Spike-related haemodynamic responses overlap with high frequency oscillations in patients with focal epilepsy

Spike-related haemodynamic responses overlap with high frequency oscillations in patients with... Abstract Simultaneous scalp EEG/functional MRI measures non-invasively haemodynamic responses to interictal epileptic discharges, which are related to the epileptogenic zone. High frequency oscillations are also an excellent indicator of this zone, but are primarily recorded from intracerebral EEG. We studied the spatial overlap of these two important markers in patients with drug-resistant epilepsy to assess if their combination could help better define the extent of the epileptogenic zone. We included patients who underwent EEG-functional MRI and later intracerebral EEG. Based on intracerebral EEG findings, we separated patients with unifocal seizures from patients with multifocal or unknown onset seizures. Haemodynamic t-maps were coregistered with the intracerebral electrode positions. Each EEG channel was classified as pertaining to one of the following categories: primary haemodynamic cluster (maximum t-value), secondary cluster (t-value > 90% of the primary cluster) or outside the primary and secondary clusters. We marked high frequency oscillations (ripples: 80–250 Hz; fast ripples: 250–500 Hz) during 1 h of slow wave sleep, and compared their rates in each haemodynamic category. After classifying channels as high- or low-rate, the proportion of high-rate channels within the primary or primary plus secondary clusters was compared to the proportion expected by chance. Twenty-five patients, 11 with unifocal and 14 with multifocal/unknown seizure onsets, were studied. We found a significantly higher median high frequency oscillation rate in the primary cluster compared to secondary cluster and outside these two clusters for the unifocal group (P < 0.0001), but not for the multifocal/unknown group. For the unifocal group, the number of high-rate channels within the primary or primary plus secondary clusters was significantly higher than expected by chance. This held only for the high-ripple-rate channels in the multifocal/unknown group. At the patient level, most patients (18/25, or 72%) had at least one high-rate channel within a primary cluster. In patients with unifocal epilepsy, the maximum haemodynamic response (primary cluster) related to scalp interictal discharges overlaps with the tissue generating high frequency oscillations at high rates. If intracranial EEG is warranted, this response should be explored. As a tentative clinical use of the combination of these techniques we propose that higher high frequency oscillation rates inside than outside the maximum response indicates that the patient has indeed a focal epileptogenic zone demarcated by this response, whereas similar rates inside and outside may indicate a widespread epileptogenic zone or an epileptogenic zone not covered by the implantation. epilepsy, EEG-fMRI, high frequency oscillations Introduction Simultaneous EEG/functional MRI (EEG-fMRI) shows non-invasively the cerebral haemodynamic activity associated with interictal epileptic discharges (IED, used as synonymous with spikes) recorded on scalp EEG by measuring the related changes in blood oxygenation level-dependent (BOLD) response (Ives et al., 1993; Gotman, 2008). These BOLD responses have been shown to be related to the epileptogenic zone, which is defined as the minimum amount of cortical tissue that must be resected to produce seizure freedom (Luders et al., 2006). For instance, complete resection of the maximum BOLD response was associated with seizure freedom (Thornton et al., 2010; An et al., 2013). Also, studies reported concordance between the maximum BOLD response and spikes (Benar et al., 2006; Aghakhani et al., 2015) or seizure onset zone defined by intracranial EEG recordings (Thornton et al., 2011; Khoo et al., 2017a). However, the BOLD responses are at times multiple or widespread, and the significance of the responses other than the maximum peak is unclear. High frequency oscillations (HFOs) are an EEG pattern described in epileptic patients that appears to be an excellent indicator of the epileptogenic zone; it reflects the seizure-generating capability of the underlying tissue and was proposed to be a biomarker of epilepsy (Zijlmans et al., 2012). HFOs are classified according to their frequency in ripples (80–250 Hz) and fast ripples (250–500 Hz). Several studies suggested that the resection of areas containing presurgically high rates of HFOs are associated with a good surgical seizure outcome (Jacobs et al., 2010; Haegelen et al., 2013; Holler et al., 2015), and may be more accurate in determining the surgical outcome than the resection of spike-generating areas (Jacobs et al., 2008) or even the seizure onset zone (Akiyama et al., 2011). The detection of HFOs, however, depends mostly on intracranial EEG recording methods. Studies have shown a high spatial correspondence between the gamma activity (>40 Hz) generated by the human cortex and the BOLD related response to a task (Mukamel et al., 2005; Lachaux et al., 2007; Nir et al., 2007). A similar correspondence between HFOs (>120 Hz) and IED-related BOLD responses has not been yet demonstrated. IED-related BOLD responses and HFOs are both defining the epileptogenic zone in some way. We therefore hypothesized that areas showing IED-related BOLD changes overlap with those generating pathological interictal HFOs. We sought to assess if the regions with maximum BOLD responses are indeed associated with those generating HFOs at high rates. We aimed to evaluate the spatial overlap between these two markers and to assess if their combination could help define the extent of the epileptogenic zone in patients with drug-resistant epilepsy. Material and methods Patients We included consecutive patients with drug-resistant epilepsy (Kwan et al., 2010) from our database of EEG-fMRI recordings obtained from April 2006 to November 2015, and who later underwent intracranial stereo-EEG for seizure onset zone identification. Most of the patients underwent EEG-fMRI as an experimental test done during their pre-surgical epilepsy evaluation, and only in some of the latest cases the information provided by this study was used for presurgical management. Also, in most of the patients the implantation sites were decided independently of the EEG-fMRI results, and based on seizure semiology, neurocognitive findings, scalp EEG and imaging [MRI, and in some cases ictal single-photon emission computed tomography (SPECT) and PET studies]. We included only patients who had intracranial electrodes both inside and outside regions of maximum EEG-fMRI BOLD response. All the patients were implanted with the aim of defining a focal generator for their seizures. Intracranial EEG confirmed the hypothesis in some cases, but in others a unique focal generator was not identified. For this reason, we classified the patients based on the intracranial EEG findings into two groups: a ‘unifocal seizures’ group (if one or two discrete seizure onset zones were found during the investigation), and a ‘multifocal or unknown seizure onset’ group [in this group, we included patients with multifocal (i.e. ≥3 seizure onset zones), with non-lateralized (e.g. bilateral synchronous supplementary motor area), with extensive bilobar or with non-localizable (e.g. clinical symptoms appeared before any intracranial EEG change) generators]. We defined ‘patients with an unknown seizure onset’ as patients in whom the clinical seizure started before any change in the invasive EEG (therefore the seizure onset was missed by the implanted electrodes), and the patients in whom a diffuse change was seen among all the implanted electrodes despite focal seizure semiology. For clarity, the group with multifocal or unknown seizure onset are hereafter called the ‘multifocal’ group. This study was approved by the Montreal Neurological Institute and Hospital Research Ethics Board. All patients gave written informed consent. EEG-functional MRI acquisition and processing Methods were as described in previous studies (An et al., 2013; Khoo et al., 2017a). EEG was continuously recorded inside a 3 T MRI scanner (Trio; Siemens) with 25 magnetic resonance-compatible scalp electrodes placed according to 10–20 (reference FCz) and 10–10 (F9, T9, P9, F10, T10, P10) electrode systems, using a Brain Amp system (Brain Products, 5 kHz sampling). Interictal epileptiform discharges with the same distribution but different morphology were grouped as one study. Thus, patients with multiple IED distributions had more than one study. A t-map was created for each study by taking, at each voxel, the maximum t-value (either positive or negative) from four combined t-maps, based on four haemodynamic response functions (with peaks at 3, 5, 7 and 9 s). The t-map was thresholded at an uncorrected t-value of 3.0, and then the corrected threshold for each subject was computed, corresponding to a whole-brain topological false discovery rate (FDR) of 0.05 (Khoo et al., 2017b). We considered both the positive (activation) and negative (deactivation) responses within the cortex. Deactivations were included, as the maximum has been shown to also reflect the epileptic focus (Pittau et al., 2012; Khoo et al., 2017a). We only excluded deactivations in areas related to the ‘default mode network’ (Raichle et al., 2001), as they have been suggested not to be indicative of the region responsible for the generation of the epileptic discharges (Laufs et al., 2007; Fahoum et al., 2013). The maximum absolute t-value, whether positive or negative, was defined as the primary peak. Absolute t-values >90% of the absolute t-value of the primary peak were defined as secondary peaks. We included in our analysis the brain regions showing significant responses with t-values <90% of the maximum, but we grouped them with the brain regions without haemodynamic responses. We named those regions as being ‘outside the primary and secondary clusters’. Stereo-EEG recording methods and channels selection Intracranial EEG recordings were obtained from electrodes (DIXI or manufactured on-site) stereotactically implanted using an image-guided system (SSN Neuronavigation System) (Olivier et al., 1994) with or without robotized surgical assistant (ROSA™, Medtech) (Hall and Khoo, 2017). The implantation sites were decided according to the presurgical hypothesis of the suspected epileptogenic regions. The EEG signal was acquired using the Harmonie EEG system (Stellate), with a low-pass filter at 500 Hz and high-pass filter at 0.3 Hz, and 2000 Hz sample rate with a referential electrode placed over the parietal lobe contralateral to the suspected epileptogenic zone. All the channels within the brain and free of artefacts were included in the analysis. HFO marking We took 1 h of interictal slow wave sleep to mark HFOs, excluding at least 1 h before and after any ictal activity. As has been done in numerous studies, we selected slow wave sleep given that HFOs are more frequent during this sleep stage than in wakefulness or REM sleep (Bagshaw et al., 2009), providing more reliable results. Slow wave sleep was selected using simultaneous subdermal scalp EEG electrodes placed in the fronto-central-parietal region, electrooculography and chin electromyography. One minute of stereo-EEG was visually marked by one reviewer (K.G.O.) using a bipolar montage made of adjacent contacts. Visual marking was carried out as previously described (Jacobs et al., 2010) by selecting events containing at least four consecutive oscillations. Two events were considered different if separated by at least two oscillations that did not fulfil HFO criteria. HFO detection during the remaining 59 mins of slow wave sleep was automatically done in MATLAB using a published detector (von Ellenrieder et al., 2016) after optimization of the threshold. An optimum threshold was selected for each channel as the value that minimized the number of errors (false detections and missed events) compared to the visual marking in the first minute of the analysed recordings. For the final HFO rate we used the 60 min marked by the detector after optimization of the threshold with the visual marks from the first minute. Classification of channels The post-implantation brain images (MRI or CT) and the functional MRI t-maps were both co-registered to the same pre-implantation T1 anatomical image in a separate process using methods published previously (Khoo et al., 2017a), to avoid possible differences in coregistration accuracy due to the use of different post-implantation imaging modalities. In brief, we first co-registered the functional MRI echo planalr image (EPI) image (which is the image in the same space as the functional MRI t-map) to the pre-implantation anatomic image acquired during the EEG-fMRI recording and then using the resulting coregistration matrix, we coregistered the functional MRI t-map to the same preimplantation anatomic image. We then co-registered the post-implantation image (MRI or CT) to the same preimplantation anatomic image in a separate process. The localization of each contact was obtained in relation to BOLD responses. The channels were classified according to their relation to the BOLD responses and to their rate of HFOs per minute: according to EEG-fMRI BOLD responses; and according to the HFO rate. EEG-functional MRI BOLD responses A channel was classified as pertaining to a ‘primary cluster’ if one of its contacts was placed within a 20-mm sphere centred at the primary peak. A channel was classified as pertaining to a ‘secondary cluster’ if one of its contacts was placed within a 20-mm sphere centred at the secondary peak. The contacts that did not fulfil these conditions were classified as being ‘outside the primary and secondary clusters’. To assign an electrode contact as pertaining to a primary or secondary cluster we chose a maximum of 20 mm from the peak in order to account for potential co-registration inaccuracies and mismatch between neuronal activity and BOLD responses. By definition we did not take into account the extension of the BOLD response but rather considered the HFOs overlap with the most significant BOLD responses: primary and secondary clusters. The HFO rate: definition of low and high HFO rate channels Given the fact that in a recent meta-analysis (Holler et al., 2015) the HFO rates were more strongly related to outcome than the simple presence of HFOs on a channel, we established a threshold to identify the channels with high rates of HFOs. We first identified channels with very low HFO rates that we assumed to be physiological channels. To determine the thresholds for channels with low HFO rates, we applied the same automatic detector (without per channel optimization) in the first NREM sleep cycle of 