Neuromodulation is approved for the treatment of drug-resistant epilepsy. It has been increasingly utilized over the past two decades with the approval of deep brain stimulation (DBS) and responsive neurostimulation (RNS) in addition to vagus nerve stimulation (VNS)-particularly in patients who are not deemed to be good resective surgical candidates or in case of patient preference. In this seminar in epileptology article, we address learning objective 3.7 "Demonstrate knowledge of indications, limitations and risks for vagal nerve stimulation and other neuromodulation techniques" of the International League Against Epilepsy curriculum for epileptologists. We will review the indications for neuromodulation, device programming, surgical considerations for implantation, side effects, effects on sleep, and mood. This is complemented by case examples that provide perspective for trainees, along with an outlook on future directions for neuromodulation in epilepsy. This article aims to serve as a learning resource for trainees in neurology, clinical neurophysiology, and epileptology by providing guidance in how to navigate the landscape of neuromodulation in epilepsy.

Numerous studies across species emphasize the importance of theta oscillations within medial temporal lobe (MTL) regions, such as the hippocampus, in relation to memory. In rodents, physical movement strongly influences theta activity, while this relationship remains more ambiguous in primates. This disparity could stem from the increased reliance on visual search in primates during navigation. To explore this, we analyzed intracranial electroencephalographic (iEEG) activity from the human MTL recorded simultaneously with body and eye movements during ambulatory navigation. We found that MTL theta power was significantly higher during periods when saccadic eye movements were taking place, and this effect was observed only during periods with overt memory demands. The largest increases occurred during saccades with more variable and exploratory gaze patterns, on trials with better memory performance, and during the early planning period of each route. The modulation was also amplified near environmental boundaries, spatial features known to anchor memory representations and guide navigation. During memory-guided navigation, theta power further tended to increase during both locomotion and stationary periods, consistent with broad engagement during active information gathering. In addition to these memory-specific effects, theta aligned its phase to saccade onset during both memory-guided and visually-guided navigation, suggesting that eye movements impose a consistent temporal structure on ongoing MTL activity. Together, these findings reveal that memory-related theta dynamics in the human MTL are tightly coupled to exploratory visual search and prospective planning during memory-guided navigation, revealing a mechanism by which saccades may help organize mnemonic computations in naturalistic settings.

Recurrent seizures, the hallmark of epilepsy, are influenced by rhythms operating over multiple timescales. Chronobiology is the study of biological timing that aims to explain temporal patterns of events like seizures. Fueled by recent advances in genetics, computational modeling, and device engineering, the chronobiology of epilepsy is now a burgeoning field poised to shed new light on mechanisms governing seizure recurrence. Although seizures were long believed to occur at random, epilepsy is now understood as a cyclical disorder, and time-varying therapeutic interventions are increasingly possible. Yet, potential barriers to progress in this field exist, such as reconciling variable experimental methodology, deconvolving coexisting rhythms, and harnessing the power of new technologies and data sharing. In this report from the International League Against Epilepsy Task Force on Chronobiology, we review these knowledge gaps and offer recommendations to help close them. By unraveling mechanisms of seizure timing, chronobiology promises to usher in a new era of personalized epilepsy management in which seizures are viewed as predictable, potentially avoidable events.

