Background Seizure unpredictability is a significant burden in the lives of people with epilepsy. Previously published approaches to seizure forecasting analysed intracranial electroencephalographic recordings (iEEG) and showed that seizures can be forecast above chance levels. Although passive observation of the brain might provide some insights, repeated active perturbation of the cortex and measuring the cortical response may provide more direct information about time-varying cortical excitability. Objective The aim of this study is to investigate whether seizures can be forecast by stimulating the cortex via intracranial electrodes and measuring cortical response from the iEEG. Methods We studied a cohort of eight patients with treatment-resistant epilepsy who were admitted to King’s College Hospital for presurgical evaluation with iEEG. During their stay, they underwent prolonged single pulse electrical stimulation for approximately one day. Stimuli were delivered every 5 minutes to a constant pair of electrodes and all patients experienced at least one clinical seizure during the period of stimulation. We extracted quantitative features from the iEEG post-stimulus response and developed a logistic regression algorithm to estimate the seizure likelihood at each stimulus. To evaluate the algorithm’s performance, we used improvement over chance (IoC), sensitivity, time spent in warning and Brier Skill score. We also compared performance with seizure prediction based on passive observation of iEEG. Results In seven out of eight patients, seizures could be forecast using the post-stimulus response above chance levels (average IoC: 0.74). In comparison, the seizure forecasting performance based on passive (unstimulated) iEEG was less good (average IoC: 0.54). Conclusions These results suggest that cortical response to electrical stimulation may aid in the development of seizure forecasting algorithms as well as in the design of novel implantable devices that deliver electrical stimulation to control seizures.

OBJECTIVE:The emergence of large language models (LLMs) and the increasing prevalence of electronic health records (EHRs) present significant opportunities for advancing health care research and practice. However, research that compares and applies LLMs to extract key epilepsy-related information from unstructured medical free text is under-explored. This study fills this gap by comparing and applying different open-source LLMs and methods to extract epilepsy information from unstructured clinic letters, thereby optimizing EHRs as a resource for the benefit of epilepsy research. We also highlight some limitations of LLMs. METHODS:Employing a dataset of 280 annotated clinic letters from King's College Hospital, we explored the efficacy of open-source LLMs (Llama and Mistral series) for extracting key epilepsy-related information, including epilepsy type, seizure type, current anti-seizure medications (ASMs), and associated symptoms. The study used various extraction methods, including direct extraction, summarized extraction, and contextualized extraction, complemented by role-prompting and few-shot prompting techniques. Performance was evaluated against a gold standard dataset, and was also compared to advanced fine-tuned models and human annotations. RESULTS:Llama 2 13b (a 13-billion-parameter LLM developed by Meta) demonstrated superior extraction capabilities across tasks by consistently outperforming other LLMs (F1 = .80 in epilepsy-type extraction, F1 = .76 in seizure-type extraction, and F1 = .90 in current ASMs extraction). Here, F1 score is a balanced metric indicating the model's accuracy in correctly identifying relevant information without excessive false positives. The study highlights the direct extraction showing consistent high performance. Comparative analysis showed that LLMs outperformed current approaches like MedCAT (Medical Concept Annotation Tool) in extracting epilepsy-related information (.2 higher in F1). SIGNIFICANCE:The results affirm the potential of LLMs in medical information extraction relating to epilepsy, offering insights into leveraging these models for detailed and accurate data extraction from unstructured texts. The study underscores the importance of method selection in optimizing extraction performance and suggests a promising avenue for enhancing medical research and patient care through advanced natural language processing technologies.