990 channels from 40 patients (34 belonged to a different cohort, and six were part of the present study) that were known to be channels without epileptic activity (von Ellenrieder et al., 2016), resulting in distributions of physiological rates. Channels with a rate in the 95th percentile of this distribution were considered ‘low HFO rate channels’. As a result, channels with an absolute rate of ripples <3/min or of fast ripples <2.3/min were classified as low rate. This was to avoid classifying channels with low absolute HFO rates in the high HFO rate category (see example of fast ripples in Fig. 1, where all the channels were classified as ‘low HFO rate’). Figure 1 View largeDownload slide Example of discordant patient in the multifocal and unknown seizure onset group. Patient 5 is a 24-year-old female with normal MRI, cutaneous pigmentary mosaicism with 5 p tetrasomy and multifocal epilepsy. The three images in the first row show a small focal primary cluster localized over the right fusiform gyrus. The contact RFug 2 was the closest to the primary cluster, but demonstrated a low rate of HFO. The black dotted line in the ripple histogram separate the high-ripple-rate channels. None of the channels in the fast ripples (FR) histogram were considered as high-fast ripples-rate. The pictures in the second row show the localization of two high-ripple-rate contacts: RIa 4 and RSMA 7 that were outside the primary cluster. The contact RCp 11 pertained to one of the channels with the highest fast ripples rate for this patient, but was not considered high-fast ripples-rate channel due to its low absolute rate, similar to the rate in normal channels (see text for details). For histogram clarity more than 50 channels with low HFO rates were excluded. Figure 1 View largeDownload slide Example of discordant patient in the multifocal and unknown seizure onset group. Patient 5 is a 24-year-old female with normal MRI, cutaneous pigmentary mosaicism with 5 p tetrasomy and multifocal epilepsy. The three images in the first row show a small focal primary cluster localized over the right fusiform gyrus. The contact RFug 2 was the closest to the primary cluster, but demonstrated a low rate of HFO. The black dotted line in the ripple histogram separate the high-ripple-rate channels. None of the channels in the fast ripples (FR) histogram were considered as high-fast ripples-rate. The pictures in the second row show the localization of two high-ripple-rate contacts: RIa 4 and RSMA 7 that were outside the primary cluster. The contact RCp 11 pertained to one of the channels with the highest fast ripples rate for this patient, but was not considered high-fast ripples-rate channel due to its low absolute rate, similar to the rate in normal channels (see text for details). For histogram clarity more than 50 channels with low HFO rates were excluded. For each patient we generated histograms in which all channels were ranked in a descending order according to absolute HFO rates, and separately for ripples and fast ripples. To identify high rate channels we first tried to visually identify a discontinuity in the distribution by differentiating some channels with high rates from many channels with low rates as described previously (Jacobs et al., 2010). This was easily done in 17 patients (example in Fig. 2), but this differentiation was not clear in eight patients (Patients 1, 3, 5, 6, 11, 14, 17 and 21; see Fig. 3 for ripples in which a clear discontinuity in the distribution of the channels was not present). To apply a uniform cut-off ratio to all patients we normalized rates by dividing each channel rate by the highest rate of each patient. We then calculated the ratio that identified channels that represent the highest 5% in HFO rates, which was 0.575. After excluding channels with low absolute HFO rate as defined above, we then considered for all patients, the channels with a ratio ≥0.575 as ‘high-HFO-rate channels’. This cut-off value was in agreement with that in patients who presented a clear discontinuity, in whom the cut-off ratio that separated high rate channels was close to 0.6. Figure 2 View largeDownload slide Example of concordant patient with unifocal epilepsy. Patient 22 is a 24-year-old female with a left frontal focal cortical dysplasia, which after resection was proved to be a focal cortical dysplasia (FCD)-IIB. The EEG-fMRI results showed a primary cluster over the left frontal mesial area (t-value: 10.6) that was concordant with the channels showing the highest rate of ripples and fast ripples (e.g. LCs2). The figure shows the coordinates of the contact LCs2. The electrode contacts are shown as green dots. The black dotted line in the ripple and fast ripple (FR) histograms separate the high-HFO-rate channels. The distributions of ripples and fast ripples show that the two high ripple rate and high-fast ripples rate channels were within the primary cluster. The patient also had two secondary clusters, one over the left anterior cingulate area and another over the left superior cingulate area (not shown). P&SC = primary and secondary cluster. Figure 2 View largeDownload slide Example of concordant patient with unifocal epilepsy. Patient 22 is a 24-year-old female with a left frontal focal cortical dysplasia, which after resection was proved to be a focal cortical dysplasia (FCD)-IIB. The EEG-fMRI results showed a primary cluster over the left frontal mesial area (t-value: 10.6) that was concordant with the channels showing the highest rate of ripples and fast ripples (e.g. LCs2). The figure shows the coordinates of the contact LCs2. The electrode contacts are shown as green dots. The black dotted line in the ripple and fast ripple (FR) histograms separate the high-HFO-rate channels. The distributions of ripples and fast ripples show that the two high ripple rate and high-fast ripples rate channels were within the primary cluster. The patient also had two secondary clusters, one over the left anterior cingulate area and another over the left superior cingulate area (not shown). P&SC = primary and secondary cluster. Figure 3 View largeDownload slide Example of concordant patient with unifocal epilepsy. Patient 17 is a 26-year-old female with right orbito-frontal (ROF) epilepsy and normal MRI. After resection of the lateral right orbito-frontal area, a FCD-IIA was found. The EEG-fMRI results show a primary cluster (t-value: 14.1) over the lateral right orbito-frontal area. The figure shows the coordinates of the contact ROF 8. The electrode contacts are shown as green dots. The black dotted line in the ripple and fast ripples (FR) histograms separate the high HFO rate channels. The distribution of ripples shows that three of four high ripple rate channels are within the primary cluster. The fast ripples distribution shows that all three high-fast ripples rate channels are within the primary cluster. Figure 3 View largeDownload slide Example of concordant patient with unifocal epilepsy. Patient 17 is a 26-year-old female with right orbito-frontal (ROF) epilepsy and normal MRI. After resection of the lateral right orbito-frontal area, a FCD-IIA was found. The EEG-fMRI results show a primary cluster (t-value: 14.1) over the lateral right orbito-frontal area. The figure shows the coordinates of the contact ROF 8. The electrode contacts are shown as green dots. The black dotted line in the ripple and fast ripples (FR) histograms separate the high HFO rate channels. The distribution of ripples shows that three of four high ripple rate channels are within the primary cluster. The fast ripples distribution shows that all three high-fast ripples rate channels are within the primary cluster. Data analysis We calculated the rate as number of ripples and fast ripples per minute for each channel. We then classified the channels according to their EEG-fMRI category (primary cluster, secondary cluster and outside the primary and secondary clusters) and later according to their HFO rate (high and low rate). HFO rates in relation to EEG-functional MRI categories We first calculated the median rate of ripples and fast ripples of the channels pertaining to the primary cluster, secondary cluster and outside the primary and secondary clusters. We performed a permutation test to determine if the difference in median rates between primary cluster versus outside the primary and secondary clusters, primary cluster versus secondary cluster, and secondary cluster versus outside the primary and secondary clusters was significantly larger than that expected by chance. One million permutations of the primary cluster, secondary cluster, and outside the primary and secondary clusters labels of each patient were carried out and the difference in the medians for the whole cohort was computed to determine the null-hypothesis distribution. All the P-values were corrected for 18 comparisons using Holm-Bonferroni procedure (ripples and fast ripples; primary cluster and outside the primary and secondary clusters, primary cluster and secondary cluster, secondary cluster and outside the primary and secondary clusters; unifocal group, multifocal group, and combined groups). EEG-functional MRI categories in relation to high HFO rate channels We then evaluated the difference between the observed number of high HFO rate channels (as definition above) pertaining to each EEG-fMRI category, and the expected number of channels pertaining to each category if there would be no association between them. Under a null hypothesis of no association between BOLD response and high HFO rate channels, the probability of a channel to have high HFO rate would be equal in the primary cluster (or primary cluster + secondary cluster) and outside primary and secondary clusters. Assuming that this probability is given by the observed ratio of high-HFO-rate channels to the total number of channels, the expected number of channels with high HFO rate follows a binomial distribution. We computed the probability P of the observed number of high HFO rate channels as predicted by this binomial distribution, and compared it to a 5% level to determine significant differences from the null hypothesis. All the P-values were corrected for 12 comparisons using Holm-Bonferroni procedure [ripples and fast ripples; primary cluster (or primary cluster + secondary cluster) and outside the primary and secondary clusters; unifocal group, multifocal group and combined groups]. Concordance between high HFO rate channels and EEG-functional MRI categories per patient For this comparison we analysed ripples and fast ripples together, considering as high HFO rate channels those that presented a high rate in both or either in the ripple or fast ripples band. Rates of both ripples and fast ripples have been linked to the epileptogenic zone and were shown to have a significant effect in the postsurgical outcome when resected (Holler et al., 2015). Therefore, we classified the patients into three groups defined by the degree of concordance between the high HFO rate channels (ripple and/or fast ripples) and the EEG-fMRI categories: ‘concordant’ if ≥1 high HFO rate channel was within a primary cluster alone or a primary and secondary cluster; ‘partially concordant’ if none of the high HFO rate channels were within a primary cluster but ≥1 was within a secondary cluster; and ‘discordant’ if none of the high HFO rate channels were within a primary or secondary cluster. Results Sixty-six patients had a stereo-EEG investigation after an EEG-fMRI study, and 41 were excluded: 15 with no active EEG during EEG-fMRI, 12 with no electrodes implanted within 20 mm from a maximum BOLD response, six with scalp EEG artefacts preventing marking abnormalities for EEG-fMRI analysis, four with no available post-implantation CT or MRI, one with artefacts in EPI signal preventing analysis, one with only ictal events during EEG-fMRI, one with stereo-EEG electrodes accidentally removed during postictal confusion and one with <1 h of slow wave sleep separated from any ictal activity during the stereo-EEG investigation. Twenty-five patients were included (16 female; mean age at stereo-EEG evaluation 29.4 ± 8.2 years, range 17–47 years). Seventeen patients had a detectable brain lesion defined either by MRI, pathology, or both; eight patients were defined as non-lesional given that they presented with normal MRI and no resection, or with gliosis or normal tissue in the post-surgical pathology. Electroclinical characteristics are summarized in Supplementary Table 1. Eleven patients belonged to the unifocal seizures group and 14 to the multifocal group, as per our definition above. For the whole group (n = 25), we studied 1600 stereo-EEG channels, of which 260 were within the primary cluster, 133 within the secondary cluster and the remaining 1207 outside the primary and secondary clusters. To evaluate the relationship between BOLD responses and the occurrence of HFOs we addressed two main questions: are there HFOs in regions of BOLD response? (HFO rates in relation to EEG-fMRI categories) and are there BOLD responses in regions of high HFOs rates? (EEG-fMRI categories in relation to high HFO rate channels). HFO rates in relation to EEG-functional MRI categories We calculated the median rate of ripples and fast ripples for each EEG-fMRI category. Figure 4 shows the rates of HFOs in each EEG-fMRI category for the unifocal and multifocal groups. We analysed the difference in the median HFO rate, for ripples and fast ripples, among channels belonging to the EEG-fMRI categories: primary cluster, secondary cluster and outside the primary and secondary clusters. Figure 4 View largeDownload slide HFO rates in relation to EEG-fMRI categories for the unifocal seizure and multifocal and unknown seizure onset groups. For the unifocal seizure group we found a higher median HFO rate (ripples and fast ripples) in the primary cluster (PC) when compared to the secondary cluster (SC) and outside the primary and secondary cluster (Outside P&SC). For the multifocal or unknown seizure onset group we found no differences in the rates among the different EEG-fMRI categories. Note the logarithmic scale. *P < 0.05. Figure 4 View largeDownload slide HFO rates in relation to EEG-fMRI categories for the unifocal seizure and multifocal and unknown seizure onset groups. For the unifocal seizure group we found a higher median HFO rate (ripples and fast ripples) in the primary cluster (PC) when compared to the secondary cluster (SC) and outside the primary and secondary cluster (Outside P&SC). For the multifocal or unknown seizure onset group we found no differences in the rates among the different EEG-fMRI categories. Note the logarithmic scale. *P < 0.05. For ripples (n = 25 patients), we found that comparing the primary cluster to outside the primary and secondary cluster, the median rates were higher in the primary cluster [primary