BACKGROUND AND OBJECTIVES:Neuromodulation therapies are approved for the treatment of focal epilepsy based on data from randomized controlled trials (RCTs). After approval of a responsive direct brain stimulation device (The RNS System for focal epilepsy), the Food and Drug Administration required a prospective study to evaluate whether real-world safety and effectiveness differed from outcomes in the RCT. METHODS:This open-labeled study enrolled adult participants who met the RNS System-approved indication for use. The primary effectiveness end point was median percent change in seizure frequency at 3 years of treatment. Interim safety is presented; the primary safety endpoint analysis will be conducted at 5 years. RESULTS:Across 32 US epilepsy centers, 324 patients (mean age 37.1, 59.6% female individuals) were implanted and 271 completed 3 years of follow-up. The median percent reduction in seizure frequency at 6 months was 62% and 82% at 3 years (p < 0.0001; Wilcoxon signed-rank test); 41% had a ≥90% reduction in seizure frequency at 3 years, 42.5% of participants had at least 1 seizure-free period of 6 months or more, and 22.0% experienced seizure freedom for 12 months or more. Observed effectiveness was similar across patients with 1 or 2 seizure onsets and across onset locations (mesial temporal, neocortical, or both mesial temporal and neocortical). No serious stimulation-related adverse events were reported. Combining data from all RNS System trials (n = 645), the sudden unexplained death in epilepsy (SUDEP) rate was 2.3/1,000 patient years, which was significantly lower than predefined comparators (p < 0.05; 1-tailed χ2). DISCUSSION:This prospective real-world study contributes to the body of evidence that adjunctive direct brain-responsive neurostimulation provides significant and sustained reductions in the frequency of focal seizures. Seizure reductions were greater and were achieved faster than in the RCT and long-term treatment trials but were similar to a more recent retrospective multicenter real-world study. As in the preapproval studies, treatment was well-tolerated and safe, and the SUDEP rate was low. The RNS System showed similar safety and improved seizure outcomes in real-world use compared with the RCT. Improvements in efficacy may reflect changes in programming practices. Future research efforts will focus on using the brain data obtained by the device to optimize detection and stimulation paradigms for each patient. TRIAL REGISTRATION INFORMATION:ClinicalTrials.gov, NCT02403843, submitted March 26, 2015. CLASSIFICATION OF EVIDENCE:This study provides Class IV evidence that in adults with refractory focal-onset seizures, direct brain-responsive neurostimulation reduces seizure frequency without serious adverse events up to 3 years.

Objective:Generating reference standards to train and evaluate the accuracy of artificial intelligence (AI) algorithms poses a significant challenge. Particularly when interpreting complex signals like electroencephalography (EEG), where interrater variability is considerable. We aimed to characterize the impact of interrater variability when evaluating the performance of an AI algorithm for detecting electrographic status epilepticus in point-of-care (POC) limited-montage EEG. Methods:We analyzed 604 EEGs collected using a POC EEG system (Ceribell Inc.). Each EEG was independently reviewed by 5-7 blinded experts, who annotated seizures and completed standardized assessments. The EEGs were later analyzed by the Clarity AI algorithm (version 7), trained on separate EEG dataset. Interrater agreement was assessed using Gwet's AC1. Sensitivity and specificity for detecting electrographic status epilepticus (ESE) were estimated for the reviewers and the AI algorithm. Multiple evaluation schemes were employed, including simple majority consensus of the full group (group majority), simple majority using leave-one-out analysis, and 2-of-3 majority with bootstrap sampling. Results:Of 604 POC EEGs, the group majority identified 8 cases (1.3%) meeting ACNS criteria for ESE and 14 (2.3%) with seizures, the rest were classified as normal/slowing (80.1%), highly epileptiform patterns (6.6%), or other findings (4.3%). Twentynine cases lacked consensus interpretation. Interrater agreement among reviewers was 0.67-0.68. Compared to the group majority, the AI algorithm showed higher sensitivity (median 100%) with lower specificity (93.5%) than individual reviewers (median sensitivity 60%, specificity 98.7%). Using all possible 3-reviewer combinations to define majority agreement, the number of EEGs identified as ESE varied widely (4-18 cases, median = 10). The AI algorithm consistently achieved significantly higher sensitivity than external human reviewers (71.4% vs. 50%, p < 0.001). However, the AI's specificity, while still high (median = 93.9%), was slightly lower than that of human reviewers (median = 98.4%, p < 0.001), though the AI's specificity had consistently narrower spread. Discussion:This study highlights the challenges of defining the correct answer for EEG interpretation, especially for ESE, and its consequences for evaluating seizure detection AI tools. Future work should explore how AI assistance impacts human interpretation, particularly in reducing interrater variability across a broader range of EEG patterns.

