Functional connectivity in electroencephalography of brain tumor patients in status epilepticus: a retrospective evaluation of prognostic value

Introduction

Status epilepticus (SE) is a life-threatening neurological emergency associated with high morbidity and mortality (1), particularly in patients with underlying brain tumors. Prognostic assessment in this population is difficult, as both tumor biology and SE contribute to outcome, and established scoring systems such as Status Epilepticus Severity Score (STESS) and Epidemiology-based Mortality Score in Status Epilepticus (EMSE) may not fully capture these interactions.

Epileptic seizures occur in up to 30% of brain tumor patients (2), with tumor-associated SE (TASE) accounting for 3–12% of adult SE cases (3). High-grade gliomas (HGGs) are especially epileptogenic, with seizures in up to 80% of patients (4), whereas meningiomas and metastases are associated with lower seizure rates (5). Tumor characteristics such as location, size, and histology influence both risk and manifestation of SE.

SE may become refractory (RSE) or progress to super-refractory SE (SRSE), which is associated with poor outcomes (6). In brain metastases, 57.9% of patients with prior seizures developed RSE or SRSE (7). Management is further complicated by oncological treatments that affect seizure thresholds (8). Adverse prognostic factors include advanced age, de novo SE, reduced consciousness, and certain seizure types (9-11). Electroencephalography (EEG) provides insights into neural dynamics during SE. Spectral and connectivity analyses may refine prognostication (12), and EEG-derived biomarkers have shown potential for indicating disease progression and treatment response (13,14). Moreover, the application of machine learning—particularly deep learning approaches—has demonstrated substantial potential in extracting clinically relevant information from complex EEG datasets, achieving classification accuracies superior to conventional methods (15). Despite these advances, the use of such techniques for outcome prediction in SE patients with brain tumors remains underexplored, and no robust biomarkers have yet been established to reliably predict individual patient trajectories (16,17).

Objectives

The aim of this study was to investigate, in a proof-of-principle framework, whether EEG-based functional connectivity can differentiate survival in brain tumor patients with SE. Our focus was on methodological feasibility and hypothesis generation, acknowledging that definitive prognostic conclusions cannot be drawn from this cohort size. Additionally, we aimed to examine whether connectivity patterns vary across tumor types, including gliomas, meningiomas, and brain metastases in our patient cohort. Through the integration of advanced EEG analytical methodologies, we aim to identify potential biomarkers that could inform clinical management and improve prognostication in this challenging patient population. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1739/rc).

Methods

Study design and setting

This study is a retrospective, monocentric cohort analysis including consecutive patients with brain tumours who were admitted with SE and underwent EEG recordings at the University Hospital Marburg between 2012 and 2022.

In this proof-of-principle, retrospective study, we sought to determine whether EEG-derived functional connectivity can, in principle, differentiate survival outcomes in patients with brain tumors and SE, potentially informing individualized therapeutic strategies. We reviewed all consecutive patients admitted to the Department of Neurology, University Hospital Marburg, Germany, between 2012 and 2022, who underwent EEG recording during an episode of SE and had a confirmed diagnosis of a brain tumor. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the regional Ethics Committee of the Department of Human Medicine at Philipps University Marburg (No. 23-187 RS) and individual consent for this retrospective analysis was waived.

Participants

This monocentric cohort comprised 37 patients. Routine EEGs were acquired using a Nihon Kohden system (10–20 electrode configuration, eyes closed, 30-minute duration) exclusively during episodes of SE, and not in the postictal state.

Comprehensive clinical data were extracted for each patient, including age at EEG, gender, tumor diagnosis and histology, Karnofsky Performance Status Scale (Karnofsky Index), Eastern Cooperative Oncology Group (ECOG) performance status, SE duration, and survival at 3 months, 1 year, 3 years, and 5 years. SE-specific parameters included the number of anti-seizure medications (ASMs) administered and the need for escalation to level 3 therapy (intubation anesthesia).

Variables

The primary outcome was survival status at 1 year following SE. Secondary outcomes included overall survival at 3, 12, 36, and 60 months. The main exposures of interest were tumor type (glioma, meningioma, metastasis, other) and EEG-derived functional connectivity across canonical frequency bands. Predictors considered comprised demographic and clinical variables such as age, sex, Karnofsky performance status, ECOG score, duration of SE, and number of ASMs administered. Potential confounders included prior tumor resection and tumor histology, while SE severity and treatment escalation to anesthesia were regarded as possible effect modifiers.

Bias

To minimize selection bias, all consecutive patients with brain tumors and SE treated at our center between 2012 and 2022 were included. EEG recordings were obtained exclusively during episodes of SE to ensure comparability of electrophysiological data. To reduce measurement bias, a standardized preprocessing pipeline with automated and manual artefact rejection was applied. Nevertheless, we acknowledge that the retrospective and monocentric design, the heterogeneous tumor entities, and the limited sample size may have introduced residual confounding.

EEG preprocessing and epoch selection

EEG data were preprocessed with a band-pass filter (0.5–100 Hz) and a 50 Hz notch filter. Artifact reduction was performed using Persyst’s blind source separation algorithm to remove muscle, electrode, ocular, and cardiac artifacts. The artifact-corrected EEG was segmented into 3-second epochs. For each epoch, relative power was calculated for the canonical frequency bands (delta, theta, alpha, beta, gamma). For each band, the 30 epochs with the highest relative power were automatically selected, balancing representativeness of activity with computational feasibility, in line with prior methodological work, maximizing the representativeness of band-specific activity. Epochs contaminated by residual artifacts were excluded. This approach allowed for the selection of distinct epochs across frequency bands, reflecting the temporal variability of band-specific activity.

