Our main finding is that HFOs reflect different grades of epileptic disease activity in FCD. As hypothesized, patients with FCD type 2 had more severe epilepsy than those with type 1. Respectively, rates of HFOs were higher in the severe FCD type 2 than in the mild FCD type 1. In addition, we could confirm in this large group of patients with FCD findings of former studies (Urrestarazu et al., 2007; Jacobs et al., 2008, 2009) in which HFOs are closely related to the site of seizure onset and the removal of HFO-generating brain tissue correlates with the postsurgical outcome (Jacobs et al., 2010a; Wu et al., 2010; Akiyama et al., 2011).
In our study, we used macroelectrodes, the standard EEG equipment for clinical purposes, to record HFOs. With use of macroelectrodes, the signal-to-noise-ratio between pathologic HFOs and the EEG background activity might be reduced. It is thus a subject of discussion whether recordings with macroelectrodes may underestimate the number of HFOs (Worrell et al., 2008; Châtillon et al., 2011). Independent of this, several studies have successfully investigated rates of HFOs using macrocontacts. In the present study the same electrode type was used for all patients. It is therefore unlikely that the contact size systematically biased the results.
HFOs were recorded using a sampling rate of 1,024 Hz, which is low compared to previous studies (Urrestarazu et al., 2007; Jacobs et al., 2008, 2009; Zelmann et al., 2009; Zijlmans et al., 2009; Jacobs et al., 2010a). This might result in limited recording of HFOs, particularly of fast ripples. However, there is evidence that HFOs of frequencies below 300 Hz sampled at <1,024 Hz are specific for areas of seizure onset (Ochi et al., 2007; Worrell et al., 2008; Crépon et al., 2010).
HFOs are detected more often during slow wave sleep (Staba et al., 2004; Bagshaw et al., 2009). In addition, artifacts caused by motor cortex activity are less frequent during sleep. As it remains unclear whether HFO activity might change in the periictal period, we tried to exclude a possible influence by using segments at least 2 h apart from any seizure. EEG segments were also chosen at the beginning of monitoring when the potential increase in seizure frequency due to drug withdrawal had not yet taken place. In our patient group, seizures during daily life were more common in patients with FCD type 2 than type 1. Therefore, we analyzed whether the seizure frequency during the recording period was significantly different for both groups and whether seizure frequency correlated with HFO rates in each patient. No interaction could be seen, which further supports our hypothesis that pathology and baseline disease activity is reflected by the difference in HFO rates between patients with FCD types 1 and 2.
HFOs were marked in an EEG segment of 3 min duration that is sufficient to ensuring stable rates (Zelmann et al., 2009). The range of frequencies set for the identification and distinction of ripples and fast ripples implies 200 Hz as lower limit for fast ripples according to Staba et al. (2002).
Because FCD is associated with frequent polyspikes, excluding a potential influence exerted by spikes on our analysis was a major concern that could be addressed only by visual marking. We therefore analyzed HFOs occurring without spikes separately and demonstrated that they led to the described results independently of spikes. To ensure stable rates, the concordance between HFOs that were visually marked by two reviewers was assessed using Cohen's kappa coefficient for each contact. After visual analysis was performed, a radial basis function neural network automatically detected HFOs in a copy of the same iEEG segment to additionally reduce reviewer bias. Our comparison between automatic detections and visual analysis provided evidence that both methods would have resulted in similar differences of HFO rates between types of FCDs (Dümpelmann et al., 2012). In principle, both automatic and visual assessments have advantages and disadvantages. For the present study, visual analysis was the preferred choice of method because automatic detection does not allow correction for channels with polyspikes, which may result in false identification of HFOs (Bénar et al., 2010; Crépon et al., 2010). Conversely, future clinical studies will rely heavily on automatic detection, as prospective studies using visual identification will be impossible owing to time limitations.
It has to be noticed that in our study the SOZ was defined before surgery by a clinical epileptologist, whereas analysis of HFOs was completed retrospectively after surgery.
Because the new International League Against Epilepsy (ILAE) classification of cortical malformations (Blümcke et al., 2011) had not yet been introduced during the time of our study, FCDs were categorized according to Palmini et al. (2004) into types 1a and b and 2a and b. However, because none of the FCDs included fulfilled the criteria of the new category FCD type 3, using the new classification would not have led to different results.
HFOs and epileptogenicity
HFOs have been investigated extensively as new markers of epileptogenicity over recent years. They are more specific indicators of the SOZ than spikes (Jacobs et al., 2008; Crépon et al., 2010) and might also identify epileptic areas outside the SOZ (Jacobs et al., 2010b). The strongest evidence for the validity of HFOs as markers of epileptogenic areas have been several independent observations that connect the surgical removal of HFO generating tissue with the postsurgical outcome (Ochi et al., 2007; Jacobs et al., 2010a; Wu et al., 2010; Akiyama et al., 2011; Nariai et al., 2011). The present results again confirm these observations and the potential value of HFOs in the presurgical evaluation. Because HFOs can be spontaneously generated by the nonepileptogenic eloquent cortex during slow-wave sleep (Blanco et al., 2011; Nagasawa et al., 2012), the specificity of HFOs as a biomarker of epileptogenicity remains to be determined. Clear differentiation between physiologic and pathologic HFOs is especially difficult in sensorimotor cortices (Fukuda et al., 2008), as in the case of patients 12 and 13.
