Epilepsy Research (2009) 86, 153—162
journal homepage: www.elsevier.com/locate/epilepsyres
A decrease in EEG energy accompanies anti-epileptic drug taper during intracranial monitoring夽 Hitten P. Zaveri a,∗, Steven M. Pincus b, Irina I. Goncharova a, Edward J. Novotny a,c,d, Robert B. Duckrow a,d, Dennis D. Spencer d, Susan S. Spencer a,d a
Department of Neurology, Yale University, 333 Cedar Street, New Haven, CT 06520, USA Guilford, CT 06437, USA c Department of Pediatrics, Yale University, New Haven, CT 06520, USA d Department of Neurosurgery, Yale University, New Haven, CT 06520, USA b
Received 5 December 2008; received in revised form 29 May 2009; accepted 7 June 2009 Available online 24 July 2009
KEYWORDS Intracranial EEG; Anti-epileptic drugs; Teager energy; Seizure generation; Cortical excitation; Cortical inhibition
Summary Objective: During intracranial EEG (icEEG) monitoring the likelihood of observing a seizure is increased by tapering anti-epileptic drugs (AEDs). Presumably AED taper results in an increase in cortical excitation which in turn promotes seizure emergence. We measured change in signal energy of icEEGs in response to AED taper to quantify changes in excitation which accompany the increased propensity for seizures. Methods: Twelve consecutive adult patients who completed intracranial monitoring were studied. Two icEEG epochs from before and after AED taper, each 1 h in duration, during wake, matched by time-of-day and removed from seizures were selected for each patient. Teager energy, a frequency weighted measure of signal energy, was estimated for both the seizure onset region as well as all other brain areas monitored. Results: Considerable changes in Teager energy, evaluated at a 1-h time-resolution, occur during intracranial monitoring. The most dominant trend is a decrease to lower values than those when the patient is on AEDs. A decrease of 35% was observed for both all the brain areas monitored and the seizure onset region.
∗
Corresponding author. Tel.: +1 203 737 5407; fax: +1 203 785 5694. E-mail address:
[email protected] (H.P. Zaveri). 夽 In Memoriam: Susan S. Spencer died on May 21, 2009. She was the head of the Yale Epilepsy Program and founder and director of the Epilepsy Monitoring Unit at the Yale-New Haven Hospital. During her career she was an ardent supporter of clinical research in epilepsy, a devoted advocate and model for physicians in training, and an uncompromising clinician. She recognized and nurtured collaboration between investigators from diverse scientific disciplines to advance our understanding of epilepsy. 0920-1211/$ — see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.eplepsyres.2009.06.002
154
H.P. Zaveri et al. Conclusions: A decrease in signal energy occurs during intracranial EEG monitoring, possibly accompanying AED taper. If the decrease is due to AED taper this would suggest that AEDs prevent seizures in ways other than reduction of cortical excitation and seizure generation may be influenced by factors other than poorly regulated cortical excitation. © 2009 Elsevier B.V. All rights reserved.
1. Introduction The past several years have witnessed an advance towards the definition of a pre-seizure state in human partial epilepsy. Seizure prediction studies have reported EEG changes suggestive of increased excitability, altered synchrony, and decreased signal complexity during a presumed pre-seizure period (Mormann et al., 2006). While these observations provide some insight into the pre-seizure state, the role of this state in ictogenesis and the mechanisms which may be at play during it remain poorly understood. Most studies of seizure prediction have been conducted in the setting of intracranial EEG monitoring. Intracranial monitoring studies of patients with intractable epilepsy being considered for resective surgery follow a well established protocol. Once the patient is stabilized following placement of intracranial electrodes, AEDs are tapered to facilitate expression of seizures. AED taper presumably results in an increase in cortical excitation that promotes seizure emergence. After a necessary number of seizures are observed AEDs are restarted. Previous studies of seizure prediction have not taken AEDs or patient state into account. In the belief that these two factors play a considerable role in changes to the background EEG as well as for the propensity for seizures we have sought to study their effect. In this study we evaluated change in signal energy and power of background intracranial EEGs (icEEGs) with AED taper during intracranial monitoring while patient state was kept fixed (Zaveri, 2007). Specifically we evaluated differences between two time points, the first towards the start of monitoring while the patient was on full or nearly full dose of AEDs, and the second after AEDs were decreased or withdrawn. We compared signal energy and power estimates of the on- and off-AEDs states to detect changes in the background icEEG which accompany the increased propensity for seizures between these two states.
2. Methods 2.1. Subjects Twelve consecutive adult patients who completed intracranial EEG and video monitoring for resective epilepsy surgery at the Yale-New Haven Hospital were included in this study. The average age of the patients was 35 and included 3 women and 9 men.
