Objective.Accurate localization of the epileptogenic zone (EZ) is crucial for epilepsy surgery, but the class imbalance of epileptogenic vs. non-epileptogenic electrode contacts in intracranial electroencephalography (iEEG) data poses significant challenges for automatic localization methods. This review evaluates methodologies for handling the class imbalance in EZ localization studies that use machine learning (ML).Approach.We systematically reviewed studies employing ML to localize the EZ from iEEG data, focusing on strategies for addressing class imbalance in data handling, algorithm design, and evaluation.Results.Out of 2,128 screened studies, 35 fulfilled the inclusion criteria. Across the studies, the iEEG contacts annotated as epileptogenic prior to automatic localization constituted a median of 18.34% of all contacts. However, many of these studies did not adequately address the class imbalance problem. Techniques such as data resampling and cost-sensitive learning were used to mitigate the class imbalance problem, but the chosen evaluation metrics often failed to account for it.Significance.Class imbalance significantly impacts the reliability of EZ localization models. More comprehensive management and innovative approaches are needed to enhance the robustness and clinical utility of these models. Addressing class imbalance in ML models for EZ localization will improve both the predictive performance and reliability of these models.
In drug-resistant focal epilepsy, planning surgical resection can involve presurgical intracranial EEG (iEEG) recordings to detect seizures and other iEEG patterns to improve postsurgical seizure outcome. We hypothesized that resection of tissue generating interictal high-frequency oscillations (HFOs, 80-500 Hz) in the iEEG predicts surgical outcome. In eight international epilepsy centres, iEEG was recorded during the presurgical evaluation of patients. The patients were of all ages, had epilepsy of all types, and underwent surgical resection of a single focus aiming at seizure freedom. In a prospective analysis, we applied a fully automated definition of HFO that was independent of the dataset. Using an observational cohort design that was blinded to postsurgical seizure outcome, we analysed HFO rates during non-rapid-eye-movement sleep. If channels had consistently high rates over multiple epochs, they were labelled the 'HFO area'. After HFO analysis, centres provided the electrode contacts located in the resected volume and the seizure outcome at follow-up ≥24 months after surgery. The study was registered at www.clinicaltrials.gov (NCT05332990). We received 160 iEEG datasets. In 146 datasets (91%), the HFO area could be defined. The patients with a completely resected HFO area were more likely to achieve seizure freedom in comparison to those without [odds ratio 2.61, 95% confidence interval (CI) 1.15-5.91, P = 0.02]. Among seizure-free patients, the HFO area was completely resected in 31 and not completely resected in 43. Among patients with recurrent seizures, the HFO area was completely resected in 14 and not completely resected in 58. When predicting seizure freedom, the negative predictive value of the HFO area (68%, CI 52-81) was higher than that for the resected volume as a predictor by itself (51%, CI 42-59, P = 4 × 10-5). The sensitivity and specificity for complete HFO area resection were 0.88 (CI 0.72-0.98) and 0.39 (CI 0.25-0.54), respectively, and the area under the curve was 0.83 (CI 0.58-0.97), indicating good predictive performance. In a blinded cohort study from independent epilepsy centres, applying a previously validated algorithm for HFO marking without the need for adjusting to new datasets allowed us to validate the clinical relevance of HFOs to plan the surgical resection.
- MeSH
- Child MeSH
- Adult MeSH
- Electroencephalography methods MeSH
- Electrocorticography * methods MeSH
- Middle Aged MeSH
- Humans MeSH
- Adolescent MeSH
- Young Adult MeSH
- Prospective Studies MeSH
- Drug Resistant Epilepsy * surgery physiopathology MeSH
- Treatment Outcome MeSH
- Seizures * surgery physiopathology diagnosis MeSH
- Check Tag
- Child MeSH
- Adult MeSH
- Middle Aged MeSH
- Humans MeSH
- Adolescent MeSH
- Young Adult MeSH
- Male MeSH
- Female MeSH
- Publication type
- Journal Article MeSH
- Multicenter Study MeSH
- Observational Study MeSH
Precise localization of the epileptogenic zone is pivotal for planning minimally invasive surgeries in drug-resistant epilepsy. Here, we present a graph neural network (GNN) framework that integrates interictal intracranial EEG features, electrode topology, and MRI features to automate epilepsy surgery planning. We retrospectively evaluated the model using leave-one-patient-out cross-validation on a dataset of 80 drug-resistant epilepsy patients treated at St. Anne's University Hospital (Brno, Czech Republic), comprising 31 patients with good postsurgical outcomes (Engel I) and 49 with poor outcomes (Engel II-IV). The GNN predictions demonstrated a significantly better (P < 0.05, Mann-Whitney-U test) area under the precision-recall curve in patients with good outcomes (area under the precision-recall curve: 0.69) compared with those with poor outcomes (area under the precision-recall curve: 0.33), indicating that the model captures clinically relevant targets in successful cases. In patients with poor outcomes, the graph neural network proposed alternative intervention sites that diverged from the original clinical plans, highlighting its potential to identify alternative therapeutic targets. We show that topology-aware GNNs significantly outperformed (P < 0.05, Wilcoxon signed-rank test) traditional neural networks while using the same intracranial EEG features, emphasizing the importance of incorporating implantation topology into predictive models. These findings uncover the potential of GNNs to automatically suggest targets for epilepsy surgery, which can assist the clinical team during the planning process.