Source space connectivity analysis

Spectral analysis was performed on each recording to select the epochs for each of the canonical EEG bands: delta, theta, alpha, beta, and gamma. The representative epochs were used to compute the functional connectivity pattern in each band separately. The procedure to go from EEG to source space connectivity is depicted in Figure 1. First, EEG source imaging with a template head model was done to extract the neuronal time series in 50 regions of interest indicated in Figure 2. The template head model was constructed from an adult male magnetic resonance imaging (MRI). The MRI was resampled to 1×1×1 mm voxels. Afterward, the MRI was segmented into six tissues: gray matter, white matter, cerebrospinal fluid, skull, scalp, and air cavities. To each 1×1×1 mm voxel, the electrical conductivity of the corresponding tissue class was assigned. The electrodes were placed on the scalp by calculating distances from selected landmarks (nasion, inion and auricular points) and placing them at the correct intersections. The finite difference method was used to calculate the leadfields. The gray matter was parcelled into 25 sublobes per hemisphere. Weighted Minimum Norm Estimation (wMNE) was used to extract the time series per sublobe. The functional connectivity analysis was run on these sublobar time series.

Figure 1 EEG connectivity analysis in source space: from MRI and routine EEG to source space functional connectome by using a template head model. EEG, electroencephalography; MRI, magnetic resonance imaging.

Figure 2 Brain parcellation into sublobes: 25×2 hemispheres =50 sublobes.

Functional connectivity between these estimated time series was calculated using the weighted Phase Lag Index (wPLI). The wPLI is a tool to assess EEG functional connectivity and plays a central role in characterizing the neural network (18-20). The synchrony of rhythmic activity in EEG captures the brain neuronal network, while graph theoretical analysis describes the network architecture and has been widely applied in epilepsy research (21,22). The wPLI is a refinement of the phase lag index that incorporates a weighting of lag magnitude to reduce the influence of small phase differences (23,24). It was chosen for its robustness against volume conduction and source leakage, factors that can introduce spurious connections, and it more accurately reflects the strength and stability of phase synchronization compared to the traditional phase synchronization index. Alternative connectivity measures such as coherence or imaginary coherence were considered, but these are more sensitive to volume conduction effects and can inflate connectivity estimates in source-reconstructed EEG. Likewise, graph-theoretical parameters could have been derived, but given the proof-of-principle design and limited sample size, we focused on wPLI as a direct and robust phase-based metric. By weighting phase differences, wPLI reduces the influence of zero-phase lag and improves accuracy in capturing true phase synchrony, even at low signal-to-noise ratios.

Statistical analysis

Group-level statistical comparisons were conducted to evaluate differences in functional connectivity according to two primary factors: survival and tumor type. For the survival analysis, patients were dichotomized into “high survival” (surviving more than 1 year following SE) and “low survival” (surviving less than 1 year). Eight patients lacking survival data at the 1-year mark were excluded from this analysis. For tumor type, patients were classified into three categories: glioma, meningioma, and brain metastases; four patients with other tumor entities were excluded from this subgroup analysis.

To assess the statistical significance of differences in functional connectivity between groups, a non-parametric permutation testing approach was employed. Specifically, 150,000 permutation iterations were performed for each comparison, and the false discovery rate (FDR) was controlled at a significance threshold of 5% to correct for multiple comparisons. This permutation-based procedure was applied to the connectomes derived from each of the five canonical EEG frequency bands.

For clinical and demographic variables, statistical analyses were tailored to the number of groups and the distributional properties of the data. Dichotomous variables (e.g., gender, tumor resection prior to SE) were analyzed using Chi-squared or Fisher’s exact tests, as appropriate, within both the survival and tumor type analyses. Logistic regression was utilized to further explore associations between these categorical variables and survival or tumor type.

Continuous, discrete, and ordinal variables (e.g., age, ECOG performance status, Karnofsky Index, SE duration, number of ASMs) were analyzed as follows: For two-group comparisons (high vs. low survival), Student’s t-test was applied for normally distributed data with homogeneous variances. If variances were unequal, Welch’s t-test was used. For non-normally distributed data with equal variances, the Mann-Whitney U test was employed, and Yuen’s test was used when both normality and variance homogeneity assumptions were violated. For three-group comparisons (glioma, meningioma, metastasis), analysis of variance (ANOVA) was used for normally distributed data, with Welch’s ANOVA applied in the presence of variance heterogeneity. Post hoc analyses were conducted using the Games-Howell test (for unequal variances) or Tukey’s test (for homogeneous variances) when significant results were observed. For non-normally distributed data, the Kruskal-Wallis test was utilized, followed by the Dwass-Steel-Critchlow-Fligner test for post hoc comparisons. In cases where neither normality nor variance homogeneity could be assumed, robust ANOVA methods were implemented.

Preliminary descriptive analyses included visual inspection of data distributions (e.g., histograms, Q-Q plots) and formal testing for normality (Shapiro-Wilk test) and homogeneity of variances (Levene’s test). All statistical tests were conducted using a two-tailed significance level of α=0.05.

Software

EEG preprocessing, source imaging, spectral analysis, and functional connectivity computations were performed in MATLAB (version 2018b) (25) using custom-developed scripts. Statistical analyses of functional connectomes were conducted in Python (26), while additional statistical evaluations were performed using jamovi (27).