cluster median ripple rate: 4.4 (interquartile range, IQR: 0.8–19.1; Min-max: 0–113.4) and outside the primary and secondary clusters median ripple rate: 1.2 (IQR: 0.1–6.2; Min-max: 0–109.4); P < 0.0001]. The median ripple rates also remained significantly higher in the primary cluster when compared to the secondary cluster [secondary cluster median ripple rate: 2.6 (IQR: 0.4–9.6; Min-max: 0–63.1]; P = 0.0004]. When separating the unifocal and multifocal groups, we found that these differences held for the group with unifocal seizures (n = 11) only. For the multifocal group (n = 14), there were no statistically significant differences. For fast ripples (n = 25 patients), comparing the primary cluster to the outside the primary and secondary cluster the median rates were higher in the primary cluster [primary cluster median fast ripples rate: 0.4 (IQR: 0.03–2.2; Min-max: 0–210.2); and outside the primary and secondary clusters median fast ripples rate: 0.1 (IQR: 0.01–0.8; Min-max: 0–82.5); P = 0.009]. The rates remained higher in the primary cluster when compared to the secondary cluster [secondary cluster median fast ripples rate: 0.1 (IQR: 0.01–0.8; Min-max: 0–51.3); P = 0.0013]. Again, these differences remained for the group with unifocal seizures, whereas for the multifocal group there were no statistically significant differences. For ripples and fast ripples all comparisons between the secondary cluster and outside the primary and secondary cluster were not significant. Supplementary Fig. 1 shows the median rate per category for each patient, illustrating the variable individual rates. To give a better idea of individual results, and demonstrate how our results could have a tentative clinical use, we calculated a ratio between the median HFO rates in the primary cluster and outside the primary and secondary cluster. We observed that when the median ripple rate in the primary cluster was >10 times higher than outside the primary and secondary clusters, the patient was in the unifocal group. This condition was true for 6/11 patients in the unifocal group, and for none in the multifocal group. For fast ripples, we calculated this ratio to be 12 times higher, which was found in 5/11 unifocal patients, and in none of the multifocal patients. Supplementary Fig. 2 shows these ratios for each patient. However, given that fast ripples are more localized, we have two patients with no fast ripples outside the primary and secondary cluster, and in consequence such a ratio could not be calculated for them. When looking at surgical outcome, we found that among the patients that had a brain resection of the presumed epileptogenic zone, 50% had Engel 1 in the unifocal group, whereas none had Engel 1 in the multifocal group. The surgical outcome of each patient is presented in Supplementary Table 1. We did not specifically aim to correlate our findings with the surgical outcome, and therefore some of the patients we studied did not have resective surgery, and for one patient we have no follow-up information. For those patients who had a resection and the outcome was available, we did not account for the several factors that could play a role in the postoperative seizure outcome, such as the extent of the resection in relation to the presumed epileptogenic zone and time of antiepileptic drugs withdrawal, among others. EEG-functional MRI categories in relation to high HFO rate channels Of the 1600 studied channels in the total cohort (n = 25), 113 were high-ripple-rate channels and 63 were high-fast ripples-rate channels (the definition of high HFO rate channels is given in the ‘Materials and methods’ section). We evaluated the proportion of high HFO rate channels that belonged to the primary cluster or primary cluster + secondary cluster (Fig. 5). Compared to the proportion of high HFO rate channels within a primary cluster or secondary cluster expected by chance (primary cluster 16.2% and primary cluster + secondary cluster 18.3%), the number of high HFO rate channels within the primary cluster [for ripples 41/113 (36.3%), P < 0.0001; for fast ripples 20/63 (31.7%), P = 0.005] and within the primary cluster + secondary cluster [for ripples 52/113 (46%), P < 0.0001; for fast ripples 25/63 (39.7%), P = 0.013] were significantly higher. This remained true for ripples and fast ripples in the unifocal group and only true for ripples in the multifocal group. Figure 5 View largeDownload slide Proportion of channels from the primary cluster and the primary plus secondary cluster that have high HFO rate channels. (A) For all patients (n = 25) and for the unifocal seizure group (n = 11), the primary cluster (PC) contained a significantly higher number of high ripple rate and high-fast ripples rate channels than the proportion expected by chance (see ‘Materials and methods’ section for details). For the multifocal and unknown onset seizure group the primary cluster contained a significantly higher number of high ripple rate channels than the proportion expected by chance, whereas the proportion of high-fast ripples rate channels was not statistically different. (B) When analysing the proportion of channels from the primary cluster + secondary cluster (PC + SC) observed to have high-ripple rate and high-fast ripple rate, the differences were similar to the ones observed for the primary cluster alone. *P < 0.05, **P < 0.0001, n.s. = not statistically significant. P-values are corrected for multiple comparisons with Holm-Bonferroni procedure. FR = fast ripple. Figure 5 View largeDownload slide Proportion of channels from the primary cluster and the primary plus secondary cluster that have high HFO rate channels. (A) For all patients (n = 25) and for the unifocal seizure group (n = 11), the primary cluster (PC) contained a significantly higher number of high ripple rate and high-fast ripples rate channels than the proportion expected by chance (see ‘Materials and methods’ section for details). For the multifocal and unknown onset seizure group the primary cluster contained a significantly higher number of high ripple rate channels than the proportion expected by chance, whereas the proportion of high-fast ripples rate channels was not statistically different. (B) When analysing the proportion of channels from the primary cluster + secondary cluster (PC + SC) observed to have high-ripple rate and high-fast ripple rate, the differences were similar to the ones observed for the primary cluster alone. *P < 0.05, **P < 0.0001, n.s. = not statistically significant. P-values are corrected for multiple comparisons with Holm-Bonferroni procedure. FR = fast ripple. Concordance between high HFO rate channels and EEG-functional MRI categories per patient Most of the patients (18/25, 72%) were concordant having at least one high HFO rate channel within a primary cluster. Two patients were partially concordant and five discordant. The results of the concordance between high HFO rate channels and primary or secondary cluster are shown in Fig. 6. Figure 6 View largeDownload slide Proportion of high HFO rate channels within the primary cluster and secondary cluster per patient, and percentage of concordance. Each bar represents a patient. All patients in the unifocal seizure group were concordant, having at least one channel within a primary cluster (PC). For the multifocal and unknown seizure onset group seven patients were concordant, two were partially concordant [having at least one high-HFO-rate channel within a secondary cluster (SC)], and five were discordant. The discordant patients are marked in the graph with the letter ‘d’. Figure 6 View largeDownload slide Proportion of high HFO rate channels within the primary cluster and secondary cluster per patient, and percentage of concordance. Each bar represents a patient. All patients in the unifocal seizure group were concordant, having at least one channel within a primary cluster (PC). For the multifocal and unknown seizure onset group seven patients were concordant, two were partially concordant [having at least one high-HFO-rate channel within a secondary cluster (SC)], and five were discordant. The discordant patients are marked in the graph with the letter ‘d’. Unifocal seizure group All 11 patients were concordant, showing at least one high HFO rate channel within the primary cluster (six patients had high ripple and fast ripples rate channels, four only high ripple rate channels and one only high-fast ripples channels within a primary cluster; Supplementary Table 2). Two patients had all high HFO rate channels within a primary cluster (Patients 2 and 22). As an example, the results of Patient 22 are shown in Fig. 2. Of the remaining nine concordant patients, seven had at least one secondary cluster explored; of those, four had at least one high HFO rate channel within a secondary cluster. Figure 3 shows the HFO distribution and primary cluster BOLD response of Patient 17, who had a primary cluster over the right orbitofrontal region, concordant with the high HFO rate channels. Multifocal and unknown seizure onset group Seven of 14 patients (50%) were concordant (Patients 3, 7, 9, 14, 18, 21 and 24), one had a secondary cluster also containing high HFO rate channels (three patients had high ripple and fast ripples rate channels and four only high ripple rate channels within a primary cluster; Supplementary Table 2). Of 14 patients, two (14%) were partially concordant (Patients 6 and 12); and five (34%) were discordant (Patients 5, 10, 13, 15, and 20) (Fig. 6). Figure 1 shows the HFO distribution and primary cluster BOLD response of Patient 5 who had multifocal epilepsy with discordance between primary cluster and high HFO rate channels. Lesional versus non-lesional patients We found that 15 of 17 (88%) lesional patients were concordant or partially concordant according to our definition above, whereas this was the case for five of eight (62.5%) non-lesional patients. Therefore, there was a mild preference for concordance among the lesional patients, although not significant (P = 0.1379, Fisher’s exact test). Given the limited number of patients, a more detailed analysis according to specific pathologies was not feasible. Discussion This study investigates the relation between two markers of the epileptogenic zone: the scalp IED-related BOLD responses and cortical epileptic HFOs recorded by intracranial EEG. We show that HFOs at high rates co-localize with the maximum scalp IED-related BOLD response in most patients (18/25, 72%) with drug-resistant epilepsy. In patients with discordant findings, intracranial EEG revealed multifocal or unknown seizure onsets. In patients with unifocal epilepsy, the regions generating HFOs at highest rates were mainly restricted to the regions showing a maximal haemodynamic response on functional MRI. Our findings can be helpful to define the extent of the epileptogenic zone in patients with difficult to localize drug-resistant epilepsy. Defining low and high HFO rate channels Ripples can be either pathological or physiological (Alkawadri et al., 2014), and fast ripples, despite having been more strongly linked to the epileptogenic zone, can also be physiological as happens during cognitive processes in humans (Kucewicz et al., 2014). There is a large difference in rates of HFOs within epilepsy patients; they vary according to types of pathology and cortical brain regions (Zijlmans et al., 2012; Ferrari-Marinho et al., 2015). However, it is currently unknown which HFO rate is high enough to be considered as clinically significant. There are brain areas showing low rates of HFOs, sometimes over widespread regions, which might not necessarily be representative of the epileptogenic zone and, hence, do not need to be resected to render a patient seizure free. High rates of HFOs have been more strongly related to outcome than considering each electrode as showing or not showing HFOs (Holler et al., 2015), emphasizing the relevance of the identification of high rate channels. For this reason, we first looked at the relation between the BOLD responses and median rates of HFOs, and then established thresholds to identify channels with low and high rates. We proposed an original solution to solve the problem of setting a threshold to identify low rate channels: we calculated the rates of ripples and fast ripples found in channels from relatively healthy brain tissue i.e. without epileptiform discharges and outside any visible lesion, and considered them as physiological HFOs. Therefore, channels showing HFO rates at what can be considered physiological levels were not labelled as high HFO rate channels. This method helps not to call ‘high rate’ a channel that may be the highest in a patient but has an absolute low rate. To establish a threshold to identify channels with high rates of HFOs, we took a rather strict approach as we only considered as high HFO rate channel the 5% of the channels with the highest HFO rates. This method resulted in a mean of 4.5 high ripple rate channels and 2.5 high fast ripple rate channels per patient. A less strict threshold would have probably resulted in additional concordance between high rate channels and BOLD responses. Unifocal seizure group The results indicate that for patients with focal epilepsy confirmed by intracranial EEG, the rates of ripples and fast ripples are higher within the area of maximum BOLD response (primary cluster) than in other brain regions outside this maximum. The primary clusters also contained higher HFO rates compared to other cortical regions with IED-related BOLD responses having values close to the maximum (secondary clusters). With our definition of high HFO rate channels, we found that for the unifocal epilepsy group of patients, the maximum BOLD response (primary cluster) alone contained a significantly high number of high HFO rate channels, both in the ripple and fast ripple rate. All 11 patients with unifocal epilepsy had at least one high HFO rate channel within the primary cluster, and in four of them, the secondary cluster also contained high HFO rate channels. Taken together, our findings indicate that in unifocal epilepsy the maximum BOLD response contains tissue that generates HFOs at higher rates than other parts of the epileptic brain, and contains a high proportion of high-ripple and high-fast ripples rate channels. This implies that in these patients the regions generating HFOs at highest rates are confined and mainly restricted to the regions showing a primary cluster on functional MRI. Multifocal and unknown seizure onset group On the other hand, in the multifocal or unknown seizure onset group the HFO rates within the maximum BOLD did not differ from the ones observed outside this area. In this circumstance, the primary cluster (maximum BOLD response) contained similar HFO rates compared to the secondary clusters and compared to brain regions