Spatial and temporal properties of intra-operatively recorded spikes and high frequency oscillations in focal cortical dysplasia

(cid:1) Spikes and ripples had high rates over resected tissue in people with focal cortical dysplasia and post-surgical seizure-freedom. (cid:1) Most of ripples and fast ripples occur before the spike peak and on the rising ﬂank. (cid:1) The ratio of ripples on spikes / all ripples was high over resected tissue and a potent marker for individual patient prediction.


Introduction
Epilepsy occurs in about 65 million people worldwide.(Kanner and Bicchi, 2022) Epilepsy surgery can cure focal epilepsy.Intraoperative electrocorticography (ioECoG) might improve the surgical outcome by delineating epileptic tissue based on display of interictal biomarkers to guide the neurosurgery on the extent of the resection.(Lesko et al., 2020) Although removal of all tissue showing interictal spikes has been related to good outcome (Tripathi et al., 2010), other research shows that seizure freedom is also achieved by incomplete removal of tissue showing spikes.(Tran et al., 1997;Roessler et al., 2019) Residual high frequency oscillations (HFOs), a newly proposed biomarker, may predict seizure recurrence better than residual spikes.(van 't Klooster et al., 2015;van 't Klooster et al., 2017) HFOs are subdivided into ripples (80-250 Hz) and fast ripples (FR;.A recent doubleblinded randomized-controlled trial demonstrated non-inferiority of HFOs to spikes in extra-temporal lobe epilepsies and not in temporal lobe epilepsies (Zweiphenning et al., 2022).Co-occurring events, like ripples and fast ripples (FR) on spikes (RoS; FRoS) are thought to be more specific for the epileptogenic focus than spikes or HFOs on their own.(Jahromi et al., 2021;Bruder et al., 2021) A study in scalp EEG suggests that RoS discriminate between clinically important sharper spikes and irrelevant propagating spikes.(van Klink et al., 2016).
Focal cortical dysplasia (FCD), the underlying pathology in 20% of people with focal epilepsy, could generate characteristic electrophysiological patterns with continuous spiking.Alternatively, spikes can be completely absent.(Ferrier et al., 2006) HFOs are often prominently present in FCD and higher HFO rates occur in patients with FCD type II compared to FCD type I. (Kerber et al., 2013;van 't Klooster et al., 2017;Straumann et al., 2022) Epilepsy surgery is successful in about 62% of people with FCD.(Baud et al., 2018) Since the border of an FCD lesion can be hard to distinguish on MRI images, more knowledge is needed to delineate the extent of resection.Improving our understanding of the relation between HFOs and spikes might increase knowledge about which tissue to resect and, thus, increase surgery success.We investigated the spatial and temporal relations of intra-operative spikes and HFOs in people with FCD with a good post-surgical outcome.The aim of this study was to find distinctive differences between resected and non-resected tissue.

Patient selection
We retrospectively selected patients from the UMCU RESPect database (Demuru et al., 2022), a database consisting of patients who underwent epilepsy surgery from 2008 till now.Patients were selected if they met the following inclusion criteria: 1) epilepsy surgery with pre-resection ioECoG (pre-ioECoG), 2) FCD pathology confirmed by histology, 3) seizure free (Engel 1A) ! 1 year after surgery, allowing us to assume that all epileptogenic tissue was incorporated in the resected area.Exclusion criteria were: 1) participant in randomized control trial 'the HFO study' (Zweiphenning et al., 2022) 2) ioECoG sample frequency < 1000 Hz, 3) long-term ECoG or SEEG monitoring performed during pre-surgical work-up.
Our institutional ethics committee waived the need for informed consent for all retrospectively collected data before 2018 À as of 2018 informed consent is required and obtained À and approved the use of pseudonymized data in the RESPect database for the purpose of retrospective research.

Clinical information
We collected clinical characteristics, including sex, age at epilepsy onset, age at surgery, anti-epileptic drugs (AED) before and after surgery, side of the lesion (left/right), anatomical location of the lesion, from RESPect database electronic data capturing system (Castor, Amsterdam, The Netherlands) -data originates from the electronic patient database (Hix, Amsterdam, The Netherlands) À. Histology reports were screened, and we used the 2011 ILAE classification of FCD subtypes to classify pathology.(Blümcke et al., 2011) Post-surgical follow-up was performed by the treating neurologist on a yearly basis.Seizure outcome was classified by the Engel classification (Engel, 1993), and seizure freedom was defined as Engel 1A.Seizure outcome at the longest follow-up, with a minimum of 1 year, was used.