OBJECTIVE:Focal epilepsy is increasingly conceptualized as a network disorder, yet the extent to which network dysfunction reflects a shared phenotype remains unknown. Spatially conserved patterns of network dysfunction may implicate a centralized mechanism underlying widespread impairment. Here, we investigate whether network connectivity disruptions are spatially similar across temporal lobe and extra-temporal lobe epilepsy cohorts and whether shared dysfunction aligns with thalamic connectivity profiles. METHODS:We retrospectively analyzed resting-state magnetoencephalographic imaging from 71 individuals with nonlesional, drug-resistant focal epilepsy (n = 45 temporal, n = 26 extratemporal), collected between 2014 and 2023, and healthy controls (n = 18). Source reconstructed time series were bandpass filtered, and long-range functional connectivity was quantified using imaginary coherence. Network disturbance maps were computed as T-score maps, comparing functional connectivity in epilepsy cohorts to controls, across topographical parcels and frequency bands. Spatial similarity of temporal and extratemporal network dysfunction maps were assessed using Pearson correlations. To infer thalamic involvement, shared network dysfunction maps were correlated with normative functional magnetic resonance imaging-derived thalamocortical connectivity profiles. RESULTS:Extra-temporal lobe epilepsy demonstrated reduced global network connectivity relative to controls in the delta (p = .012), alpha (p = .034), and gamma (p < .001) frequency bands. Across all frequencies, the spatial patterns of network disturbances between temporal and extratemporal cohorts were significantly correlated (r = .287-.717, all p < .001), indicating a shared network dysfunction. Shared spatial maps of network dysfunction correlated with normative thalamocortical connectivity profiles, with significant correlations in the anterior, pulvinar, and dorsomedial thalamus. SIGNIFICANCE:Nonlesional focal epilepsy exhibits a common, frequency-dependent pattern of cortical network dysfunction that is spatially aligned with thalamic connectivity, supporting a thalamic hub contribution to widespread network impairment.

Abstract Mood fluctuations in major depressive disorder are difficult to anticipate. The biological neural rhythms that organize mood dynamics over days to weeks remain unknown. In individuals implanted with a chronic neural sensing and stimulation device for treatment-resistant depression, we collected years-long intracranial neural recordings alongside daily mood ratings. Both mood and limbic neural activity fluctuated cyclically with multiday (multidien) periodicities of 2–34 days. An individual’s daily phase position within mood cycles tracked depression severity, distinguishing whether symptoms were rising, peaking, or resolving. Neural rhythms led mood cycles and forecast an individual’s mood trajectory up to 30 days in advance, outperforming models based on raw neural activity. Electrical stimulation reshaped these rhythms, shifting individuals away from the peak-depression phase of their multidien cycle. Our results identify multidien rhythms as an organizing principle of mood in depression and a forecastable, modifiable target for chronotherapeutic neuromodulation.

Responsive neurostimulation (RNS) is an implanted device that delivers direct brain stimulation for drug-resistant focal epilepsy. Individual responses are highly variable, and no validated framework exists to predict outcome or guide lead placement before implantation. We hypothesized that this variability is partly explained by lead placement in relation to patterns of functional connectivity in brain networks. Fourty-nine patients with drug-resistant focal epilepsy who underwent pre-implantation intracranial EEG (iEEG) and RNS implantation across three independent epilepsy centers were retrospectively studied. We developed a composite functional connectivity score, based on simple Spearman correlation, combining the standard deviation and kurtosis of interictal iEEG connectivity distributions to predict the response outcome in a training cohort (HUP, n=18) and validated in two independent cohorts (NYU, n=17; UCSF, n=14). We accounted for a spatial mismatch between iEEG and RNS electrodes with a distance-based correction. The score was extended to generate patient-specific 3D maps of predicted RNS efficacy across 200 simulated, or "virtual RNS", lead configurations. Accuracy of the score in predicting clinical outcome was 72% at the group level, 61% at the individual patient level, and, after distance-based optimization, 100% in patients with RNS electrodes placed close to location of iEEG electrodes. Applied to the validation cohort, the same score reached 68% accuracy (71% balanced accuracy, 55% sensitivity, 88% specificity). The spatial combination of the scores at different SEEG contacts localization gives a spatial score for each patient. Responders showed significantly higher spatial scores than non-responders, supporting that actual RNS lead placement in responders was located in map-identified favorable regions. Interictal iEEG functional connectivity predicts individual RNS response across independent epilepsy centers, and patient-specific 3D maps derived from this biomarker could prospectively guide lead implantation toward favorable network regions, opening a promising avenue toward network-informed RNS surgical planning.

