Ripple classification helps to localize the seizure-onset zone in neocortical epilepsy

Authors


Address correspondence to Norman K. So, Epilepsy Center, Desk S51, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, U.S.A. E-mail: son2@ccf.org

Summary

Purpose:  Fast ripples are reported to be highly localizing to the epileptogenic or seizure-onset zone (SOZ) but may not be readily found in neocortical epilepsy, whereas ripples are insufficiently localizing. Herein we classified interictal neocortical ripples by associated characteristics to identify a subtype that may help to localize the SOZ in neocortical epilepsy. We hypothesize that ripples associated with an interictal epileptiform discharge (IED) are more pathologic, since the IED is not a normal physiologic event.

Methods:  We studied 35 patients with epilepsy with neocortical epilepsy who underwent invasive electroencephalography (EEG) evaluation by stereotactic EEG (SEEG) or subdural grid electrodes. Interictal fast ripples and ripples were visually marked during slow-wave sleep lasting 10–30 min. Neocortical ripples were classified as type I when superimposed on epileptiform discharges such as paroxysmal fast, spike, or sharp wave, and as type II when independent of epileptiform discharges.

Key Findings:  In 21 patients with a defined SOZ, neocortical fast ripples were detected in the SOZ of only four patients. Type I ripples were detected in 14 cases almost exclusively in the SOZ or primary propagation area (PP) and marked the SOZ with higher specificity than interictal spikes. In contrast, type II ripples were not correlated with the SOZ. In 14 patients with two or more presumed SOZs or nonlocalizable onset pattern, type I but not type II ripples also occurred in the SOZs. We found the areas with only type II ripples outside of the SOZ (type II-O ripples) in SEEG that localized to the primary motor cortex and primary visual cortex.

Significance:  Neocortical fast ripples and type I ripples are specific markers of the SOZ, whereas type II ripples are not. Type I ripples are found more readily than fast ripples in human neocortical epilepsy. Type II-O ripples may represent spontaneous physiologic ripples in the human neocortex.

Interictal high-frequency oscillations (HFOs, 80–500 Hz) are electroencephalography (EEG) signals that provide additional diagnostic information in interpretation of the intracranial EEG in patients with intractable epilepsy (Engel et al., 2009; Gotman, 2010). Recently HFO research has attracted much attention because the signal can be readily detected by commercially available macroelectrodes and has been shown to help localize the seizure onset or epileptogenic zone (Urrestarazu et al., 2007; Worrell et al., 2008; Crépon et al., 2010; Jacobs et al., 2010; Wu et al., 2010; Akiyama et al., 2011). HFOs are differentiated into “ripples” (80–250 Hz) and “fast ripples” (250–500 Hz). The fast ripple is considered more pathologic because of its close association with regions capable of generating spontaneous seizures (Engel et al., 2009). The ripple is found to be less specific whether in mesial temporal (Urrestarazu et al., 2007) or neocortical epilepsy (Akiyama et al., 2011). One possible reason could be that the brain is capable of generating spontaneous physiologic ripples.

Many studies in animals and humans showed that the hippocampus can generate spontaneous physiologic ripple activity that is associated with memory consolidation (Girardeau & Zugaro, 2011). In contrast, spontaneous physiologic ripple activity in the neocortex is seldom reported. Spontaneous ripple activity has been observed in sensory and association cortices of the cat, more often during slow-wave sleep (Grenier et al., 2001). Spontaneous physiologic ripples were recently identified in human visual cortex (Nagasawa et al., 2012). There are further reports of HFO activity induced in specific neocortical areas by sensory stimulation or during language and motor tasks in humans (see Discussion). Sensory-evoked fast ripples with a frequency up to 600 Hz can be recorded in the animal neocortex (Kandel & Buzsáki, 1997). Such physiologic fast activities were time-locked to stimulation or task and identified by power spectrum analysis after averaging.

Fast ripples are infrequently detected in adult neocortical epilepsy (Jacobs et al., 2008; Crépon et al., 2010; Blanco et al., 2011; Cho et al., 2012). Here we classified ripples in neocortical epilepsy based on associated characteristics to identify “more pathologic” ripples, which can better localize the SOZ. The majority of ripples or fast ripples occur simultaneously with an interictal epileptiform discharge (IED) in the conventional EEG (Urrestarazu et al., 2007; Jacobs et al., 2008;  Crépon et al., 2010). Because the IED is not a normal physiologic event, we take the simple premise that HFO activity that occurs simultaneously with an IED could be more pathologic.