2.2. Intracranial EEG protocol The intracranial studies were performed as the final step in a multi-step evaluation protocol which is similar to that performed at other epilepsy centers (Spencer et al., 2009). Surgery to place electrodes is performed on a week day morning, following which the patient is stabilized in neuro-intensive care unit overnight. Neuroimaging with both CT scan and MRI is performed the morning
after surgery to allow precise localization of intracranial electrode contacts. The patient is transferred to an intracranial EEG and video recording suite usually the day following placement of intracranial electrodes. Monitoring is begun on this day which we refer to as the first day of the study. AED reduction proceeds within 24 h. Based on the literature and our own experience, we do not withdraw barbiturates and benzodiazepines because atypical seizures may ensue. Other AEDs are withdrawn one by one by dose reductions of 50—100% per 24 h. Discontinued AEDs are restarted once spontaneous seizures have been recorded and sufficient information is obtained, usually at their original doses except when enzyme deinduction dictates a slower schedule is advisable. The electrodes are then removed in the operating room.
2.3. Intracranial EEG data collection Intracranial electrodes consist of various combinations of subdural strips and grids and intracerebral depth electrodes (AdTech Medical, Racine, WI) as required. The icEEGs were recorded with 128-channel clinical icEEG and video monitoring equipment (BioLogic Systems Corp., Mundelein, IL). The icEEGs were sampled at 256 Hz. Between 65 and 261 intracranial electrode contacts, with a mean of 176 contacts, were employed (see Table S1 in the supplemental data section for information on electrode placement). When more than 128 intracranial electrode contacts were employed, a subset of the contacts, up to the capacity of the recording equipment, were monitored at a given time.
2.4. Definition of EEG epochs for analysis Intracranial EEG epochs, 1 h in duration, at least 6 h removed from a seizure, were selected for analysis. All measurements were made during wake and an attempt was made to match the on- and off-AEDs epochs by time-of-day. The epoch selection was performed in the following manner. First a candidate time range for the on- and off-AEDs epochs was selected by considering the medication record and the time of seizures. Next, epochs where the patient was clearly awake were selected by examination of video recordings. If the patient was asleep or drowsy during the preferred time of 9—10 a.m., alternate time periods were considered, first between 8 a.m. and 12 noon, and second between 2 and 6 p.m. The majority of patients were studied between 8 a.m. and 12 noon and a few were studied between 2 and 6 p.m. The on-AEDs epoch was typically selected from day 2 or day 3 of the monitoring. Seizures are a confounding factor for this study and it was considered useful to evaluate the effect of AED taper without the influence of seizures where possible. To achieve this two sets of on- and off-AEDs epochs were obtained (Sets A and B). The first set of on- and off-AEDs epochs (Set A) was selected to measure change to the point of maximal AED taper. In several instances this included data epochs after the occurrence of seizures. All 12 patients in this study were included in Set A. Of the 12 patients studied, in Set A, 1 experienced a seizure before the on-AEDs epoch and 7 experienced seizures between the on- and off-AEDs epochs. A second set of data epochs (Set B) was selected to measure change to a point after the start of AED taper but before the occurrence of seizures. Eight of the 12 patients were studied in Set B. The patients studied in Set B did not include the patient in Set A who experienced a seizure before the on-AEDs epoch. The patients
A decrease in EEG energy accompanies anti-epileptic drug taper during intracranial monitoring Table 1
155
Seizure onset localization, AEDs, and days between measurements.
Patient
RISE
Epoch 1 AEDs
Epoch 2 AEDs (Set A)
Days between epochs (Set A)
Epoch 2 AEDs (Set B)
Days between epochs (Set B)
1
L Anterior Medial Frontal
CBZ 800 LEV 2250 ZNS 500
CBZ 800 Discontinued Discontinued
3
CBZ 800 LEV 750 Discontinued
3
2
L Medial Temporal
OXC 1800 GPN 1800 LEV 500 VPA 250
Discontinued Discontinued Discontinued Discontinued
8
OXC 900 Discontinued Discontinued Discontinued
5
3
L Superior Parietal
CZP 1.5 PHT 200 CBZ 1600
CZP 1.5 Discontinued Discontinued
6
CZP 1.5 Discontinued Discontinued
6
4
L Medial Temporal
OXC 900 PHT 330
Discontinued Discontinued
8
Discontinued Discontinued
6
5
L Medial Temporal
ZNS 400 CBZ 600 PHT 100
ZNS 400 CBZ 300 Discontinued
2
ZNS 400 CBZ 300 Discontinued
2
6
R Anterior Superior Lateral Temporal
LTG 400 CBZ 800
LTG 400 Discontinued
4
LTG 400 Discontinued
