Epilepsy Research

Significance: There is potential for generative artificial intelligence (AI) to perform "full stack" academic publishing, including comprehensive manuscript fabrication, "peer"-review, and editorial decision making. This study evaluated the benefits and cautions in using AI for full stack academic publishing in clinical research about seizures. Methods: Three epileptologists each selected 1 published manuscript in Epilepsia and produced an AI-fabricated manuscript about seizures. AI tools were then used to perform "peer"-review and make editorial decisions with minimal human supervision. The rates of identifying AI-fabricated manuscripts and editorial decisions among the epileptologists was analyzed, and epileptologists' observations on how AI differed from typical human "peer"-review and editorial decisions was collected. Results: Four of 6 (67%) pairs of AI-fabricated and human-generated manuscripts were correctly identified. Only 1 AI-fabricated manuscript was editorially rejected, whereas another had a subtle indicator of fabrication. When compared to human content, AI generated content was comprehensive, detailed, and based on the highest standard of academic publication. However, generated content was more verbose, used more structured responses, and more critical feedback. Conclusions: These results demonstrate that the quality of AI-fabricated manuscripts about clinical research in seizures was sufficiently high that the current methods of peer and editorial review could not easily distinguish fabricated content compared to human-written content. Moreover, AI review and editorial decisions were high quality and appeared reasonable in many cases. AI tools have matured to the point that there should be thoughtful policies that permit AI assistance but prohibit insufficiently supervised autonomous scientific submissions, reviews, and editorial decisions.

BackgroundSerotonergic signaling plays an important role in the modulation of absence seizures (ASs), and activation of serotonin 2A receptors (5-HT2ARs) with TCB-2 suppresses spontaneous absence seizures (ASs) in Genetic Absence Epilepsy Rats from Strasbourg (GAERS). Lisuride is a pharmacologically complex serotonergic and dopaminergic ligand with low 5-HT2AR signaling efficacy and a receptor profile distinct from that of TCB-2. We therefore investigated its effects on spontaneous ASs and the contribution of 5-HT2ARs to its actions. MethodsThe effects of systemic lisuride (0.1 and 0.5 mg/kg) on spontaneous ASs were examined in freely moving GAERS over 120 min by quantifying time spent in seizure, mean seizure duration and seizure number. The contribution of 5-HT2ARs was investigated by administering lisuride (0.5 mg/kg) following pretreatment with the selective 5-HT2AR antagonist MDL11,939 (0.5 mg/kg). ResultsLisuride produced marked dose- and time-dependent effects on ASs. At 0.1 mg/kg, lisuride elicited modest pro-absence effect. In contrast, 0.5 mg/kg produced a pronounced biphasic response, with a marked reduction in seizure activity during the first 40 min, followed by rebound increase during the second hour. The early suppression involved reductions in both seizure number and duration, whereas the delayed pro-absence effect was predominantly associated with prolonged ASs. MDL11,939 did not prevent the early seizure suppression but blocked the increase in seizure duration while producing a transient increase in seizure number. ConclusionsLisuride is a dose- and time-dependent bidirectional modulator of ASs in GAERS. The persistence of the profound early anti-absence effect during 5-HT2AR blockade indicates that this response does not require 5-HT2AR activation, whereas the modification of the delayed response by MDL11,939 supports a 5-HT2AR-dependent component in the later phase. These findings distinguish lisuride pharmacologically from TCB-2 and suggest that its biphasic effects arise from the dynamic engagement of multiple receptor mechanisms, potentially including dopaminergic signaling.