Methods

Patients and sampling of invasive EEG

This studied was approved by the Cleveland Clinic Institution Review Board (Study number 10-171). The 35 patients with medically intractable neocortical focal epilepsy underwent invasive EEG by either stereotactic EEG (SEEG, 21 patients) or subdural grid and supplementary depth electrodes (14 patients) from July 2007 to September 2011. The depth electrodes (Integra, Plainsboro, NJ, U.S.A.) were composed of 8–16 cylindrical platinum contacts 2.5 mm long, 0.8–1.1 mm in diameter, and 4–5 mm apart center to center. Fourteen patients were implanted with subdural grid and strip electrodes (stainless steel), with 0–4 depth electrodes (same depth electrodes used in SEEG). Each subdural contact had a diameter of 4 mm with a center-to-center separation of 10 mm. A three-dimensional reconstruction map was done to show the location of electrode contacts after implantation.

Intracranial EEG was analyzed by at least two epileptologists using routine methods. The areas showing the earliest intracranial EEG change in a seizure such as onset of repetitive spikes, background suppression, or paroxysmal fast activity was defined as the seizure-onset zone (SOZ). The primary propagation area (PP) was defined as the cortex next involved in the seizure within 3 s after EEG onset. Twenty-one patients were classified as focal neocortical epilepsy (group 1, Table S1), since the SOZ was consistently localized in one neocortical region by SEEG in 11, and subdural EEG in 10. The other 14 patients were defined as multiregional or nonlocalizable (group 2), since the seizure onset was found at two or more different neocortical regions or the onset pattern was too diffuse to permit a resective strategy. In group 2, 10 patients were investigated by SEEG and four by subdural EEG (Table 1).

Table 1.   Summary of Epilepsy types and invasive EEG methods in all patients
 Seizure onset patternSEEGGrid and depthSubtotal
Group 1One defined SOZ111021
Group 2Non-localizable/multiregional10414
Total 211435

After several seizures were recorded routinely on reduced or no antiepileptic medications (band width 0.08–500 Hz), the digital sampling rate was increased to 2,000 Hz (band width, 0.016–2,000 Hz) for an additional 2–7 days to obtain another seizure wherever possible while still on reduced to no medications. This entailed limiting recording to 64 channels, as this is the maximal capacity at 2,000 Hz sampling for our EEG system (Nihon Kohden, Foothill Ranch, CA, U.S.A.). Up to 62 contacts were selected to record from the SOZ, PP, and other cortical areas. The contacts in the SOZ and PP were selected with priority (Results and Table S2), but more than two thirds of contacts recorded from neighboring or more distant areas. A few channels were used for scalp electrodes and electrocardiography. For SEEG evaluation, we sacrificed the electrode contacts in the white mater to maximize cortical coverage.

Visual inspection of interictal HFOs assisted by time-frequency analysis

Intracranial EEG was exported to the Multiview analysis software (Nihon Kohden). Multiview digital filters are of Butterworth type. The analysis was performed in a referential montage using a scalp reference electrode. We selected artifact-free interictal EEG in slow-wave sleep as revealed on the scalp leads lasting 10–30 min. The selected EEG segments were at least 4 h after a seizure to avoid postictal change. The EEG was displayed at 1 s per page, and high-pass filter cutoff was sequentially set to 1.6, 80, and 250 Hz (zero phase shift) to mark the epileptiform discharge (spike, polyspike, sharp wave, or paroxysmal fast activity), ripple, and fast ripple. The HFOs were identified according to the methods of Jacobs (2008). A candidate HFO contains at least five consecutive oscillations (Figs. 1 and S1), and two events were defined when separated by at least two non-HFO oscillations. A candidate ripple was marked if an event was distinct from the background on the 80-Hz cutoff but not on the 250-Hz cutoff. A candidate HFO has to be recurrent and with a frequency no lower than 2 per 10 min.

Figure 1.


Illustration of type I and II ripples and their time frequency analysis. Top panel shows three types of HFO events. Note different amplitude calibrations. Type I ripple is superimposed on a spike or a fast activity. The fast activity here is sharply contoured beta waves found interictally and which evolved into an ictal pattern in a patient with parietal lobe epilepsy. Type II ripple occurs independently of epileptiform discharges. The middle panel shows the signals after high-pass filtering. Events are manually marked and its parameters automatically calculated by Multiview software, as exemplified in the type II ripple. Bottom panels show the time frequency analysis for all three ripples, demonstrating isolated peaks at 100–150 Hz range.