Results

Patient characteristics, descriptive data

A total of 37 patients met the inclusion criteria, with a mean age of 68.1 years (range, 41–88 years). The cohort comprised 25 females (67.6%) and 12 males (32.4%). Tumor diagnoses included 12 adult-type diffuse gliomas [9 glioblastomas, 2 astrocytomas, 1 oligodendroglioma), 11 intracranial meningiomas, and 11 brain metastases [originating from lung (n=4), melanoma (n=4), breast (n=1), renal cell carcinoma (n=1), and rectal carcinoma (n=1)]. Single cases of primary B-cell lymphoma, vestibular schwannoma, and pineal tumor were also present (Table 1). Tumor diagnosis was histologically confirmed in 22 patients; the remainder were diagnosed radiologically. The mean duration of SE was 41.8 hours (range, 1–120 hours). Prior tumor resection had been performed in 20 patients (54.1%), while 17 (45.9%) had not undergone resection. At 1-year follow-up, 12 patients (32.4%) survived, 17 (45.9%) had died, and 8 (21.6%) had unknown outcomes due to loss to follow-up.

Survival group analysis

Of the 29 patients with available 1-year survival data, 12 (41.4%) survived beyond 1 year (high-survival group), while 17 (58.6%) died within 1 year (low-survival group). The mean age was 62.7 years (range, 41–78 years) in the high-survival group and 67.9 years (range, 47–88 years) in the low-survival group. Gender distribution was balanced in the high-survival group (50% male, 50% female), whereas females predominated in the low-survival group (70.6% female, 29.4% male). Prior tumor resection was more frequent in the high-survival group (75%) compared to the low-survival group (47.1%). Tumor type distribution in the high-survival group was evenly split among glioblastoma, meningioma, and brain metastases (each 25%), whereas in the low-survival group, brain metastases (35.3%) and glioblastoma (29.4%) were most common (Table 1). The mean ECOG performance status was 1.92 in the high-survival group and 2.24 in the low-survival group. The mean Karnofsky Index was higher in the high-survival group (69.2) compared to the low-survival group (62.9). SE duration was shorter among high-survival patients (mean: 29.6 hours) than low-survival patients (mean: 59.5 hours). The mean number of ASM administered was 2.42 in the high-survival group and 2.94 in the low-survival group (Tables 2,3).

Tumor type group analysis

Thirty-three patients were included in the tumor type analysis, with 11 patients each in the glioma, meningioma, and brain metastasis groups. The mean age was 65.0 years (glioma), 79.0 years (meningioma), and 61.5 years (brain metastases). Gender distribution was balanced in the glioma group (54.5% female), predominantly female in the meningioma group (90.9%), and predominantly male in the brain metastases group (54.5%).

Prior tumor resection was reported in 54.5% (glioma), 63.6% (meningioma), and 45.5% (brain metastases) of patients. The mean ECOG performance status was 2.18 (glioma), 2.00 (meningioma), and 1.82 (brain metastases). The mean Karnofsky Index was 63.6% (glioma), 69.1% (meningioma), and 68.3% (brain metastases). SE duration averaged 55.1 hours (glioma), 27.5 hours (meningioma), and 44.7 hours (brain metastases). One-year survival rates were 36.4% (glioma), 27.3% (meningioma), and 27.3% (brain metastases), with a substantial proportion of meningioma cases (45.5%) lost to follow-up. The mean number of ASMs administered was 2.72 (glioma), 3.00 (meningioma), and 2.91 (brain metastases) (Tables S1,S2).

Functional connectivity analysis in source space

No statistically significant differences in functional connectivity, as measured by the wPLI, were observed between survival groups across any frequency band (Figure 3). While higher source-space connectivity in the alpha and beta bands was noted among patients with higher survival rates, these differences did not reach statistical significance.

Figure 3 wPLI connectomes in patients stratified by survival. Each panel displays the source-space functional connectivity pattern for high-survival (>1 year) and low-survival (<1 year) groups across the five canonical EEG frequency bands (delta, theta, alpha, beta, gamma). The difference maps (right column) highlight connections that were stronger in one group compared to the other. Color coding represents connectivity strength (blue = lower, red = higher). The bottom-right panel summarizes statistically significant group differences; in this cohort, no connections reached significance after correction for multiple comparisons. EEG, electroencephalography; wPLI, weighted Phase Lag Index.

Similarly, analysis by tumor type revealed no significant differences in wPLI-based connectivity among patients with glioma, meningioma, or brain metastases (Figure 4). Notably, patients with brain metastases exhibited higher delta band connectivity compared to those with glioblastoma or meningioma, although this trend was not statistically significant.

Figure 4 wPLI connectomes in the delta band stratified by tumor type. Pairwise comparisons are shown between glioma, meningioma, and brain metastasis groups. Each panel illustrates the difference in source-space connectivity between two tumor types, with edges indicating relative connectivity strength. Color coding represents group differences (blue = lower connectivity, red = higher connectivity). Although patients with brain metastases showed a trend toward higher delta-band connectivity, no differences reached statistical significance after correction. wPLI, weighted Phase Lag Index.

Discussion

The principal aim of this study was to evaluate the potential of EEG-derived functional connectivity as a prognostic biomarker for survival following SE in patients with brain tumors. To this end, we stratified patients according to survival duration (“high survival” >1 year, “low survival” <1 year) and by tumor type (glioma, meningioma, brain metastases), and systematically compared connectivity metrics between these groups.