Deletion of the microtubule-associated protein, tau reduces seizure burden in rodent models of genetic epilepsies and suppresses evoked seizures. In the intrahippocampal kainate (IHK) model of acquired temporal lobe epilepsy (TLE), mice lacking tau expression develop spontaneous seizures, but at a lower rate than controls. Mechanisms by which tau deletion modifies the development of acquired TLE are not well understood, but a modification of the synaptic reorganization that accompanies epileptogenesis may contribute to modified seizure expression. Increased inhibitory synaptic input to dorsal dentate granule cells (DGCs) contralateral to the focal kainate injection results after development of TLE in tau-/- mice, but inhibitory synaptic reorganization distal to the kainate lesion is undefined in this model. We therefore investigated how development of TLE modifies excitability and synaptic transmission in DGCs of the ventral dentate gyrus, distant from the kainate lesion. Our results show that development of TLE increases evoked action potential firing of ventral DGCs (vDGCs) in mice that develop TLE, demonstrating an effect of TLE development on excitability of DGCs distant from the focal injury. After development of TLE, DGCs from tau-/- mice received increased inhibitory synaptic input bilaterally, suggesting that tau expression suppresses inhibitory synaptic reorganization in DGCs. Optogenetic stimulation of interneurons in tau-/- mice revealed increased inhibitory synaptic input to vDGCs and an increase in inhibitory convergence onto vDGCs once epileptogenesis occurred. Therefore, tau deletion modifies spontaneous seizure expression after epileptogenesis in mice, potentially involving increased synaptic inhibition of DGCs throughout the dentate gyrus.

Epilepsy is increasingly recognized as a disorder involving both neuronal and vascular dysfunction. While connexin signaling has been implicated in epileptogenesis, the contribution of vascular connexins to seizure associated cerebrovascular pathology remains poorly understood. Connexin40 (Cx40) is an endothelial gap junction protein that plays a crucial role in vascular communication and blood-flow regulation. Seizures induce dynamic changes in cerebral perfusion and oxygenation, including prolonged postictal hypoperfusion/hypoxia. To determine whether Cx40 influences postictal hypoxia following focal seizures, we examined seizure characteristics and postictal oxygen dynamics in Cx40 knockout (Cx40-/-) mice using an established focal hippocampal seizure model. Electrically kindled seizures were elicited in wild-type and Cx40-/- mice, and local hippocampal tissue oxygenation was continuously monitored before and after seizure induction. Seizure duration did not differ between genotypes, indicating comparable seizure severity. Interestingly, Cx40 deletion altered the temporal pattern of postictal oxygen recovery, producing greater early hypoxia and a delayed secondary rebound in pO2 despite similar peak oxygen levels and overall hypoxic burden. These findings demonstrate that loss of Cx40 selectively alters the temporal profile of postictal oxygen dynamics without affecting seizure duration. Taken together, the results suggest that endothelial gap junctional communication contributes to postictal vascular recovery and identify Cx40 as a potential modulator of seizure associated neurovascular dysfunction.

Intrahippocampal kainate (IHK) is a well-established mouse model of temporal lobe epilepsy, in which focal hippocampal injection of kainic acid induces excitotoxic injury followed by the development of chronic spontaneous seizures. Commercial sources of kainic acid (KA) have increasingly transitioned from naturally derived to synthetic KA, which has introduced a potential source of variability to the implementation of the IHK model. Here, we evaluated injection parameters for synthetic KA, with the goal of identifying conditions that reliably produce spontaneous seizures while minimizing mortality. We found that intrahippocampal injection of 200nL of 5mM synthetic KA resulted in low mortality compared to 50-100nL of 20mM synthetic KA (a dosing range commonly used with naturally derived KA), while robustly inducing TLE-like hippocampal sclerosis and recurrent spontaneous electroclinical seizures. We further demonstrated high inter-animal variability in both seizure frequency and temporal clustering of seizures across animals receiving either injection protocol (200nL of 5mM or 50nL of 20mM). Together, these findings provide a framework for implementing the IHK model with synthetic KA and support its continued use to advance our understanding of the cellular and network mechanisms underlying temporal lobe epilepsy.

Based on the observation that in normal EEG rhythms, the low-voltage beta frequencies dominate in the anterior region while characteristic high(er)-amplitude slower rhythms manifest more posteriorly, we studied these gradients across normal subjects and patients with different monogenic epilepsies. From large data sets (long term records of 374 controls and 241 patients) we selected 60s eyes-open rest-EEG epochs to determine the spectral power of the EEG bands. To help manage the large data sets and facilitate analyses, we employed AI to preselect and assist in coding. While both preselection and coding was extensively verified and validated by the authors, which was a time-consuming effort, the AI contribution reduced the net effort of evaluating the large dataset. We developed a metric that reflects the anterior-posterior gradients of the beta rhythm and the so-called posterior dominant rhythm. Our results indicate that this simple gradient measure across 1-10 year-age groups may provide a robust quantitative biomarker of cerebral dysfunction in monogenic epilepsy. Because EEG is noninvasive, widely available, repeatable, and relatively inexpensive, such a biomarker could ultimately provide a practical means of assessing the cerebral effects of emerging targeted therapies across multiple genetic epilepsies.