outside these two clusters. In other words, HFOs were not restricted to the maximum BOLD response, but were rather more widespread. We found, however, that the primary cluster alone or in combination with the secondary cluster contained a significantly higher proportion of the high ripple rate channels, but not of fast ripples. The primary cluster was indicative of the region with highest HFO rates in a good proportion of patients: in 7 of 14 patients (50%) the primary cluster contained at least one high HFO rate channel. Interestingly, in two patients only the secondary cluster showed high HFO rate channels. Therefore, secondary clusters could be of additional help in identifying epileptogenic brain regions in some patients with multifocal epilepsy. Previous studies showed that IED-related BOLD responses discordant with the seizure onset zone, are associated with poor post-surgical seizure outcome, probably representing complex epilepsy processes (Thornton et al., 2011; An et al., 2013). Our results are to some extent in agreement with that conclusion, as the seizure onset zone has been related to the highest HFO rates, and the presence of a maximum BOLD response discordant with the high HFO rate channels was predominant in patients with multifocal epilepsy in our study. This discordance may be of practical importance in pointing to a complex epilepsy problem rather than a unifocal one. Clinical significance One of the challenges of newer techniques such as EEG-fMRI and the detection of high frequency oscillations when considering their use as clinical tools and to inform surgical planning, is our ability to better interpret the information that they provide. Our results demonstrate that the maximum scalp IED-related BOLD response indicates the region with high HFOs rates in most patients with drug resistant epilepsy. This may be of practical clinical importance if invasive EEG is considered: when planning an implantation, the maximum haemodynamic response should be a target. Besides, as a tentative clinical use of the combination of these techniques we propose the co-localization of HFOs and maximum haemodynamic responses could help to define the extent of the epileptogenic zone. If the electrodes targeting the maximum BOLD response shows the highest rate of HFO, it is highly possible that the patient has a well-defined and localized generator (a unifocal epilepsy), and the BOLD response is revealing this epileptogenic tissue. On the other hand, if the maximum BOLD response does not contain the highest rate of HFOs, then either the patient has a widespread epileptogenic zone or the seizure-generating area has not been covered by the implantation. Although encouraging, we think that applying our results for individual patients should be done with caution, given they are based on a limited number of patients. Therefore, prospective studies should be carried out to validate our observations. Limitations and methodological considerations Some issues may limit our ability to understand the overlap between HFOs and BOLD responses. First, HFOs are small neurophysiological signals locally generated, and without evidence of propagation. Besides, the intracranial EEG suffers from a limited spatial coverage, as implantation of stereo-EEG electrodes are not without risk and the practical number of electrodes that can be implanted is limited. Second, our patients were most often not implanted taking into account the EEG-fMRI results, often limiting the exploration of the primary clusters. Third, EEG-fMRI studies were obtained during wakefulness or light sleep, and HFOs were marked during slow wave sleep. Studies have shown that both HFOs and spikes are more widespread during slow wave sleep, when compared with wakefulness and REM sleep (Sammaritano et al., 1991; von Ellenrieder et al., 2017). However, we are mainly looking at regions with peak HFO rates and it has been shown that the regions with most of these oscillations are the same in the different sleep/wake states (Bagshaw et al., 2009). Therefore we do not expect an important influence of the vigilance state in our comparisons. Finally, HFO activity was recorded at the time of the intracranial EEG whereas the scalp EEG-fMRI study was done prior to implantation. We ignore the relationship in time between EEG-fMRI responses and intracranial EEG HFOs measures. For that reason, we do not study here a physio-pathological relation between BOLD responses and HFOs, but rather showed that the IED-related BOLD response and the HFOs have a close spatial overlap. Conclusions Maximum scalp IED-related BOLD responses (primary clusters) co-localize with HFOs at high rates in most patients with drug-resistant epilepsy. Hence, if intracranial EEG is warranted, we propose that this maximum response should be explored. Moreover, analysing the concordance between the maximum BOLD response and HFO rate could be helpful to define the extent of the epileptogenic zone in patients with difficult-to-localize drug-resistant epilepsy. If the maximum BOLD response overlaps with the tissue that generates ripples and fast ripples at high rates, it is highly possible that the patient has indeed a focal epileptogenic zone marked by the haemodynamic response. If the region of maximum BOLD response does not show the highest HFO rates, the patient may have a widespread epileptogenic zone or an epileptogenic zone not covered by the implantation. These observations are true for our group of patients, and our results are primarily with respect to the relationship between haemodynamic responses and HFOs. Further research is needed to evaluate if these observations can be used for individual patients. Secondary BOLD responses apart from the maximum response may be of additional help in identifying epileptogenic tissue. Acknowledgements We thank Dr Alexei Yankovsky, Assistant Professor, Department of Internal Medicine (Neurology) at the University of Manitoba, Canada, for his help with the outcome information on one of the reported patients. Funding This study was supported by the Canadian Institutes of Health Research grant FDN 143208. K.G.O. was supported by the Frederick Andermann Clinical/Research fellowship in Epileptology and EEG of the Montreal Neurological Institute (Canada). H.M.K. is supported by Mark Rayport and Shirley Ferguson Rayport fellowship in epilepsy surgery and was supported by the Preston Robb fellowship of the Montreal Neurological Institute (Canada), research fellowship of the Uehara Memorial Foundation (Japan), travel grants from Osaka Medical Research Foundation for Intractable Diseases (Japan) and Japan Epilepsy Research Foundation (Japan). Supplementary material Supplementary material is available at Brain online. Abbreviations Abbreviations BOLD blood oxygenation level-dependent EEG-fMRI simultaneous EEG/functional MRI HFO high frequency oscillations IED interictal epileptic discharges References Aghakhani Y, Beers CA, Pittman DJ, Gaxiola-Valdez I, Goodyear BG, Federico P. Co-localization between the BOLD response and epileptiform discharges recorded by simultaneous intracranial EEG-fMRI at 3 T. Neuroimage Clin  2015; 7: 755– 63. Google Scholar CrossRef Search ADS PubMed  Akiyama T, McCoy B, Go CY, Ochi A, Elliott IM, Akiyama M, et al.   Focal resection of fast ripples on extraoperative intracranial EEG improves seizure outcome in pediatric epilepsy. Epilepsia  2011; 52: 1802– 11. Google Scholar CrossRef Search ADS PubMed  Alkawadri R, Gaspard N, Goncharova II, Spencer DD, Gerrard JL, Zaveri H, et al.   The spatial and signal characteristics of physiologic high frequency oscillations. Epilepsia  2014; 55: 1986– 95. Google Scholar CrossRef Search ADS PubMed  An D, Fahoum F, Hall J, Olivier A, Gotman J, Dubeau F. Electroencephalography/functional magnetic resonance imaging responses help predict surgical outcome in focal epilepsy. Epilepsia  2013; 54: 2184– 94. Google Scholar CrossRef Search ADS PubMed  Bagshaw AP, Jacobs J, LeVan P, Dubeau F, Gotman J. Effect of sleep stage on interictal high-frequency oscillations recorded from depth macroelectrodes in patients with focal epilepsy. Epilepsia  2009; 50: 617– 28. Google Scholar CrossRef Search ADS PubMed  Benar CG, Grova C, Kobayashi E, Bagshaw AP, Aghakhani Y, Dubeau F, et al.   EEG-fMRI of epileptic spikes: concordance with EEG source localization and intracranial EEG. Neuroimage  2006; 30: 1161– 70. Google Scholar CrossRef Search ADS PubMed  Fahoum F, Zelmann R, Tyvaert L, Dubeau F, Gotman J. Epileptic discharges affect the default mode network–FMRI and intracerebral EEG evidence. PLoS One  2013; 8: e68038. Google Scholar CrossRef Search ADS PubMed  Ferrari-Marinho T, Perucca P, Mok K, Olivier A, Hall J, Dubeau F, et al.   Pathologic substrates of focal epilepsy influence the generation of high-frequency oscillations. Epilepsia  2015; 56: 592– 8. Google Scholar CrossRef Search ADS PubMed  Gotman J. Epileptic networks studied with EEG-fMRI. Epilepsia  2008; 49 (Suppl 3): 42– 51. Google Scholar CrossRef Search ADS PubMed  Haegelen C, Perucca P, Chatillon CE, Andrade-Valenca L, Zelmann R, Jacobs J, et al.   High-frequency oscillations, extent of surgical resection, and surgical outcome in drug-resistant focal epilepsy. Epilepsia  2013; 54: 848– 57. Google Scholar CrossRef Search ADS PubMed  Hall JA, Khoo HM. Robotic-assisted and image-guided MRI-compatible stereoelectroencephalography. Can J Neurol Sci  2017; 1–9. doi: 10.1017/cjn.2017.240. Holler Y, Kutil R, Klaffenbock L, Thomschewski A, Holler PM, Bathke AC, et al.   High-frequency oscillations in epilepsy and surgical outcome. A meta-analysis. Front Hum Neurosci  2015; 9: 574. Google Scholar CrossRef Search ADS PubMed  Ives JR, Warach S, Schmitt F, Edelman RR, Schomer DL. Monitoring the patient's EEG during echo planar MRI. Electroencephalogr Clin Neurophysiol  1993; 87: 417– 20. Google Scholar CrossRef Search ADS PubMed  Jacobs J, LeVan P, Chander R, Hall J, Dubeau F, Gotman J. Interictal high-frequency oscillations (80-500 Hz) are an indicator of seizure onset areas independent of spikes in the human epileptic brain. Epilepsia  2008; 49: 1893– 907. Google Scholar CrossRef Search ADS PubMed  Jacobs J, Zijlmans M, Zelmann R, Chatillon CE, Hall J, Olivier A, et al.   High-frequency electroencephalographic oscillations correlate with outcome of epilepsy surgery. Ann Neurol  2010; 67: 209– 20. Google Scholar CrossRef Search ADS PubMed  Khoo HM, Hao Y, von Ellenrieder N, Zazubovits N, Hall J, Olivier A, et al.   The haemodynamic response to interictal epileptic discharges localizes the seizure-onset zone. Epilepsia  2017a; 58: 811– 23. Google Scholar CrossRef Search ADS   Khoo HM, von Ellenrieder N, Zazubovits N, Dubeau F, Gotman J. Epileptic networks in action: synchrony between distant hemodynamic responses. Ann Neurol  2017b; 82: 57– 66. Google Scholar CrossRef Search ADS   Kucewicz MT, Cimbalnik J, Matsumoto JY, Brinkmann BH, Bower MR, Vasoli V, et al.   High frequency oscillations are associated with cognitive processing in human recognition memory. Brain  2014; 137 (Pt 8): 2231– 44. Google Scholar CrossRef Search ADS   Kwan P, Arzimanoglou A, Berg AT, Brodie MJ, Allen Hauser W, Mathern G, et al.   Definition of drug resistant epilepsy: consensus proposal by the ad hoc task force of the ILAE Commission on Therapeutic Strategies. Epilepsia  2010; 51: 1069– 77. Google Scholar CrossRef Search ADS PubMed  Lachaux JP, Fonlupt P, Kahane P, Minotti L, Hoffmann D, Bertrand O, et al.   Relationship between task-related gamma oscillations and BOLD signal: new insights from combined fMRI and intracranial EEG. Hum Brain Mapp  2007; 28: 1368– 75. Google Scholar CrossRef Search ADS PubMed  Laufs H, Hamandi K, Salek-Haddadi A, Kleinschmidt AK, Duncan JS, Lemieux L. Temporal lobe interictal epileptic discharges affect cerebral activity in ‘default mode’ brain regions. Hum Brain Mapp  2007; 28: 1023– 32. Google Scholar CrossRef Search ADS PubMed  Luders HO, Najm I, Nair D, Widdess-Walsh P, Bingman W. The epileptogenic zone: general principles. Epileptic disorders: international epilepsy journal with videotape. Epileptic Disord  2006; 8 (Suppl 2): S1– 9. Google Scholar PubMed  Mukamel R, Gelbard H, Arieli A, Hasson U, Fried I, Malach R. Coupling between neuronal firing, field potentials, and FMRI in human auditory cortex. Science  2005; 309: 951– 4. Google Scholar CrossRef Search ADS PubMed  Nir Y, Fisch L, Mukamel R, Gelbard-Sagiv H, Arieli A, Fried I, et al.   Coupling between neuronal firing rate, gamma LFP, and BOLD fMRI is related to interneuronal correlations. Curr Biol  2007; 17: 1275– 85. Google Scholar CrossRef Search ADS PubMed  Olivier A, Germano IM, Cukiert A, Peters T. Frameless stereotaxy for surgery of the epilepsies: preliminary experience. Technical note. J Neurosurg  1994; 81: 629– 33. Google Scholar CrossRef Search ADS PubMed  Pittau F, Dubeau F, Gotman J. Contribution of EEG/fMRI to the definition of the epileptic focus. Neurology  2012; 78: 1479– 87. Google Scholar CrossRef Search ADS PubMed  Raichle ME, MacLeod AM, Snyder AZ, Powers WJ, Gusnard DA, Shulman GL. A default mode of brain function. Proc Natl Acad Sci USA  2001; 98: 676– 82. Google Scholar CrossRef Search ADS PubMed  Sammaritano M, Gigli GL, Gotman J. Interictal spiking during wakefulness and sleep and the localization of foci in temporal lobe epilepsy. Neurology  1991; 41 (2 Pt 1): 290– 7. Google Scholar CrossRef Search ADS   Thornton R, Laufs H, Rodionov R, Cannadathu S, Carmichael DW, Vulliemoz S, et al.   EEG correlated functional MRI and postoperative outcome in focal epilepsy. J Neurol Neurosurg Psychiatry  2010; 81: 922– 7. Google Scholar CrossRef Search ADS PubMed  Thornton R, Vulliemoz S, Rodionov R, Carmichael DW, Chaudhary UJ, Diehl B, et al.   Epileptic networks in focal cortical dysplasia revealed using electroencephalography-functional magnetic resonance imaging. Ann Neurol  2011; 70: 822– 37. Google Scholar CrossRef Search ADS PubMed  von Ellenrieder N, Dubeau F, Gotman J, Frauscher B. Physiological and pathological high-frequency oscillations have distinct sleep-homeostatic properties. Neuroimage Clin  2017; 14: 566– 73. Google Scholar CrossRef Search ADS PubMed  von Ellenrieder N, Frauscher B, Dubeau F, Gotman J. Interaction with slow waves during sleep improves discrimination of physiologic and pathologic high-frequency oscillations (80-500 Hz). Epilepsia  2016; 57: 869– 78. Google Scholar CrossRef Search ADS PubMed  Zijlmans M, Jiruska P, Zelmann R, Leijten FS, Jefferys JG, Gotman J. High-frequency oscillations as a new biomarker in epilepsy. Ann Neurol  2012; 71: 169– 78. Google Scholar CrossRef Search ADS PubMed  © The Author(s) (2018). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Brain Oxford University Press

Spike-related haemodynamic responses overlap with high frequency oscillations in patients with focal epilepsy

Loading next page...