Statistical differences in demographic data

Statistically significant differences were observed in age between meningioma and metastasis groups, and in SE duration between survival groups. The association between prolonged SE and poorer survival is consistent with prior literature, underscoring the detrimental impact of sustained seizure activity on neurological outcomes. However, the observed age differences between tumor types did not yield an immediately apparent causal relationship in our cohort.

Functional connectivity and survival

Accurate prognostication following SE is of considerable clinical importance, informing therapeutic decision-making and patient counseling. While the current standard for outcome prediction relies heavily on underlying aetiology (28), non-invasive and cost-effective EEG-based biomarkers could offer substantial added value if proven reliable.

Previous studies have explored prognostic factors in SE (28-32)—including RSE, SRSE and non-convulsive SE (NCSE) (32-36)—with aetiology consistently emerging as the most robust predictor.

In our study, functional connectivity analysis based on the wPLI did not reveal significant differences between survival groups. We specifically chose wPLI because it is less susceptible to spurious connections caused by volume conduction and source leakage, which are major concerns in source-reconstructed EEG data. Furthermore, wPLI provides a reliable estimate of true phase synchronization even in noisy clinical recordings. Nevertheless, wPLI may not fully capture higher-order network dynamics. Complementary approaches such as graph theoretical measures (e.g., clustering coefficient, path length) or spectral coherence could provide additional insights in future investigations. Importantly, although our main results are negative, they provide valuable information for the field: feasibility studies that yield null findings help clarify methodological challenges, prevent premature conclusions, and inform the design of future prospective multicenter trials. In particular, our findings highlight the difficulty of disentangling the effects of tumor biology, treatment history, and SE itself on EEG connectivity, and stress the need for adequately powered studies integrating clinical, molecular, and electrophysiological data. Reporting such negative findings is essential to counteract publication bias and to ensure a balanced and cumulative scientific evidence base.

This likely reflects the multifactorial nature of outcomes in brain tumor patients, encompassing tumor histology, location, size, comorbidities, and molecular or genetic factors. While more aggressive tumor types, such as glioblastoma and metastases, are generally associated with poorer outcomes relative to meningioma, our findings suggest that EEG connectivity metrics alone are insufficient for individualized prognostication post-SE in this population. Future research should consider larger cohorts and focus on more narrowly defined subgroups—such as patients with HGG and SE, or those stratified by molecular tumor characteristics—to better elucidate potential prognostic signatures. The limited sample size in our study (n=37, with 1-year survival data available for 29 patients) precluded such subgroup analyses.

Importantly, future studies should investigate whether EEG connectivity metrics can be integrated with established clinical scoring systems such as STESS or EMSE to enhance prognostic accuracy.

Functional connectivity and tumor type

We further investigated whether functional connectivity differed among SE patients with glioma, meningioma, or brain metastases. Such distinctions, if present, could inform both therapeutic strategies and aetiological classification, particularly in cases lacking histological confirmation. Prior studies have demonstrated that brain tumors can significantly alter functional connectivity, although the direction and magnitude of these effects are variable (14,37-43).

Moreover, changes in connectivity have been reported across different phases of SE, including pre-ictal, ictal, and post-ictal periods.

van Dellen et al. 2012 studied functional connectivity in patients with a history of seizures and diagnosed brain lesions, including HGG, low-grade glioma (LGG) and non-glioma lesions, and healthy controls (44). Other studies, such as that by Douw 2008, also showed complex changes in connectivity in patients with brain tumors before and after tumor resection. These results were found to be robust to patient- and tumor-specific data (45).

However, to our knowledge, no previous studies have specifically examined connectivity in patients with both SE and brain tumors—a population that remains underrepresented in the literature (46).

Our analyses did not identify significant connectivity differences between tumor types in the context of SE. This finding is notable given the heterogeneity of tumor-related network alterations reported in the literature. Previous work has primarily contrasted tumor patients with healthy controls, yielding inconsistent results regarding the directionality of connectivity changes (37,38,42,43,47,48).

For example, Bartolomei et al. [2006a] reported increased delta band synchrony, whereas Bosma et al. [2009] observed both increases and decreases in interregional delta synchrony. Divergent findings have also been reported for gamma band connectivity. Importantly, our study differs in that we compared tumor subtypes within a cohort of SE patients, rather than against healthy controls, highlighting the complexity of network alterations in this dual pathology (37,47).

The inconclusive and sometimes contradictory results across studies underscore the intricate interplay between tumor biology, epileptogenicity, and network dynamics. Factors such as tumor resection status, molecular subtype [e.g., isocitrate dehydrogenase (IDH) mutation], and peri-ictal state further complicate the landscape (41,45,49). Our findings suggest that, at least in the acute setting of SE, the synchronizing effects of prolonged seizure activity may override tumor-specific connectivity signatures.

Secondary findings

Within the tumor type stratification, notable differences emerged in the duration of SE. Patients with meningioma exhibited a shorter mean SE duration (27.5 hours) compared to those with glioma (55.1 hours) or brain metastases (44.7 hours). This observation may suggest a differential response to SE treatment in meningioma patients, potentially attributable to the more circumscribed, displacing growth pattern of meningiomas, as opposed to the infiltrative nature of gliomas.

In the brain metastasis subgroup, a disproportionately high proportion of cases (36.4%) originated from malignant melanoma. Given the documented association between melanoma metastases and both intracerebral hemorrhage and seizure tendency (50,51), this finding is consistent with the known clinical behavior of melanoma brain metastases.