ObjectiveThe current International League Against Epilepsy (ILAE) guidelines for Lennox-Gastaut syndrome (LGS) mandate the presence of generalized paroxysmal fast activity (GPFA) on EEG. However, GPFA is not necessarily restricted to LGS and has been described in, among others, idiopathic generalized epilepsies. A clear definition for GPFA is currently lacking, hampering research into its utility as a diagnostic or treatment biomarker. Here, we assessed expert agreement in the interpretation of GPFA to explore feasibility of a consensus-based definition. MethodsWe internationally distributed a two-part survey on GPFA to EEG experts. First, participants completed thirteen questions gauging their preferred clinical and EEG criteria for GPFA, including duration, frequency, amplitude and topography. Second, participants classified the presence versus absence of GPFA in eleven EEG clips, while blinded to clinical context, from patients with LGS, other epilepsies, and physiological variants. Results162 respondents, predominantly from North America and Europe, completed the survey. 87% considered GPFA to also occur in epilepsies other than LGS. Considerable heterogeneity was present on defining duration, spectral frequency, amplitude and topographic distribution of GPFA. Particularly poor agreement was seen for lower limits of duration and frequency, with no single response option receiving a majority consensus (36% and 30% agreement, respectively). Similarly, the degree of agreement on interpretations was highly variable across real-world EEG clips. SignificanceMarked discrepancies exist in the EEG identification and clinical interpretation of GPFA, undermining its utility as a potential biomarker in research and clinical trials of LGS and other epilepsies. The findings also highlight the need to establish a consensus-based and eventually evidence-informed definition to formally operationalize the clinical and EEG criteria for GPFA. Key pointsO_LIGeneralized paroxysmal fast activity (GPFA) is variably defined in the literature, limiting its utility as a diagnostic or treatment biomarker C_LIO_LIIn this international survey (n = 162 respondents, 25 countries) we found limited agreement on GPFAs defining features and on interpretation of fast activity on EEG clips C_LIO_LI87% of respondents considered GPFA to occur in epilepsies other than Lennox-Gastaut syndrome C_LIO_LIPending further elucidation of the pathophysiology underlying GPFA or its computational characterization (i.e., to inform a biologically-or evidence-based definition), we underscore the need for an interim consensus-based definition C_LI

Resistance to anti-seizure medications (ASMs) remains a major clinical challenge in the management of epilepsy. Neuroinflammation has been implicated as a contributing mechanism and, consistent with this, antagonists of the ATP-gated P2X7 receptor (P2X7R) have been shown to enhance ASM efficacy in animal models. It remains unclear, however, if P2X7R antagonism is effective in human models of drug-resistant epilepsy. Here, we assessed the seizure-suppressive potential of the P2X7R antagonists AFC-5128 and JNJ-47965567 using electrophysiological recordings in acute resected brain slices from patients with epilepsy using artificial cerebrospinal fluid containing low Mg2+ and high K+ to evoked seizure-like events. P2X7R antagonists alone did not suppress seizure-like activity. When co-administered, however, P2X7R antagonists significantly enhanced the anti-seizure efficacy of carbamazepine and lorazepam. Notably, P2X7R antagonist treatment lowered Interleukin-1{beta} levels, and betaine, a drug targeting Interleukin-1{beta} release, mimicked the effects of P2X7R antagonists when combined with carbamazepine. Finally, mice with microglia-specific P2X7R depletion showed improved responsiveness to carbamazepine, suggesting P2X7R-mediated effects are partially mediated via their functions in microglia. These findings suggest that P2X7R-based therapies may represent an effective add-on approach for treating drug-resistant seizures associated with neuroinflammation.