2
7
Neocortical Unlocalized
LTG 300 CBZ 1200 TPM 75
LTG 50 Discontinued Discontinued
5
LTG 50 Discontinued Discontinued
5
8
Bilateral Medial Temporal (L > R)
CBZ 1200 ZNS 200
CBZ 600 Discontinued
3
9
R Inferior Temporal
OXC 1950 LEV 4000
OXC 1200 LEV 1200
5
10
Neocortical Diffuse Onset
CBZ 1200 GPN 600
Discontinued Discontinued
9
11
R Medial Temporal
TPM 400 LEV 1500
Discontinued Discontinued
8
12
R Parietal
PB 150 CBZ 600 LTG 300
PB 150 CBZ 400 Discontinued
2
PB 150 CBZ 400 Discontinued
2
RISE location and AEDs during on- and off-AEDs states for both Set A and Set B epochs, and the number of days between these epochs. The RISE location was medial temporal in 5, neocortical in 5, and poorly localized in 2 patients. The patients were on multiple AEDs at admission and during the course of the study one or more AEDs were tapered. The abbreviations used for AEDs are: CBZ, carbamazepine; CZP, clonazepam; GPN, gabapentin; LEV, levetiracetam; LTG, lamotrigine; OXC, oxcarbazepine; PB, Phenobarbital; PHT, phenytoin; TPM, topiramate; VPA, valproic acid; ZNS, zonisamide. studied also did not include 3 of the 7 patients who had seizures between on- and off-AEDs epochs, because in these 3 patients an off-AEDs measurement time could not be defined prior to seizure occurrence. Set A was the primary data for this study and Set B was included to control for the effect of seizures. In Set A, 8 of 12 patients experienced seizures before the on-AEDs epoch or between the on- and off-AEDs epochs. In Set B, none of the 8 patients studied experienced seizures before the on- or off-AEDs epochs. The average mismatch in time-of-day between on- and off-AEDs epochs, for both Sets A and B, was less than 20 min. For the purpose of this study, the area where the initial seizure expression is observed is called the region of initial seizure
expression (RISE). The RISE locations, AEDs during the on- and off-AEDs epochs, and number of days between the on- and off-AEDs epochs are listed in Table 1.
2.5. Measurement of intracranial EEG energy and power The signal energy of icEEGs was measured with Teager energy (Kaiser, 1990). The conventional measure of signal energy weights contributions from all frequencies of a signal equally. Teager energy is a weighted measure of signal energy (E ∝ w 2 A2 , where A is signal amplitude and w is frequency) such that high-frequency signals
156
H.P. Zaveri et al.
contribute more than low frequency signals. Analysis was repeated with the traditional measure of signal power to allow a comparison to this more conventional measure. To minimize the effect of noise the icEEG recordings were filtered with a cutoff at 50 Hz. The unfiltered and filtered icEEGs were examined and marked for artifacts at a 1-s time-resolution. Teager energy and power estimates were obtained for artifact free filtered 1-s segments for each electrode contact and averaged over the 1-h epoch to obtain average measures for each contact. Since the electrode contacts being monitored could change over the course of intracranial monitoring, only matched contacts between the two epochs of a set were studied. Change in Teager energy was calculated in the following manner. For each patient the average off-AEDs energy of each electrode contact was divided by the average on-AEDs energy for that contact. This normalized energy was averaged separately over the RISE contacts and over all electrode contacts studied for each patient. These values were subsequently averaged over all patients. For a few example patients energy and power were evaluated for every hour of recording and for every electrode contact studied. These evaluations were employed for illustrative purposes only.
example signal is a constant amplitude linear up-chirp function, that is, a signal whose amplitude is constant but whose frequency is increasing linearly. Teager energy values of this signal start with relatively low estimates. As the frequency of the chirp signal increases the energy ascribed to it increases reflecting the relationship E ∝ w 2 A2 , that is, energy is proportional to both amplitude and frequency. Here, because amplitude is held constant, the energy increases as the square of the frequency of the chirp signal.
2.6. Statistical analysis
3.3. Characterization of AEDs during study
Statistical comparisons were performed with a two-sided Student’s t-test with significance set at p < 0.05.
3. Results 3.1. Teager energy In Fig. 1 we illustrate the relationship between Teager energy and signal frequency with an example signal. The
3.2. RISE localization The RISE was localized in 10 of 12 patients (see Table 1). In these patients seizure onset was of medial temporal origin in five and of neocortical origin in five. In 2 patients, who were considered to have seizures of neocortical origin, the RISE was poorly localized or diffuse.