Seizure forecasting has progressed from theoretical aspiration to a rapidly advancing research domain, yet clinical translation remains limited. Over the past decades, advances in algorithm development, chronic electroencephalography (EEG), wearable sensors, and the characterization of seizure cycles have demonstrated that seizure risk is not random but fluctuates according to identifiable biological rhythms and patient-specific patterns. Forecasting algorithms have shown promising performance across diverse retrospective datasets, including intracranial EEG, subscalp recordings, wearable physiological signals, and even self-reported diaries. However, the clinical value of forecasting is contentious. Prospective real-world validation and regulatory approval of patient-facing forecasting systems remain rare. This review incorporates perspectives presented at the 5th International Congress on Mobile Health and Digital Technology in Epilepsy (2025). We examine barriers impeding clinical translation and current attempts to address them. Crucially, forecasting performance cannot be evaluated in isolation from intended use. Applications range from low-risk uses, like scheduling diagnostic monitoring or visualization of historical trends, to higher risk interventions, including medication titration and adaptive neuromodulation. Each application entails distinct performance thresholds, ethical considerations, and regulatory requirements. Translational challenges include reliable seizure annotation, nonstationarity dynamics of biological cycles, and practical constraints for real-time deployment. Ethical concerns center on miscalibrated reliance on low-risk states, potential anxiety associated with high-risk advisories, and the heterogeneity of patient preferences and risk tolerance. Regulatory pathways are likely to depend on clearly defined use cases and clinically meaningful endpoints, which may extend beyond seizure counts to include quality of life, anxiety, locus of control, and other patient-reported outcomes. Ultimately, translation will require rigorous prospective evaluation against transparent benchmarks, sustainable scientific-commercial partnerships, and integration of probabilistic risk information into clinical workflows. With careful implementation, seizure forecasting may evolve from proof-of-concept research into a clinically meaningful component of epilepsy management, and we remain cautiously optimistic.

BackgroundFor patients with drug-resistant focal epilepsy, surgical resection of the epileptogenic zone (EZ) is an effective treatment to control seizures. Accurate localization of the EZ is crucial and is typically achieved through comprehensive presurgical approaches such as seizure semiology interpretation, electroencephalography (EEG), magnetic resonance imaging (MRI), and intracranial EEG (iEEG). However, interpreting seizure semiology is challenging because it heavily relies on expert knowledge. The semiologies are often inconsistent and incoherent, leading to variability and potential limitations in presurgical evaluation. To overcome these challenges, advanced technologies like large language models (LLMs)—with ChatGPT being a notable example—offer valuable tools for analyzing complex textual information, making them well-suited to interpret detailed seizure semiology descriptions and accurately localize the EZ. ObjectiveThis study evaluates the clinical value of ChatGPT for interpreting seizure semiology to localize EZs in presurgical assessments for patients with focal epilepsy and compares its performance with that of epileptologists. MethodsWe compiled 2 data cohorts: a publicly sourced cohort of 852 semiology-EZ pairs from 193 peer-reviewed journal publications and a private cohort of 184 semiology-EZ pairs collected from Far Eastern Memorial Hospital (FEMH) in Taiwan. ChatGPT was evaluated to predict the most likely EZ locations using 2 prompt methods: zero-shot prompting (ZSP) and few-shot prompting (FSP). To compare the performance of ChatGPT, 8 epileptologists were recruited to participate in an online survey to interpret 100 randomly selected semiology records. The responses from ChatGPT and epileptologists were compared using 3 metrics: regional sensitivity (RSens), weighted sensitivity (WSens), and net positive inference rate (NPIR). ResultsIn the publicly sourced cohort, ChatGPT demonstrated high RSens reliability, achieving 80% to 90% for the frontal and temporal lobes; 20% to 40% for the parietal lobe, occipital lobe, and insular cortex; and only 3% for the cingulate cortex. The WSens, which accounts for biased data distribution, consistently exceeded 67%, while the mean NPIR remained around 0. These evaluation results based on the private FEMH cohort are consistent with those from the publicly sourced cohort. A group t test with 1000 bootstrap samples revealed that ChatGPT-4 significantly outperformed epileptologists in RSens for the most frequently implicated EZs, such as the frontal and temporal lobes (P<.001). Additionally, ChatGPT-4 demonstrated superior overall performance in WSens (P<.001). However, no significant differences were observed between ChatGPT and the epileptologists in NPIR, highlighting comparable performance in this metric. ConclusionsChatGPT demonstrated clinical value as a tool to assist decision-making during epilepsy preoperative workups. With ongoing advancements in LLMs, their reliability and accuracy are anticipated to improve.