 
/lp/ou_press/spike-related-haemodynamic-responses-overlap-with-high-frequency-sogobsPayH
Publisher
Oxford University Press
Copyright
© The Author(s) (2018). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: journals.permissions@oup.com
ISSN
0006-8950
eISSN
1460-2156
D.O.I.
10.1093/brain/awx383
Publisher site
See Article on Publisher Site

Abstract

Abstract Simultaneous scalp EEG/functional MRI measures non-invasively haemodynamic responses to interictal epileptic discharges, which are related to the epileptogenic zone. High frequency oscillations are also an excellent indicator of this zone, but are primarily recorded from intracerebral EEG. We studied the spatial overlap of these two important markers in patients with drug-resistant epilepsy to assess if their combination could help better define the extent of the epileptogenic zone. We included patients who underwent EEG-functional MRI and later intracerebral EEG. Based on intracerebral EEG findings, we separated patients with unifocal seizures from patients with multifocal or unknown onset seizures. Haemodynamic t-maps were coregistered with the intracerebral electrode positions. Each EEG channel was classified as pertaining to one of the following categories: primary haemodynamic cluster (maximum t-value), secondary cluster (t-value > 90% of the primary cluster) or outside the primary and secondary clusters. We marked high frequency oscillations (ripples: 80–250 Hz; fast ripples: 250–500 Hz) during 1 h of slow wave sleep, and compared their rates in each haemodynamic category. After classifying channels as high- or low-rate, the proportion of high-rate channels within the primary or primary plus secondary clusters was compared to the proportion expected by chance. Twenty-five patients, 11 with unifocal and 14 with multifocal/unknown seizure onsets, were studied. We found a significantly higher median high frequency oscillation rate in the primary cluster compared to secondary cluster and outside these two clusters for the unifocal group (P < 0.0001), but not for the multifocal/unknown group. For the unifocal group, the number of high-rate channels within the primary or primary plus secondary clusters was significantly higher than expected by chance. This held only for the high-ripple-rate channels in the multifocal/unknown group. At the patient level, most patients (18/25, or 72%) had at least one high-rate channel within a primary cluster. In patients with unifocal epilepsy, the maximum haemodynamic response (primary cluster) related to scalp interictal discharges overlaps with the tissue generating high frequency oscillations at high rates. If intracranial EEG is warranted, this response should be explored. As a tentative clinical use of the combination of these techniques we propose that higher high frequency oscillation rates inside than outside the maximum response indicates that the patient has indeed a focal epileptogenic zone demarcated by this response, whereas similar rates inside and outside may indicate a widespread epileptogenic zone or an epileptogenic zone not covered by the implantation. epilepsy, EEG-fMRI, high frequency oscillations Introduction Simultaneous EEG/functional MRI (EEG-fMRI) shows non-invasively the cerebral haemodynamic activity associated with interictal epileptic discharges (IED, used as synonymous with spikes) recorded on scalp EEG by measuring the related changes in blood oxygenation level-dependent (BOLD) response (Ives et al., 1993; Gotman, 2008). These BOLD responses have been shown to be related to the epileptogenic zone, which is defined as the minimum amount of cortical tissue that must be resected to produce seizure freedom (Luders et al., 2006). For instance, complete resection of the maximum BOLD response was associated with seizure freedom (Thornton et al., 2010; An et al., 2013). Also, studies reported concordance between the maximum BOLD response and spikes (Benar et al., 2006; Aghakhani et al., 2015) or seizure onset zone defined by intracranial EEG recordings (Thornton et al., 2011; Khoo et al., 2017a). However, the BOLD responses are at times multiple or widespread, and the significance of the responses other than the maximum peak is unclear. High frequency oscillations (HFOs) are an EEG pattern described in epileptic patients that appears to be an excellent indicator of the epileptogenic zone; it reflects the seizure-generating capability of the underlying tissue and was proposed to be a biomarker of epilepsy (Zijlmans et al., 2012). HFOs are classified according to their frequency in ripples (80–250 Hz) and fast ripples (250–500 Hz). Several studies suggested that the resection of areas containing presurgically high rates of HFOs are associated with a good surgical seizure outcome (Jacobs et al., 2010; Haegelen et al., 2013; Holler et al., 2015), and may be more accurate in determining the surgical outcome than the resection of spike-generating areas (Jacobs et al., 2008) or even the seizure onset zone (Akiyama et al., 2011). The detection of HFOs, however, depends mostly on intracranial EEG recording methods. Studies have shown a high spatial correspondence between the gamma activity (>40 Hz) generated by the human cortex and the BOLD related response to a task (Mukamel et al., 2005; Lachaux et al., 2007; Nir et al., 2007). A similar correspondence between HFOs (>120 Hz) and IED-related BOLD responses has not been yet demonstrated. IED-related BOLD responses and HFOs are both defining the epileptogenic zone in some way. We therefore hypothesized that areas showing IED-related BOLD changes overlap with those generating pathological interictal HFOs. We sought to assess if the regions with maximum BOLD responses are indeed associated with those generating HFOs at high rates. We aimed to evaluate the spatial overlap between these two markers and to assess if their combination could help define the extent of the epileptogenic zone in patients with drug-resistant epilepsy. Material and methods Patients We included consecutive patients with drug-resistant epilepsy (Kwan et al., 2010) from our database of EEG-fMRI recordings obtained from April 2006 to November 2015, and who later underwent intracranial stereo-EEG for seizure onset zone identification. Most of the patients underwent EEG-fMRI as an experimental test done during their pre-surgical epilepsy evaluation, and only in some of the latest cases the information provided by this study was used for presurgical management. Also, in most of the patients the implantation sites were decided independently of the EEG-fMRI results, and based on seizure semiology, neurocognitive findings, scalp EEG and imaging [MRI, and in some cases ictal single-photon emission computed tomography (SPECT) and PET studies]. We included only patients who had intracranial electrodes both inside and outside regions of maximum EEG-fMRI BOLD response. All the patients were implanted with the aim of defining a focal generator for their seizures. Intracranial EEG confirmed the hypothesis in some cases, but in others a unique focal generator was not identified. For this reason, we classified the patients based on the intracranial EEG findings into two groups: a ‘unifocal seizures’ group (if one or two discrete seizure onset zones were found during the investigation), and a ‘multifocal or unknown seizure onset’ group [in this group, we included patients with multifocal (i.e. ≥3 seizure onset zones), with non-lateralized (e.g. bilateral synchronous supplementary motor area), with extensive bilobar or with non-localizable (e.g. clinical symptoms appeared before any intracranial EEG change) generators]. We defined ‘patients with an unknown seizure onset’ as patients in whom the clinical seizure started before any change in the invasive EEG (therefore the seizure onset was missed by the implanted electrodes), and the patients in whom a diffuse change was seen among all the implanted electrodes despite focal seizure semiology. For clarity, the group with multifocal or unknown seizure onset are hereafter called the ‘multifocal’ group. This study was approved by the Montreal Neurological Institute and Hospital Research Ethics Board. All patients gave written informed consent. EEG-functional MRI acquisition and processing Methods were as described in previous studies (An et al., 2013; Khoo et al., 2017a). EEG was continuously recorded inside a 3 T MRI scanner (Trio; Siemens) with 25 magnetic resonance-compatible scalp electrodes placed according to 10–20 (reference FCz) and 10–10 (F9, T9, P9, F10, T10, P10) electrode systems, using a Brain Amp system (Brain Products, 5 kHz sampling). Interictal epileptiform discharges with the same distribution but different morphology were grouped as one study. Thus, patients with multiple IED distributions had more than one study. A t-map was created for each study by taking, at each voxel, the maximum t-value (either positive or negative) from four combined t-maps, based on four haemodynamic response functions (with peaks at 3, 5, 7 and 9 s). The t-map was thresholded at an uncorrected t-value of 3.0, and then the corrected threshold for each subject was computed, corresponding to a whole-brain topological false discovery rate (FDR) of 0.05 (Khoo et al., 2017b). We considered both the positive (activation) and negative (deactivation) responses within the cortex. Deactivations were included, as the maximum has been shown to also reflect the epileptic focus (Pittau et al., 2012; Khoo et al., 2017a). We only excluded deactivations in areas related to the ‘default mode network’ (Raichle et al., 2001), as they have been suggested not to be indicative of the region responsible for the generation of the epileptic discharges (Laufs et al., 2007; Fahoum et al., 2013). The maximum absolute t-value, whether positive or negative, was defined as the primary peak. Absolute t-values >90% of the absolute t-value of the primary peak were defined as secondary peaks. We included in our analysis the brain regions showing significant responses with t-values <90% of the maximum, but we grouped them with the brain regions without haemodynamic responses. We named those regions as being ‘outside the primary and secondary clusters’. Stereo-EEG recording methods and channels selection Intracranial EEG recordings were obtained from electrodes (DIXI or manufactured on-site) stereotactically implanted using an image-guided system (SSN Neuronavigation System) (Olivier et al., 1994) with or without robotized surgical assistant (ROSA™, Medtech) (Hall and Khoo, 2017). The implantation sites were decided according to the presurgical hypothesis of the suspected epileptogenic regions. The EEG signal was acquired using the Harmonie EEG system (Stellate), with a low-pass filter at 500 Hz and high-pass filter at 0.3 Hz, and 2000 Hz sample rate with a referential electrode placed over the parietal lobe contralateral to the suspected epileptogenic zone. All the channels within the brain and free of artefacts were included in the analysis. HFO marking We took 1 h of interictal slow wave sleep to mark HFOs, excluding at least 1 h before and after any ictal activity. As has been done in numerous studies, we selected slow wave sleep given that HFOs are more frequent during this sleep stage than in wakefulness or REM sleep (Bagshaw et al., 2009), providing more reliable results. Slow wave sleep was selected using simultaneous subdermal scalp EEG electrodes placed in the fronto-central-parietal region, electrooculography and chin electromyography. One minute of stereo-EEG was visually marked by one reviewer (K.G.O.) using a bipolar montage made of adjacent contacts. Visual marking was carried out as previously described (Jacobs et al., 2010) by selecting events containing at least four consecutive oscillations. Two events were considered different if separated by at least two oscillations that did not fulfil HFO criteria. HFO detection during the remaining 59 mins of slow wave sleep was automatically done in MATLAB using a published detector (von Ellenrieder et al., 2016) after optimization of the threshold. An optimum threshold was selected for each channel as the value that minimized the number of errors (false detections and missed events) compared to the visual marking in the first minute of the analysed recordings. For the final HFO rate we used the 60 min marked by the detector after optimization of the threshold with the visual marks from the first minute. Classification of channels The post-implantation brain images (MRI or CT) and the functional MRI t-maps were both co-registered to the same pre-implantation T1 anatomical image in a separate process using methods published previously (Khoo et al., 2017a), to avoid possible differences in coregistration accuracy due to the use of different post-implantation imaging modalities. In brief, we first co-registered the functional MRI echo planalr image (EPI) image (which is the image in the same space as the functional MRI t-map) to the pre-implantation anatomic image acquired during the EEG-fMRI recording and then using the resulting coregistration matrix, we coregistered the functional MRI t-map to the same preimplantation anatomic image. We then co-registered the post-implantation image (MRI or CT) to the same preimplantation anatomic image in a separate process. The localization of each contact was obtained in relation to BOLD responses. The channels were classified according to their relation to the BOLD responses and to their rate of HFOs per minute: according to EEG-fMRI BOLD responses; and according to the HFO rate. EEG-functional MRI BOLD responses A channel was classified as pertaining to a ‘primary cluster’ if one of its contacts was placed within a 20-mm sphere centred at the primary peak. A channel was classified as pertaining to a ‘secondary cluster’ if one of its contacts was placed within a 20-mm sphere centred at the secondary peak. The contacts that did not fulfil these conditions were classified as being ‘outside the primary and secondary clusters’. To assign an electrode contact as pertaining to a primary or secondary cluster we chose a maximum of 20 mm from the peak in order to account for potential co-registration inaccuracies and mismatch between neuronal activity and BOLD responses. By definition we did not take into account the extension of the BOLD response but rather considered the HFOs overlap with the most significant BOLD responses: primary and secondary clusters. The HFO rate: definition of low and high HFO rate channels Given the fact that in a recent meta-analysis (Holler et al., 2015) the HFO rates were more strongly related to outcome than the simple presence of HFOs on a channel, we established a threshold to identify the channels with high rates of HFOs. We first identified channels with very low HFO rates that we assumed to be physiological channels. To determine the thresholds for channels with low HFO