The EEG segments containing the candidate HFO further underwent time frequency analysis using Morlet wavelet decomposition. The analysis was performed using MATLAB 6.9 (The Mathworks Inc., Natick, MA, U.S.A.). The input of the time-frequency analysis is unfiltered data. The power of the oscillation is depicted by a color scale code. The time-frequency map of each candidate HFO had to show a primary isolated peak in the frequency range of 80–500 Hz to qualify as a true HFO event. If a candidate HFO showed broad band increase in the frequency range, it was not confirmed because it might represent “false HFO” caused by filtering artifact (Bénar et al., 2010).

Quantitative analysis and classification of ripples

The frequency, amplitude, and the duration of each confirmed HFO was calculated in Multiview software. The amplitude is defined as the peak-to-peak value of the highest oscillation in the HFO. In addition, the number of HFOs is counted for each channel. If the rate of HFO was <1 per min in one channel, a longer EEG was reviewed. The HFO rate was normalized to the number of events per 10 min. All neocortical ripples were classified into two types: type I ripples superimposed on an IED, and type II ripples, which occurred independent of IED.

Statistical analysis

By using the Mann-Whitney U test, we compared (1) the rate and density of the events in different regions, (2) the proportion of type I ripples and spikes in different regions, (3) the difference of parameters between type I and type II ripples, and (4) age difference between fast ripple (or type I ripple) positive and negative patients. These analyses were done using SPSS 17.0 (IBM, Armonk, NY, U.S.A.). The significance level was set to p < 0.05.

Results

HFO distribution in Group 1

Three of the 21 patients with localized onset did not proceed with surgery because of a high risk of functional deficit. The other 18 patients underwent resective surgery and were followed up for 5–36 months (median 12 months). The postsurgical pathology showed focal cortical dysplasia in 14 patients (77.7%). Thirteen patients (72%) were seizure free after surgery. The mean number of all electrode contacts in a patient is 134.5 ± 4.9 (Table S2), including 8.0 ± 1.3 contacts in SOZ, 10.0 ± 2.4 contacts in PP, and 116.5 ± 6.1 contacts in other areas. During 2,000 Hz EEG sampling, we recorded from 59.1 ± 0.6 contacts in each patient, which included 8.0 ± 1.3 contacts in SOZ, 9.8 ± 2.4 contacts in PP, and 41.3 ± 3.1 contacts in other areas.

Neocortical fast ripples were detected in the SOZ of four patients (19%), with a median rate of 78 per 10 min (range, 9–162), frequency of 339 Hz (256–500 Hz), duration of 31 msec (16–70 msec), and amplitude of 6.1 μV (3.6–15 μV). Neocortical fast ripples (96.5%) were associated with epileptiform discharges; others occurred independently. Neocortical fast ripples (68.1%) were superimposed on ripples. The four patients showing fast ripples were younger in age and age at seizure onset as compared to the other 17 patients without fast ripples (Fig. S2). But the duration of seizure history, postsurgical pathology, and seizure outcome were not different. In one patient, fast ripples were detected in the mesial temporal structures, which were in the PP, but no fast ripples were detected in the SOZ.

Type I ripples accounted for 56.9% of all ripples and were identified in 14 cases (66.7%). Type II ripples accounted for the other 43.1% of ripples. Type I ripples were in 76.8% associated with spike, polyspike, or sharp wave activity, and in 23.2% associated with paroxysmal fast activity. A single ripple event could be recorded in up to one to three neighboring contacts in depth electrodes and one to two neighboring contacts on subdural grid electrodes.

To correct for the bias that electrodes recording from the SOZ (100%) and PP (98.9%) were always included during 2,000 Hz sampling compared with other areas (35.4%, Table S2), we calculated the event density (the rate of an event in each area divided by the number of assigned electrodes for the area) in addition to the event rate (the total number of events/10 min from all electrodes in the area). The rate and event density of both type I ripple and spikes were significantly higher in the SOZ (Fig. 2) and PP than other areas, and higher in SOZ than PP. Type I ripples were more specifically confined to the SOZ as compared to spikes (Fig. 3). By contrast, neither the rate nor the event density of type II ripples was significantly different among the different areas.

Figure 2.


The distribution of type I ripples, type II ripples, and spikes in SOZ, PP, and other areas in group 1. Event rate is number of events per 10 min. Event density is event rate divided by the number of electrode contacts in the region. Data were presented in box plot. The extreme values are denoted by blank circles beyond the whiskers; n = 14 for type I ripples, n = 18 for type II ripples, and n = 21 for spikes. Mann-Whitney U test was used to compare group difference. The level of statistical significance is shown as *p < 0.05 and **p < 0.01 for comparison between the SOZ/PP and other areas, and as #p < 0.05 and ##p < 0.01 for comparison between the SOZ and PP.