Childhood absence epilepsy (CAE), the most common childhood epilepsy syndrome, is associated with bursts of generalized thalamocortical spike-wave discharges and cognitive impairment, yet the physiological mechanisms linking disease activity, treatment, and cognition remain unclear. Sleep spindles, generated by thalamocortical circuits, are associated with cognitive function and may be disrupted by epileptic activity. Here, we retrospectively studied 87 EEGs from 53 children with active and resolved CAE and 87 age- and sex-matched controls to evaluate sleep spindle activity across disease states and its association with antiseizure medication (ASM) exposure. We found that children with active CAE had reduced spindle rates across all cortical regions compared to those with resolved CAE and controls, with the largest reduction in the frontal region, where epileptic spike activity was also most prominent. Spindle rate was inversely correlated with epileptic spike rate, consistent with a shared thalamocortical circuitry. The association between ASM exposure and spindle rate varied by disease state. In active CAE, treated children had higher spindle rates than untreated children. In resolved CAE, treated children had lower spindle rates than untreated children. In contrast, ASM-related differences in spike rate were more limited, although a medication-specific reduction was observed in active CAE. These findings support sleep spindle rate as a candidate physiological marker of disease state and ASM exposure in children with CAE. Given the established role of sleep spindles in memory consolidation, these findings support testing of whether spindle disruption provides a mechanistic link between epileptic thalamocortical activity and cognitive vulnerability in CAE.

Absence epilepsy is a generalized non-convulsive epilepsy characterized by recurrent absence seizures (ASs) and enhanced GABAA receptor-mediated tonic inhibition of thalamocortical (TC) neurons has been shown to lead to ASs. Serotonergic signalling is an important modulator of these seizures, yet the contribution of specific 5-HT receptor subtypes remains incompletely understood. Here, we investigated whether endogenous serotonergic signalling through 5-HT2A receptors (5-HT2ARs) contributes to pathological tonic inhibition and ASs expression. Patch-clamp recordings were obtained from ventrobasal (VB) TC neurons of Genetic Absence Epilepsy Rats from Strasbourg (GAERS) and non-epileptic control (NEC) rats. The selective 5-HT2AR antagonist MDL11,939 had no effect on tonic GABA-A currents in NEC TC neurons but significantly reduced the enhanced tonic current in GAERS neurons, revealing a disease-dependent contribution of endogenous 5-HT2AR signalling. MDL11,939 administered bilaterally into the VB of freely moving GAERS by reverse microdialysis did not alter ASs. This dissociation between the modulation of a disease-associated cellular mechanism and seizure outcome suggests that the previously reported reduction of ASs by systemic 5-HT2AR blockade may involve mechanisms outside the VB.

Background and Objectives Interictal spikes on intraoperative electrocorticography have been used to localize the epileptogenic zone but have limited specificity. We hypothesized that identifying spatially distributed, spike-associated very high-frequency activity (vHFA; 500-2,000 Hz) would improve the ability of spikes to localize the epileptogenic zone. We aimed to characterize spike-associated vHFA and to determine whether a seizure outcome prediction model incorporating vHFA could be externally validated across independent centers. Methods This retrospective two-center cohort study included patients with drug-resistant focal epilepsy who underwent electrocorticography and subsequent resection between January 2016 and May 2025. Spikes occurring across contacts within 50 ms were classified as co-occurring; all others were classified as isolated. We determined whether vHFA was associated with co-occurring spikes. We then assessed whether contacts enriched in vHFA-positive co-occurring spikes were more likely to be located within the resected region among seizure-free patients. Finally, we determined whether the rate of such spikes remaining outside the resection predicted seizure outcomes across the two centers. Results Among 56 patients (mean age: 21.2 years; 26 females), we identified 132,359 spikes, of which 1,925 were vHFA-positive. The vHFA peak preceded the associated spike peak by 8.0 ms (95%CI: 5.0-16.0). vHFA-positive spikes were more likely to be classified as co-occurring, even after adjustment for anatomy and inter-electrode distance (OR: 3.43; 95%CI: 2.82-4.17). Among the 35 seizure-free patients, contacts with higher rates of vHFA-positive co-occurring spikes were more likely to be resected (OR: 11.92 per occurrence/min; 95%CI: 5.62-25.32). This association was stronger than that observed for all spikes (OR: 1.09 per occurrence/min; 95%CI: 1.06-1.11). Twenty patients experienced postoperative seizure recurrence; a higher proportion of vHFA-positive co-occurring spikes in preserved areas was associated with recurrence (OR: 3.22 per 0.01 increase in proportion; 95%CI: 1.14-9.12). In bidirectional cross-center validation, models incorporating vHFA-positive co-occurring spikes yielded areas under the receiver operating characteristic curves of 0.80 and 0.69, with better performance when the model was developed using the larger sample. Discussion Spike-associated vHFA may identify a biologically meaningful subset of interictal spikes involved in propagation across the epileptogenic zone. vHFA-positive co-occurring spikes may represent a generalizable intraoperative biomarker of the epileptogenic zone.