This study investigates the feasibility of using a two-channel subcutaneous EEG device (SubQ) to detect and monitor PGES. The SubQ device, developed by UNEEG Medical A/S, offers a minimally invasive alternative to scalp EEG, enabling ultra-long-term monitoring and remote data analysis. We used annotated scalp EEG data and data from the SubQ device. The pre-processing pipeline included channel reduction, resampling, filtering, and feature extraction. A Variational Auto-Encoder (VAE) was employed for anomaly detection, trained to identify PGES instances, and post-processing was applied to predict their duration. The VAE achieved a 100% detection rate for PGES in both scalp and SubQ datasets. However, the predicted durations had an average offset of 35.67 s for scalp EEG and 26.42 s for SubQ data. The model’s false positive rate (FPR) was 59% for scalp EEG and 56% for SubQ data, indicating a need for further refinement to reduce false alarms. This study demonstrates the potential of subcutaneous EEG as a valuable tool in the study of epilepsy and the monitoring of PGES, ultimately contributing to a better understanding and management of SUDEP risk.

OBJECTIVE:Novel subcutaneous electroencephalography (sqEEG) systems enable prolonged, near-continuous cerebral monitoring in real-world conditions. Nevertheless, the feasibility, acceptability and overall clinical utility of these systems remain unclear. We report on the longest observational study using ultra-long-term sqEEG to date. METHODS:We conducted a 15-month prospective, observational study including 10 adult people with treatment-resistant epilepsy. After device implantation, patients were asked to record sqEEG, to use an electronic seizure diary, and to complete acceptability and usability questionnaires. sqEEG seizures were annotated visually, aided by automated detection. Individualized temporal patterns of seizure occurrence were assessed via circadian circular statistics and via Fano factor analysis. RESULTS:Over a median duration of 438 days, 10 patients recorded a median 18.8 h/day, totaling 71 984 h of real-world sqEEG data. Adherence and acceptability remained high throughout the study. Although 754 sqEEG seizures were recorded across patients, more than half (52%) of these were not reported in the patient diary. Of the 140 (27%) diary reports not associated with an identifiable sqEEG seizure, the majority (68%) were reported as seizures with preserved awareness. The sqEEG to diary F1 agreement score was highly variable, ranging from .06 to .97. Patient-specific patterns of circadian seizure occurrence and seizure clustering were found, including several relevant discrepancies between sqEEG and diary. SIGNIFICANCE:We demonstrate feasibility and high acceptability of ultra-long-term (months-years) sqEEG monitoring. These systems help provide real-world, more objective seizure counting compared to patient diaries. It is possible to objectively monitor individual temporal fluctuations of seizure occurrence.

OBJECTIVE:This study reports development and validation of a deep learning seizure detection algorithm for two-channel subscalp electroencephalographic (EEG) recordings. Ultra-long-term monitoring of people with epilepsy may produce new insights into the timing and pattern of their seizures and may pave the way for novel therapeutic options. Although EEG is the accepted standard for epilepsy monitoring, ultra-long-term EEG recordings generate a massive quantity of data and are not reasonably reviewable in full by human readers. METHODS:The convolutional neural network-bidirectional long short-term memory (CNN-BiLSTM) hybrid algorithm uses nine layers, operating on channel spectrograms 5 min in length with 50% overlap. Retrospective subscalp EEG data from 16 patients recorded at three centers were available for algorithm development and testing. EEG was recorded for a median of 63 days (range = 9-508), and a median of 17 seizures (range = 0-96) were recorded. Training data were augmented with scalp EEG seizures, and performance was benchmarked against a conventional spectral power classifier algorithm. We also evaluated an intrapatient training-testing approach where the algorithm was trained on the first 45% of data. RESULTS:The CNN-BiLSTM trained on scalp EEG and subscalp EEG achieved an area under the receiver operating characteristic curve (AUROC) of .98 and an area under the precision-recall curve (AUPRC) of .50, which correspond to 94% sensitivity with 1.11 false detections per day. The same model trained on iEEG achieved only AUROC = .94 and AUPRC = .36. The conventional spectral band power detector achieved AUROC = .93 and AUPRC = .38. The CNN-BiLSTM detector trained on the earliest half of the subscalp EEG data achieved AUROC = .93 and AUPRC = .37, corresponding to 87% sensitivity and 5.9 false detections per day. SIGNIFICANCE:High sensitivity and specificity are possible in automated seizure detection in two-channel subscalp EEG data using a CNN-BiLSTM framework. Performance of the detector is superior using subcutaneous EEG data for training rather than intracranial EEG, but addition of scalp EEG seizures for training was helpful.