Figure 3.


The proportion of type I ripples and spikes distributed in the SOZ, PP, and other areas in group 1. The extreme values are denoted by blank circles beyond the whiskers; n = 14 for type I ripples and n = 21 for spikes. Mann-Whitney U test was used to compare group difference. The level of statistical significance is shown as $p < 0.05 for comparison between type I ripple and spike. Distribution in each area was also analyzed independently for type I ripples and spikes: **p < 0.01 for comparison between the SOZ and other areas, ##p < 0.01 for comparison between the SOZ and PP.

Type I ripple showed higher frequency (p = 0.019), shorter duration (p = 0.000), and higher amplitude (p = 0.000) compared with type II ripple (Fig. 4). But the two types of ripples greatly overlapped in each parameter; and no cut-off value of a parameter could separate the majority of one ripple type from another.

Figure 4.


The normalized histograms of the percentage of all type I (gray bar) and II ripples (blank bar) sorted by frequency, duration, and amplitude in group 1 patients. The medians for type I and type II ripples were marked by a gray arrow or blank arrow, respectively.

HFO distribution in Group 2

Most of these 14 patients did not receive resective surgery, since they did not have a single discrete SOZ. Neocortical fast ripples were recorded in the areas showing the earliest EEG change in two patients. In four patients, fast ripples were found in mesial temporal structures, which were involved by early seizure spread, but no neocortical fast ripples were detected. Type I ripples were detected in 8 of 14 patients (57.1%), appearing in several widely separated areas in four of them. The event density of type II ripples was similar in the presumed SOZ, PP, and other areas.

Type II-O ripple in SEEG

To find potential physiologic ripple with less epileptogenicity, we identified type II-O (outside) ripple by the following two criteria: (1) it occurred at an electrode contact that had only type II and no type I ripple; and (2) at sites away from and noncontiguous to electrodes recording from SOZ and PP. Our observation is restricted to SEEG cases because in SEEG evaluation the electrodes were usually widely separated, and more electrodes were far away from the SOZ as compared with subdural grid evaluation. Type II-O ripples were detected in 13 of 21 patients from both groups 1 and 2 evaluated by SEEG. They were consistently detected in the precentral gyrus and pericalcarine area, in the paracentral lobule, lateral occipital region with less consistency, and less commonly in the posterior part of the premotor area, postcentral gyrus, and precuneus (Fig. 5). No type II-O ripple was identified in other areas sampled by multiple electrodes.

Figure 5.


Anatomic distribution of type II-O ripples in all patients evaluated by SEEG. The location of electrode contacts was projected to a schematic brain atlas. If ripples were detected simultaneously in the neighboring contacts, only the location of contacts with the highest ripple amplitude was shown. The location of electrode contacts showing type II-O ripples: red circles in the motor and premotor areas, green circle in the postcentral gyrus, yellow circles in the occipital region. The blank circles denote location of electrode contacts without ripples in the above regions. Other electrode sites sampling from many other cortical regions showed no type II-O ripples and were not displayed.

Nine patients showed type II-O ripples in the motor or premotor cortex, which was confirmed by functional mapping using cortical stimulation (Fig. S3). Type II-O ripples were consistently detected in the primary motor cortex (Table 1), in which the ripples occurred more frequently in the hand/arm cortex (49.6 ± 7.8/10 min) than the foot/leg cortex (21.5 ± 3.5/10 min, p = 0.027).

Six patients had type II-O ripples in the occipital region. Type II-O ripples were consistently detected in the cuneus, which was confirmed as primary visual cortex by cortical stimulation in four patients (Fig. S4). The rate of type II-O ripples was higher in the cuneus (28.5 ± 2.1/10 min) than in other occipital regions (15.1 ± 1.5/10 min, p = 0.042).

Discussion

Neocortical fast ripples existed only in the SOZ, in agreement with prior reports, but were uncommonly detected in our group 1 patients. By contrast, interictal fast ripples were found in most pediatric patients with neocortical epilepsy using electrodes similar to ours (Wu et al., 2010; Akiyama et al., 2011). Technical issues such as electrode size or HFO identification method should not be the primary factors responsible for the low rate of fast ripple detection in our study, as we found fast ripples in 9 of 10 patients with medial temporal epilepsy (authors’ unpublished data). In addition, some of our patients showed fast ripples in the mesial temporal structures outside of the SOZ. Our data suggest that epileptic neocortex cannot readily support fast ripple generation as in mesial temporal structures, or that the generators of neocortical fast ripples may be too small to be sampled easily. Another possibility is that the patient’s age plays a role in fast ripple generation. The patients with fast ripples were younger in age or age at seizure onset compared to the other patients without fast ripples. Others also found that fast ripples were rarely detected (Blanco et al., 2011; Cho et al., 2012) or not detected (Crépon et al., 2010) in adult neocortical epilepsy.