The on- and off-AEDs epochs in Set A were separated by between 2 and 9 days, with a median separation of 5 days. At the time of the on-AEDs epoch, one of the patients was receiving 4 AEDs, 5 were receiving 3 AEDs and 6 were receiving 2 AEDs. At the time of the off-AEDs epoch 3 of the patients were receiving 2 AEDs, 5 were receiving 1 AED, and AEDs were fully discontinued in 4 patients. The 12 patients were receiving 31 AEDs at the time of the on-AEDs epoch and 11 AEDs, of which the dose of 6 were reduced, at the
Figure 1 (a) A constant amplitude chirp signal and (b) the Teager energy of this signal. This example seeks to demonstrate the relationship between Teager energy and signal frequency. The conventional measure of energy is proportional to amplitude while Teager energy is proportional to both amplitude and frequency. If given two signals with identical amplitudes but different frequency content, the Teager energy evaluation of the signal with higher frequency content would be greater than that of the signal with lower frequency content. In this example the amplitude of the signal is held constant while the frequency is increased linearly from 0 to 10 Hz. The Teager energy evaluation demonstrates that the energy ascribed to the signal increases with signal frequency.
A decrease in EEG energy accompanies anti-epileptic drug taper during intracranial monitoring time of the off-AEDs epoch. The icEEG was analyzed from between 61 and 117 matched electrode contacts in the individual patients with an average of 103 matched electrode contacts (Table S1, Supplementary data). The on- and off-AEDs epochs in Set B were separated by between 2 and 6 days, with a median separation of 4 days. At the time of the on-AEDs epoch, one of the patients was receiving 4 AEDs, 5 were receiving 3 AEDs and 2 were receiving 2 AEDs. At the time of the off-AEDs epoch 3 of the patients were receiving 2 AEDs, 4 were receiving 1 AED, and AEDs were fully discontinued in 1 patient. The 8 patients were receiving 23 AEDs at the time of the on-AEDs epoch and 10 AEDs, of which the dose of 5 were reduced, at the time of the off-AEDs epoch. In Set B between 99 and 114 matched electrode contacts with an average of 108 matched electrode contacts were studied.
3.4. Changes in energy with AED taper The mean and standard deviation of normalized Teager energy for all electrode contacts studied and RISE contacts are shown in Fig. 2(a). Energy decreased between the onand off-AEDs states by 35% both for all electrode contacts and for the RISE. The Teager energy measurements of onand off-AEDs states were significantly different for all electrode contacts (p < 0.001) and for the RISE (p < 0.05). The mean normalized signal energy for each patient, for all contacts and the RISE, are shown in Fig. 2(b). There was a decrease in energy measured for all electrode contacts in all patients and for the RISE in 9 of 10 patients. The change in energy ranged from −58% to −8% for all electrode contacts and from −71% to 12% for RISE contacts. The power decreased by 54% for all electrode contacts and by 47% for the RISE. The signal power measurements of onand off-AEDs states were significantly different for all electrode contacts (p < 0.0001) and for the RISE (p < 0.01). For epochs in Set B, that is the epochs which were recorded before the expression of any seizures, energy decreased by 30% for all electrode contacts and by 38% for the RISE. The
157
Teager energy measurements of on- and off-AEDs states, for Set B, were significantly different for all electrode contacts (p < 0.001) and for the RISE (p < 0.05). Furthermore, for epochs in Set B, signal power decreased by 46% for all electrode contacts and by 37% for the RISE. The signal power measurements of on- and off-AEDs states, for Set B, were significantly different for all electrode contacts (p < 0.001) and for the RISE (p < 0.01).
3.5. Temporal changes in energy A decrease in energy from all electrode contacts in all 12 patients and from the RISE in 9 of 10 patients is highly significant. To explore the nature of the decrease we examined the temporal structure of signal energy over the course of intracranial monitoring. Example evaluations of average Teager energy of all electrode contacts and from a selected RISE contact for patient 3 are shown in Fig. 3(a) and (b), respectively. Patient 3 was admitted for intracranial monitoring with 3 AEDs, one of which was maintained while the other two were decreased and then stopped. Changes in the hourly estimates of signal energy (Fig. 3(a)) are evident during the course of the study. There is an apparent decrease in energy in response to each decrease in AED. There is also a period of time when the patient is on AEDs where the values are relatively steady, though with diurnal changes. This is followed by a decrease over 3 days when the second AED is tapered and terminated. After this there is a period of low energy. The patient experienced eight seizures on day 11. On day 13 a separate study on the safety of a neuro-stimulation device was initiated and data from this period have not been included here. The icEEG monitoring study ended on day 14. In Fig. 3(b) the energy measures for a RISE contact located in the left parietal lobe, for the same patient, is shown. This energy profile is similar to that observed for the average of all electrode contacts, though with greater variability. Energy values are high, on day 2, at the start of the study, and decrease at a time after AED withdrawal. Relatively low values are recorded after the second AED is tapered and
Figure 2 (a) Normalized Teager energy for the off-AEDs state averaged over 12 patients, for all locations studied and the RISE. There was a 35% decrease (p < 0.001) over all contacts and a 35% decrease at the RISE (p < 0.05). Normalized values are obtained by dividing the energy estimate for each electrode contact during the off-AEDs state by the estimate during the on-AEDs state and averaging across contacts. (b) Normalized Teager energy for the off-AEDs state for each of the 12 patients, both averaged over all electrode contacts and RISE electrode contacts. There was a decrease in Teager energy in all patients from the on- to off-AEDs state when measured from all electrode contacts and in 9 of 10 patients when measured from the RISE. Energy change at the RISE was not calculated for the 2 patients in whom the RISE was not localized.
