Hrtonová, Valentina

Abstract Objective Interictal epileptiform spikes (IESs) are widely used biomarkers of epileptogenic tissue in presurgical epilepsy evaluation. However, their reliability for seizure onset zone (SOZ) localization in high‐density intracranial electroencephalography (iEEG) remains inconsistent. It is underexplored whether this variability reflects methodological differences across detection algorithms or a limitation of IESs as a biomarker. This study evaluates whether IES detection alone provides sufficient spatial specificity for accurate SOZ localization in iEEG. Methods We systematically benchmarked eight established IES detection algorithms using iEEG recordings from 35 patients with drug‐resistant focal epilepsy. To provide a clinically meaningful reference, algorithm performance was evaluated against the clinically defined SOZ. To refine detection outputs and move beyond only IES detection, we applied masking strategies to isolate informative IES subsets based on rate, amplitude, frequency, and co‐occurrence with high‐frequency oscillations (HFOs). Conceptually, masking acts as a spatial refinement strategy that progressively restricts diffuse IES activity to more focal, pathologically relevant event populations. Results Despite substantial variability in detection characteristics, SOZ localization performance (ratio of events originating from SOZ divided by total number of events) converged within a narrow range (.23–.51), suggesting that IESs exhibit an intrinsic ceiling in SOZ localization performance when used alone. Amplitude‐based masking yielded modest improvements, whereas frequency‐based masking reduced performance. In contrast, IES‐HFO co‐occurrence markedly improved localization, particularly when combined with rate‐based masking, achieving a median SOZ localization ratio of .89, exceeding IESs (.37) and HFOs (.63) evaluated independently. Significance These findings demonstrate that IES‐based SOZ localization may require moving beyond detection‐only approaches. The integration of informed, biomarker‐driven masking strategies provides a more robust framework for identifying the activity most relevant to the epileptogenic network and paves the way for future incorporation into the clinical decision‐making pipeline.

Abstract 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.