Proportion of resected seizure onset zone contacts in pediatric stereo-EEG-guided resective surgery does not correlate with outcome

OBJECTIVE
We aimed to determine whether the proportion of putative seizure onset zone (SOZ) contacts resected associates with seizure outcome in a cohort of children undergoing stereoelectroencephalography (SEEG)-guided resective epilepsy surgery.


METHODS
Patients who underwent SEEG-guided resective surgery over a six-year period were included. The proportion of SOZ contacts resected was determined by co-registration of pre- and post-operative imaging. Outcome was classified as seizure free (SF, Engel class I) or not seizure-free (NSF, Engel class II-IV) at last clinical follow-up.


RESULTS
Twenty-nine patients underwent resection of whom 22 had sufficient imaging data for analysis (median age at surgery of 10 years, range 5-18). Fifteen (68.2%) were SF at median follow-up of 19.5 months (range 12-46). On univariate analysis, histopathology, was the only significant factor associated with SF (p < 0.05). The percentage of defined SOZ contacts resected ranged from 25-100% and was not associated with SF (p = 0.89). In a binary logistic regression model, it was highly likely that histology was the only independent predictor of outcome.


CONCLUSIONS
The percentage of SOZ contacts resected was not associated with SF in children undergoing SEEG-guided resective epilepsy surgery.


SIGNIFICANCE
Factors such as spatial organisation of the epileptogenic zone, neurophysiological biomarkers and the prospective identification of pathological tissue may therefore play an important role.


Introduction
For carefully selected children with drug-resistant focal epilepsy, resective surgery is an established treatment, with up to 70% achieving seizure freedom (SF) (Barba et al., 2020). To delineate the resective target, a careful pre-surgical evaluation must be carried out and, in select candidates, this can involve the use of intracranial electroencephalography (iEEG) including stereoelectroencephalography (SEEG).
In cases that proceed directly to resective surgery (without iEEG), it has been shown that factors including complete resection of the magnetic resonance imaging (MRI)-visible lesion and histopathological diagnosis are key determinants of seizure freedom (Lamberink et al., 2020). The factors determining SF following SEEG-guided resective surgery have not been extensively studied. Correct delineation and subsequent resection of the putative seizure onset zone (SOZ) are potentially important factors. In adults, there is evidence that other markers such as interictal high frequency oscillations (HFOs) (Thomschewski, Hincapié and Frauscher, 2019) and ictal phase-locked high gamma (PLHG) (Weiss et al., 2015) may be better markers than the putative SOZ contacts, although these have largely been in patients undergoing subdural grid and strip recordings. The main aims of this study were to (a) quantify the proportion of SEEG-defined putative SOZ contacts resected by co-registering pre-and post-operative imaging and (b) identify factors, including the proportion of these contact resected, associated with post-operative SF in paediatric patients undergoing SEEG-guided resective epilepsy surgery at a single centre.

Setting
This was a single-centre, retrospective, observational study. STROBE guidelines were adhered to throughout this study (von Elm et al., 2007). The project was registered with the Great Ormond Street Hospital (GOSH) R&D Office (19BI26). As it involved only retrospective use of routinely collected clinical data, formal ethical approval was not required.

Participants
Paediatric patients (aged 18 years) who underwent SEEG at GOSH between 2014 and 2020 and subsequent resective epilepsy surgery were eligible for inclusion. Patients who had undergone previous epilepsy surgery, patients with tuberous sclerosis and patients with large structural abnormalities on MRI or computerised tomography (CT) imaging that would affect robust coregistration were excluded. As previously published, patients are selected for SEEG and subsequent surgical treatment based on a multidisciplinary team decision (UK Children's Epilepsy Surgery Collaboration, 2021). This cohort overlaps partially with this previously published cohort and the technical details of the SEEG procedure and the clinical workflow are outlined elsewhere (Narizzano et al., 2017;Sharma et al., 2019; (UK Children's Epilepsy Surgery Collaboration, 2021).

Data collection
Demographic, pre-surgical evaluation and SEEG variables were collected from electronic patient notes via a piloted proforma. Seizure onset patterns (SOP), which have been known to associate with post-surgical seizure outcome, were classified according to the methodology by Lagarde et al (Lagarde et al., 2019). Better prognosis has been reported with the presence of low voltage fast activity (LVFA) and for statistical analysis, the 8 patterns were dichotomised based on either the presence or absence of LVFA on SEEG (Lagarde et al., 2019). The main outcome measure was the Engel classification at last follow-up, dichotomised into SF (Engel class I) and not-seizure free (NSF, Engel classes II-IV). Assessment within 30 days of one year were classified as sufficient for one year follow-up. The SOZ contacts were taken as the ictal onset contacts as defined by the consultant neurophysiologist in the formal SEEG report; this was a descriptive definition following assessment of the SEEG electrophysiological and video data and encompasses the contacts where there was initial seizure activity, prior to onset of clinical seizure manifestation.

