Stereo-electroencephalography (SEEG) is the gold standard to delineate surgical targets in focal drug-resistant epilepsy. SEEG uses electrodes placed directly into the brain to identify the seizure-onset zone (SOZ). However, its major constraint is limited brain coverage, potentially leading to misidentification of the ‘true’ SOZ. Here, we propose a framework to assess adequate SEEG sampling by coupling epileptic biomarkers with their spatial distribution and measuring the system’s response to a perturbation of this coupling. We demonstrate that the system’s response is strongest in well-sampled patients when virtually removing the measured SOZ. We then introduce the spatial perturbation map, a tool that enables qualitative assessment of the implantation coverage. Probability modelling reveals a higher likelihood of well-implanted SOZs in seizure-free patients or non-seizure free patients with incomplete SOZ resections, compared to non-seizure-free patients with complete resections. This highlights the framework’s value in sparing patients from unsuccessful surgeries resulting from poor SEEG coverage.
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 temporal lobe epilepsy (DR-TLE), some seizures are clinically symptomatic while others remain electrographic despite arising from the same seizure-onset zone (SOZ), suggesting that clinical expression reflects network dynamics beyond seizure origin. We examined whether thalamo-cortical network engagement differs between clinical (CS) and non-clinical seizures (NCS). We analyzed 286 seizures from 62 DR-TLE patients recorded using stereoelectroencephalography with pulvinar and/or anterior thalamic group coverage. Thalamo-cortical synchrony during the first 20 seconds of seizure onset was quantified as correlation between SEEG-derived time-frequency patterns in thalamic nuclei and the cortical SOZ. Synchrony was significantly stronger during CS than NCS (p < 0.0001, Cliff's δ > 0.6), including within-subject analyses, and was independent of other clinical variables, including postsurgical outcome. Linear mixed-effects modeling identified seizure type as the sole independent predictor of synchrony (p = 0.0069). These findings identify thalamo–cortical synchrony as a network-level feature associated with clinical seizure expression in DR-TLE.
Cortical oscillations across sleep-wake cycles are essential for coordinating functional brain dynamics. High-density electroencephalography (HDEEG) combined with electrical source imaging (ESI) provides a noninvasive approach to map cortical dynamics; however, its ability to capture spatial ongoing oscillations across different vigilance states remains uncertain. Here, we directly compared HDEEG source imaging by comparing it to a normative intracranial EEG (iEEG) atlas from 110 epilepsy patients with electrodes in healthy brain regions (https://mni-open-ieegatlas.research.mcgill.ca/). Wavelet-based Maximum Entropy on the Mean (wMEM) was applied to localize oscillatory patterns using overnight HDEEG recordings from 35 healthy adults (16 females, mean age 31.1 ± 6.3 years). Virtual iEEG (ViEEG) signals were estimated by applying an iEEG forward model to wMEM sources to examine oscillatory patterns across 5 frequency bands, 38 regions, and 4 vigilance states. We found that HDEEG source imaging exhibited comparable spectral patterns of iEEG in low frequencies but overestimated oscillatory activities at high frequencies. Lateral cortical regions exhibited more accurate source estimation than medial regions (p < 0.05). After removing the aperiodic components, the spectral alignment between ViEEG and iEEG significantly improved except for N3 sleep (p < 0.05). Oscillatory peak patterns in ViEEG reflect state-dependent dynamics that are broadly consistent with iEEG peaks (p < 0.05). HDEEG-derived ViEEG and magnetoencephalography-derived ViEEG approximated iEEG spectral features, showing complementary correspondence. These findings reveal that vigilance states significantly shape cortical oscillations by altering their spectral and spatial profiles. Our results establish HDEEG as a powerful tool for large-scale, noninvasive investigations of human sleep neurophysiology and brain network dynamics.
The thalamus coordinates sleep-dependent brain function including memory consolidation, sensory gating, and network stability through spindle oscillations. Thalamocortical circuits are also affected in many neurological and neuropsychiatric disorders. Yet how human thalamocortical circuits respond to pathological disruptions remain poorly understood. Here, we leveraged interictal epileptiform discharges (IEDs) as spontaneous perturbations to probe spindle-generating circuits and thalamocortical dynamics in vivo. Using multi-night intracranial recordings spanning multiple thalamic nuclei and cortical regions in 55 individuals with epilepsy, we investigated interactions between sleep spindles and epileptic activity across timescales from milliseconds to days. Sleep spindles were associated with IED suppression, whereas IEDs increased subsequent spindle probability, revealing bidirectional interactions between physiological and pathological activity. Nights with lower IEDs corresponded to increased spindle occurrence, longer duration, and faster frequency, reflecting a more stable thalamocortical state. These findings propose sleep spindles as potential regulators of thalamocortical networks that actively respond to and regulate pathological activity.