OBJECTIVE:Responsive Neurostimulation (RNS) is a closed-loop neuromodulation therapy approved for treating drug resistant epilepsy (DRE) with 1 or 2 seizure foci, but its potential utility for treating complex seizure networks, such as in focal cortical dysplasia (FCD), remains uncertain. This review and commentary discuss the current practice of RNS use in focal cortical dysplasia-related drug-resistant epilepsy(FCD-DRE), and the potential of individualized approaches. METHODS:Our scoping review followed a search to identify relevant studies on epilepsy and RNS across MEDLINE, Embase, and Web of Science, yielding 674, 1,255, and 579 results, respectively followed by abstract and full text review to include FCD-DRE. Data on history, imaging, intracranial EEG, RNS implantation and programming strategies were recorded. RESULTS:78 patients with FCD-DRE across 25 studies were included. The most common lead configuration was two depth electrodes in 53 % (19/36). The median seizure reduction was 85 % [IQR = 66, 96] with a median follow up of 17 months., including 6 patients (7.6 %) achieving seizure freedom for a median 15 months. In 17 patients with resections and RNS implantation, median seizure frequency reduction was 87 % (N = 15), not significantly different from the group with RNS only. 8 patients with cortical and thalamic leads had median seizure frequency reduction of 87 % [IQR = 51, 92]. RNS was effective when used in refractory status epilepticus associated with FCDs. SIGNIFICANCE:RNS is a flexible therapy that effectively reduces seizures in FCD-DRE. Electrographic and imaging signatures can potentially be leveraged. Hybrid resection with RNS approaches and the role in refractory status epilepticus associated with FCD is highlighted. Future studies are necessary to optimize RNS therapy in FCD-DRE.

Recent strategies to predict neonatal seizure risk using machine learning (ML) in combination with quantitative electroencephalography (QEEG) have primarily focused on subject-level predictions over several days during the postnatal period. Time-dependent neonatal preictal state classification with high temporal resolution remains unexplored. In this study, we utilized QEEG feature engineering for ML classification of preictal states and compared it to an end-to-end ML approach. We used two publicly available EEG seizure datasets with a total of 132 neonates containing a total of 281 h of EEG data and segmented data into preictal and interictal epochs of 20 s duration. We employed the Boruta algorithm with Shapley values for QEEG feature selection into ML models. The performance of ML models was assessed with cross-validation with area under the receiver operator characteristic curve (AUROC), area under the precision-recall curve (AUPRC), Matthews Correlation Coefficient (MCC), and F1 score. Feature selection demonstrated statistical moments, spectral power, and recurrence quantification analysis features as robust predictors of preictal states. QEEG feature selection combined with convolutional LSTM outperformed other ML models at preictal versus interictal classification, with AUROC 0.678, AUPRC 0.218, MCC 0.255, and F1 0.334. Our results demonstrate the feasibility of applying ML to facilitate prediction and understanding of neonatal preictal states.