The survival profile among meningioma patients was also atypical. Of the six patients with available 1-year outcome data, survival was evenly split between those surviving more or less than 1 year. This contrasts with the generally favorable prognosis of non-malignant meningiomas, which typically exhibit 5-year survival rates at 88% (52). The adverse outcomes observed here may reflect selection bias toward larger, more aggressive, or unfavorably located meningiomas predisposing to SE, or may be influenced by perioperative complications, as most of these patients had undergone tumor resection prior to SE onset. However, these findings are based on small sample sizes and lack statistical significance, warranting cautious interpretation.

Limitations

Several limitations must be acknowledged. The most important limitation is the small sample size (n=37, with 1-year outcomes available for only 29 patients), which strongly limits statistical power and increases the risk of type II error. In addition, heterogeneity in both tumor biology and treatment history—including tumor entity (glioma, meningioma, metastasis), molecular markers (e.g., IDH mutation status, which was unavailable in our cohort), and meningioma grades—as well as whether patients had undergone tumor resection prior to SE and EEG recording, may have influenced connectivity patterns, as prior work has demonstrated alterations in network dynamics following tumor resection (45,53). Future studies would benefit from stratification by molecular characteristics and treatment modalities to reduce this confounding effect. The real-world clinical design of this study, while enhancing generalizability, introduces potential confounders related to treatment variability.

Moreover, both SE and brain tumors independently modulate connectivity, and their co-occurrence may produce complex effects that cannot be disentangled in this dataset. Finally, although rigorous artifact reduction protocols were applied, residual noise—particularly in the delta band—cannot be completely excluded. Future prospective multicenter studies with larger and more homogeneous cohorts are needed to validate and extend these preliminary observations.

Conclusions

In summary, this proof-of-principle study found no significant differences in EEG-derived functional connectivity between brain tumor patients with SE who survived more than 1 year and those who did not, nor between different tumor types. These findings suggest that, at present, connectivity analysis alone does not provide a reliable basis for individualized prognostic assessment in this patient population. The results may also indicate a synchronizing effect of SE on brain network dynamics, potentially overriding tumor-specific connectivity signatures.

The multifactorial nature of brain connectivity in the context of SE and brain tumors is underscored, with potential contributions from tumor type, resection status, anatomical location, comorbidities, and treatment regimens. Our secondary analyses suggest that SE duration may differ by tumor type, possibly reflecting differences in tumor biology or treatment response.

Objective prognostication in this complex clinical scenario remains a major challenge. Future studies should aim to recruit larger, well-characterized cohorts and focus on more homogeneous subgroups—such as patients with specific tumor histology, molecular profiles, or anatomical localizations—to better elucidate potential prognostic biomarkers. Such efforts are essential to advance individualized management and improve outcomes for patients with brain tumors experiencing SE.

Future research should build on this proof-of-principle study through larger multicenter prospective trials, ideally integrating EEG-based metrics with molecular tumor features and clinical prognostic scores. Machine learning approaches may also enhance prediction by combining multimodal data streams.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1739/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1739/dss