- Publication type
- Journal Article MeSH
INTRODUCTION: Precise localization of the epileptogenic zone is critical for successful epilepsy surgery. However, imbalanced datasets in terms of epileptic vs. normal electrode contacts and a lack of standardized evaluation guidelines hinder the consistent evaluation of automatic machine learning localization models. METHODS: This study addresses these challenges by analyzing class imbalance in clinical datasets and evaluating common assessment metrics. Data from 139 drug-resistant epilepsy patients across two Institutions were analyzed. Metric behaviors were examined using clinical and simulated data. RESULTS: Complementary use of Area Under the Receiver Operating Characteristic (AUROC) and Area Under the Precision-Recall Curve (AUPRC) provides an optimal evaluation approach. This must be paired with an analysis of class imbalance and its impact due to significant variations found in clinical datasets. CONCLUSIONS: The proposed framework offers a comprehensive and reliable method for evaluating machine learning models in epileptogenic zone localization, improving their precision and clinical relevance. SIGNIFICANCE: Adopting this framework will improve the comparability and multicenter testing of machine learning models in epileptogenic zone localization, enhancing their reliability and ultimately leading to better surgical outcomes for epilepsy patients.
- MeSH
- Adult MeSH
- Electrocorticography methods standards MeSH
- Middle Aged MeSH
- Humans MeSH
- Adolescent MeSH
- Young Adult MeSH
- Drug Resistant Epilepsy * surgery physiopathology MeSH
- Machine Learning * MeSH
- Check Tag
- Adult MeSH
- Middle Aged MeSH
- Humans MeSH
- Adolescent MeSH
- Young Adult MeSH
- Male MeSH
- Female MeSH
- Publication type
- Journal Article MeSH
Objective.The proportion of patients becoming seizure-free after epilepsy surgery has stagnated. Large multi-center stereo-electroencephalography (SEEG) datasets can allow comparing new patients to past similar cases and making clinical decisions with the knowledge of how cases were treated in the past. However, the complexity of these evaluations makes the manual search for similar patients impractical. We aim to develop an automated system that electrographically and anatomically matches seizures to those in a database. Additionally, since features that define seizure similarity are unknown, we evaluate the agreement and features among experts in classifying similarity.Approach.We utilized 320 SEEG seizures from 95 consecutive patients who underwent epilepsy surgery. Eight international experts evaluated seizure-pair similarity using a four-level similarity score. As our primary outcome, we developed and validated an automated seizure matching system by employing patient data marked by independent experts. Secondary outcomes included the inter-rater agreement (IRA) and features for classifying seizure similarity.Main results.The seizure matching system achieved a median area-under-the-curve of 0.76 (interquartile range, 0.1), indicating its feasibility. Six distinct seizure similarity features were identified and proved effective: onset region, onset pattern, propagation region, duration, extent of spread, and propagation speed. Among these features, the onset region showed the strongest correlation with expert scores (Spearman's rho = 0.75,p< 0.001). Additionally, the moderate IRA confirmed the practicality of our approach with an agreement of 73.9% (7%), and Gwet's kappa of 0.45 (0.16). Further, the interoperability of the system was validated on seizures from five centers.Significance.We demonstrated the feasibility and validity of a SEEG seizure matching system across patients, effectively mirroring the expertise of epileptologists. This novel system can identify patients with seizures similar to that of a patient being evaluated, thus optimizing the treatment plan by considering the results of treating similar patients in the past, potentially improving surgery outcome.