Zero-phase-delay synchrony between the activity of distant neural populations has been robustly observed. Nevertheless, contemporary electroencephalography and magnetencephalography functional connectivity analyses typically exclude zero-phase-delay functional connections, assuming that they are predominantly artefactual. However, the effects of excluding them on the performance of functional connectivity metrics as potential biomarkers are unknown. Here, we showed that most cortico-cortical functional connections occur with zero- or near-zero phase-delay, even where such connectivity was unlikely to be artefactual. Including, rather than excluding, zero-phase-delay connectivity increased the reliability, concordance with structural connectivity, and predictive validity (for longitudinal changes in cognition) of functional connectivity metrics. We found that excluding zero-phase-delay connections penalised functional connectivity strength between the strongest structurally connected regions: stronger structural connections led to functional connections with phase-delays closer to zero, mediated by a shorter signal propagation time. Our findings challenge generally accepted assumptions that zero-phase-exclusive methods are superior to zero-phase-inclusive methods. ### Competing Interest Statement The authors have declared no competing interest.

OBJECTIVE:Seizure unpredictability can be debilitating and dangerous for people with epilepsy. Accurate seizure forecasters could improve quality of life for those with epilepsy but must be practical for long-term use. This study presents the first validation of a seizure-forecasting system using ultra-long-term, non-invasive wearable data. METHODS:Eleven participants with epilepsy were recruited for continuous monitoring, capturing heart rate and step count via wrist-worn devices and seizures via electroencephalography (average recording duration of 337 days). Two hybrid models-combining machine learning and cycle-based methods-were proposed to forecast seizures at both short (minutes) and long (up to 44 days) horizons. RESULTS:The Seizure Warning System (SWS), designed for forecasting near-term seizures, and the Seizure Risk System (SRS), designed for forecasting long-term risk, both outperformed traditional models. In addition, the SRS reduced high-risk time by 29% while increasing sensitivity by 11%. SIGNIFICANCE:These improvements mark a significant advancement in making seizure forecasting more practical and effective.

OBJECTIVE:Epilepsy is a common condition associated with significant morbidity, mortality, and costs. Poor documentation of seizures is a major challenge in epilepsy care. Objective seizure counting with mobile devices may mitigate this challenge and improve patient management. The aim of this study is to investigate whether ultra long-term subcutaneous EEG improves seizure documentation and disease monitoring in people with drug-resistant epilepsy. METHODS:Real World Testing and Cost-effectiveness Analysis of Subcutaneous EEG (REAL-ASE) is a UK-based multi-centre prospective interventional study with an expected duration of 7 months. Thirty-three adult participants will be implanted with 24/7 EEG SubQ and collect 2-channel EEG data for 6 months. Data will be reviewed and annotated by experts, and a summary sent to the treating clinician weekly. The treating clinician will communicate information from this annotation report to participants monthly. Changes in management can be made at the treating clinician's discretion. REAL-ASE was approved by the Bromley Research Ethics Committee in July 2023 (reference number 23/LO/0419). RESULTS:We anticipate that subcutaneous EEG may improve seizure documentation and be a well-accepted addition to epilepsy care by clinicians and patients. We believe it will be associated with better quality of life and will be a cost-effective solution. SIGNIFICANCE:If our study demonstrates improved seizure documentation with subcutaneous EEG, this may contribute to patient safety, improved quality of life, and a reduction in healthcare costs. PLAIN LANGUAGE SUMMARY:We present a clinical trial protocol for a prospective cohort study of 33 people with epilepsy across the UK. The study aims to assess whether an EEG implant placed under the skin: (1) is more accurate than patient-reported seizure diary, (2) is feasible and acceptable to patients and clinicians, (3) reduces the impact of epilepsy, and (4) is beneficial to the National Health Service (NHS).