Visual prostheses have targeted early structures in the visual pathway, including the retina, optic nerve, lateral geniculate nucleus, and visual cortex, to generate visual percepts. Stimulation of higher-order visual regions, which may better encode complex visual features, remains relatively unexplored. We tested whether electrical stimulation of the medial and inferior pulvinar elicits visual percepts in a patient with medically refractory epilepsy undergoing invasive seizure monitoring with intracranial electroencephalography. Bipolar stimulation (30 biphasic pulses, 0.5 ms per phase, 3 mA, 1 Hz) was delivered between adjacent pulvinar contacts while eye movements were tracked. Stimulation of different electrode pairs elicited visual percepts at distinct retinotopic locations (p = 0.010). Stimulation also altered cortical activity, producing an early power increase followed by gamma desynchronization in occipital and parietal regions (p < 0.001) and beta desynchronization in the lateral temporal region (p < 0.01). These findings suggest that pulvinar stimulation can generate visual percepts and modulate distributed cortical networks, supporting further investigation of the pulvinar as a target for vision restoration.

Activity in the gamma frequency range (>30Hz) serves as a neural hallmark of stimulus encoding across multiple sensory modalities. Based on its distinct properties, this activity could be classified as broadband or narrowband oscillations. Visual stimuli, such as a grating, induce narrowband oscillations, which have shown promise as diagnostic markers for neurological disorders. However, such tasks are discomforting for the subjects. These can be made subject-friendly by swapping the stimulus for an auditory one; however, most auditory stimuli elicit a broadband activity. In search of an ideal candidate to induce auditory narrowband oscillations, we previously recorded responses to auditory gratings in electroencephalography (EEG) recordings (Gulati & Ray, 2024) using a passive listening paradigm. We observed a focal sustained broadband increase in high gamma activity (70-150Hz), which was not selective towards any stimulus attributes. Nevertheless, since EEG represents activity of a large population, it remained unclear whether observed activity represented a broadband increase or comprised multiple narrowband rhythms, and whether it was indeed untuned. To elucidate this, we recorded invasively from the regions on and around the superior temporal gyrus, closer to Heschl's gyrus of seven patients undergoing epilepsy resection surgery. Patients were either under anesthesia or fully awake while we presented auditory gratings through loudspeakers. We found a transient increase in broadband gamma activity (30-100Hz) followed by a sustained suppression of beta rhythm (14-28Hz). This broadband gamma was selective for slower spectro-temporal modulations. Our result compares the responses obtained across different recording scales and corroborates that auditory gratings produce broadband but not narrowband gamma activity.

Background and Objective: The volume of patient portal messages has grown substantially and artificial intelligence (AI) can support clinical staff and improve the timeliness of patient care. Our objective is to test if an AI tool for managing the electronic inbox In-Basket reduces response times to patient portal messages in neurology. Methods: We conducted an implementation study from January to May 2026 in the neurology department at two clinical sites within a regional medical system. A single AI tool was implemented to auto-categorize and auto-route all patient messages received in the electronic inbox. The tool also generates optional draft replies to messages. Primary process measures were the median weekly time (hours) from message creation to the first action and to closure. Impact was analyzed using a controlled interrupted time-series (ITS) design, comparing message-volume-matched control divisions. Results: Among 24,142 messages analyzed, pre-intervention median weekly time to first action was 11.3 hours [IQR: 5.6, 15.1] (Site A) and 16.1 hours [IQR: 13.8, 19.0] (Site B), falling post-intervention to 1.0 hour [IQR: 0, 3.1] (Site A) and 0.6 hours [IQR: 0.2, 2.4] (Site B). Controlled interrupted time-series analysis showed a significant immediate differential level change in median weekly hours to first action at Site A (-92.5%, 95% CI: -97.5% to -78.0%), and a significant downward trend at Site B (-43.5%, 95% CI: -65.2% to -8.2%). Changes in time to closure were directionally favorable but not statistically significant at either site. There was low use of the generated draft replies, only 0.5% at Site A and 3% at Site B. Discussion: Implementation of AI-assisted message categorization and routing was associated with a substantially reduced initial action time for patient In-Basket messages in Epilepsy. The utilization of AI-generated draft-replies to messages was low. These findings support the feasibility and effectiveness of AI-augmented workflows to improve inbox management.