Intra-operative ECoG recordings
IoECoG was recorded during surgery with a 2x4 grid, 4x5 grid and/or 1x8 strip (AD-Tech) placed directly on the cortex.The grids and strips had platinum electrodes, which were embedded in silicon, with a contact surface of 4.2 mm 2 and 10 mm inter-electrode distance.Recordings were made with a 64-channel EEG system (Systemplus, Micromed, Treviso, Italy) sampled at 1024 Hz or 2048 Hz using a common reference montage.During recordings, propofol anesthesia was temporarily paused to avoid burstsuppression patterns in the ioECoG and to avoid suppression of epileptic activity.To make reliable clinical decisions based on the io-ECoG recording, we wait until the burst-suppression is gone and then we record for an additional 2-3 min.We selected the last minute of each recording to minimize the propofol, and thus burstsuppression, in the recordings.Included recordings were selected after cessation of burst-suppression patterns.(Zijlmans et al., 2012;Maccabeo et al., 2022) Multiple recordings were done by placing the grid on different locations on the cortex before (pre-ioECoG) and after the resection.Tailoring was based on interictal spikes in the ioECoG.Per patient, the following three data preprocessing steps were taken: 1) we selected the pre-ioECoG recording where the grid covered the resected and the nonresected tissue.We chose to select one grid location -including both resected and non-resected channels without taking into account anatomical location nor functional areas -instead of all available grid locations, because many patients have grids placed on multiple locations during pre-ioECoG recording outside the ultimately resected tissue.Including all these grid locations would have merely led to an increase in the number of non-resected channels, recorded in different time epochs, while we wanted to directly compare resected and non-resected channels.2) we selected the last minute of the pre-ioECoG recording to limit bias due to burst suppression and 3) converted the data to the Brain Imaging Data Structure (BIDS) format.(Demuru et al., 2022).

Channel labels
We matched the intra-operatively taken photos of the grid location on the cortical surface with a photo of the final resected tissue (Fig. 1A).The researcher identified channels covering resected and non-resected tissue while being blinded for identified ECoG events.Subsequently we labelled all the unipolar pre-resection grid channels as follows: 1.) resected (covering the resection), 2.) edge (within 5 mm of the resection border), or 3.) non-resected (>5mm from the resection border).Bipolar labels, required for HFO analysis, were derived from the unipolar channel labels as follows (Fig. 1B); -resected, in case both channels were labeled resected, or one resected and the other on edge; -edge, in case both channels were on edge, or one was on edge and the other not-resected, or one was resected and the other not-resected; -non-resected, in case both channels were labeled not-resected.