rates, we applied the same automatic detector (without per channel optimization) in the first NREM sleep cycle of 990 channels from 40 patients (34 belonged to a different cohort, and six were part of the present study) that were known to be channels without epileptic activity (von Ellenrieder et al., 2016), resulting in distributions of physiological rates. Channels with a rate in the 95th percentile of this distribution were considered ‘low HFO rate channels’. As a result, channels with an absolute rate of ripples <3/min or of fast ripples <2.3/min were classified as low rate. This was to avoid classifying channels with low absolute HFO rates in the high HFO rate category (see example of fast ripples in Fig. 1, where all the channels were classified as ‘low HFO rate’). Figure 1 View largeDownload slide Example of discordant patient in the multifocal and unknown seizure onset group. Patient 5 is a 24-year-old female with normal MRI, cutaneous pigmentary mosaicism with 5 p tetrasomy and multifocal epilepsy. The three images in the first row show a small focal primary cluster localized over the right fusiform gyrus. The contact RFug 2 was the closest to the primary cluster, but demonstrated a low rate of HFO. The black dotted line in the ripple histogram separate the high-ripple-rate channels. None of the channels in the fast ripples (FR) histogram were considered as high-fast ripples-rate. The pictures in the second row show the localization of two high-ripple-rate contacts: RIa 4 and RSMA 7 that were outside the primary cluster. The contact RCp 11 pertained to one of the channels with the highest fast ripples rate for this patient, but was not considered high-fast ripples-rate channel due to its low absolute rate, similar to the rate in normal channels (see text for details). For histogram clarity more than 50 channels with low HFO rates were excluded. Figure 1 View largeDownload slide Example of discordant patient in the multifocal and unknown seizure onset group. Patient 5 is a 24-year-old female with normal MRI, cutaneous pigmentary mosaicism with 5 p tetrasomy and multifocal epilepsy. The three images in the first row show a small focal primary cluster localized over the right fusiform gyrus. The contact RFug 2 was the closest to the primary cluster, but demonstrated a low rate of HFO. The black dotted line in the ripple histogram separate the high-ripple-rate channels. None of the channels in the fast ripples (FR) histogram were considered as high-fast ripples-rate. The pictures in the second row show the localization of two high-ripple-rate contacts: RIa 4 and RSMA 7 that were outside the primary cluster. The contact RCp 11 pertained to one of the channels with the highest fast ripples rate for this patient, but was not considered high-fast ripples-rate channel due to its low absolute rate, similar to the rate in normal channels (see text for details). For histogram clarity more than 50 channels with low HFO rates were excluded. For each patient we generated histograms in which all channels were ranked in a descending order according to absolute HFO rates, and separately for ripples and fast ripples. To identify high rate channels we first tried to visually identify a discontinuity in the distribution by differentiating some channels with high rates from many channels with low rates as described previously (Jacobs et al., 2010). This was easily done in 17 patients (example in Fig. 2), but this differentiation was not clear in eight patients (Patients 1, 3, 5, 6, 11, 14, 17 and 21; see Fig. 3 for ripples in which a clear discontinuity in the distribution of the channels was not present). To apply a uniform cut-off ratio to all patients we normalized rates by dividing each channel rate by the highest rate of each patient. We then calculated the ratio that identified channels that represent the highest 5% in HFO rates, which was 0.575. After excluding channels with low absolute HFO rate as defined above, we then considered for all patients, the channels with a ratio ≥0.575 as ‘high-HFO-rate channels’. This cut-off value was in agreement with that in patients who presented a clear discontinuity, in whom the cut-off ratio that separated high rate channels was close to 0.6. Figure 2 View largeDownload slide Example of concordant patient with unifocal epilepsy. Patient 22 is a 24-year-old female with a left frontal focal cortical dysplasia, which after resection was proved to be a focal cortical dysplasia (FCD)-IIB. The EEG-fMRI results showed a primary cluster over the left frontal mesial area (t-value: 10.6) that was concordant with the channels showing the highest rate of ripples and fast ripples (e.g. LCs2). The figure shows the coordinates of the contact LCs2. The electrode contacts are shown as green dots. The black dotted line in the ripple and fast ripple (FR) histograms separate the high-HFO-rate channels. The distributions of ripples and fast ripples show that the two high ripple rate and high-fast ripples rate channels were within the primary cluster. The patient also had two secondary clusters, one over the left anterior cingulate area and another over the left superior cingulate area (not shown). P&SC = primary and secondary cluster. Figure 2 View largeDownload slide Example of concordant patient with unifocal epilepsy. Patient 22 is a 24-year-old female with a left frontal focal cortical dysplasia, which after resection was proved to be a focal cortical dysplasia (FCD)-IIB. The EEG-fMRI results showed a primary cluster over the left frontal mesial area (t-value: 10.6) that was concordant with the channels showing the highest rate of ripples and fast ripples (e.g. LCs2). The figure shows the coordinates of the contact LCs2. The electrode contacts are shown as green dots. The black dotted line in the ripple and fast ripple (FR) histograms separate the high-HFO-rate channels. The distributions of ripples and fast ripples show that the two high ripple rate and high-fast ripples rate channels were within the primary cluster. The patient also had two secondary clusters, one over the left anterior cingulate area and another over the left superior cingulate area (not shown). P&SC = primary and secondary cluster. Figure 3 View largeDownload slide Example of concordant patient with unifocal epilepsy. Patient 17 is a 26-year-old female with right orbito-frontal (ROF) epilepsy and normal MRI. After resection of the lateral right orbito-frontal area, a FCD-IIA was found. The EEG-fMRI results show a primary cluster (t-value: 14.1) over the lateral right orbito-frontal area. The figure shows the coordinates of the contact ROF 8. The electrode contacts are shown as green dots. The black dotted line in the ripple and fast ripples (FR) histograms separate the high HFO rate channels. The distribution of ripples shows that three of four high ripple rate channels are within the primary cluster. The fast ripples distribution shows that all three high-fast ripples rate channels are within the primary cluster. Figure 3 View largeDownload slide Example of concordant patient with unifocal epilepsy. Patient 17 is a 26-year-old female with right orbito-frontal (ROF) epilepsy and normal MRI. After resection of the lateral right orbito-frontal area, a FCD-IIA was found. The EEG-fMRI results show a primary cluster (t-value: 14.1) over the lateral right orbito-frontal area. The figure shows the coordinates of the contact ROF 8. The electrode contacts are shown as green dots. The black dotted line in the ripple and fast ripples (FR) histograms separate the high HFO rate channels. The distribution of ripples shows that three of four high ripple rate channels are within the primary cluster. The fast ripples distribution shows that all three high-fast ripples rate channels are within the primary cluster. Data analysis We calculated the rate as number of ripples and fast ripples per minute for each channel. We then classified the channels according to their EEG-fMRI category (primary cluster, secondary cluster and outside the primary and secondary clusters) and later according to their HFO rate (high and low rate). HFO rates in relation to EEG-functional MRI categories We first calculated the median rate of ripples and fast ripples of the channels pertaining to the primary cluster, secondary cluster and outside the primary and secondary clusters. We performed a permutation test to determine if the difference in median rates between primary cluster versus outside the primary and secondary clusters, primary cluster versus secondary cluster, and secondary cluster versus outside the primary and secondary clusters was significantly larger than that expected by chance. One million permutations of the primary cluster, secondary cluster, and outside the primary and secondary clusters labels of each patient were carried out and the difference in the medians for the whole cohort was computed to determine the null-hypothesis distribution. All the P-values were corrected for 18 comparisons using Holm-Bonferroni procedure (ripples and fast ripples; primary cluster and outside the primary and secondary clusters, primary cluster and secondary cluster, secondary cluster and outside the primary and secondary clusters; unifocal group, multifocal group, and combined groups). EEG-functional MRI categories in relation to high HFO rate channels We then evaluated the difference between the observed number of high HFO rate channels (as definition above) pertaining to each EEG-fMRI category, and the expected number of channels pertaining to each category if there would be no association between them. Under a null hypothesis of no association between BOLD response and high HFO rate channels, the probability of a channel to have high HFO rate would be equal in the primary cluster (or primary cluster + secondary cluster) and outside primary and secondary clusters. Assuming that this probability is given by the observed ratio of high-HFO-rate channels to the total number of channels, the expected number of channels with high HFO rate follows a binomial distribution. We computed the probability P of the observed number of high HFO rate channels as predicted by this binomial distribution, and compared it to a 5% level to determine significant differences from the null hypothesis. All the P-values were corrected for 12 comparisons using Holm-Bonferroni procedure [ripples and fast ripples; primary cluster (or primary cluster + secondary cluster) and outside the primary and secondary clusters; unifocal group, multifocal group and combined groups]. Concordance between high HFO rate channels and EEG-functional MRI categories per patient For this comparison we analysed ripples and fast ripples together, considering as high HFO rate channels those that presented a high rate in both or either in the ripple or fast ripples band. Rates of both ripples and fast ripples have been linked to the epileptogenic zone and were shown to have a significant effect in the postsurgical outcome when resected (Holler et al., 2015). Therefore, we classified the patients into three groups defined by the degree of concordance between the high HFO rate channels (ripple and/or fast ripples) and the EEG-fMRI categories: ‘concordant’ if ≥1 high HFO rate channel was within a primary cluster alone or a primary and secondary cluster; ‘partially concordant’ if none of the high HFO rate channels were within a primary cluster but ≥1 was within a secondary cluster; and ‘discordant’ if none of the high HFO rate channels were within a primary or secondary cluster. Results Sixty-six patients had a stereo-EEG investigation after an EEG-fMRI study, and 41 were excluded: 15 with no active EEG during EEG-fMRI, 12 with no electrodes implanted within 20 mm from a maximum BOLD response, six with scalp EEG artefacts preventing marking abnormalities for EEG-fMRI analysis, four with no available post-implantation CT or MRI, one with artefacts in EPI signal preventing analysis, one with only ictal events during EEG-fMRI, one with stereo-EEG electrodes accidentally removed during postictal confusion and one with <1 h of slow wave sleep separated from any ictal activity during the stereo-EEG investigation. Twenty-five patients were included (16 female; mean age at stereo-EEG evaluation 29.4 ± 8.2 years, range 17–47 years). Seventeen patients had a detectable brain lesion defined either by MRI, pathology, or both; eight patients were defined as non-lesional given that they presented with normal MRI and no resection, or with gliosis or normal tissue in the post-surgical pathology. Electroclinical characteristics are summarized in Supplementary Table 1. Eleven patients belonged to the unifocal seizures group and 14 to the multifocal group, as per our definition above. For the whole group (n = 25), we studied 1600 stereo-EEG channels, of which 260 were within the primary cluster, 133 within the secondary cluster and the remaining 1207 outside the primary and secondary clusters. To evaluate the relationship between BOLD responses and the occurrence of HFOs we addressed two main questions: are there HFOs in regions of BOLD response? (HFO rates in relation to EEG-fMRI categories) and are there BOLD responses in regions of high HFOs rates? (EEG-fMRI categories in relation to high HFO rate channels). HFO rates in relation to EEG-functional MRI categories We calculated the median rate of ripples and fast ripples for each EEG-fMRI category. Figure 4 shows the rates of HFOs in each EEG-fMRI category for the unifocal and multifocal groups. We analysed the difference in the median HFO rate, for ripples and fast ripples, among channels belonging to the EEG-fMRI categories: primary cluster, secondary cluster and outside the primary and secondary clusters. Figure 4 View largeDownload slide HFO rates in relation to EEG-fMRI categories for the unifocal seizure and multifocal and unknown seizure onset groups. For the unifocal seizure group we found a higher median HFO rate (ripples and fast ripples) in the primary cluster (PC) when compared to the secondary cluster (SC) and outside the primary and secondary cluster (Outside P&SC). For the multifocal or unknown seizure onset group we found no differences in the rates among the different EEG-fMRI categories. Note the logarithmic scale. *P < 0.05. Figure 4 View largeDownload slide HFO rates in relation to EEG-fMRI categories for the unifocal seizure and multifocal and unknown seizure onset groups. For the unifocal seizure group we found a higher median HFO rate (ripples and fast ripples) in the primary cluster (PC) when compared to the secondary cluster (SC) and outside the primary and secondary cluster (Outside P&SC). For the multifocal or unknown seizure onset group we found no differences in the rates among the different EEG-fMRI categories. Note the logarithmic scale. *P < 0.05. For ripples (n = 25 patients), we found that comparing the primary cluster to outside the primary