158
H.P. Zaveri et al.
Figure 3 An evaluation of Teager energy for every hour of monitoring over 12 days in patient 3. (a) Average energy of all electrode contacts. (b) Energy of a selected RISE electrode contact. AED taper is indicated at the top of each figure by the line drawn next to each AED listed. The solid line corresponds to the AED dose indicated, and increasingly dashed lines depict increasingly lower doses. The patient was admitted on 3 AEDs, 2 of which were tapered. Energy values decreased during the course of monitoring to a low point by day 10 and the patient experienced 8 seizures on day 11. Seizure occurrence is indicated by a cross-shaped marker at the top of each figure. Following these seizures AEDs were restarted and the patient was enrolled in a safety study of a neuro-stimulation device. This resulted in the introduction of noise in the icEEGs, and these data have not been included in this figure. The times of the on- and off-AEDs epochs are indicated by arrows placed next to the time-axis.
before and after seizures are recorded. The energy evaluations in Patient 3 demonstrate reasonably steady values during the on- and off-AEDs states, and a relatively steady transition between the two states. There is a considerable decrease in Teager energy which precedes the expression of seizures. There is also a considerable diurnal variation in energy values, which is more clearly observed in the average of all electrode contacts.
Longitudinal energy measures from the RISE for three more patients are shown in Fig. 4. The three selected studies were relatively long in duration; 19 days for patient 11 (Fig. 4(a)), 21 days for patient 10 (Fig. 4(b)) and 14 days for patient 2 (Fig. 4(c)). The signal energy estimates in Fig. 4(a) peaks thrice, at the start, middle and end of monitoring. Similar to patient 3, illustrated in Fig. 3, the seizures tend to cluster. In contrast to patient 3, though,
A decrease in EEG energy accompanies anti-epileptic drug taper during intracranial monitoring the clusters of seizures occur during times of relatively high energy. In the two other patients we observe a considerable decrease over the course of the intracranial study. The energy profile in Fig. 4(a) demonstrates that trends other than decreases occur. The profile in Fig. 4(b) demonstrates a considerable diurnal modulation of a declining energy trend. The examples presented in Figs. 3 and 4 indicate that though there are temporal changes and differences between patients, the dominant trend is a slow decrease of energy over days apparently coincident with the taper of AEDs.
4. Discussion We have observed a decrease in Teager energy between on- and off-AEDs states during intracranial monitoring. A decrease of 35% was observed over all brain areas studied, and a decrease of 35% was observed at the RISE. These observations of decreases in Teager energy are supported by a decrease in signal power, a more conventional measure, at the RISE and all brain areas studied. These observations are also supported by a decrease in Teager energy and signal power measurements in Set B, that is for the data epochs obtained from 8 patients before the expression of any seizures. These observations are novel because a comparison of icEEG energy estimates between on- and off-AEDs time points has not been reported previously, and a considerable and spatially widespread decrease between them is unexpected.
4.1. Teager energy The signal energy of icEEGs was measured with Teager energy. As indicated above, the conventional measure of signal energy weights contributions from all frequencies of a signal equally. While this is an appropriate measure of energy for electrical circuits, it may not be accurate for many physical systems as these systems expend greater energy when operating at higher frequencies than when operating at lower frequencies. Starting with this observation Kaiser introduced an alternate measure of signal energy, based on the work of Teager, which weights contributions of different frequencies non-uniformly, emphasizing higher frequencies over lower frequencies by a square law weighting (Kaiser, 1990). Teager’s algorithm is attractive because it is simple, physically intuitive, computationally efficient, and amenable to time-varying signals. Importantly to the study of EEGs the weighting performed by Teager’s algorithm may correct the low contribution of higher frequencies in the EEG, which is a possibly result of the process by which EEGs are generated (Zaveri, 1993). In previous reports we have demonstrated that Teager energy is superior to the conventional measure of energy for detecting and demarcating seizures suggesting Teager energy is a better correlate of excitation (Zaveri, 1993; Zaveri et al., 1993). Teager energy has been shown to be positively associated with extracellular glutamate (Pan et al., 2008) and has found utility for detecting interictal activity (Mukhopadhyay and Ray, 1998) lending further support for this measure to serve as a correlate of excitation. Since the underlying hypothesis of this study is that the taper
159
of AEDs should lead to greater excitation we sought to use a measure which was a correlate of excitation. We observe that Teager energy measurements decreased to a lesser extent, between on- and off-AEDs states, than signal power measurements. This may suggest that there is a lower decrease in high-frequencies in comparison to low frequencies. An increase in high-frequency power with AED taper has recently been reported, though for frequencies which are much higher than those studied here and for very short data epochs during slow wave sleep (Zijlmans et al., 2009).