Although fast ripples were nonetheless closely linked to epileptic areas in both studies of microelectrodes and macroelectrodes (Bragin et al., 2002; Staba et al., 2007; Engel et al., 2009), ripples were found to be pathologic only in the most recent studies (Jacobs et al., 2008; Worrell et al., 2008; Jacobs et al., 2010a). The results from the studies discussed earlier suggest that physiologic as well as pathologic ripples may exist (Engel et al., 2009; Wang et al., 2012) and differentiation by frequency band is not sufficient to distinguish between both. In the present study, ripples occurred over more widespread areas than fast ripples. Therefore, ripples correlated with the postsurgical outcome to a lesser extent than fast ripples. However, the majority of the described significant differences were observed for both ripples and fast ripples.
In addition, several studies suggest that rates of HFOs may be dependent on the epileptic disease activity of the underlying tissue and vary largely between patients (Jacobs et al., 2008; Akiyama et al., 2011). Moreover, rates of HFOs increase when antiepileptic drugs are reduced (Zijlmans et al., 2009). From a clinical perspective it is important to understand the variability of rates of HFOs so that they may eventually become a useful predictive marker of postoperative outcome. FCD is a frequent finding in patients with refractory epilepsy, and depending on the pathology of FCD the extent of dysplastic tissue and cortical disorganization may vary largely. For this reason we have chosen a homogenous group of patients with different types of FCD to evaluate whether patients with more severe dysplastic changes and thus higher epileptogenic disease activity generate more HFOs than those with less severe lesions.
Epileptogenicity in FCD
There is evidence that the different types of FCD classified according to Palmini show different grades of epileptogenicity (Kloss et al., 2002; Boonyapisit et al., 2003; Kral et al., 2003; Lawson et al., 2005; Widdess-Walsh et al., 2005; Lerner et al., 2009; Palmini et al., 2010). FCD type 2 occurs at an earlier stage during ontogenesis than type 1 (Cepeda et al., 2006), explaining the more undifferentiated and dysmorphic cells. Histologic abnormalities in FCD type 2 resulting in a loss of normal cell physiology are more clearly visible in MRI than in FCD type 1: FCD type 1 reveals only subtle signal changes being thereby frequently overseen, whereas FCD type 2 presents with increased cortical thickness and pronounced blurring of the gray and white matter junction (Colombo et al., 2009; Krsek et al., 2009; Lerner et al., 2009). In the present patient group, the diagnosis of FCD type 1 could be made from MRI findings in only half of the patients, whereas 85% of patients with FCD type 2 had a visible MRI lesion. The first clinical manifestation of epilepsy occurs at an earlier age in FCD type 2 and with a higher frequency of seizures (Lawson et al., 2005; Lerner et al., 2009; Palmini et al., 2010). Our study confirms these findings due to the observation of earlier mean first-time manifestation as well as higher rates of seizures in FCD type 2. In iEEG, higher rates of repetitive spiking patterns were found in patients with FCD type 2a than in type 1 (Boonyapisit et al., 2003).
FCD and HFOs
It was not the aim of this study to differentiate between lesional and nonlesional areas and their overlap with the SOZ. Even though we could confirm previous findings that HFOs are more linked to the SOZ than the lesion (Jacobs et al., 2009), the main aim was to investigate whether rates of HFOs in general were higher in patients with FCD type 2 compared to type 1. Studies analyzing HFOs in patients with FCD are rare and inconsistent (Jacobs et al., 2009; Brázdil et al., 2010). In a previous study, one patient with FCD with high rates of HFOs stood out of a heterogeneous group of patients owing to high rates of fast ripples outside the SOZ (Urrestarazu et al., 2007). Jacobs et al. (2009) found low rates of HFOs for four patients with FCD, but could see a significantly higher rate of fast ripples inside than outside the SOZ, whereas ripples did not show a significant difference. In contrast, Brázdil et al. (2010) saw ripples to occur more frequently in the SOZ, whereas fast ripples were less reliable markers in four patients with FCD. The differences of our results are most likely because of the larger number of patients and a greater variety of different types of FCD in our study.
To our knowledge no study looked at HFOs in different types of FCD. A link between the degree of underlying tissue change and the rates of HFOs has been shown only in patients with mesial temporal sclerosis (Staba et al., 2007; Ogren et al., 2009). The present results for the first time describe a correlation between the histologic findings and HFOs in neocortical epilepsy. Moreover, the results suggest that HFOs can reflect the epileptic disease activity of the underlying lesion and that the latter is higher in FCD type 2. It can be concluded that HFOs do not solely represent the localization of epileptogenic areas but distinguish between areas with lower and higher epileptogenicity. For clinical epileptology this important finding highlights that for the clinical use of HFOs it is not only important to look at the localization of HFOs but also at their rate. The use of HFOs as a prognostic biomarker is promising, but thorough prospective investigations in various types of epilepsies are needed prior to their widespread clinical use.