and secondary cluster, the median rates were higher in the primary cluster [primary cluster median ripple rate: 4.4 (interquartile range, IQR: 0.8–19.1; Min-max: 0–113.4) and outside the primary and secondary clusters median ripple rate: 1.2 (IQR: 0.1–6.2; Min-max: 0–109.4); P < 0.0001]. The median ripple rates also remained significantly higher in the primary cluster when compared to the secondary cluster [secondary cluster median ripple rate: 2.6 (IQR: 0.4–9.6; Min-max: 0–63.1]; P = 0.0004]. When separating the unifocal and multifocal groups, we found that these differences held for the group with unifocal seizures (n = 11) only. For the multifocal group (n = 14), there were no statistically significant differences. For fast ripples (n = 25 patients), comparing the primary cluster to the outside the primary and secondary cluster the median rates were higher in the primary cluster [primary cluster median fast ripples rate: 0.4 (IQR: 0.03–2.2; Min-max: 0–210.2); and outside the primary and secondary clusters median fast ripples rate: 0.1 (IQR: 0.01–0.8; Min-max: 0–82.5); P = 0.009]. The rates remained higher in the primary cluster when compared to the secondary cluster [secondary cluster median fast ripples rate: 0.1 (IQR: 0.01–0.8; Min-max: 0–51.3); P = 0.0013]. Again, these differences remained for the group with unifocal seizures, whereas for the multifocal group there were no statistically significant differences. For ripples and fast ripples all comparisons between the secondary cluster and outside the primary and secondary cluster were not significant. Supplementary Fig. 1 shows the median rate per category for each patient, illustrating the variable individual rates. To give a better idea of individual results, and demonstrate how our results could have a tentative clinical use, we calculated a ratio between the median HFO rates in the primary cluster and outside the primary and secondary cluster. We observed that when the median ripple rate in the primary cluster was >10 times higher than outside the primary and secondary clusters, the patient was in the unifocal group. This condition was true for 6/11 patients in the unifocal group, and for none in the multifocal group. For fast ripples, we calculated this ratio to be 12 times higher, which was found in 5/11 unifocal patients, and in none of the multifocal patients. Supplementary Fig. 2 shows these ratios for each patient. However, given that fast ripples are more localized, we have two patients with no fast ripples outside the primary and secondary cluster, and in consequence such a ratio could not be calculated for them. When looking at surgical outcome, we found that among the patients that had a brain resection of the presumed epileptogenic zone, 50% had Engel 1 in the unifocal group, whereas none had Engel 1 in the multifocal group. The surgical outcome of each patient is presented in Supplementary Table 1. We did not specifically aim to correlate our findings with the surgical outcome, and therefore some of the patients we studied did not have resective surgery, and for one patient we have no follow-up information. For those patients who had a resection and the outcome was available, we did not account for the several factors that could play a role in the postoperative seizure outcome, such as the extent of the resection in relation to the presumed epileptogenic zone and time of antiepileptic drugs withdrawal, among others. EEG-functional MRI categories in relation to high HFO rate channels Of the 1600 studied channels in the total cohort (n = 25), 113 were high-ripple-rate channels and 63 were high-fast ripples-rate channels (the definition of high HFO rate channels is given in the ‘Materials and methods’ section). We evaluated the proportion of high HFO rate channels that belonged to the primary cluster or primary cluster + secondary cluster (Fig. 5). Compared to the proportion of high HFO rate channels within a primary cluster or secondary cluster expected by chance (primary cluster 16.2% and primary cluster + secondary cluster 18.3%), the number of high HFO rate channels within the primary cluster [for ripples 41/113 (36.3%), P < 0.0001; for fast ripples 20/63 (31.7%), P = 0.005] and within the primary cluster + secondary cluster [for ripples 52/113 (46%), P < 0.0001; for fast ripples 25/63 (39.7%), P = 0.013] were significantly higher. This remained true for ripples and fast ripples in the unifocal group and only true for ripples in the multifocal group. Figure 5 View largeDownload slide Proportion of channels from the primary cluster and the primary plus secondary cluster that have high HFO rate channels. (A) For all patients (n = 25) and for the unifocal seizure group (n = 11), the primary cluster (PC) contained a significantly higher number of high ripple rate and high-fast ripples rate channels than the proportion expected by chance (see ‘Materials and methods’ section for details). For the multifocal and unknown onset seizure group the primary cluster contained a significantly higher number of high ripple rate channels than the proportion expected by chance, whereas the proportion of high-fast ripples rate channels was not statistically different. (B) When analysing the proportion of channels from the primary cluster + secondary cluster (PC + SC) observed to have high-ripple rate and high-fast ripple rate, the differences were similar to the ones observed for the primary cluster alone. *P < 0.05, **P < 0.0001, n.s. = not statistically significant. P-values are corrected for multiple comparisons with Holm-Bonferroni procedure. FR = fast ripple. Figure 5 View largeDownload slide Proportion of channels from the primary cluster and the primary plus secondary cluster that have high HFO rate channels. (A) For all patients (n = 25) and for the unifocal seizure group (n = 11), the primary cluster (PC) contained a significantly higher number of high ripple rate and high-fast ripples rate channels than the proportion expected by chance (see ‘Materials and methods’ section for details). For the multifocal and unknown onset seizure group the primary cluster contained a significantly higher number of high ripple rate channels than the proportion expected by chance, whereas the proportion of high-fast ripples rate channels was not statistically different. (B) When analysing the proportion of channels from the primary cluster + secondary cluster (PC + SC) observed to have high-ripple rate and high-fast ripple rate, the differences were similar to the ones observed for the primary cluster alone. *P < 0.05, **P < 0.0001, n.s. = not statistically significant. P-values are corrected for multiple comparisons with Holm-Bonferroni procedure. FR = fast ripple. Concordance between high HFO rate channels and EEG-functional MRI categories per patient Most of the patients (18/25, 72%) were concordant having at least one high HFO rate channel within a primary cluster. Two patients were partially concordant and five discordant. The results of the concordance between high HFO rate channels and primary or secondary cluster are shown in Fig. 6. Figure 6 View largeDownload slide Proportion of high HFO rate channels within the primary cluster and secondary cluster per patient, and percentage of concordance. Each bar represents a patient. All patients in the unifocal seizure group were concordant, having at least one channel within a primary cluster (PC). For the multifocal and unknown seizure onset group seven patients were concordant, two were partially concordant [having at least one high-HFO-rate channel within a secondary cluster (SC)], and five were discordant. The discordant patients are marked in the graph with the letter ‘d’. Figure 6 View largeDownload slide Proportion of high HFO rate channels within the primary cluster and secondary cluster per patient, and percentage of concordance. Each bar represents a patient. All patients in the unifocal seizure group were concordant, having at least one channel within a primary cluster (PC). For the multifocal and unknown seizure onset group seven patients were concordant, two were partially concordant [having at least one high-HFO-rate channel within a secondary cluster (SC)], and five were discordant. The discordant patients are marked in the graph with the letter ‘d’. Unifocal seizure group All 11 patients were concordant, showing at least one high HFO rate channel within the primary cluster (six patients had high ripple and fast ripples rate channels, four only high ripple rate channels and one only high-fast ripples channels within a primary cluster; Supplementary Table 2). Two patients had all high HFO rate channels within a primary cluster (Patients 2 and 22). As an example, the results of Patient 22 are shown in Fig. 2. Of the remaining nine concordant patients, seven had at least one secondary cluster explored; of those, four had at least one high HFO rate channel within a secondary cluster. Figure 3 shows the HFO distribution and primary cluster BOLD response of Patient 17, who had a primary cluster over the right orbitofrontal region, concordant with the high HFO rate channels. Multifocal and unknown seizure onset group Seven of 14 patients (50%) were concordant (Patients 3, 7, 9, 14, 18, 21 and 24), one had a secondary cluster also containing high HFO rate channels (three patients had high ripple and fast ripples rate channels and four only high ripple rate channels within a primary cluster; Supplementary Table 2). Of 14 patients, two (14%) were partially concordant (Patients 6 and 12); and five (34%) were discordant (Patients 5, 10, 13, 15, and 20) (Fig. 6). Figure 1 shows the HFO distribution and primary cluster BOLD response of Patient 5 who had multifocal epilepsy with discordance between primary cluster and high HFO rate channels. Lesional versus non-lesional patients We found that 15 of 17 (88%) lesional patients were concordant or partially concordant according to our definition above, whereas this was the case for five of eight (62.5%) non-lesional patients. Therefore, there was a mild preference for concordance among the lesional patients, although not significant (P = 0.1379, Fisher’s exact test). Given the limited number of patients, a more detailed analysis according to specific pathologies was not feasible. Discussion This study investigates the relation between two markers of the epileptogenic zone: the scalp IED-related BOLD responses and cortical epileptic HFOs recorded by intracranial EEG. We show that HFOs at high rates co-localize with the maximum scalp IED-related BOLD response in most patients (18/25, 72%) with drug-resistant epilepsy. In patients with discordant findings, intracranial EEG revealed multifocal or unknown seizure onsets. In patients with unifocal epilepsy, the regions generating HFOs at highest rates were mainly restricted to the regions showing a maximal haemodynamic response on functional MRI. Our findings can be helpful to define the extent of the epileptogenic zone in patients with difficult to localize drug-resistant epilepsy. Defining low and high HFO rate channels Ripples can be either pathological or physiological (Alkawadri et al., 2014), and fast ripples, despite having been more strongly linked to the epileptogenic zone, can also be physiological as happens during cognitive processes in humans (Kucewicz et al., 2014). There is a large difference in rates of HFOs within epilepsy patients; they vary according to types of pathology and cortical brain regions (Zijlmans et al., 2012; Ferrari-Marinho et al., 2015). However, it is currently unknown which HFO rate is high enough to be considered as clinically significant. There are brain areas showing low rates of HFOs, sometimes over widespread regions, which might not necessarily be representative of the epileptogenic zone and, hence, do not need to be resected to render a patient seizure free. High rates of HFOs have been more strongly related to outcome than considering each electrode as showing or not showing HFOs (Holler et al., 2015), emphasizing the relevance of the identification of high rate channels. For this reason, we first looked at the relation between the BOLD responses and median rates of HFOs, and then established thresholds to identify channels with low and high rates. We proposed an original solution to solve the problem of setting a threshold to identify low rate channels: we calculated the rates of ripples and fast ripples found in channels from relatively healthy brain tissue i.e. without epileptiform discharges and outside any visible lesion, and considered them as physiological HFOs. Therefore, channels showing HFO rates at what can be considered physiological levels were not labelled as high HFO rate channels. This method helps not to call ‘high rate’ a channel that may be the highest in a patient but has an absolute low rate. To establish a threshold to identify channels with high rates of HFOs, we took a rather strict approach as we only considered as high HFO rate channel the 5% of the channels with the highest HFO rates. This method resulted in a mean of 4.5 high ripple rate channels and 2.5 high fast ripple rate channels per patient. A less strict threshold would have probably resulted in additional concordance between high rate channels and BOLD responses. Unifocal seizure group The results indicate that for patients with focal epilepsy confirmed by intracranial EEG, the rates of ripples and fast ripples are higher within the area of maximum BOLD response (primary cluster) than in other brain regions outside this maximum. The primary clusters also contained higher HFO rates compared to other cortical regions with IED-related BOLD responses having values close to the maximum (secondary clusters). With our definition of high HFO rate channels, we found that for the unifocal epilepsy group of patients, the maximum BOLD response (primary cluster) alone contained a significantly high number of high HFO rate channels, both in the ripple and fast ripple rate. All 11 patients with unifocal epilepsy had at least one high HFO rate channel within the primary cluster, and in four of them, the secondary cluster also contained high HFO rate channels. Taken together, our findings indicate that in unifocal epilepsy the maximum BOLD response contains tissue that generates HFOs at higher rates than other parts of the epileptic brain, and contains a high proportion of high-ripple and high-fast ripples rate channels. This implies that in these patients the regions generating HFOs at highest rates are confined and mainly restricted to the regions showing a primary cluster on functional MRI. Multifocal and unknown seizure onset group On the other hand, in the multifocal or unknown seizure onset group the HFO rates within the maximum BOLD did not differ from the ones observed outside this area. In this circumstance, the primary cluster (maximum BOLD response) contained similar HFO rates compared to