4.2. Confounding factors and limitations It is possible that surgery to place intracranial electrodes alters excitability. Indeed, patients may experience seizures in the immediate post-surgical period and these seizures are commonly discounted during the evaluation to locate the RISE as they are considered unrepresentative. In the series of patients studied here 1 patient experienced a seizure shortly after surgery to place electrodes. Another factor that could confound the measurements is a bioreactive effect to intracranial electrodes. Research conducted on silicon micromachined electrodes employed in brain computer interface (BCI) endeavors report both short-term and long-term bioreactive effects to these electrodes, with long-term effects lasting days to weeks (Bjornsson et al., 2006; Szarowski et al., 2003; Turner et al., 1999). The bioreactive effect to intracranial electrodes, if any, is not known. There are essential differences between electrodes used in BCI endeavors and those employed for intracranial monitoring for epilepsy surgery. The former are introduced within the cortex. The latter, with the exception of depth electrodes are placed subdurally. Here, most of the electrodes used were subdural strip or grid electrodes, with depth electrode contacts representing a small fraction of the total contacts. Another difference includes the now considerable long-term international experience in the use of intracranial EEG electrodes, as compared to the essentially research use of silicon micromachined electrodes. A third factor could be the relatively large number of electrode contacts employed at our center. We do not know the effect of surgery, the effect of anesthesia, and the effect of invasive electrodes on the intracranial EEG. Our lack of knowledge of these factors constitute limitations due to the setting of intracranial monitoring. An additional limitation of this study is that we did not measure serum levels of AEDs. During the design of the study this step was considered unwarranted because we wished to compare the patients at two points during their monitoring. Further, we did not wish to relate EEG changes to exact serum levels but rather we sought to evaluate the difference between two relative levels of AEDs. While it is clear that there is a considerable difference in AED dose between the on- and off-AEDs time points, it remains an assumption of this study that this difference would be reflected in serum levels. While these limitations restrict the general applicability of findings useful observations are still possible in this setting and with this experimental design.
160
H.P. Zaveri et al.
Figure 4 Longitudinal evaluation of energy from a selected RISE contact in patients (a) 11, (b) 10, and (c) 2. The average energy measured for each hour over 19, 21 and 14 days are presented. AED taper is indicated at the top of each figure by the line drawn next to each AED listed. The solid line corresponds to the AED dose indicated, and increasingly dashed lines depict increasingly lower doses. Seizure occurrence is indicated by a cross-shaped marker at the top of each figure. There is considerable diurnal modulation
A decrease in EEG energy accompanies anti-epileptic drug taper during intracranial monitoring
4.3. Energy, AEDs, cortical excitation and seizure generation The results presented in 2, 3, 4(b) and 4(c) demonstrate that a decrease in signal energy continues for several days after surgery. That is, the decrease is not necessarily linked to the initial part of the intracranial study. Fig. 4(a) indicates time-trends other than a steady decrease occur and suggests the decrease in signal energy may not be due to surgery to place intracranial electrodes. One manifestation of a possible bioreactive response could be a change in the interface between intracranial electrode contacts and tissue. If the impedance increases, for example due to gliosis, then the amplitude of the icEEG would decrease over the course of the intracranial monitoring period, and this would result in decreased energy measurements. Impedance is not routinely measured during intracranial studies as it involves the intracranial introduction of current. In a separate study on the safety of a neuro-stimulation device, we monitored the impedance of intracranial contacts over the course of intracranial monitoring. The patients in that study overlapped with some of the patients in this study. The impedance values were found to remain reasonably constant through the course of intracranial monitoring (Duckrow and Tcheng, 2007). Though not performed for all patients and contacts studied here, this suggests that change in impedance is not a factor. Lastly, Shah and coworkers reported a lack of correlation between intracranial pressure and the number of intracranial electrode contacts in a study on 16 children. Though the number of contacts used in that study (Shah et al., 2007) was lower than those used here and the measurand is different, this suggests that the relatively large number of electrode contacts employed in our studies may not be a factor in the decrease observed. The use of signal energy to detect seizures or pre-seizure cursors is based in part upon a long-standing premise in epilepsy that seizures arise from a lack of balance between excitation and inhibition at the RISE, such that excitation overwhelms inhibition and results in seizure. In a separate study we studied intracranial interictal spike counts and observed a striking decrease in spike counts, between 24 h on- and off-AEDS periods, which is similar in some manners to the decrease observed here. The decrease in spikes spanned several days, and occurred for spike counts from all intracranial electrode contacts pooled together. The main difference from this study is that, whereas a spike is a discrete feature observed at some intracranial electrode contacts, signal energy is a continuous valued measure which can be estimated for all electrode contacts. One of the explanations we advanced to explain the decrease of spikes with AED taper is that spikes may be a marker of inhibition and not excitation (Spencer et al., 2008). Evidence from some in vitro animal model and human studies support this interpretation (Avoli, 2001; Bragdon et al., 1992; Engel and Ackermann, 1980; Staley and Dudek,