Event detection
We used an algorithm to automatically detect spikes (Gaspard et al., 2014) and an HFO detector (Fedele et al., 2016;Boran et al., 2019) to automatically detect ripples and FRs in Matlab 2021b (The MathWorks, Inc., Natick, Massachusetts, United States).We used a train set consisting of four patients from the total population to determine the required thresholds to optimize both detectors for our ioECoG data.A unipolar referential montage was used for spikes (settings: 0.16-80 Hz IIR filter, 800 lV ampli-tude, 15sec/page) and a bipolar montage for ripples (settings: 80-1000 Hz FIR filter, 50 lV amplitude, 1sec/page) and FRs (250-1000 Hz FIR filter, 20 lV amplitude, 1sec/page).FR detection could not be performed in recordings with a sample frequency of 1024 Hz and/or those recorded with an LTM 64 Express headbox (this yielded high-frequency noise).The output of the spike detector was the time point of the peak of the spike on the channel of occurrence per detected event.The output of the HFO detector included the start and end time on the channel of occurrence per detected HFO.All automatically detected events were visually checked in BrainQuick (Micromed, Treviso, Italy) by two expert reviewers (ES, MvtK).Missing (e.g., false negative) events were added and false positive detections were discarded.
We evaluated the following co-occurring events, defined as overlap in time of occurrence: spikes with ripples (SwR), spikes with FRs (SwFR) ripples on spikes (RoS), and FRs on spikes (FRoS).We calculated the events not occurring with another event by subtracting the number of co-occurring events from the total number of events.This resulted in spikes without ripples (SwoR), spikes without FRs (SwoFR), ripples not on spikes (RnoS), and FRs not on spikes (FRnoS).We checked unipolar channels for spikes and their co-occurrence with the HFO on the adjacent bipolar channels (Fig. 1D).In case of two unipolar spikes that co-occurred during the same bipolar HFO, the first in time occurring spike was considered most relevant and was therefore used.SwR/spikeand SwoR/spike-ratios per unipolar channel were calculated by: Fig. 1.Method for channel labeling and event detection.A) Intra-operative photos are used to determine channel positions and resected tissue.B) Visualization of the grid with unipolar and bipolar channels with channel labels for each channel: resected, edge or non-resected.Detected events are spike in unipolar channels, HFO (ripples and fast ripples) on bipolar channels.D) Visualization of the method to find co-occurring events for 1. Unipolar to bipolar (SwR as example) and 2. Bipolar to unipolar (RoS as example).Note that the number of unipolar channels is different from the number of bipolar channels and the number of spikes is different than the number of HFOs, which causes a different number of SwR and RoS.E) Equations of the calculation of the ratios per channel.SwR / (SwR + SwoR) per channel and SwoR / (SwR + SwoR) per channel (Fig. 1E).RoS/ripple-and RnoS/ripple-ratios per bipolar channel were calculated by: RoS / (RoS + RnoS) per channel and RnoS / (RoS + RnoS) per channel.

Event timing
We visually marked the start time, end time, and time of the maximum for all spikes.We calculated time-based properties of the spike: the duration of the spike -defined as the time between the start of the spike and the end of the spike À, the duration of the rising flank À defined as the time between the start of the spike and the peak À, and the duration of the descending flank À defined as the time between the peak and the end of the spike.We calculated time-based properties of the HFOs: the duration of HFOsdefined as the time between the start and the end time.We calculated time-based properties of the co-occurring events: the percentages of co-occurring HFOs before the onset of spike, on the rising flank, on the descending flank and after the spike offset.

Statistical analysis
Medians and interquartile ranges (IQR) were calculated for continuous variables.We calculated the following spatial properties: the total number of events (spikes, ripples, FR, RoS, RnoS, FRoS, FRnoS) on channels covering resected and non-resected tissue and the median and IQR of number of events per channel.We compared total numbers, spike-and ripple ratios and time-based properties between all channels covering resected and non-resected tissue using a logistic mixed model: e patient represented the residuals.We performed subgroup analyses, only when ! 10 patients, for anatomical location of the resection and FCD subtypes using afore mentioned statistics.Statistics were performed in RStudio 2021.09.1 (PBC, Boston, United States).We considered a p-value < 0.05 as significant.We corrected for multiple comparison by the false discovery rate (FDR) method.(Benjamini and Hochberg, 1995)

Patient-wise tailoring
We evaluated the clinical usability of the analyzed event detections and time-based properties, per patient.For all properties in which we found a difference between the resected and nonresected tissue at group level, we calculated the differences per patient.If the directionality of the difference matched the direction at group level, it was a matching property for that patient.For example; number of RoS is positively associated with resected tissue, for each patient with a positive association between RoS and resected tissue, this is a matching property.Subsequently, we calculated the percentage of matching properties per patient and the percentage of patients with a match for each property.

Clinical characteristics
A total of 22 patients (9 female) met the inclusion criteria.Median age at surgery was 14 (range 0 -55), median age at epilepsy onset was 2 years (range 0-20 years).Two patients used AED after surgery, the rest was seizure free (Engel 1A) with complete AED withdrawal.The anatomical region of the resection was frontal in ten, parietal in five and temporal in seven patients.Four temporal lobe resections included hippocampectomies, but histopathology revealed no hippocampal abnormalities.Based on histopathology results, two patients had FCD subtype 1A, six had FCD subtype 2A, 14 had FCD subtype 2B (Table 1).