Background: Working memory, the cognitive ability to maintain and manipulate a finite amount of information, is fundamental in reasoning, learning, and decision-making. Working memory erodes in conditions from Alzheimer's disease to stroke to healthy aging. Transcutaneous auricular vagus nerve stimulation (taVNS) is a promising neuromodulation method for improving working memory. By stimulating the auricular branch of the vagus nerve, taVNS is thought to engage ascending arousal systems such as the locus coeruleus-norepinephrine (LC-NE) system. However, the physiological link between LC-NE system modulation via taVNS and improved working memory remains largely unknown. This mechanistic gap limits the rational optimization of the stimulation protocol and the identification of the cognitive functions and patient populations that are likely to benefit. Objective: We test two competing hypotheses for how taVNS may improve working memory: sensory amplification or enhancement of cortical efficiency. The sensory-amplification hypothesis predicts increased cortical responses to stimuli, whereas the cortical-efficiency hypothesis predicts reduced cortical responses. Methods: Twenty healthy adults completed N-back working memory tasks during three counterbalanced sessions: baseline, active vibrotactile taVNS (vtaVNS), and sham stimulation, while we simultaneously recorded electroencephalography, electrocardiography, eye-tracking, and behavioral responses. To study the effect of vtaVNS on neuronal activity, we also recruited 20 patients with epilepsy undergoing invasive monitoring. Of these, 19 fully completed the spatial working-memory task. We quantified cortical responses to visual stimuli using event-related potentials (ERPs) in scalp EEG and high-gamma activity in intracranial local field potentials. To examine prediction of the cortical-efficiency hypothesis, we further explored the effect of vtaVNS on distance to criticality and heartbeat-evoked potentials. Results: Active vtaVNS reduced both scalp-recorded visual ERP and high-gamma response in temporal cortices, relative to sham stimulation. Moreover, smaller visual event-related potentials were associated with successful encoding during the most demanding working-memory condition (i.e., 4-back task). Exploratory analyses showed that vtaVNS shifted cortical dynamics closer to criticality. Conclusion: Our results converge on the conclusion that vtaVNS does not uniformly amplify cortical responses to external stimuli. Instead, vtaVNS enhances cortical efficiency, supporting more effective allocation of cognitive resources between exteroceptive and mnemonic processes. These findings provide a framework for predicting when taVNS will be most beneficial and for optimizing stimulation protocols across cognitive and clinical applications.

Pharmacological treatments targeting core autistic features remain limited. In neurotypical adults, cannabidiol (CBD) alters mismatch negativity (MMN), an electrophysiological marker of automatic auditory deviance detection that has been associated with autistic traits. The effects of CBD on MMN, as measured using electroencephalography (EEG), in autistic children remain unknown. This secondary analysis of a pilot randomised, double-blind, placebo-controlled crossover trial examined the effects of 12 weeks of oral CBD oil on MMN during a duration-deviant auditory oddball task. We hypothesised that CBD would produce more negative MMN amplitudes and shorter peak latencies relative to placebo. Nineteen autistic children (6-12 years, 8 female) received CBD (dose) and placebo in randomised order, separated by an 8-week washout. Age-adjusted linear mixed-effects models tested Time (baseline, follow-up) x Treatment (CBD, placebo) effects on mean MMN amplitude and peak latency separately at Fz and FCz. Exploratory Spearman correlations examined baseline associations between MMN and Developmental Behaviour Checklist-2 (DBC-2) Social Relating scores. At FCz, latency decreased following CBD but increased following placebo but did not survive FDR correction (p = .029, pFDR = .058) and no other interactions were significant. Higher baseline DBC-2 Social Relating scores were associated with more negative MMN amplitude at FCz (rho = -.58, pFDR = .028) and Fz (rho = -.59, pFDR = .028). These findings provide tentative evidence that CBD may modulate MMN latency in autistic children. Larger trials should clarify CBDs effects on MMN and further examine the association between MMN amplitude and social-relating difficulties.