The National Association of Epilepsy Centers first published the guidelines for epilepsy centers in 1990, which were last updated in 2010. Since that update, epilepsy care and the science of guideline development have advanced significantly, including the importance of incorporating a diversity of stakeholder perspectives such as those of patients and their caregivers. Currently, despite extensive published data examining the efficacy of treatments and diagnostic testing for epilepsy, there remain significant gaps in data identifying the essential services needed for a comprehensive epilepsy center and the optimal manner for their delivery. The trustworthy consensus-based statements (TCBS) process produces unbiased, scientifically valid guidelines through a transparent process that incorporates available evidence and expert opinion. A systematic literature search returned 5937 relevant studies from which 197 articles were retained for data extraction. A panel of 41 stakeholders with diverse expertise evaluated this evidence and drafted recommendations following the TCBS process. The panel reached consensus on 52 recommendations covering services provided by specialized epilepsy centers in both the inpatient and outpatient settings in major topic areas including epilepsy monitoring unit care, surgery, neuroimaging, neuropsychology, genetics, and outpatient care. Recommendations were informed by the evidence review and reflect the consensus of a broad panel of expert opinions.

The National Association of Epilepsy Centers first published the guidelines for epilepsy centers in 1990, which were last updated in 2010. Since that update, epilepsy care and the science of guideline development have advanced significantly, including the importance of incorporating a diversity of stakeholder perspectives such as those of patients and their caregivers. Currently, despite extensive published data examining the efficacy of treatments and diagnostic testing for epilepsy, there remain significant gaps in data identifying the essential services needed for a comprehensive epilepsy center and the optimal manner for their delivery. The trustworthy consensus-based statements (TCBS) process produces unbiased, scientifically valid guidelines through a transparent process that incorporates available evidence and expert opinion. A systematic literature search returned 5937 relevant studies from which 197 articles were retained for data extraction. A panel of 41 stakeholders with diverse expertise evaluated this evidence and drafted recommendations following the TCBS process. The panel reached consensus on 52 recommendations covering services provided by specialized epilepsy centers in both the inpatient and outpatient settings in major topic areas including epilepsy monitoring unit care, surgery, neuroimaging, neuropsychology, genetics, and outpatient care. Recommendations were informed by the evidence review and reflect the consensus of a broad panel of expert opinions.

Objective. This paper aims to investigate the possibility of detecting tonic-clonic seizures (TCSs) with behind-the-ear, two-channel wearable electroencephalography (EEG), and to evaluate its added value to non-EEG modalities in TCS detection.Methods. We included 27 participants with a total of 44 TCSs from the European multicenter study SeizeIT2. The wearable Sensor Dot (Byteflies) was used to measure behind-the-ear EEG, electromyography (EMG), electrocardiography, accelerometry (ACC) and gyroscope. We evaluated automatic unimodal detection of TCSs, using sensitivity, precision, false positive rate (FPR) and F1-score. Subsequently, we fused the different modalities and again assessed performance. Algorithm-labeled segments were then provided to two experts, who annotated true positive TCSs, and discarded false positives.Results. Wearable EEG outperformed the other single modalities with a sensitivity of 100% and a FPR of 10.3/24 h. The combination of wearable EEG and EMG proved most clinically useful, delivering a sensitivity of 97.7%, an FPR of 0.4/24 h, a precision of 43%, and an F1-score of 59.7%. The highest overall performance was achieved through the fusion of wearable EEG, EMG, and ACC, yielding a sensitivity of 90.9%, an FPR of 0.1/24 h, a precision of 75.5%, and an F1-score of 82.5%.Conclusions. In TCS detection with a wearable device, combining EEG with EMG, ACC or both resulted in a remarkable reduction of FPR, while retaining a high sensitivity.Significance. Adding wearable EEG could further improve TCS detection, relative to extracerebral-based systems.