Segmentation & image registration
Individual electrode contact points from SEEG electrodes, localised using a CT scan and identified using SEEG assistant (Narizzano et al., 2017), were assigned voxel spaces and registered to the pre-operative imaging using reg_aladin (http://cmictig.cs. ucl.ac.uk/wiki/index.php/Reg_aladin). Post-operative volumetric MRI scans were used to manually delineate the area of resection using ITK-SNAP (v.3.X) and registered to the pre-operative imaging using ANTS (https://antspy.readthedocs.io/en/latest/index.html); the resection volume was excluded during this procedure to minimise distortion and account for brain collapse into the resection cavity, a common event following brain resections ( Fig. 1) (Wellmer et al., 2002;Ozawa et al., 2009;Fan et al., 2018). The resected contacts were subsequently manually identified using FSLeyes to determine electrode overlap with the resection volume ( Fig. 1).
Following identification of patients without completely resected SOZ contacts, we classified the reasons for this into the following to assess whether there were differences in the reasons in the seizure free and non-seizure free groups.: 1) plan to resect, however not executed during resective surgical procedure 2) plan to not resect due to electrophysiological reasons (distant or disconnected SOZ contacts and clear decision beforehand to not resect these); 3) likely resected however confirmation of resection was limited by pre-to post-operative image registration; 4) All contacts resected.

Statistical analysis
Appropriate statistical tests (chi-sq for categorical variables & Mann-Whitney U-test for continuous variables) were used to assess association between variables of interest and seizure freedom. A post-hoc, residual analysis was performed to assess individual sub-categorical significance if predictors with more than two categories were significant, using Bonferroni-adjusted pvalues to determine significance (García-pérez and Núñez-antón, 2003). A binary logistic regression model was fitted to assess for independent predictors for the post-operative Engel outcome.
Patients without sufficient imaging (either pre-operative or post-operative) were excluded from the primary analysis but were included in a sensitivity analysis (identical to the primary analysis but excluding variables determined by image analysis) to determine the robustness of the primary analysis.
All statistical tests were performed on SPSS v27. Images were created using GraphPad Prism 9.1. Statistical significance was taken at p-values < 0.05.
From the 29 patients undergoing resective surgery, 22 had sufficient imaging data for the primary analysis (Table 1); all 29 were Fig. 1. Co-registration of pre-and post-operative magnetic resonance images with stereoelectroencephalography (SEEG) electrodes' computerised tomography images and segmentation images. Magnetic resonance imaging (MRI) reconstructions of the resected seizure onset zone (SOZ) based on SEEG recordings, with co-registered coronal (left hand side panels) and axial (right hand side panels) scans. Four separate depth electrodes can be seen: electrode a targeting the right parietal operculum (light blue), electrode b targeting the right posterior superior temporal gyrus (STG) (red), electrode v targeting the left STG (green) and electrode d yellow targeting the right posterior hippocampus (yellow). Each electrode contact point is represented by a circle. Starting from the deepest one, contacts for each electrode are labelled numerically. In this example, pre-operative SEEG ictal recordings revealed a putative SOZ in the right STG (particularly the temporal operculum), represented by contacts b [6,7,8,9,10] (A and B). These contacts are subsequently resected, as indicated in the post-operative scans (C and D). The resection area is then manually delineated to calculate the proportion of SOZ contacts resected, with the resection cavity marked in dark blue (E and F). In this case, 50% of the SEEG-defined SOZ contacts have been resected (note that other electrodes in the SEEG-defined, putative SOZ cannot be seen in these MRI planes. These MRI planes depicting 4 electrodes have only been chosen for representation purposes).
included in the subsequent sensitivity analysis (Supplementary Table 1).
The percentage of SOZ contacts resected did not significantly associate with seizure outcome (p = 0.89), although there was a trend showing that the number of identified SOZ contacts was lower in the SF group (Fig. 3B-C). There was also not a significant difference when looking at the distribution of reasons for incomplete resection of the SOZ contacts (p = 0.87) ( Table 2).
A binary logistic regression model with backwards elimination was fitted to ascertain the effects of pre-operative & SEEG factors on post-operative seizure outcome. Variables with a p 0.3 on univariate analysis were selected, resulting in a statistically significant model (p < 0.05). The model explained 56% of the variance in seizure outcome and correctly classified 77.3% of cases. The model parameter estimates failed to converge due to pseudo-complete separation of the seizure outcome data with the histopathological classification as all the FCD & HS patients had a favourable outcome. Although this resulted in the p-value for histopathology being uninterpretable, it is highly likely that the variable was an independent significant predictor in the model (Heinze and Schemper, 2002). All other variables were not independent predictors of seizure outcome.

Sensitivity analysis
In a sensitivity analysis including a total of 29 patients, all the above univariate and binary logistic regression statistical outcomes were confirmed, indicating robust outcomes and no systematic bias in the patients that did/did not have adequate imaging (Supplementary Table 1).