Background: Software tools can support MRI diagnosis of focal cortical dysplasias (FCDs) - subtle, surgically-treatable lesions that underlie drug-resistant focal epilepsy (DRFE). MELD-graph, a widely used open-source tool combining surface-based morphometry with a convolutional neural network, was developed using 1.5T and 3T MRI data. 7T MRI is increasingly used for lesion detection in DRFE. This study investigated whether MELD-graph is robust to domain shift at 7T. Methods: We collated paired 3T and 7T T1-weighted MRI, and FLAIR where available, from 23 patients with FCD (6 histologically-verified, 17 identified radiologically) and 39 healthy controls at four epilepsy centres. Manual lesion labels were used as ground truth. MELD-graph was run on 3T images, on 7T images with the default pipeline, and on 7T images with adaptations for 7T MRI (SPM bias-field correction, SynthStrip skull-stripping, sub-millimetric FreeSurfer reconstruction), with and without ComBat harmonisation. FCD detection performance and surface morphometry features were compared. Results: Sensitivity for FCD detection was 65% (15/23) at 3T, 48% (11/23) for 7T default and 52% (12/23) for 7T adapted, with Combat harmonisation. Among radiologically 3T-negative patients, 2/8 were detected at 3T and 3/8 at 7T; and among histologically confirmed FCD, 3/6 at 3T and 1/6 at 7T. Specificity was significantly lower at 7T (default 39%, adapted 45%) than 3T (76%). False-positives predominantly occurred in the basal temporal lobes. ComBat harmonisation improved specificity in all conditions and largely compensated for field-strength dependent differences in surface morphometry features. Adding FLAIR at 7T increased false-positive predictions. Conclusion: Sensitivity is comparable but specificity of MELD-graph is lower at 7T compared to 3T. Adaptations for 7T MRI can reduce segmentation errors but do not significantly improve FCD detection, while harmonisation improves performance. The adapted pipeline is released open-source. 7T MELD-graph outputs may inform radiologist review in challenging cases but require cautious interpretation.

BackgroundClinicians increasingly use AI systems to search the medical literature, but current benchmarks do not jointly test whether a response identifies the originating paper and its supporting passage. LitBench evaluates four dimensions: stated versus interpretation-requiring facts, the number and composition of competing papers, one-versus two-paper evidence, and refusal when evidence is absent. MethodsLitBench comprises 1,980 open-access papers (1,000 epilepsy papers holding the answers, 980 decoys from unrelated fields) and 9,472 human-reviewed facts, giving 2,188 single-paper questions and 170 requiring a fact from each of two papers. Difficulty rose by burying the answer among up to 2,000 distractors, then again over live PubMed Central. The same questions were then asked with the answering paper removed, so that refusal was the correct response. Four systems were tested: Gemma-4B, Gemma-12B, Sonnet-5, and DeepSeek-V4-Flash (refusal conditions only). Three model judges scored each answer by majority; two-paper questions counted only when both facts were found. ResultsAcross fixed-corpus single-paper conditions, accuracy ranged from 55.9% to 71.9% for Gemma-4B, 61.2% to 75.5% for Gemma-12B, and 90.0% to 91.3% for Sonnet-5. Similar epilepsy papers were harder than mixed candidate sets for both Gemma configurations. Two-paper accuracy was 5.7%, 17.1%, and 37.8%, respectively. With evidence absent, Gemma-12B refusal fell from 88.3% to 39.9% as candidate sets grew; Gemma-4B almost never refused. Sonnet-5 refused on 94.9% of sampled single-paper and all sampled two-paper cells. DeepSeek-V4-Flash also refused frequently, but on 80.3% of matched answer-present controls. ConclusionsNo system was reliable across every axis tested: accuracy fell as competing papers were added, interpretation-requiring facts were harder than stated ones, two-paper synthesis was rarely achieved, and only some systems recognized absent evidence. What counts as "successful" literature search carries considerable nuance; LitBench can evaluate proposed tools at a fine-grained level.