- MeSH
- Child MeSH
- Adult MeSH
- Electroencephalography * methods MeSH
- Epilepsy * surgery diagnosis physiopathology MeSH
- Clinical Decision-Making * methods MeSH
- Middle Aged MeSH
- Humans MeSH
- Adolescent MeSH
- Young Adult MeSH
- Reproducibility of Results MeSH
- Stereotaxic Techniques MeSH
- Seizures diagnosis surgery physiopathology MeSH
- Check Tag
- Child MeSH
- Adult MeSH
- Middle Aged MeSH
- Humans MeSH
- Adolescent MeSH
- Young Adult MeSH
- Male MeSH
- Female MeSH
- Publication type
- Journal Article MeSH
- Research Support, Non-U.S. Gov't MeSH
OBJECTIVE: Evidence suggests that the most promising results in interictal localization of the epileptogenic zone (EZ) are achieved by a combination of multiple stereo-electroencephalography (SEEG) biomarkers in machine learning models. These biomarkers usually include SEEG features calculated in standard frequency bands, but also high-frequency (HF) bands. Unfortunately, HF features require extra effort to record, store, and process. Here we investigate the added value of these HF features for EZ localization and postsurgical outcome prediction. METHODS: In 50 patients we analyzed 30 min of SEEG recorded during non-rapid eye movement sleep and tested a logistic regression model with three different sets of features. The first model used broadband features (1-500 Hz); the second model used low-frequency features up to 45 Hz; and the third model used HF features above 65 Hz. The EZ localization by each model was evaluated by various metrics including the area under the precision-recall curve (AUPRC) and the positive predictive value (PPV). The differences between the models were tested by the Wilcoxon signed-rank tests and Cliff's Delta effect size. The differences in outcome predictions based on PPV values were further tested by the McNemar test. RESULTS: The AUPRC score of the random chance classifier was .098. The models (broad-band, low-frequency, high-frequency) achieved median AUPRCs of .608, .582, and .522, respectively, and correctly predicted outcomes in 38, 38, and 33 patients. There were no statistically significant differences in AUPRC or any other metric between the three models. Adding HF features to the model did not have any additional contribution. SIGNIFICANCE: Low-frequency features are sufficient for correct localization of the EZ and outcome prediction with no additional value when considering HF features. This finding allows significant simplification of the feature calculation process and opens the possibility of using these models in SEEG recordings with lower sampling rates, as commonly performed in clinical routines.
- MeSH
- Child MeSH
- Adult MeSH
- Electroencephalography * methods MeSH
- Epilepsy surgery physiopathology diagnosis MeSH
- Middle Aged MeSH
- Humans MeSH
- Adolescent MeSH
- Young Adult MeSH
- Drug Resistant Epilepsy surgery physiopathology diagnosis MeSH
- Stereotaxic Techniques MeSH
- Treatment Outcome MeSH
- Check Tag
- Child MeSH
- Adult MeSH
- Middle Aged MeSH
- Humans MeSH
- Adolescent MeSH
- Young Adult MeSH
- Male MeSH
- Female MeSH
- Publication type
- Journal Article MeSH
INTRODUCTION: Epilepsy surgery is the only curative treatment for patients with drug-resistant focal epilepsy. Stereoelectroencephalography (SEEG) is the gold standard to delineate the seizure-onset zone (SOZ). However, up to 40% of patients are subsequently not operated as no focal non-eloquent SOZ can be identified. The 5-SENSE Score is a 5-point score to predict whether a focal SOZ is likely to be identified by SEEG. This study aims to validate the 5-SENSE Score, improve score performance by incorporating auxiliary diagnostic methods and evaluate its concordance with expert decisions. METHODS AND ANALYSIS: Non-interventional, observational, multicentre, prospective study including 200 patients with drug-resistant epilepsy aged ≥15 years undergoing SEEG for identification of a focal SOZ and 200 controls at 22 epilepsy surgery centres worldwide. The primary objective is to assess the diagnostic accuracy and generalisability of the 5-SENSE in predicting focality in SEEG in a prospective cohort. Secondary objectives are to optimise score performance by incorporating auxiliary diagnostic methods and to analyse concordance of the 5-SENSE Score with the expert decisions made in the multidisciplinary team discussion. ETHICS AND DISSEMINATION: Prospective multicentre validation of the 5-SENSE score may lead to its implementation into clinical practice to assist clinicians in the difficult decision of whether to proceed with implantation. This study will be conducted in accordance with the Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans (2014). We plan to publish the study results in a peer-reviewed full-length original article and present its findings at scientific conferences. TRIAL REGISTRATION NUMBER: NCT06138808.