Funding: This study was supported by the Foundation P.E. Kempkes (reference number VB 1.2-5.45.08.03.01).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1739/coif). P.K. received a €3,000 educational fellowship from Deutsche Gesellschaft für Neurologie (DGfE) for a 6-week research internship at Epilog (Ghent, Belgium) to learn Python coding applications in epileptology. This grant covered travel and accommodation expenses for training purposes and had no direct relationship to the current study. L.H. received a traveling grant from Jazz Pharma and Angelini Pharma, as well as a clinical fellowship from DGfE. P.v.M. is co-founder and shareholder at Epilog, Clouds of Care NV, Ghent, Belgium. P.v.M. is co-founder and shareholder of Epilog, Clouds of Care NV (Ghent, Belgium), a company developing EEG analysis software and epilepsy-related technologies. This represents both equity interest and leadership role in an entity with potential relevance to epilepsy research. K.M. received speaker honoraria from Jazz pharmaceuticals for educational and consulting activities related to epilepsy treatment. C.N. is consultant for Brainlab and received speaker honoraria from Aesculap, BKmedical, and Brainlab. M.G. declares that she has no competing interest. A.K. received consulting fees from Phenox, Penumbra, and Stryker for medical device-related consulting services in interventional neuroradiology and neurosurgical procedures. P.S.Z. has received a lecture honorarium from Bristol-Myers Squibb. She is funded by the Mildred Scheel Career Center Frankfurt (Deutsche Krebshilfe) and by the Ministry of Higher Education, Research and the Arts of the State of Hesse (HMWK) within the LOEWE Center Frankfurt Cancer Institute (FCI). The Dr. Senckenberg Institute of Neurooncology is supported by the Dr. Senckenberg Foundation. J.P.S. has received honoraria for lectures, advisory board participation, consulting, and travel grants from Abbvie, Roche, Boehringer, Bristol-Myers Squibb, Medac, Mundipharma, and UCB. L.T. received consulting fees from Boston Scientific and speaker honoraria from Boston Scientific, AbbVie, Novartis, Neuraxpharm, Teva, Movement Disorders Society, and DIAPLAN (May 2021-May 2023). The author’s institution received funding from Boston Scientific, German Research Foundation, German Ministry of Education and Research, Otto-Loewi-Foundation, and Deutsche Parkinson Vereinigung. L.T. serves as unpaid president of the German Neurological Society. The author declares no stock holdings, patents, or financial interests in any mentioned companies or competitors for himself or family members. A.S. reports personal fees and grants from Angelini Pharma, Biocodex, Desitin Arzneimittel, Eisai, Jazz (GW) Pharmaceuticals, Marinus Pharma, Precisis, Takeda, UCB (Zogenix) Pharma, and UNEEG Medical. These relationships included consulting fees, speaker honoraria, and research support for clinical studies and educational activities in epilepsy and neurological disorders. S.K. received speaker honoraria from Arvelle Pharma, Bial, Eisai, Epilog, Desitin, Merck Serono, Precisis AG, UCB Pharma, and Zogenix for educational lectures and presentations on epilepsy and neurological disorders. These payments were for educational activities outside the submitted work and had no direct relationship to the current study. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the regional Ethics Committee of the Department of Human Medicine at Philipps University Marburg (No. 23-187 RS) and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Ascoli M, Ferlazzo E, Gasparini S, Mastroianni G, Citraro R, Roberti R, Russo E. Epidemiology and Outcomes of Status Epilepticus. Int J Gen Med 2021;14:2965-73. [Crossref] [PubMed]
2. Hauser WA, Annegers JF, Kurland LT. Incidence of epilepsy and unprovoked seizures in Rochester, Minnesota: 1935-1984. Epilepsia 1993;34:453-68. [Crossref] [PubMed]
3. Giovannini G, Pasini F, Orlandi N, Mirandola L, Meletti S. Tumor-associated status epilepticus in patients with glioma: Clinical characteristics and outcomes. Epilepsy Behav 2019;101:106370. [Crossref] [PubMed]
4. Saviuk M, Sleptsova E, Redkin T, Turubanova V. Unexplained Causes of Glioma-Associated Epilepsies: A Review of Theories and an Area for Research. Cancers (Basel) 2023;15:5539. [Crossref] [PubMed]
5. Pauletto G, Nilo A, Pez S, Zonta ME, Bagatto D, Isola M, Verriello L, Valente M, Skrap M, Ius T. Meningioma-Related Epilepsy: A Happy Ending? J Pers Med 2023;13:1124. [Crossref] [PubMed]
6. Rosenow F, Weber J. on behalf of Deutsche Gesellschaft für Neurologie (DGN) & Österreichische Gesellschaft für Neurologie (ÖGN). S2k-Leitlinie: Status Epilepticus im Erwachsenenalter. Nervenarzt 2021;92:1002-30. [Crossref] [PubMed]
7. Fox J, Ajinkya S, Greenblatt A, Houston P, Lekoubou A, Lindhorst S, Cachia D, Olar A, Kutluay E. Clinical characteristics, EEG findings and implications of status epilepticus in patients with brain metastases. J Neurol Sci 2019;407:116538. [Crossref] [PubMed]
8. Hauff NS, Storstein A. Seizure Management and Prophylaxis Considerations in Patients with Brain Tumors. Curr Oncol Rep 2023;25:787-92. [Crossref] [PubMed]