Background Elevated resting gamma power is a reproducible, robust neurophysiology marker across species in fragile X syndrome (FXS), a major known cause of autism spectrum disorders (ASD). In FXS, this neurophysiological phenotype is well described and particularly prominent in males harboring fully methylated full CGG repeat expansions. However, gamma power evaluations in idiopathic ASD have yielded variable results. As a single-gene model of ASD, FXS hold promise to biologically model some of the autism spectrum. Exaggerated gamma power in FXS is thought to reflect increased cortical excitability, a common feature of ASD, and has been shown to correlate with symptoms in both human FXS individuals and the Fmr1 knockout mouse model of FXS. A recent clinical trial of the BK channel positive modulator, Codabakalner, provided encouraging signs that normalization of gamma power in FXS may provide therapeutic benefit [14]. Interventional and biomarker studies in ASD provide support that this may also be the case in ASD. We utilized gamma band power to evaluate neurophysiological overlap between ASD and FXS seeking to determine the prevalence and clinical characteristics of those with ASD who exhibit a FXS-like elevated resting state gamma phenotype. Methods This cross-sectional resting EEG analysis included 159 individuals with non-syndromic ASD and 41 males with fully methylated full mutation FXS. Relative gamma power was calculated as 30-100 Hz power divided by 1-100 Hz power across 10-20-equivalent electrodes, excluding 55-65 Hz. We defined the cutoff for elevated resting gamma band power in ASD as relative gamma power of greater than or equal to the median minus one standard deviation of gamma power in our FXS male sample. Results A FXS-derived relative-gamma threshold of 0.185 or greater was met by 22% of participants with ASD (n=35). As expected, the mean ASD group resting state gamma was lower than the male FXS sample (p < 0.001). ASD participant threshold proportions by age group were 3/37 (8.1%) below 13 years, 9/50 (18.0%) age 13-18 years, and 23/72 (31.9%) in the 18+ years group. Within the ASD group, there were no observable clinical characteristics that readily differentiated participants with an elevated gamma phenotype from those without this phenotype. Conclusions Approximately one fifth of all participants and slightly over 30 percent of adults with ASD met the FXS-derived elevated resting gamma band power criterion. This represents the first investigation using a translational brain neurophysiology marker in a syndromic form of autism to parse a clinical idiopathic ASD sample to identify potential overlapping brain features. The results will work to support potential targeting of ASD subgroups in the future utilizing therapeutic approaches initially established in a single-gene form of developmental disability.

Objective: Cortico-cortical evoked potentials (CCEPs) are used during intracranial EEG (iEEG) monitoring to probe effective connectivity and guide epilepsy surgery. However, the factors governing spatial distribution and variability of stimulation-evoked responses across patients remain incompletely understood, making it difficult to distinguish physiological variation from pathological responses. We aimed to develop interpretable models of CCEP responses that integrate normative and patient-specific features, and to assess how these relationships are influenced by distance, brain-state, and epileptogenicity. Methods: We analyzed single-pulse stimulation data from 34 patients with drug-resistant epilepsy undergoing stereo-EEG monitoring. Significant CCEP responses were defined using N1 peak amplitude. For each stimulation-response pair, we derived multimodal features capturing Euclidean distance, effective connectivity priors from F-TRACT, a published population atlas of CCEP response probabilities, normative structural connectivity and communication metrics, and patient-specific functional connectivity estimated from iEEG during baseline and pre-stimulation periods. Lasso-regularized logistic regression models were trained to predict response using leave-one-patient-out cross-validation, and feature importance was examined across spatial scales and seizure onset zone (SOZ) involvement. Results: Performance to predict stimulation-evoked responses improved with the integration of multiple feature domains, with the full model achieving the highest accuracy (median AUC = 0.84). Feature relevance varied with spatial scale: short-range responses were dominated by distance and normative priors, whereas long-range responses depended strongly on patient-specific functional connectivity, especially when measured from pre-stimulation periods. Prediction errors were significantly higher for stimulation-response pairs within the SOZ, indicating deviations from normative connectivity-response relationships in epileptogenic networks. Significance: CCEP responses reflect an interaction between spatial constraints, network architecture, and patient-specific brain state. Multimodal modeling of CCEPs provides a principled framework to interpret stimulation responses and may offer complementary markers for identifying epileptogenic networks during presurgical evaluation.