Event numbers
A total of 3344 spikes were found in 21 out of 22 patients, in 291 out of 416 unipolar channels (11 (4) spikes per channel in resected tissue and 1.5 (5) in non-resected tissue).SwR, 20% of all spikes, were found in 19 out of 22 patients, in 78 out of 416 channels (0 (1) SwR per channel in resected tissue and 0 (0) in non-resected tissue).SwoR, 80% of all spikes, were found in 21 out of 22 patients, in 286 out of 416 channels (3 (8) SwoR per channel in resected tissue and 1 (5) in non-resected tissue).A total of 1296 ripples were found in 20 out of 22 patients, in 134 out of 321 bipolar channels (0 (5) ripples per channel in resected tissue and 0 (1) in non-resected tissue).RoS,77% of total number of ripples, were found in 20 out of 22 patients, in 97 out of 321 channels (0 (2) RoS per channel in resected tissue and 0 (0) in non-resected tissue).RnoS, 23% of total number of ripples, were found in 19 out of 22 patients, in 95 out of 321 channels (0 (1) RnoS per channel in resected tissue and 0 (0) in non-resected tissue).
FR detection could not be performed in eleven patients (three with sample frequency of 1024 Hz, eight with LTM64 headbox).A total of 214 FR were found in seven out of the eleven analyzed patients, in 21 out of 182 channels.These eleven patients and channels were included in further analysis.All FRs were found in resected channels.FRoS and FRoR were found in all seven patients with FRs.94% of FRs co-occurred with spikes.Of these, on average 76% of the fast ripple span preceded the spike peaks (Fig. 3).Subsequent mixed model statistics could not be performed due to low FR counts.

Event timing
The differences in total durations of spikes and the rising and descending flanks of spikes can be found in Table 3.For all spikes, a non-significant negative trend was observed between the duration of the rising flank and resected tissue (b = À12.07,p = .07),and a positive association was seen between the descending flank and resected tissue (b = 17.41, p < .001).For SwR, a negative association was observed between resected tissue and total duration (b = À75.97,p < .001)and duration of the rising flank (b = À102.77,p < .001).A positive association was observed between resected tissue and duration of the descending flank (b = 11.88,p < .001).Comparing SwR and SwoR, we found a shorter total duration (Wilcoxon, p < .001)and shorter duration of the rising flank (Wilcoxon, p < .001) in SwR than SwoR.No difference was observed for the duration of the descending flank in SwR and SwoR (Wilcoxon, p = .47).
The median duration of ripples was 47.0 (21.0) ms in resected and 44.0 (18.8) ms in non-resected tissue (b = 3.85, p = .70;Table 3).The duration of RoS was longer than the duration of RnoS (Wilcoxon, p < .001).A positive association was observed between resected tissue and the start time of the ripple minus the start time of the spike (b = 5.25, p < .001),and a negative association was observed for the duration of RnoS and resected tissue (b = À5.71,p < .001,).The averaged time span of ripples preceded the spike peak with 62% for the resected and 56% for the non-resected tissue (Fig. 3).A positive trend was observed between resected tissue and percentage of the ripple before the spike started, while a negative association was observed between resected tissue and percentage of the ripple after the spike ended.

Table 2
Logistic mixed model results for the analysis of the total population and subgroup analysis frontal lobe and FCD 2B.Light grey box: positive significant association between resected channels and property.Dark grey box: negative significant association between resected channels and properties.b is the fixed effect of the mixed model for that type of event.p-values are the FDR adjusted p-values.Abbreviations: SwR = spike co-occurring with ripples, SwR ratio = SwR / spikes per channel, SwoR = spikes not co-occurring with ripple, SwoR ratio = SwoR / spikes per channel, RoS = ripples co-occurring with a spike, RoS ratio = RoS / ripples per channel, RnoS = ripples not co-occurring with a spike, RnoS = RnoS / ripples per channel.