9. Roberg LE, Monsson O, Kristensen SB, Dahl SM, Ulvin LB, Heuser K, Taubøll E, Strzelczyk A, Knake S, Bechert L, Rosenow F, Beier D, Beniczky S, Krøigård T, Beier CP. Prediction of Long-term Survival After Status Epilepticus Using the ACD Score. JAMA Neurol 2022;79:604-13. [Crossref] [PubMed]
10. Sairanen JJ, Kantanen AM, Hyppölä HT, Kälviäinen RK. Outcome of status epilepticus and the predictive value of the EMSE and STESS scores: A prospective study. Seizure 2020;75:115-20. [Crossref] [PubMed]
11. Loddenkemper T. Detect, predict, and prevent acute seizures and status epilepticus. Epilepsy Behav 2023;141:109141. [Crossref] [PubMed]
12. Cao J, Zhao Y, Shan X, Wei HL, Guo Y, Chen L, Erkoyuncu JA, Sarrigiannis PG. Brain functional and effective connectivity based on electroencephalography recordings: A review. Hum Brain Mapp 2022;43:860-79. [Crossref] [PubMed]
13. Hadjiabadi DH, Pung L, Zhang J, Ward BD, Lim WT, Kalavar M, Thakor NV, Biswal BB, Pathak AP. Brain tumors disrupt the resting-state connectome. Neuroimage Clin 2018;18:279-89. [Crossref] [PubMed]
14. Ladisich B, Rampp S, Trinka E, Weisz N, Schwartz C, Kraus T, Sherif C, Marhold F, Demarchi G. Network topology in brain tumor patients with and without structural epilepsy: a prospective MEG study. Ther Adv Neurol Disord 2023;16:17562864231190298. [Crossref] [PubMed]
15. Choudhary K, DeCost B, Chen C, Jain A, Tavazza F, Cohn R, Park CW, Choudhary A, Agrawal A, Billinge SJ, Holm E, Ong SP, Wolverton C. Recent advances and applications of deep learning methods in materials science. npj Comput Mater 2022;8:1-26.
16. Nikam RM, Yue X, Kaur G, Kandula V, Khair A, Kecskemethy HH, Averill LW, Langhans SA. Advanced Neuroimaging Approaches to Pediatric Brain Tumors. Cancers (Basel) 2022;14:3401. [Crossref] [PubMed]
17. di Noia C, Grist JT, Riemer F, Lyasheva M, Fabozzi M, Castelli M, Lodi R, Tonon C, Rundo L, Zaccagna F. Predicting Survival in Patients with Brain Tumors: Current State-of-the-Art of AI Methods Applied to MRI. Diagnostics (Basel) 2022;12:2125. [Crossref] [PubMed]
18. Kanai S, Oguri M, Okanishi T, Miyamoto Y, Maeda M, Yazaki K, Matsuura R, Tozawa T, Sakuma S, Chiyonobu T, Hamano SI, Maegaki Y. Predictive modeling based on functional connectivity of interictal scalp EEG for infantile epileptic spasms syndrome. Clin Neurophysiol 2024;167:37-48. [Crossref] [PubMed]
19. Imperatori LS, Betta M, Cecchetti L, Canales-Johnson A, Ricciardi E, Siclari F, Pietrini P, Chennu S, Bernardi G. EEG functional connectivity metrics wPLI and wSMI account for distinct types of brain functional interactions. Sci Rep 2019;9:8894. [Crossref] [PubMed]
20. Ortiz E, Stingl K, Münssinger J, Braun C, Preissl H, Belardinelli P. Weighted phase lag index and graph analysis: preliminary investigation of functional connectivity during resting state in children. Comput Math Methods Med 2012;2012:186353. [Crossref] [PubMed]
21. Pegg EJ, McKavanagh A, Bracewell RM, Chen Y, Das K, Denby C, Kreilkamp BAK, Laiou P, Marson A, Mohanraj R, Taylor JR, Keller SS. Functional network topology in drug resistant and well-controlled idiopathic generalized epilepsy: a resting state functional MRI study. Brain Commun 2021;3:fcab196. [Crossref] [PubMed]
22. Pegg EJ, Taylor JR, Laiou P, Richardson M, Mohanraj R. Interictal electroencephalographic functional network topology in drug-resistant and well-controlled idiopathic generalized epilepsy. Epilepsia 2021;62:492-503. [Crossref] [PubMed]
23. Fang C, Li X, Na M, Jiang W, He Y, Wei A, Huang J, Zhou M. Epilepsy lesion localization method based on brain function network. Front Hum Neurosci 2024;18:1431153. [Crossref] [PubMed]
24. Vinck M, Oostenveld R, van Wingerden M, Battaglia F, Pennartz CM. An improved index of phase-synchronization for electrophysiological data in the presence of volume-conduction, noise and sample-size bias. Neuroimage 2011;55:1548-65. [Crossref] [PubMed]
25. The MathWorks Inc. MATLAB version: 9.5.0 (R2018b). Natick, Massachusetts: The MathWorks Inc., 2018. Available online: https://www.mathworks.com/products/matlab.html
26. Python Software Foundation. Python. Available online: https://www.python.org (Computer software).
27. The jamovi project (2025). jamovi (Version 2.6) [Computer Software]. Available online: https://www.jamovi.org
28. Sculier C, Gaínza-Lein M, Sánchez Fernández I, Loddenkemper T. Long-term outcomes of status epilepticus: A critical assessment. Epilepsia 2018;59:155-69. [Crossref] [PubMed]
29. Tuppurainen KM, Ritvanen JG, Mustonen H, Kämppi LS. Predictors of mortality at one year after generalized convulsive status epilepticus. Epilepsy Behav 2019;101:106411. [Crossref] [PubMed]
30. Sirikarn P, Pattanittum P, Sawanyawisuth K, Tiamkao S. Causes of death in patients with status epilepticus. Epilepsy Behav 2019;101:106372. [Crossref] [PubMed]
31. Leitinger M, Trinka E, Giovannini G, Zimmermann G, Florea C, Rohracher A, Kalss G, Neuray C, Kreidenhuber R, Höfler J, Kuchukhidze G, Granbichler C, Dobesberger J, Novak HF, Pilz G, Meletti S, Siebert U. Epidemiology of status epilepticus in adults: A population-based study on incidence, causes, and outcomes. Epilepsia 2019;60:53-62. [Crossref] [PubMed]
32. Giovannini G, Monti G, Polisi MM, Mirandola L, Marudi A, Pinelli G, Valzania F, Girardis M, Nichelli PF, Meletti S. A one-year prospective study of refractory status epilepticus in Modena, Italy. Epilepsy Behav 2015;49:141-5. [Crossref] [PubMed]
