constantino and E. Rodin Journal of Clinical Neurophysiology Volume 29, Number 4, August 2012 FIG. 4. A, Same patient, another 20-minute interictal epoch during early morning sleep. Calibration bar ¼ 500 mv. High-pass filter set to 1 Hz and low-pass filter set to 3.3 Hz; the tracings appear unremarkable. B, Same epoch but high-pass filter open and low-pass filter 0.1 Hz. Calibration signal ¼ 1 mv. Marked infraslow activity increase is apparent, which initially involves the lateral temporal strip electrodes but subsequently spreads to the most anterior subtemporal strip. Maximum amplitude in channel 18 is 2 mv and in channel 24 is 2.9 mv. 304 Copyright Ó 2012 by the American Clinical Neurophysiology Society

8 Journal of Clinical Neurophysiology Volume 29, Number 4, August 2012 Interictal, Intracranial Infraslow Activity FIG. 5. Interictal 20-minute epoch from patient 5. High-pass filter open and low-pass filter set to 0.1 Hz. Calibration bar ¼ 200 mv. Same electrode array as in patient 1 but the additional channels labeled clear record activity from a lateral inferior frontal gyrus strip. Infraslow activity buildup is most pronounced in channel 6, 580 mv, but present with lower amplitudes also in other strips. FIG. 6. Another 20-minute interictal epoch from the same patient with the same amplifications and same filter settings. Infraslow activity increase is highest in the most mesial contact of the most anterior subtemporal strip, 600 mv, but involves in addition portions of all the other strips. Copyright Ó 2012 by the American Clinical Neurophysiology Society 305

9 T. Constantino and E. Rodin Journal of Clinical Neurophysiology Volume 29, Number 4, August 2012 TABLE 2. Comparison of ISA Power Spectra for Ictal and Interictal State (Waking and Asleep) on a Day When a Seizure Had Occurred and Interictal Waking and Sleep Activity on a Seizure-Free Day Clinical State Power in mvsq Frequency Range Patient 1 Ictal day Awake Hz Ictal Sleep Interictal day Awake Sleep Patient 2 Ictal day Awake Ictal Sleep Interictal day Awake Sleep Patient 4 Ictal day Awake Ictal Sleep 01: Interictal day Awake Asleep 00: The first column shows the clinical state. In the second column, the top line shows the power for the ranges between 0 and 100 mvsq, 101 and 999 mvsq, 1000 and 9999 mvsq, and.10,000 mvsq. The numbers underneath show how many channels of the record these values were present, during that epoch. The third column gives the ISA frequency range from minimum to maximum. In all instances, power is always highest and more widely distributed in the ictal than interictal state and lower during sleep than awake. addition, it provided new information, on which further investigations can be based. The most important observation was that currently used EEG systems, even with an input filter of 0.05 Hz, can reveal aspects of ISA that are not apparent when only the conventional frequency band is used. As mentioned in the Introduction, this fact was known for ictal activity but had not previously been demonstrated for interictal data. The study has also shown that from a lower-amplitude interictal ISA background, higher-amplitude waves in the infraslow range can arise, which are not merely in proximity to the epileptogenic zone, as defined by seizure onset in the conventional frequency band, but can also occur in more distant areas. Although one may have to disregard a buildup as artifact when only one electrode contact is involved, this is unlikely when it occurs in neighboring electrodes of the same or neighboring strips. These observations are most reliable for intracranial data and may well reflect a larger potentially epileptogenic network (Bertram, 2003). This may be one of the reasons for the previously mentioned decay of surgical results when long-term follow-up studies are performed. But this cannot be appreciated with the conventional frequency bands and is only detected when ISA is added to the evaluation of the data. The observation that background ISA was decreased during sleep, rather than increased, was likewise unexpected. Yet it was not artifact due to small sample size but constituted a genuine finding because it has also been seen in another case where bilateral hippocampal depth electrodes had been inserted via occipital burr holes (Shi et al., 2012). Furthermore, Picchioni et al. (2011) have shown in a combined EEG/functional magnetic resonance imaging study that sleep ISA did not correspond to what was seen in the delta and subdelta (defined as 0.55 to 0.99 Hz) bands. A similar 306 Copyright Ó 2012 by the American Clinical Neurophysiology Society