- Publication type
- Journal Article MeSH
Well-documented sleep datasets from healthy adults are important for sleep pattern analysis and comparison with a wide range of neuropsychiatric disorders. Currently, available sleep datasets from healthy adults are acquired using low-density arrays with a minimum of four electrodes in a typical sleep montage. The low spatial resolution is thus prohibitive for the analysis of the spatial structure of sleep. Here we introduce an open-access sleep dataset from 29 healthy adults (13 female, aged 32.17 ± 6.30 years) acquired at the Montreal Neurological Institute. The dataset includes overnight polysomnograms with high-density scalp electroencephalograms incorporating 83 electrodes, electrocardiogram, electromyogram, electrooculogram, and an average of electrode positions using manual co-registrations and sleep scoring annotations. Data characteristics and group-level analysis of sleep properties were assessed. The database can be accessed through ( https://doi.org/10.17605/OSF.IO/R26FH ). This is the first high-density electroencephalogram open sleep database from healthy adults, allowing researchers to investigate sleep physiology at high spatial resolution. We expect that this database will serve as a valuable resource for studying sleep physiology and for benchmarking sleep pathology.
OBJECTIVE: This study was undertaken to develop a standardized grading system based on expert consensus for evaluating the level of confidence in the localization of the epileptogenic zone (EZ) as reported in published studies, to harmonize and facilitate systematic reviews in the field of epilepsy surgery. METHODS: We conducted a Delphi study involving 22 experts from 18 countries, who were asked to rate their level of confidence in the localization of the EZ for various theoretical clinical scenarios, using different scales. Information provided in these scenarios included one or several of the following data: magnetic resonance imaging (MRI) findings, invasive electroencephalography summary, and postoperative seizure outcome. RESULTS: The first explorative phase showed an overall interrater agreement of .347, pointing to large heterogeneity among experts' assessments, with only 17% of the 42 proposed scenarios associated with a substantial level of agreement. A majority showed preferences for the simpler scale and single-item scenarios. The successive Delphi voting phases resulted in a majority consensus across experts, with more than two thirds of respondents agreeing on the rating of each of the tested single-item scenarios. High or very high levels of confidence were ascribed to patients with either an Engel class I or class IA postoperative seizure outcome, a well-delineated EZ according to all available invasive EEG (iEEG) data, or a well-delineated focal epileptogenic lesion on MRI. MRI signs of hippocampal sclerosis or atrophy were associated with a moderate level of confidence, whereas a low level was ascribed to other MRI findings, a poorly delineated EZ according to iEEG data, or an Engel class II-IV postoperative seizure outcome. SIGNIFICANCE: The proposed grading system, based on an expert consensus, provides a simple framework to rate the level of confidence in the EZ reported in published studies in a structured and harmonized way, offering an opportunity to facilitate and increase the quality of systematic reviews and guidelines in the field of epilepsy surgery.
OBJECTIVE: Interictal biomarkers of the epileptogenic zone (EZ) and their use in machine learning models open promising avenues for improvement of epilepsy surgery evaluation. Currently, most studies restrict their analysis to short segments of intracranial EEG (iEEG). METHODS: We used 2381 hours of iEEG data from 25 patients to systematically select 5-minute segments across various interictal conditions. Then, we tested machine learning models for EZ localization using iEEG features calculated within these individual segments or across them and evaluated the performance by the area under the precision-recall curve (PRAUC). RESULTS: On average, models achieved a score of 0.421 (the result of the chance classifier was 0.062). However, the PRAUC varied significantly across the segments (0.323-0.493). Overall, NREM sleep achieved the highest scores, with the best results of 0.493 in N2. When using data from all segments, the model performed significantly better than single segments, except NREM sleep segments. CONCLUSIONS: The model based on a short segment of iEEG recording can achieve similar results as a model based on prolonged recordings. The analyzed segment should, however, be carefully and systematically selected, preferably from NREM sleep. SIGNIFICANCE: Random selection of short iEEG segments may give rise to inaccurate localization of the EZ.
- MeSH
- Time Factors MeSH
- Adult MeSH
- Electroencephalography * methods MeSH
- Electrocorticography methods standards MeSH
- Epilepsy * physiopathology diagnosis MeSH
- Middle Aged MeSH
- Humans MeSH
- Adolescent MeSH
- Young Adult MeSH
- Brain physiopathology MeSH
- Sleep Stages physiology MeSH
- Machine Learning * MeSH
- Check Tag
- Adult MeSH
- Middle Aged MeSH
- Humans MeSH
- Adolescent MeSH
- Young Adult MeSH
- Male MeSH
- Female MeSH
- Publication type
- Journal Article MeSH
- Research Support, Non-U.S. Gov't MeSH