Event timing
Subgroup analysis of anatomical location frontal lobe epilepsy (N = 10) showed results that were different from the total populations: significantly negative associations between resected tissue and the duration of the rising flank of all spikes, the duration of ripples and the duration of RnoS.The associations for the durations of SwR that were all in a different direction compared to the total population (Table 3).Subgroup analysis for FCD subtype 2B (N = 14) were mostly alike the total group, showing a negative association between resected tissue and total durations of SwR and RoS, and the duration of the rising flank of SwR, while a What can be seen is that especially the ratio of ripples on spikes compared to all ripples and the ratio of ripples not on spikes compared to all ripples are related to resected/non-resected channels in all individual patients.p-values are the FDR adjusted p-values.Abbreviations: RoS = ripples co-occurring with a spike, RoS-ratio = RoS / ripples per channel, RnoS = ripples not co-occurring with a spike, RnoS-ratio = RnoS / ripples per channel.positive association was found for descending flank of SwR, the total duration of RnoS and the start time of the ripple minus the start time of the spike.(Table 3).

Patient-wise tailoring
The properties used to evaluate clinical per-patient usability were the total number of spikes, SwR, SwoR, RoS/ripple-ratio, RnoS/ripple-ratio, duration of the rising flank of spike, duration of the descending flank of spike, duration of RoS, duration of RnoS, and the time between the start of the ripple and the peak of the spike.The highest percentage of matching properties per patient was 90.0% (Median: 40%, IQR: 28%)).14 out of 22 patients had at least four matching properties.In 20 patients (91%) we found at least one matching property.The percentage of patients with a match was highest for RoS/ripple-ratio (14 patients; 64%).