33. Kantanen AM, Reinikainen M, Parviainen I, Kälviäinen R. Long-term outcome of refractory status epilepticus in adults: A retrospective population-based study. Epilepsy Res 2017;133:13-21. [Crossref] [PubMed]
34. Gaspard N, Foreman BP, Alvarez V, Cabrera Kang C, Probasco JC, Jongeling AC, Meyers E, Espinera A, Haas KF, Schmitt SE, Gerard EE, Gofton T, Kaplan PW, Lee JW, Legros B, Szaflarski JP, Westover BM, LaRoche SM, Hirsch LJCritical Care EEG Monitoring Research Consortium (CCEMRC). New-onset refractory status epilepticus: Etiology, clinical features, and outcome. Neurology 2015;85:1604-13. [Crossref] [PubMed]
35. Kim HJ, Lee SA, Kim HW, Kim SJ, Jeon SB, Koo YS. The timelines of MRI findings related to outcomes in adult patients with new-onset refractory status epilepticus. Epilepsia 2020;61:1735-48. [Crossref] [PubMed]
36. Madžar D, Knappe RU, Reindl C, Giede-Jeppe A, Sprügel MI, Beuscher V, Gollwitzer S, Hamer HM, Huttner HB. Factors associated with occurrence and outcome of super-refractory status epilepticus. Seizure 2017;52:53-9. [Crossref] [PubMed]
37. Bosma I, Reijneveld JC, Klein M, Douw L, van Dijk BW, Heimans JJ, Stam CJ. Disturbed functional brain networks and neurocognitive function in low-grade glioma patients: a graph theoretical analysis of resting-state MEG. Nonlinear Biomed Phys 2009;3:9. [Crossref] [PubMed]
38. Bartolomei F, Bosma I, Klein M, Baayen JC, Reijneveld JC, Postma TJ, Heimans JJ, van Dijk BW, de Munck JC, de Jongh A, Cover KS, Stam CJ. How do brain tumors alter functional connectivity? A magnetoencephalography study. Ann Neurol 2006;59:128-38. [Crossref] [PubMed]
39. D'Souza S, Hirt L, Ormond DR, Thompson JA. Retrospective analysis of hemispheric structural network change as a function of location and size of glioma. Brain Commun 2021;3:fcaa216. [Crossref] [PubMed]
40. Semmel ES, Quadri TR, King TZ. Correction to: Graph Theoretical Analysis of Brain Network Characteristics in Brain Tumor Patients: A Systematic Review. Neuropsychol Rev 2022;32:676. [Crossref] [PubMed]
41. Derks J, Kulik S, Wesseling P, Numan T, Hillebrand A, van Dellen E, de Witt Hamer PC, Geurts JJG, Reijneveld JC, Stam CJ, Klein M, Douw L. Understanding cognitive functioning in glioma patients: The relevance of IDH-mutation status and functional connectivity. Brain Behav 2019;9:e01204. [Crossref] [PubMed]
42. Metwali H, Raemaekers M, Ibrahim T, Samii A. Inter-Network Functional Connectivity Changes in Patients With Brain Tumors: A Resting-State Functional Magnetic Resonance Imaging Study. World Neurosurg 2020;138:e66-71. [Crossref] [PubMed]
43. Manglore S, Bharath RD, Panda R, George L, Thamodharan A, Gupta AK. Utility of resting fMRI and connectivity in patients with brain tumor. Neurol India 2013;61:144-51. [Crossref] [PubMed]
44. van Dellen E, Douw L, Hillebrand A, Ris-Hilgersom IH, Schoonheim MM, Baayen JC, De Witt Hamer PC, Velis DN, Klein M, Heimans JJ, Stam CJ, Reijneveld JC. MEG network differences between low- and high-grade glioma related to epilepsy and cognition. PLoS One 2012;7:e50122. [Crossref] [PubMed]
45. Douw L, Baayen H, Bosma I, Klein M, Vandertop P, Heimans J, Stam K, de Munck J, Reijneveld J. Treatment-related changes in functional connectivity in brain tumor patients: a magnetoencephalography study. Exp Neurol 2008;212:285-90. [Crossref] [PubMed]
46. Rickel JK, Zeeb D, Knake S, Urban H, Konczalla J, Weber KJ, et al. Status epilepticus in patients with brain tumors and metastases: A multicenter cohort study of 208 patients and literature review. Neurol Res Pract 2024;6:19. [Crossref] [PubMed]
47. Bartolomei F, Bosma I, Klein M, Baayen JC, Reijneveld JC, Postma TJ, Heimans JJ, van Dijk BW, de Munck JC, de Jongh A, Cover KS, Stam CJ. Disturbed functional connectivity in brain tumour patients: evaluation by graph analysis of synchronization matrices. Clin Neurophysiol 2006;117:2039-49. [Crossref] [PubMed]
48. Zimmermann MLM, Breedt LC, Centeno EGZ, Reijneveld JC, Santos FAN, Stam CJ, van Lingen MR, Schoonheim MM, Hillebrand A, Douw L. The relationship between pathological brain activity and functional network connectivity in glioma patients. J Neurooncol 2024;166:523-33. [Crossref] [PubMed]
49. Kesler SR, Noll K, Cahill DP, Rao G, Wefel JS. The effect of IDH1 mutation on the structural connectome in malignant astrocytoma. J Neurooncol 2017;131:565-74. [Crossref] [PubMed]
50. Lamba N, Catalano PJ, Cagney DN, Haas-Kogan DA, Bubrick EJ, Wen PY, Aizer AA. Seizures Among Patients With Brain Metastases: A Population- and Institutional-Level Analysis. Neurology 2021;96:e1237-50. [Crossref] [PubMed]
51. Zoga E, Wolff R, Ackermann H, Meissner M, Rödel C, Tselis N, Chatzikonstantinou G. Factors Associated with Hemorrhage of Melanoma Brain Metastases after Stereotactic Radiosurgery in the Era of Targeted/Immune Checkpoint Inhibitor Therapies. Cancers (Basel) 2022;14:2391. [Crossref] [PubMed]
52. Ostrom QT, Cioffi G, Gittleman H, Patil N, Waite K, Kruchko C, Barnholtz-Sloan JS. CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2012-2016. Neuro Oncol 2019;21:v1-v100. [Crossref] [PubMed]
53. Douw L, van Dellen E, de Groot M, Heimans JJ, Klein M, Stam CJ, Reijneveld JC. Epilepsy is related to theta band brain connectivity and network topology in brain tumor patients. BMC Neurosci 2010;11:103. [Crossref] [PubMed]