10 Journal of Clinical Neurophysiology Volume 29, Number 4, August 2012 Interictal, Intracranial Infraslow Activity TABLE 3. Comparison of ISA Power Spectra for Ictal and Interictal State (Waking and Sleep) in Patients 3 and 5 Clinical State Power in mvsq Frequency Range Patient 3 Awake Ictal Sleep Patient 5 Awake Hz Ictal Sleep The format is the same as in Table 2, but for unknown reasons (technical?), the power values were apparently reduced by a factor of 10 in contrast to those of Table 2. Nevertheless, the relationship in terms of highest values and number of channels involved in the ictal state and the decrease of power during sleep is preserved. discrepancy was also observed for background multiple unit activity in experimental animals. It was increased in the waking state and more markedly during seizures, whereas it was decreased during slow wave sleep and increased again in the rapid eye movement sleep stage (Rodin et al., 1973). These observations provide additional evidence that the extreme low- and extreme high-frequency bands reflect different neurophysiologic mechanisms from those that are responsible for the conventional frequencies (Vanhatalo et al., 2010; Voipio et al., 2003). The data may also be analogous to the observation by Caspers and Schulze (1959) in regard to the steady cortical potential, recorded with DC amplifiers. It became more negative in the waking state, increased further during activity, and even more so during seizures; it became positive with increasing depth of sleep. The finding that ISA preceded seizure onset in case 2 by 11 minutes is additional confirmation of previous literature reports. Ren et al. (2011) had noted in three patients that ISA buildup had preceded seizure onset from 8 to 22 minutes, and Federico et al. (2005) had detected functional magnetic resonance imaging changes in 3 cases ranging from 3 to 11 minutes before the seizure. Litt et al. (2001) had reported in five patients that burst activity could be seen seven hours before a seizure and accumulated energy increased 50 minutes before seizure onset. These data also have relevance for the problem of the prediction when a given seizure may occur, which, as mentioned previously, has recently received a great deal of attention. Up to now, studies on seizure onset prediction have been based on the conventional frequency band and have not yielded reliable, reproducible results. Since, as presented here, ISA power shows marked regional waxing and waning in the interictal state lasting at times for more than an hour, which did not lead to a seizure, future algorithms for seizure prediction should include the assessment of ISA. It might be advisable to proceed initially with intracranial data, which have fewer artifacts, and they can subsequently be validated on scalp recordings. Scalp data have to be used, however, with considerable circumspection. Not only because of excess of movement artifact, and especially eye movement artifact, but even during sleep lateral eye movements can persist, which can make interpretation of sleep recordings difficult. This applies especially to studies that rely on frequency spectra because these cannot distinguish polarity. Nevertheless, when raw data were examined, it was apparent that although out-of-phase activity did occur in the anterior temporal regions, which could extend to the T7/T8, A1/A2 contacts, in phase discharges were likewise seen with higher amplitudes on one side than the other. Although in the current study, these patients were regarded as having had unilateral left temporal lobe disease, shifting interictal higher amplitude right temporal ISA was also observed in three of the cases and in one instance independent frontal lobe involvement could be seen. To what extent this finding may be important clinically will depend on long-term follow-up of patients whose presumed epileptogenic area, based on conventional frequencies, has been surgically removed. Inasmuch as our data were not obtained from a specifically designed research project but resulted from inspection of clinically obtained material, they had, by necessity, several limitations. Of these, the most important was intracranial electrode coverage. It had been based on previous clinical information, and in all instances electrodes had been implanted unilaterally. They were usually centered on