Discussion
A combination of spatial and time-based properties of ripples co-occurring with spikes, seems to distinguish best between resected and non-resected tissue in people with focal cortical dysplasia.Where other research focuses on the individual numbers of HFOs (Kerber et al., 2013;van 't Klooster et al., 2017;Hussain et al., 2017), we concentrate on the number and time-relations of ripples co-occurring with spikes.Van Klink et al. (2016) demonstrated a first time-relation between ripples and spikes in scalp EEG and Shi et al. (2024) showed that the majority of spike ripples were removed in implanted intracranial EEG.We are the first to explore this phenomenon in ioECoG recordings in patients with FCD.We found more ripples on spikes, earlier starts of the ripples cooccurring with the spikes, and longer ripple durations in resected tissue compared to non-resected tissue.We found that nonresected tissue contained higher SwoR/spike-ratio and RnoS/ ripple-ratio than resected tissue.Subgroup analysis revealed that these properties are clearly distinctive in patients with FCD subtype 2B.These results suggest that combining spikes and HFO properties can help delineate resected tissue, which could help to achieve seizure freedom in people with FCD.
This study emphasizes the importance of the rising flank of the spikes with a co-occurring ripple.RoS occurred earlier on the spike in resected tissue compared to non-resected tissue.The morphology of the spike can be explained by the duration of rising and descending flank.Building on literature acknowledging distinct morphologies (Engel et al., 2009), our findings align with shorter rising and longer descending flanks in resected tissue, i.e. sharper spikes are more specific for the epileptogenic zone.The earlier manifestation of RoS within resected tissue, prompting a  Table 3 Logistic mixed model results for the analysis of the total population and subgroup analysis frontal lobe and FCD 2B.Light grey box: positive significant association between resected channels and property.Dark grey box: negative significant association between resected channels and properties.b is the fixed effect of the mixed model for that timebased property.p-values are the FDR adjusted p-values.Abbreviations: SwR = spike co-occurring with ripples, SwoR = spikes not co-occurring with ripple, RoS = ripples cooccurring with a spike, RnoS = ripples not co-occurring with a spike.hypothesis that ripples and spikes originate from independent neural networks (Mölle et al., 2006).The temporal disparity between resected and non-resected tissue may be attributed to the faster spread of spikes.While spreading dynamics were not explicitly explored, this avenue holds promise for future investigations into underlying neurobiological mechanisms.While most research on RoS was conducted in scalp EEG (Jahromi et al., 2021;Bruder et al., 2021), one study in intracranial stereo-EEG similarly showed that RoS have a higher correlation with seizure onset zone (SOZ) than ripples or spikes alone.(Weiss et al., 2016) A study on ictal onset patterns suggested that spikes preceded by FR occur more often in the SOZ.(Weiss et al., 2023); A post-hoc analysis confirmed that 13 out of 17 FRnoS precede a spike within 300 ms.Another post-hoc analysis showed that all FR co-occur with a ripple within 100 ms.Thus, spikes superimposed simultaneously with ripples and FR provide no additional information to SwFR.
It is often indicated in literature (Engel et al., 2009) that RnoS should be considered as physiological and have a longer duration than pathological ripples.We found a negative association between resected tissue and RnoS/ripples-ratio.In other words, the more RnoS were seen compared to RoS in one channel, the more likely it is that the tissue underneath that channels was not resected.We found longer durations of RoS than RnoS.This contrasts with the assumption that RnoS are physiological ripples and have a longer duration.The RnoS were shorter in resected tissue than in non-resected tissue and part of RnoS might thus be pathological ripples as well.(Alkawadri et al., 2014) The process of selecting both resected and non-resected tissue was deliberately conducted without being influenced by any knowledge of anatomical location or functional areas, but researchers could not be totally blinded.While we aimed to minimize biases associated with these factors, it is important to acknowledge that our data selection methodology may have inadvertently introduced biases in the pursuit of achieving statistical significance.
In all patients with FCD type 2B that were included for analyzing FR, FR were present.We found low numbers of ripples and no FR in patients with FCD type 1A.These results are similar to findings in long-term intracranial EEG, where higher numbers of HFOs were found in FCD types 2 than FCD types 1. (Kerber et al., 2013) It is important to note that our study cohort includes a limited number of patients with FCD type 1, which might impact the generalizability of this comparison.A study by Abdijadid et al. (2015) investigating the mechanisms of epileptogenisis and it electrographic correlates, in FCD type I and II, demonstrated that not the balloon cells, but the dysmorphic and immature neurons play a crucial role in initiation and spread of epileptic spikes.It is interesting what the role of these neurons play in the initiation and spread of HFOs.The lack of specific histological findings beneath each electrode in this retrospective study hinders this extra analysis.Our intra-operative recordings, and thereby potentially also the number of HFO events, were influenced by propofol anesthesia.Anesthesia reduces the number of HFOs in comparison to HFOs during non-REM sleep, while it does not reduce the number of interictal spikes.(Weiss et al., 2021;Zijlmans et al., 2012) This could explain the fewer occurrences of HFOs, compared to other long-term intracranial studies.(Bagshaw et al., 2009;Kerber et al., 2013).