Type I neocortical ripples were more abundant than fast ripples and detected in 66.7% of group 1 patients. They marked a high level of epileptogenicity and were significantly correlated to the SOZ or PP, whereas type II neocortical ripples were not. Type I ripple was a better marker of the SOZ than interictal spikes alone (Figs 2 and 3) because it was more specifically confined to the SOZ or PP as compared to spikes that were often found outside. Nevertheless interictal spikes showed a stronger correlation to the SOZ or PP than type II ripple. This comparison was made of ripples and spikes during slow-wave sleep. Because slow-wave sleep is known to increase the spatial distribution of spikes, an analysis of spikes in wakefulness may turn out to have a better correlation to the SOZ. The electrode contacts showing type I ripple, if outside the motor, somatosensory, and visual cortices, had only a small proportion of type II ripples (10.5%). Both fast ripple and type I ripple activity could also be found in group 2 patients who did not have a single localized SOZ. They were found in presumed primary propagation areas and in multiple regions. Therefore, although they are good markers of epileptogenicity, they cannot be taken out of context and considered in isolation without other electrophysiologic data.

Our data are in accordance with previous studies that interictal ripples are often (60–95% of all ripples) nested in spikes (Urrestarazu et al., 2007; Schevon et al., 2009; Bragin et al., 2010; Crépon et al., 2010). If pathologic HFO reflects hyperactive population spike firing (Grenier et al., 2003; Bragin et al., 2007) and the IED reflects hypersynchronous postsynaptical potential, the concurrence of HFO and IED suggests that the neurons in the recording area are capable of both receiving hypersynchronous inputs and generating burst firing that can be transmitted to postsynaptic targets (Bragin et al., 2010). This property may be crucial for the generation and propagation of seizure activity.

Type II-O ripple was identified for the purpose of our study when seen in an area outside the SOZ, which showed only type II and no type I ripples. It represents the ripple with the least level of epileptogenicity. Of interest, type II-O ripples were most consistently located in the primary motor cortex and primary visual cortex. Their specific anatomic distribution may suggest they are related to motor and sensory physiology rather than an extension of the epileptic network. In general, type II-O ripples had a longer duration (up to 300 msec) and smaller amplitude compared with type I ripples. Our findings are supported by a recent report (Nagasawa et al., 2012) that spontaneous physiologic ripples with a mean duration of 300 msec were detected by subdural grid electrodes in the human visual cortex. In addition, visual task activation increased ripple power over a larger area of the visual cortex including more rostral sites, suggesting that the ripples were involved in visual information processing. The recognition of potential spontaneous physiologic ripples should help improve the analysis of HFO in invasive EEG.

EEG oscillations in the high gamma or low ripple frequency range, so-called “high gamma oscillation” (HGO), can be evoked in human subjects by sensory stimulation, motor task, or language task in the corresponding cortical areas: visual cortex (Asano et al., 2009), auditory cortex (Edwards et al., 2005), motor cortex (Darvas et al., 2010), and language cortex (Sinai et al., 2005). The HGOs usually involved a larger cortical area, whereas type II-O ripples seem confined to a smaller area during slow wave sleep in the absence of motor output and sensory input. The cellular mechanism of HGO may be related to increased cellular and synaptic activity. Type II-O ripples may share a similar generator with HGOs, since there is no absolute “activated state” and “resting state” of brain.

Our findings suggest that classification of interictal ripples helps to localize the SOZ in neocortical epilepsy. However, our study had several limitations. First, our results may be specific to pathology, since most of our patients had cortical dysplasia with no or subtle magnetic resonance imaging (MRI) lesion. Second, temporal sampling of EEG was limited, since visual inspection of HFO events is a time-consuming task. Third, our EEG was sampled after seizures and major reduction or discontinuation of antiepileptic drugs, a condition that is known to increase spikes and HFO but can do so in ways making a complex interaction possible. Further study is required to examine the physiologic nature of type II-O ripples.

Acknowledgments

We thank Dr. Chong H. Wong, Aleksandar J. Ristić, Rei Enatsu, Richard Burgess, and Tim O’Connor and all the EEG technicians for their help with this study. We also thank David Jones for his contribution to the time frequency analysis.

Disclosure

None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Ancillary