Brain-responsive neurostimulation is firmly ensconced among treatment options for drug-resistant focal epilepsy, but over a quarter of patients treated with the RNS System do not experience meaningful seizure reduction. Initial titration of RNS therapy is typically similar for all patients, raising the possibility that treatment response might be enhanced by consideration of patient-specific variables. Indeed, small, single-center studies have yielded preliminary evidence that RNS System effectiveness depends on the brain state during which stimulation is applied. The generalizability of these findings remains unclear, however, and it is unknown whether state-dependent effects of responsive neurostimulation are also stratified by location of the seizure onset zone where stimulation is delivered. We aimed to determine whether state-dependent effects of the RNS System are evident in the large, diverse, multi-center cohort of RNS System clinical trial participants and to test whether these effects differ between mesiotemporal and neocortical epilepsies. Eighty-one of 256 patients who were treated with the RNS System across 31 centers during clinical trials met criteria for inclusion in this retrospective study. Risk states were defined in relation to phases of daily and multi-day cycles of interictal epileptiform activity that are thought to determine seizure likelihood. We found that the probabilities of risk state transitions depended on the stimulation parameter being changed, the starting seizure risk state, and the stimulated brain region. Changes in two commonly adjusted stimulation parameters, charge density and stimulation frequency, produced opposite effects on risk state transitions depending on seizure localization. Greater variance in acute risk state transitions was explained by state-dependent responsive neurostimulation for bipolar stimulation for neocortical epilepsies and for monopolar stimulation for mesiotemporal epilepsies. Variability in effectiveness of RNS System therapy across individuals may relate, at least partly, to the fact that current treatment paradigms do not account fully for fluctuations in brain states or locations of simulation sites. State-dependence of electrical brain stimulation may inform development of next-generation closed-loop devices that can detect changes in brain state and deliver adaptive, localization-specific patterns of stimulation to maximize therapeutic effects.

The ability to form episodic memories and later imagine them is integral to the human experience, influencing our recollection of the past and envisioning of the future. While rodent studies suggest the medial temporal lobe, especially the hippocampus, is involved in these functions, its role in human imagination remains uncertain. In human participants, imaginations can be explicitly instructed and reported. Here we investigate hippocampal theta oscillations during real-world and imagined navigation using motion capture and intracranial electroencephalographic recordings from individuals with chronically implanted medial temporal lobe electrodes. Our results revealed intermittent theta dynamics, particularly within the hippocampus, encoding spatial information and partitioning navigational routes into linear segments during real-world navigation. During imagined navigation, theta dynamics exhibited similar patterns despite the absence of external cues. A statistical model successfully reconstructed real-world and imagined positions, providing insights into the neural mechanisms underlying human navigation and imagination, with implications for understanding memory in real-world settings.

Strategies to predict neonatal seizure risk have typically focused on long-term static predictions with prediction horizons spanning days during the acute postnatal period. Higher temporal resolution or short-horizon neonatal seizure prediction, on the time-frame of minutes, remains unexplored. Here, we investigated quantitative electroencephalography (QEEG) based deep learning (DL) for short-horizon seizure prediction. We used two publicly available EEG seizure datasets with a total of 132 neonates containing a total of 281 hours of EEG data. We benchmarked current state-of-the-art time-series DL methods for seizure prediction, identifying convolutional LSTM (ConvLSTM) as having the strongest performance at preictal state classification. We assessed ConvLSTM performance in a seizure alarm system over varying short-range (1-7 minutes) seizure prediction horizons (SPH) and seizure occurrence periods (SOP) and identified optimal performance at SPH 3 min and SOP 7 min, with AUROC 0.8. At 80% sensitivity, false detection rate was 0.68 events/hour with time-in-warning of 0.36. Model calibration was moderate, with an expected calibration error of 0.106. These findings establish the feasibility of short-horizon neonatal seizure prediction and warrant the need for further validation.

SUMMARY:Over the past 20 years, responsive neurostimulation (RNS), a closed-loop device for treating certain forms of drug-resistant focal epilepsy, has become ensconced in the epileptologist's therapeutic armamentarium. Through neuromodulatory effects, RNS therapy gradually reduces seizures over years, providing diagnostically valuable intracranial recordings along the way. However, the neuromodulatory potential of RNS therapy has not been fully harnessed. Seizure reduction is often slow, outcomes vary across individuals and defy prognostication, seizure freedom is uncommon, and many patients do not derive significant benefit. These limitations may stem from the "black box" nature of RNS therapy. The antiseizure mechanism(s) of RNS remain poorly understood, and, in the absence of first principles to inform selection of the candidates most likely to benefit, the ideal brain regions to target, and the most effective stimulation parameters, contemporary use of RNS therapy is largely empiric. Fortunately, recent advances in neuroimaging, neurophysiology, artificial intelligence, and engineering have made the goal of rational, personalized neurostimulation a near-term reality. Here, we review recent progress toward this goal, focusing on novel approaches to patient selection, brain network topology, state-dependent effects, and stimulation parameter optimization. By considering the who, where, when, and how of RNS, we highlight emerging paradigm shifts that will help usher in a new age of RNS therapy that is more personalized and more effective.