the secondary clusters and compared to brain regions outside these two clusters. In other words, HFOs were not restricted to the maximum BOLD response, but were rather more widespread. We found, however, that the primary cluster alone or in combination with the secondary cluster contained a significantly higher proportion of the high ripple rate channels, but not of fast ripples. The primary cluster was indicative of the region with highest HFO rates in a good proportion of patients: in 7 of 14 patients (50%) the primary cluster contained at least one high HFO rate channel. Interestingly, in two patients only the secondary cluster showed high HFO rate channels. Therefore, secondary clusters could be of additional help in identifying epileptogenic brain regions in some patients with multifocal epilepsy. Previous studies showed that IED-related BOLD responses discordant with the seizure onset zone, are associated with poor post-surgical seizure outcome, probably representing complex epilepsy processes (Thornton et al., 2011; An et al., 2013). Our results are to some extent in agreement with that conclusion, as the seizure onset zone has been related to the highest HFO rates, and the presence of a maximum BOLD response discordant with the high HFO rate channels was predominant in patients with multifocal epilepsy in our study. This discordance may be of practical importance in pointing to a complex epilepsy problem rather than a unifocal one. Clinical significance One of the challenges of newer techniques such as EEG-fMRI and the detection of high frequency oscillations when considering their use as clinical tools and to inform surgical planning, is our ability to better interpret the information that they provide. Our results demonstrate that the maximum scalp IED-related BOLD response indicates the region with high HFOs rates in most patients with drug resistant epilepsy. This may be of practical clinical importance if invasive EEG is considered: when planning an implantation, the maximum haemodynamic response should be a target. Besides, as a tentative clinical use of the combination of these techniques we propose the co-localization of HFOs and maximum haemodynamic responses could help to define the extent of the epileptogenic zone. If the electrodes targeting the maximum BOLD response shows the highest rate of HFO, it is highly possible that the patient has a well-defined and localized generator (a unifocal epilepsy), and the BOLD response is revealing this epileptogenic tissue. On the other hand, if the maximum BOLD response does not contain the highest rate of HFOs, then either the patient has a widespread epileptogenic zone or the seizure-generating area has not been covered by the implantation. Although encouraging, we think that applying our results for individual patients should be done with caution, given they are based on a limited number of patients. Therefore, prospective studies should be carried out to validate our observations. Limitations and methodological considerations Some issues may limit our ability to understand the overlap between HFOs and BOLD responses. First, HFOs are small neurophysiological signals locally generated, and without evidence of propagation. Besides, the intracranial EEG suffers from a limited spatial coverage, as implantation of stereo-EEG electrodes are not without risk and the practical number of electrodes that can be implanted is limited. Second, our patients were most often not implanted taking into account the EEG-fMRI results, often limiting the exploration of the primary clusters. Third, EEG-fMRI studies were obtained during wakefulness or light sleep, and HFOs were marked during slow wave sleep. Studies have shown that both HFOs and spikes are more widespread during slow wave sleep, when compared with wakefulness and REM sleep (Sammaritano et al., 1991; von Ellenrieder et al., 2017). However, we are mainly looking at regions with peak HFO rates and it has been shown that the regions with most of these oscillations are the same in the different sleep/wake states (Bagshaw et al., 2009). Therefore we do not expect an important influence of the vigilance state in our comparisons. Finally, HFO activity was recorded at the time of the intracranial EEG whereas the scalp EEG-fMRI study was done prior to implantation. We ignore the relationship in time between EEG-fMRI responses and intracranial EEG HFOs measures. For that reason, we do not study here a physio-pathological relation between BOLD responses and HFOs, but rather showed that the IED-related BOLD response and the HFOs have a close spatial overlap. Conclusions Maximum scalp IED-related BOLD responses (primary clusters) co-localize with HFOs at high rates in most patients with drug-resistant epilepsy. Hence, if intracranial EEG is warranted, we propose that this maximum response should be explored. Moreover, analysing the concordance between the maximum BOLD response and HFO rate could be helpful to define the extent of the epileptogenic zone in patients with difficult-to-localize drug-resistant epilepsy. If the maximum BOLD response overlaps with the tissue that generates ripples and fast ripples at high rates, it is highly possible that the patient has indeed a focal epileptogenic zone marked by the haemodynamic response. If the region of maximum BOLD response does not show the highest HFO rates, the patient may have a widespread epileptogenic zone or an epileptogenic zone not covered by the implantation. These observations are true for our group of patients, and our results are primarily with respect to the relationship between haemodynamic responses and HFOs. Further research is needed to evaluate if these observations can be used for individual patients. Secondary BOLD responses apart from the maximum response may be of additional help in identifying epileptogenic tissue. Acknowledgements We thank Dr Alexei Yankovsky, Assistant Professor, Department of Internal Medicine (Neurology) at the University of Manitoba, Canada, for his help with the outcome information on one of the reported patients. Funding This study was supported by the Canadian Institutes of Health Research grant FDN 143208. K.G.O. was supported by the Frederick Andermann Clinical/Research fellowship in Epileptology and EEG of the Montreal Neurological Institute (Canada). H.M.K. is supported by Mark Rayport and Shirley Ferguson Rayport fellowship in epilepsy surgery and was supported by the Preston Robb fellowship of the Montreal Neurological Institute (Canada), research fellowship of the Uehara Memorial Foundation (Japan), travel grants from Osaka Medical Research Foundation for Intractable Diseases (Japan) and Japan Epilepsy Research Foundation (Japan). Supplementary material Supplementary material is available at Brain online. Abbreviations Abbreviations BOLD blood oxygenation level-dependent EEG-fMRI simultaneous EEG/functional MRI HFO high frequency oscillations IED interictal epileptic discharges References Aghakhani Y, Beers CA, Pittman DJ, Gaxiola-Valdez I, Goodyear BG, Federico P. Co-localization between the BOLD response and epileptiform discharges recorded by simultaneous intracranial EEG-fMRI at 3 T. Neuroimage Clin  2015; 7: 755– 63. Google Scholar CrossRef Search ADS PubMed  Akiyama T, McCoy B, Go CY, Ochi A, Elliott IM, Akiyama M, et al.   Focal resection of fast ripples on extraoperative intracranial EEG improves seizure outcome in pediatric epilepsy. Epilepsia  2011; 52: 1802– 11. Google Scholar CrossRef Search ADS PubMed  Alkawadri R, Gaspard N, Goncharova II, Spencer DD, Gerrard JL, Zaveri H, et al.   The spatial and signal characteristics of physiologic high frequency oscillations. Epilepsia  2014; 55: 1986– 95. Google Scholar CrossRef Search ADS PubMed  An D, Fahoum F, Hall J, Olivier A, Gotman J, Dubeau F. Electroencephalography/functional magnetic resonance imaging responses help predict surgical outcome in focal epilepsy. Epilepsia  2013; 54: 2184– 94. Google Scholar CrossRef Search ADS PubMed  Bagshaw AP, Jacobs J, LeVan P, Dubeau F, Gotman J. Effect of sleep stage on interictal high-frequency oscillations recorded from depth macroelectrodes in patients with focal epilepsy. Epilepsia  2009; 50: 617– 28. Google Scholar CrossRef Search ADS PubMed  Benar CG, Grova C, Kobayashi E, Bagshaw AP, Aghakhani Y, Dubeau F, et al.   EEG-fMRI of epileptic spikes: concordance with EEG source localization and intracranial EEG. Neuroimage  2006; 30: 1161– 70. Google Scholar CrossRef Search ADS PubMed  Fahoum F, Zelmann R, Tyvaert L, Dubeau F, Gotman J. Epileptic discharges affect the default mode network–FMRI and intracerebral EEG evidence. PLoS One  2013; 8: e68038. Google Scholar CrossRef Search ADS PubMed  Ferrari-Marinho T, Perucca P, Mok K, Olivier A, Hall J, Dubeau F, et al.   Pathologic substrates of focal epilepsy influence the generation of high-frequency oscillations. Epilepsia  2015; 56: 592– 8. Google Scholar CrossRef Search ADS PubMed  Gotman J. Epileptic networks studied with EEG-fMRI. Epilepsia  2008; 49 (Suppl 3): 42– 51. Google Scholar CrossRef Search ADS PubMed  Haegelen C, Perucca P, Chatillon CE, Andrade-Valenca L, Zelmann R, Jacobs J, et al.   High-frequency oscillations, extent of surgical resection, and surgical outcome in drug-resistant focal epilepsy. Epilepsia  2013; 54: 848– 57. Google Scholar CrossRef Search ADS PubMed  Hall JA, Khoo HM. Robotic-assisted and image-guided MRI-compatible stereoelectroencephalography. Can J Neurol Sci  2017; 1–9. doi: 10.1017/cjn.2017.240. Holler Y, Kutil R, Klaffenbock L, Thomschewski A, Holler PM, Bathke AC, et al.   High-frequency oscillations in epilepsy and surgical outcome. A meta-analysis. Front Hum Neurosci  2015; 9: 574. Google Scholar CrossRef Search ADS PubMed  Ives JR, Warach S, Schmitt F, Edelman RR, Schomer DL. Monitoring the patient's EEG during echo planar MRI. Electroencephalogr Clin Neurophysiol  1993; 87: 417– 20. Google Scholar CrossRef Search ADS PubMed  Jacobs J, LeVan P, Chander R, Hall J, Dubeau F, Gotman J. Interictal high-frequency oscillations (80-500 Hz) are an indicator of seizure onset areas independent of spikes in the human epileptic brain. Epilepsia  2008; 49: 1893– 907. Google Scholar CrossRef Search ADS PubMed  Jacobs J, Zijlmans M, Zelmann R, Chatillon CE, Hall J, Olivier A, et al.   High-frequency electroencephalographic oscillations correlate with outcome of epilepsy surgery. Ann Neurol  2010; 67: 209– 20. Google Scholar CrossRef Search ADS PubMed  Khoo HM, Hao Y, von Ellenrieder N, Zazubovits N, Hall J, Olivier A, et al.   The haemodynamic response to interictal epileptic discharges localizes the seizure-onset zone. Epilepsia  2017a; 58: 811– 23. Google Scholar CrossRef Search ADS   Khoo HM, von Ellenrieder N, Zazubovits N, Dubeau F, Gotman J. Epileptic networks in action: synchrony between distant hemodynamic responses. Ann Neurol  2017b; 82: 57– 66. Google Scholar CrossRef Search ADS   Kucewicz MT, Cimbalnik J, Matsumoto JY, Brinkmann BH, Bower MR, Vasoli V, et al.   High frequency oscillations are associated with cognitive processing in human recognition memory. Brain  2014; 137 (Pt 8): 2231– 44. Google Scholar CrossRef Search ADS   Kwan P, Arzimanoglou A, Berg AT, Brodie MJ, Allen Hauser W, Mathern G, et al.   Definition of drug resistant epilepsy: consensus proposal by the ad hoc task force of the ILAE Commission on Therapeutic Strategies. Epilepsia  2010; 51: 1069– 77. Google Scholar CrossRef Search ADS PubMed  Lachaux JP, Fonlupt P, Kahane P, Minotti L, Hoffmann D, Bertrand O, et al.   Relationship between task-related gamma oscillations and BOLD signal: new insights from combined fMRI and intracranial EEG. Hum Brain Mapp  2007; 28: 1368– 75. Google Scholar CrossRef Search ADS PubMed  Laufs H, Hamandi K, Salek-Haddadi A, Kleinschmidt AK, Duncan JS, Lemieux L. Temporal lobe interictal epileptic discharges affect cerebral activity in ‘default mode’ brain regions. Hum Brain Mapp  2007; 28: 1023– 32. Google Scholar CrossRef Search ADS PubMed  Luders HO, Najm I, Nair D, Widdess-Walsh P, Bingman W. The epileptogenic zone: general principles. Epileptic disorders: international epilepsy journal with videotape. Epileptic Disord  2006; 8 (Suppl 2): S1– 9. Google Scholar PubMed  Mukamel R, Gelbard H, Arieli A, Hasson U, Fried I, Malach R. Coupling between neuronal firing, field potentials, and FMRI in human auditory cortex. Science  2005; 309: 951– 4. Google Scholar CrossRef Search ADS PubMed  Nir Y, Fisch L, Mukamel R, Gelbard-Sagiv H, Arieli A, Fried I, et al.   Coupling between neuronal firing rate, gamma LFP, and BOLD fMRI is related to interneuronal correlations. Curr Biol  2007; 17: 1275– 85. Google Scholar CrossRef Search ADS PubMed  Olivier A, Germano IM, Cukiert A, Peters T. Frameless stereotaxy for surgery of the epilepsies: preliminary experience. Technical note. J Neurosurg  1994; 81: 629– 33. Google Scholar CrossRef Search ADS PubMed  Pittau F, Dubeau F, Gotman J. Contribution of EEG/fMRI to the definition of the epileptic focus. Neurology  2012; 78: 1479– 87. Google Scholar CrossRef Search ADS PubMed  Raichle ME, MacLeod AM, Snyder AZ, Powers WJ, Gusnard DA, Shulman GL. A default mode of brain function. Proc Natl Acad Sci USA  2001; 98: 676– 82. Google Scholar CrossRef Search ADS PubMed  Sammaritano M, Gigli GL, Gotman J. Interictal spiking during wakefulness and sleep and the localization of foci in temporal lobe epilepsy. Neurology  1991; 41 (2 Pt 1): 290– 7. Google Scholar CrossRef Search ADS   Thornton R, Laufs H, Rodionov R, Cannadathu S, Carmichael DW, Vulliemoz S, et al.   EEG correlated functional MRI and postoperative outcome in focal epilepsy. J Neurol Neurosurg Psychiatry  2010; 81: 922– 7. Google Scholar CrossRef Search ADS PubMed  Thornton R, Vulliemoz S, Rodionov R, Carmichael DW, Chaudhary UJ, Diehl B, et al.   Epileptic networks in focal cortical dysplasia revealed using electroencephalography-functional magnetic resonance imaging. Ann Neurol  2011; 70: 822– 37. Google Scholar CrossRef Search ADS PubMed  von Ellenrieder N, Dubeau F, Gotman J, Frauscher B. Physiological and pathological high-frequency oscillations have distinct sleep-homeostatic properties. Neuroimage Clin  2017; 14: 566– 73. Google Scholar CrossRef Search ADS PubMed  von Ellenrieder N, Frauscher B, Dubeau F, Gotman J. Interaction with slow waves during sleep improves discrimination of physiologic and pathologic high-frequency oscillations (80-500 Hz). Epilepsia  2016; 57: 869– 78. Google Scholar CrossRef Search ADS PubMed  Zijlmans M, Jiruska P, Zelmann R, Leijten FS, Jefferys JG, Gotman J. High-frequency oscillations as a new biomarker in epilepsy. Ann Neurol  2012; 71: 169– 78. Google Scholar CrossRef Search ADS PubMed  © The Author(s) (2018). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: journals.permissions@oup.com

Journal

BrainOxford University Press

Published: Mar 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off