one lobe and although the basal temporal areas were explored in all instances, no information was available from the basofrontal regions or the island of Reil. Nevertheless, these structures are known to be of importance in epileptogenesis, and because ISA has a smaller electrical field than conventional frequencies (Gumnit and Takahashi, 1965), valuable information may have been missed. Even in scalp recordings, only T1/2 rather than the complete inferior row of the 10/10 system had been available although infraorbital electrodes had been placed, which provided some information on basofrontal activity. Another potential limitation was the input filter of the amplifiers at 0.05 Hz, which attenuated slower activity. As such, the values for amplitudes and power are to be regarded as minima. The values in the three cases where there were no technical problems agree, however, with previously published ones obtained on a Nihon Kohden system and even with those reported by Kim et al. (2009) who had assessed ictal onset baseline shifts with DC amplifiers. In their study of 11 patients, the baseline shift values ranged from 800 mv to 3.4 mv, with one outlier at 10 mv. The software that was used in this study also had limitations not only in regard to the viewing window, which covered at maximum 20 minutes, but also in regard to filter settings. For raw data viewing, the lower limit was Hz before the system automatically reset to the default value of 0.5 Hz. For spectral analysis, the lower limit was 0.00 Hz but as can be seen in this study, as well as a previous one (Rodin and Modur, 2008), power spectra are at times maximal at the lowest point, indicating the presence of still slower frequencies. In addition there is the problem of electrode polarization. Although Ag/AgCl electrode material for scalp electrodes and platinum for intracranial recordings reduce polarization effects, they may not abolish them and this is why intermittent slow wave activity, which is not clearly electrode artifact, may or may not be trustworthy when it appears in only one channel at a time. Yet as mentioned above, despite these limitations, this investigation can be regarded as a starting point for future studies, which deal with the area(s) of potential epileptogenicity in patients who have no demonstrable anatomic lesion. As mentioned, the clinical significance of the larger than expected extent of intermittent interictal ISA buildup is at present unclear but may well be important in relation to long-term seizure freedom after surgical resections of suspected epileptogenic brain tissue. Since these data are archived in laboratories that carry out intensive Copyright Ó 2012 by the American Clinical Neurophysiology Society 307

11 T. Constantino and E. Rodin Journal of Clinical Neurophysiology Volume 29, Number 4, August 2012 monitoring of seizure patients, long-term follow-up studies of operated patients should now be undertaken to investigate the presumed epileptogenic zone not only in relation to the conventional frequency band but also ISA. Furthermore, the fact that ISA can be recorded on a longterm basis from intracranial recordings, with conventional EEG amplifiers, should encourage basic scientists to include these frequencies in their studies of experimental animals not only in regard to seizure genesis but also their relationship to underlying mechanisms. Inasmuch as ISA is a physiologic concomitant of cerebral electrical activity, its inclusion in the study of normal electromagnetic phenomena, and other pathologic entities, apart from epilepsy, is also warranted. REFERENCES Bertram EH. Why does surgery fail to cure limbic epilepsy? Seizure functional anatomy may hold the answer. Epilepsy Res 2003;56: Bragin A, Wilson C, Fields T, et al. Analysis of seizure onset on the basis of wideband EEG recordings. Epilepsia 2005;46(suppl 5): Carney PR, Myers S, Geyer JD. Seizure prediction: methods. Epilepsy Behav 2011;22:S95 S101. Caspers H, Schulze H. Die Veränderungen der corticalen Gleichspannung während der natürlichen Schlaf-Wach-Perioden beim freibeweglichen Tier. Pflügers Archiv 1959;270: de Tisi J, Bell GS, Peacock JH, et al. The long-term outcome of adult epilepsy surgery, patterns of seizure remission, and relapse: a cohort study. Lancet 2011;278: Federico P, Abbott D, Briellmann R, et al. Functional MRI of the pre-ictal state. Brain 2005;128: Fell J, Fritz N, Burr L, et al. Human neocortical and hippocampal near-dc shifts are interconnected. Hippocampus 