Discussion
With post-operative SF rates of up 70% (Barba et al., 2020), epilepsy surgery is an accepted treatment option for children with focal drug-resistant epilepsy. Numerous studies over the years have assessed predictors of post-operative SF in these patients. The importance of the extent of resection of the MRI-visible lesion is well known, with complete resection being a significant factor determining SF (Lamberink et al., 2020). However, comparatively little is known about the corollary of this 'extent of resection' in the context of SEEG-guided resective epilepsy surgery. In our series, we identify that high rate of seizure freedom (66% of the entire cohort) is possible following SEEG-guided resective surgery. We identified that the percentage of SEEG contacts resected did not associate statistically with SF both in univariate and binary logistic regression analyses. Histology was the only significant factor predicting seizure outcome, with FCD being associated with SF status and ND with NSF status.
These findings highlight the limitations of current neurophysiological paradigms in delineating the SOZ to guide resective surgery in children undergoing SEEG. The different SOP were not associated with seizure outcome. Interestingly, there was a trend to suggest that a lower number of SOZ contacts associated with seizure freedom suggesting that a more focal SOZ may be more favourable than widespread network onset, in agreement with the findings  of Lagarde et al (Lagarde et al., 2019). Perhaps, a smaller number of electrode contacts indicates a compact, localised SOZ which is more likely to indicate a discrete focal brain abnormality (Bartolomei et al., 2017). Unsurprisingly, although not significant, we also found a general trend consistent with the literature when analysing SF rates between our temporal and extratemporal patients, with SF rates of 83.3% and 62.5% respectively (p = 0.35).
As previously suggested, this could be attributed to the difficulty of SOZ localisation via SEEG in extratemporal epilepsy, where the SOZ is usually more widespread (Zentner et al., 1996;Sinclair et al., 2004). If the resection involves an FCD, seizure freedom is highly likely irrespective of the percentage of neurophysiologically-defined SOZ contacts resected whilst a ND histology is associated with less favourable outcomes. In these cases, it could be that the SEEG resulted in mis-localisation of the actual pathological abnormality (e.g., an FCD elsewhere) or there is no clear abnormality, both of which have been associated with poorer outcomes. Alternatively, it could be that there are alternative pathological entities that are newly being identified such as mild malformation of cortical development with oligodendroglial hyperplasia and epilepsy (MOGHE), associated with poorer outcomes (Seetharam et al., 2021).
The results also highlight the difficulty of ensuring that all intended contacts are resected following SEEG. Whilst some of the variability may be down to registration error, resecting intended contacts may be limited by functional boundaries or geographically separated contacts which are not all amenable to being resected. Advances in intraoperative navigation (including adding the SEEG electrode targets) may aid achieving the intended resections.
The results also highlight the expressed need for novel computational analyses and perspectives on interpreting SEEG data (Bartolomei et al., 2017). This has been acknowledged for a number of years and, recently, multiple novel approaches including identification of novel SOZ biomarkers (ictal high gamma activity and interictal HFOs) (Weiss et al., 2015;Thomschewski, Hincapié and Frauscher, 2019), comparison with normalised connectivity atlases (Frauscher et al., 2018;Taylor et al., 2021) and novel network synchronizability metrics (Khambhati et al., 2016) have been used to try and explain surgical failures although these computational analyses are not yet, to our knowledge, being used to guide routine clinical practice. Prospective multicentre evaluation of these technologies is crucial to prove efficacy prior to widespread adoption.

Limitations
There are several limitations to this study. Despite robust statistical outcomes following sensitivity analyses, this study is retrospective and its generalisability to other centres is unknown. Furthermore, the neurophysiological definition of SOZ based on the clinical report was not interrogated further and taken at face value. This was performed in a two-step process. Firstly, descriptive findings of the ictal EEG traces from neurophysiology reports were analysed and thereafter, the SOP were classified in accordance to the methodology by Lagarde et al (Lagarde et al., 2019). We aimed to evaluate current standard practice workflows and, therefore, quantitative metrics were not assessed. As with all SEEG studies, this technique suffers from an important limitation of sparse sampling of brain tissue; however, this is a uniform issue across patients and the density of sampling is affected by the pre-implantation hypotheses, which are difficult to compare between patients.

Conclusion
In this single centre study analysing 22 patients undergoing SEEG-guided resective epilepsy surgery, we found that the propor-tion of SEEG-defined SOZ contacts resected is not a significant predictor of post-operative seizure outcome. The histopathology was the only significant predictor of seizure outcome; all patients with a diagnosis of FCD were seizure-free at last follow-up. All other pre-operative, operative and post-operative factors did not significantly associate with seizure outcome.
Disclosures AC is supported by a Great Ormond Street Hospital Children's Charity Surgeon Scientist Fellowship. This work was supported by the National Institute of Health Research-Great Ormond Street Hospital Biomedical Research Centre.
JHC hold as endowed chair at UCL Great Ormond Street Institute of Child Health; she holds grants from NIHR, EPSRC, GOSH Charity, ERUK, the Waterloo Foundation and the Great Ormond Street Hospital Biomedical Research Centre. She has acted as an investigator for studies with GW Pharma, Zogenix, Vitaflo, Stoke Therapeutics and Marinius. She has been a speaker and on advisory boards for GW Pharma, Zogenix, and Nutricia; all remuneration has been paid to her department.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.