OBJECTIVES:Generalized and multifocal forms of drug-resistant epilepsy are highly prevalent but have limited treatment options. Centromedian nucleus (CMN) thalamic neuromodulation has emerged as an effective treatment for these epilepsies, but head-to-head neuromodulation modality trials do not exist, and optimal stimulation parameters are not established. MATERIALS AND METHODS:Using Preferred Reporting Items for Systematic reviews and Meta-Analyses guidelines, we conducted a systematic review by searching PubMed and Embase for peer-reviewed studies of bilateral CMN deep brain stimulation (DBS) or responsive neurostimulation (RNS) for generalized and/or multifocal epilepsy. From studies that met the inclusion criteria, we extracted individual patient data and used a mixed-effects model to compare seizure frequency reduction (SR) of modalities and for DBS, across different stimulation frequencies. RESULTS:A total of 25 studies with 192 total patients were included. DBS and RNS yielded comparable SR (76.8% vs 66.7%, respectively, p = 0.1). Within the DBS cohort, high-frequency (>100 Hz) stimulation was more effective for SR by 20.16% (CI: 6.49-33.83) than low-frequency stimulation. Patients with Lennox-Gastaut Syndrome (LGS) had 13.37% lower SR (CI: -26.17 to -0.56) than did those without this diagnosis. Longer follow-up duration (≥18 months) was associated with 12.60% greater SR (CI: 2.53-22.68). CONCLUSIONS:For CMN thalamic neuromodulation in patients with multifocal or generalized epilepsy, modality type (RNS vs DBS) may matter less for SR than may underlying diagnosis (LGS vs not LGS), stimulation parameters (high- vs low-frequency), and treatment duration. These findings have implications for the therapeutic mechanism(s) of thalamic neuromodulation and motivate further study of optimal stimulation approaches.

BACKGROUND:Successful seizure onset zone (SOZ) localisation for epilepsy surgery often relies upon intracranial recordings. Accurate delineation requires anatomical detail yet influences of intracranial electrode density on clinical variables have not been systematically studied. METHODS:In this experimental study we compared SOZ localisation between spontaneously captured seizures on higher-density depth and grid electrode arrays (4-5 mm inter-electrode spacing) vs. lower-density resampled versions of those same seizures (8-10 mm spacing). Since traditional review of channel traces would reveal density conditions, we instead projected seizure activity data as heatmaps on patient brain reconstructions and hid electrode locations. Using a single-blinded randomised crossover design, six attending-level epileptologists viewed these visualisations from ten patients under both higher-density and lower-density conditions (n = 120 observations) and digitally annotated SOZs. FINDINGS:Inter-rater agreement between epileptologists on annotated margins was moderate (average Cohen's kappa: 0.47) and lower for the lower-density condition (p = 0.021, mixed effects model). Scorer confidence ratings did not differ between higher- and lower-density conditions (p = 0.410). The spatial extents of annotated SOZs for higher-density recordings were 25.4% larger on average (p = 0.011) and always closer to true SOZ extents in computer simulations, relative to lower-density. INTERPRETATION:Epileptologists using higher-density depth and subdural intracranial EEG recordings had higher inter-rater agreement and identified larger extents of SOZs compared to lower-density recordings. While further studies assessing surgical outcomes in more patients are needed, these results suggest higher densities of electrodes on already-implanted hardware may reveal sub-centimetre extensions and clearer functional contiguity of the SOZ(s) for better appraisals of pathophysiological margins in epilepsy surgery. FUNDING:This work was supported by the National Institutes of Health through NINDS grant K23NS110920 and through a UCSF Weill Institute for Neurosciences Pilot Award.