2007;17: Geller EB, Devinsky O. Outcomes of extratemporal epilepsy surgery. In: Miller JW, Silbergeld DL, eds. Epilepsy surgery. Principles and controversy. New York: Taylor & Francis Group, 2006: Gross D, Gotman J, Quesney L, et al. Intracranial EEG with very low frequency activity fails to demonstrate an advantage over conventional recordings. Epilepsia 1999;40: Gumnit RJ, Takahashi T. Changes in direct current activity during experimental focal seizures. Electroencephalogr Clin Neurophysiol 1965;19: Hughes J, Fino J, Patel K. A newly described ictal pattern: the initial ictal slow shift. Clin EEG Neurosci 2005;36: Hughes JR. Progress in predicting seizure episodes with nonlinear methods. Epilepsy Behav 2008;12: Ikeda A, Taki W, Kunieda T, et al. Focal ictal direct current shifts in human epilepsy as studied by subdural and scalp recording. Brain 1999;122: Ikeda A, Terada K, Mikuni N, et al. Subdural recording of ictal DC shifts in neocortical seizures in humans. Epilepsia 1996;37: Keogan M, McMackin D, Peng S, et al. Temporal Neocorticectomy in management of intractable epilepsy: long-term outcome and predictive factors. Epilepsia 1992;33: Kim W, Miller J, Ojemann J, Miller K. Ictal localization by invasive recording of infraslow activity with DC-coupled amplifiers. J Clin Neurophysiol 2009;26: Litt B, Esteller R, Echauz J, et al. Epileptic seizures may begin hours in advance of clinical onset. Neuron 2001;30: Mader EC Jr, Fisch BJ, Carey ME, Villemarette-Pitman NR. Ictal onset slow potential shifts recorded with hippocampal depth electrodes. Neurol Clin Neurophysiol 2005;4:1 12. Minasyan G, Chatten J, Chatten M, Harner R. Patient-specific early seizure detection from scalp electroencephalogram. J Clin Neurophysiol 2010;27: Moretti Ojemann L, Jung M. Outcomes of temporal lobe epilepsy surgery. In: Miller JW, Silbergeld DL, eds. Epilepsy surgery. Principles and Controversy. New York: Taylor & Francis Group, 2006: Mormann F, Andrzejak RG, Elger C, Lehnertz K. Seizure prediction: the long and winding road. Brain 2007;1030: Picchioni D, Horovitz SG, Fukunaga M et al. Infraslow oscillation organize largescale cortical-subcortical interactions during sleep: a combined EEG/fMRI study. Brain Res 2011;1374: Ren L, Terada K, Baba K. Ictal very low frequency oscillations in human epilepsy patients. Ann Neurol 2011;69: Rodin E, Constantino T, Rampp S, Modur P. Seizure onset determination. J Clin Neurophysiol 2009;26:1 12. Rodin E, Constantino T, vanorman C, House P. EEG Infraslow activity in absence and partial seizures. Clin EEG Neurosci 2008;39: Rodin E, Constantino T, Van Orman C, et al. Optimal evaluation of digital electroencephalograms. Clin EEG Neurosci 2006;37: Rodin E, Modur P. Ictal Intracranial Infraslow EEG Activity. Clin Neurophysiol 2008;119: Rodin E, Wasson S, Triana E, Rodin M. Hochfrequenzableitungen: Wert und Grenzen der Methode. Z Elektroenzephalogr Elektromyogr Verwandte Geb 1973;4:9 16. Schmidt D, Baumgartner C, Löscher W. Seizure recurrence after planned discontinuation of antiepileptic drugs in seizure-free patients after epilepsy surgery: a review of current clinical experience. Epilepsia 2004;45: Shi J, Rodin E, Gupta V, Wharen R. Signal characteristics of intraventricular electrodes recording in human epilepsy. Clin EEG Neurosci 2012;43: Tanriverdi T, Olivier A, Poulin N, et al. Long-term seizure outcome after mesial temporal lobe epilepsy surgery:corticalamygdalohippocamppectomy versus selective amygdalohippocampectomy. J Neurosurg 2008;108: Vanhatalo S, Holmes MD, Tallgren P, et al. Very slow EEG responses lateralize temporal lobe seizures. An evaluation of non-invasive DC-EEG. Neurology 2003;60: Vanhatalo S, Palva JM, Holmes MD, et al. Infraslow oscillations modulate excitability and interictal epileptic activity in the human cortex during sleep. Proc Natl Acad Sci U S A 2004;101: Vanhatalo S, Voipio J, Kaila K. Infraslow EEG activity. In: Schomer D, Lopes da Silva F, eds. Niedermeyer s Electroencephalography: Basic Principles, Clinical Applications, and Related Fields. 6th ed. Chapter 36. Philadelphia: Wolters Kluwer/Lippincott Williams &Wilkins, 2010: Voipio J, Tallgren P, Heinonen E, et al. Millivolt-scale DC shifts in the human scalp EEG: evidence for a nonneuronal generator. J Neurophysiol 2003;89: Wieser HG, Ortega M, Friedman A, Yonekawa Y. Long-term seizure outcomes following amygdalohippocampectomy. J Neurosurg 2003;98: Copyright Ó 2012 by the American Clinical Neurophysiology Society