161
2006; Swartzwelder et al., 1987). The icEEG signal is a composite expression of brain activity, and not just excitatory activity. In this study, as in previous studies, we considered signal energy to represent excitability. If signal energy is re-interpreted to represent inhibition and not excitation, then the decrease in signal energy with AED taper and subsequent expression of seizures is better explained. This re-interpretation of signal energy as a measure of inhibition would suggest that previous attempts to detect and predict seizures with this measure may have been based on inhibition and not excitation. The re-interpretation of signal energy as a surrogate measure of inhibition has not been suggested before and in the absence of other supporting evidence it can only be advanced with caution. We note the time when the patients have been maximally tapered off AEDs represents the time during icEEG monitoring when the patient’s brain is closest to its innate AED-free and seizure-prone state. This represents a time which is also relatively distant from the time of electrode placement. The preliminary indications based on the results presented here are that this state corresponds to a point of relatively low signal energy. We note two factors which support this interpretation. First, the relatively long duration of some of the studies and second, 4 of the patients were fully tapered off AEDs and the remaining patients all had decreased AEDs by the time of the off-AEDs measurement. While this does not imply the effect of AEDs has been fully removed, and further cannot control for a change in tissue response to intracranial electrodes, it suggests a reasonable measure of the off-AEDs state may have been obtained. If low energy is representative of the off-AEDs state, this suggests two surprising and counterintuitive corollaries: (1) AEDs contribute to an increase in icEEG signal energy and (2) seizures can arise from a state of low, and not high, signal energy. In a very general manner we may describe a seizure as an expression of brain instability. Further, we may consider the brain to be more stable when on AEDs and less stable when off AEDs, as in this latter state the patient is more likely to suffer seizures. We suggest that the following states during intracranial monitoring represent ever increasing instability in the sense that the probability of a seizure is increasingly higher or the patient is in seizure: (1) patient is on AEDs, (2) patient is off AEDs, (3) pre-seizure, and (4) seizure. Here we have focused on the first two states. By comparing these two states we sought to examine if cortical excitation, as measured by signal energy, increases with AED taper. We observed a large dynamic in energy values marking different points of an activation curve, at times in the presence of considerable diurnal variation, during intracranial monitoring such that seizures are more likely from low and not high signal energy. We interpret the observed changes, under the qualifications stated above, to reflect a decrease in brain excitation between the on- and off-AEDs states in the RISE and other brain areas. Further we consider this decrease to mark the increased instability of the
of icEEG energy in patient 10, and to a lesser extent in patients 11 and 2. The icEEG energy profile of patient 11 is considerably different from those of patients 3, 10 and 2, who all show a decrease over days. In patient 11 we observe three periods of high energy which are coincident with clusters of seizures. The times of the on- and off-AEDs epochs for these patients are indicated by arrows placed next to the time-axis. In patient 2, the second and third arrows placed next to the time-axis indicate the time of the off-AEDs epoch for Set B and Set A, respectively.
162 off-AEDs state and increased propensity for seizures during this state. If these observations and interpretations are confirmed then they suggest a more complex route to seizures than poorly controlled excitation. These findings suggest the use of energy as a quantitative measure for increased propensity for seizures and may suggest a role in seizure generation for the hypometabolism defined by FDG-PET studies (Vinton et al., 2007; Willmann et al., 2007). Because the icEEG energy change between the on- and off-AEDs states is large, spatially widespread and occurs over several days it should be measurable by modalities other than the icEEG.
5. Conclusions We report a decrease in icEEG signal energy with AED taper in the RISE and over all brain areas monitored. The decrease in energy was observed for a number of AED types, AED combinations, AED tapers and RISE locations, and thus may suggest some level of independence of these factors. The observation of a decrease in icEEG energy with AED taper leads to two surprising corollary observations. First, that AEDs are positively correlated with signal energy, and thus with excitability. Second, that seizures may occur from a state of low and not high signal energy. These observations suggest AEDs may prevent seizures in other ways than reduction of excitation and seizure generation may be influenced by factors other than poorly regulated cortical excitation.
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.eplepsyres. 2009.06.002.