Whereas in most intracranial EEG studies the events are compared to the SOZ, we compared the events to the resected tissue.Labelling the SOZ requires ictal information and subject to nonobjective interpretation by the epileptologists, and it's value as gold standard is debated as complete removal of SOZ tissue is no guaranty for seizure freedom (Téllez-Zenteno et al., 2010).Similar to the rationale of Cimbalnik et al. (2019) we labeled resected tissue as epileptic tissue, which represents the most practical method to label epileptic tissue in our study population in absence of SOZ information.By selecting a population based on strict definition of seizure-freedom, we assumed that all epileptic tissue was incorporated in the resected tissue.A consequence of this choice is that resected tissue might contain both epileptogenic and nonepileptogenic tissue.Channels could be labeled as resected, even if there is no epileptogenic tissue underneath.This inclusion of resected non-epileptogenic channels in our results may have an impact.According to our anticipation, these channels may possess properties similar to non-resected tissue, thereby complicating the differentiation between resected and non-resected regions.Another limitation of our study is the exclusion of patients with unfavorable outcome.In our patient selection, with 91% of patient long-term Engel 1A without any AED, strengthens our hypothesis that the epileptic tissue was incorporated in the resected tissue.
A limitation is that we could not analyze FR in all patients, because of the high frequency noise produced by the LTM64 headbox (Micromed, Italy).This is an example of problems that can be encountered in clinical practice.A lot of devices are present in the operation room, which yield noise in the ioECoG recordings.It is imported to be aware of this to accurately interpret findings in the recordings.Another limitation is the choice of one recording per patient.With this choice, we lowered the number of channels covering non-resected tissue, and thus improved the resected versus non-resected channels ratio.
Although our included patients with resections of FCD still yield a heterogeneous group with differences in the anatomical location of the resection, subgroup analysis (n > 10) could be performed for frontal-lobe resection.The HFO trial, a randomized controlled trial, suggests that extra-temporal resections might benefit more from HFO analysis than temporal resections.(Zweiphenning et al., 2022) We confirm the results of use of HFOs in the subgroup of frontal-lobe resections (n = 10), but not for TLE due to a too small group size (N = 7).Please note that the HFO trial was conducted in the same center, but there was no overlap in patients for this study.In this retrospective study, we have defined anatomical locations based on the position of the FCD lesion as identified on the MRI.It is crucial to note that the exact electrode positions were not determined in accordance with gyrus anatomy due to the inherent retrospective nature of our study.As we move forward, the development of an intra-operative HFO atlas could potentially address these challenges, and we remain vigilant about incorporating such advancements in future studies.
In clinical practice, the underlying pathology is rarely confirmed prior to the surgery.Often, an FCD pathology is suggested by the radiologist based on MRI imaging pre-surgery.A next step would be to see if the pathology can be predicted by the ioECoG signal, a recent study did this on stereo-EEG data.(Sklenarova et al., 2023) This knowledge about the underlying pathology might help the neurosurgeon during epilepsy surgery.Future research focusing on biomarkers in ioECoG should examine variations in the fundamental pathologies causing epilepsy.Pathology-specific biomarkers in the ioECoG could further help delineate the epileptic tissue during epilepsy surgery.was supported by the EpilepsieNL under Award Number NEF17-07 (DvB).The project was co-funded by the PPP Allowance made available by Health$Holland, Top Sector Life Sciences & Health, to stimulate public-private partnerships.
Fig.1.Method for channel labeling and event detection.A) Intra-operative photos are used to determine channel positions and resected tissue.B) Visualization of the grid with unipolar and bipolar channels with channel labels for each channel: resected, edge or non-resected.Detected events are spike in unipolar channels, HFO (ripples and fast ripples) on bipolar channels.D) Visualization of the method to find co-occurring events for 1. Unipolar to bipolar (SwR as example) and 2. Bipolar to unipolar (RoS as example).Note that the number of unipolar channels is different from the number of bipolar channels and the number of spikes is different than the number of HFOs, which causes a different number of SwR and RoS.E) Equations of the calculation of the ratios per channel.F) Patient example (Pat.nr.6, FCD subtype 2B) of raw and filtered ioECoG data; F1) Ten seconds of raw data in AVG montage (0-70 Hz, IRR filtered) F2) Spike markings in one-second zoom of data in AVG montage/ (0-70 Hz, IRR filtered).F3) Ripple markings in bipolar montage (80-250 Hz FIR filtered), F4) FR markings in bipolar montage (250-500 Hz FIR filtered).Abbreviations: HFO = high frequency oscillation, SwR = spike co-occurring with a ripple, SwoR = spike not co-occurring with a ripple, RoS = ripple co-occurring with a spike, RnoS = ripples not co-occuring with a spike, ioECoG = intra-operative electrocorticography.
resection patient was the binary dependent variable (resected / nonresected), x property the spatial or time-based property, a the fixed intercept, b the average relation between the dependent variable and the x property .a patient and c patient x property represented the random intercept and slope for each patient.The random effect (d patient electrode) was added when analyzing time-based properties.

Fig. 2 .
Fig. 2. Relation between total number of events or ratio and resected tissue.Colored lines are the patient specific logistic regression lines.The bold black line represents the fixed effect of the model.The fixed coefficient b represents the average relation between the dependent variable and resected channels.What can be seen is that especially the ratio of ripples on spikes compared to all ripples and the ratio of ripples not on spikes compared to all ripples are related to resected/non-resected channels in all individual patients.p-values are the FDR adjusted p-values.Abbreviations: RoS = ripples co-occurring with a spike, RoS-ratio = RoS / ripples per channel, RnoS = ripples not co-occurring with a spike, RnoS-ratio = RnoS / ripples per channel.

Fig. 3 .
Fig. 3. Mean percentages of the ripple (top) and fast ripple (bottom) with respect to the time of the co-occurring spike.A) resected tissue, B) non-resected tissue.Note that there were no fast ripples present in non-resected tissue.Abbreviations: a: start of the spike, b: peak of the spike, c: end of the spike.HFO = high frequency oscillation.