References Avoli, M., 2001. Do interictal discharges promote or control seizures? Experimental evidence from an in vitro model of epileptiform discharge. Epilepsia 42 (Suppl. 3), 2—4. Bjornsson, C.S., Oh, S.J., Al-Kofahi, Y.A., Lim, Y.J., Smith, K.L., Turner, J.N., De, S., Roysam, B., Shain, W., Kim, S.J., 2006. Effects of insertion conditions on tissue strain and vascular damage during neuroprosthetic device insertion. Journal of Neural Engineering 3 (3), 196—207. Bragdon, A.C., Kojima, H., Wilson, W.A., 1992. Suppression of interictal bursting in hippocampus unleashes seizures in entorhinal cortex: a proepileptic effect of lowering [K+]o and raising [Ca2+]o. Brain Research 590 (1—2), 128—135. Duckrow, R.B., Tcheng, T.K., 2007. Daily variation in an intracranial EEG feature in humans detected by a responsive neurostimulator system. Epilepsia 48 (8), 1614—1620. Engel Jr., J., Ackermann, R.F., 1980. Interictal EEG spikes correlate with decreased, rather than increased, epilepto-
H.P. Zaveri et al. genicity in amygdaloid kindled rats. Brain Research 190 (2), 543—548. Kaiser, J.F., 1990. On a simple algorithm to calculate the energy of a signal. In: Proceedings of ICASSP, pp. 381—384. Mormann, F., Elger, C.E., Lehnertz, K., 2006. Seizure anticipation: from algorithms to clinical practice. Current Opinion in Neurology 19 (2), 187—193. Mukhopadhyay, S., Ray, G.C., 1998. A new interpretation of nonlinear energy operator and its efficacy in spike detection. IEEE Transactions on Biomedical Engineering 45 (2), 180—187. Pan, J.W., Williamson, A., Cavus, I., Hetherington, H.P., Zaveri, H.P., Petroff, O.A.C., Spencer, D.D., 2008. Neurometabolism in human epilepsy. Epilepsia 49 (Suppl. 3), 31—41. Shah, A.K., Fuerst, D., Sood, S., Asano, E., Ahn-Ewing, J., Pawlak, C., Chugani, H.T., 2007. Seizures lead to elevation of intracranial pressure in children undergoing invasive EEG monitoring. Epilepsia 48 (6), 1097—1103. Spencer, S.S., Goncharova, I.I., Duckrow, R.B., Novotny, E.J., Zaveri, H.P., 2008. Interictal spikes on intracranial recording— –behavior, physiology and implications. Epilepsia 49 (11), 1881—1892. Spencer, S.S., Nguyen, D., Duckrow, R.B., 2009. Invasive EEG in presurgical evaluation of epilepsy. In: Shorvon, S., Perucca, E., Engel Jr., J. (Eds.), The Treatment of Epilepsy. Blackwell, Oxford, pp. 767—798. Staley, K., Dudek, F., 2006. Interictal spikes and epileptogenesis. Epilepsy Currents 6, 199—202. Swartzwelder, H.S., Lewis, D.V., Anderson, W.W., Wilson, W.A., 1987. Seizure-like events in brain slices: suppression by interictal activity. Brain Research 410 (2), 362—366. Szarowski, D.H., Andersen, M.D., Retterer, S., Spence, A.J., Isaacson, M., Craighead, H.G., Turner, J.N., Shain, W., 2003. Brain responses to micro-machined silicon devices. Brain Research 983 (1—2), 23—35. Turner, J.N., Shain, W., Szarowski, D.H., Andersen, M., Martins, S., Isaacson, M., Craighead, H., 1999. Cerebral astrocyte response to micromachined silicon implants. Experimental Neurology 156 (1), 33—49. Vinton, A.B., Carne, R., Hicks, R.J., Desmond, P.M., Kilpatrick, C., Kaye, A.H., O’Brien, T.J., 2007. The extent of resection of FDGPET hypometabolism relates to outcome of temporal lobectomy. Brain 130 (Pt 2), 548—560. Willmann, O., Wennberg, R., May, T., Woermann, F.G., PohlmannEden, B., 2007. The contribution of 18F-FDG PET in preoperative epilepsy surgery evaluation for patients with temporal lobe epilepsy A meta-analysis. Seizure 16 (6), 509—520. Zaveri, H.P., 1993. Generation and propagation of seizures in temporal lobe epilepsy. Ann Arbor, University of Michigan. Zaveri, H.P., 2007. Designing a prediction algorithm: definitions and caveats. Third International Workshop on Seizure Prediction, Freiburg, Germany. Zaveri, H.P., Williams, W.J., Sackellares, J.C., 1993. Energy based detection of seizures. In: Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, USA, IEEE. Zijlmans, M., Jacobs, J., Zelmann, R., Dubeau, F., Gotman, J., 2009. High-frequency oscillations mirror disease activity in patients with epilepsy. Neurology 72 (11), 979—986.