BackgroundA new class of subcutaneous electroencephalography has enabled ultra long-term monitoring of people with epilepsy. The objective of this paper is to describe surgeons' experiences in an early series of implantations as well as discomfort or complications experienced by the participants.MethodsWe included 38 implantation procedures from two trials on people with epilepsy and healthy adults. Questionnaires to assess surgeons' and participants' experience were analyzed as well as all recorded adverse events occurring up to 21 days post-surgery.ResultsWith training, the implantation could be performed in approximately 15 min. Overall, the implantation procedure was considered easy to perform with only 2 episodes where the implant got fixated in the introducing needle and a new implant had to be used. The explantation procedure was considered effortless. In 2 cases the silicone sheath covering the lead was damaged during the explantation, but it was possible to remove the entire implant without leaving any foreign body under the skin. Especially in the trial on healthy participants, a proportion experienced adverse events in the form of headache or implant-pain up to 21 days post-operatively. In 6 cases, adverse events contributed to the decision to explant and discontinue the study: Four of these cases involved implant pain or headache; One case involved a post-operative local infection; and in one case superficial lead placement resulted in skin perforation a few weeks after implantation.ConclusionThe implantation and explantation procedures are considered swift and easy to perform by both neurosurgeons and ENT surgeons. The implant is well tolerated by most participants. However, headache or pain around the implant can occur for up to 21 days post-operatively as anticipated with any such surgery. The expected benefits from the implant should always outweigh the potential disadvantages.

Neuronal activity propagates through the network during seizures, engaging brain dynamics at multiple scales. Such propagating events can be described through the avalanches framework, which can relate spatiotemporal activity at the microscale with global network properties. Interestingly, propagating avalanches in healthy networks are indicative of critical dynamics, where the network is organized to a phase transition, which optimizes certain computational properties. Some have hypothesized that the pathologic brain dynamics of epileptic seizures are an emergent property of microscale neuronal networks collectively driving the brain away from criticality. Demonstrating this would provide a unifying mechanism linking microscale spatiotemporal activity with emergent brain dysfunction during seizures. Here, we investigated the effect of drug-induced seizures on critical avalanche dynamics, using in vivo whole-brain two-photon imaging of GCaMP6s larval zebrafish (males and females) at single neuron resolution. We demonstrate that single neuron activity across the whole brain exhibits a loss of critical statistics during seizures, suggesting that microscale activity collectively drives macroscale dynamics away from criticality. We also construct spiking network models at the scale of the larval zebrafish brain, to demonstrate that only densely connected networks can drive brain-wide seizure dynamics away from criticality. Importantly, such dense networks also disrupt the optimal computational capacities of critical networks, leading to chaotic dynamics, impaired network response properties and sticky states, thus helping to explain functional impairments during seizures. This study bridges the gap between microscale neuronal activity and emergent macroscale dynamics and cognitive dysfunction during seizures.SIGNIFICANCE STATEMENT Epileptic seizures are debilitating and impair normal brain function. It is unclear how the coordinated behavior of neurons collectively impairs brain function during seizures. To investigate this we perform fluorescence microscopy in larval zebrafish, which allows for the recording of whole-brain activity at single-neuron resolution. Using techniques from physics, we show that neuronal activity during seizures drives the brain away from criticality, a regime that enables both high and low activity states, into an inflexible regime that drives high activity states. Importantly, this change is caused by more connections in the network, which we show disrupts the ability of the brain to respond appropriately to its environment. Therefore, we identify key neuronal network mechanisms driving seizures and concurrent cognitive dysfunction.

Background Interoception describes the predictive representation and control of the internal physiological state of the body. Disturbances in interoception have been demonstrated in people with functional seizures. Little is known about interoception in epilepsy. Aim To compare dimensions of interoception across three groups - people with functional non-epileptic seizures (FS), people with epileptic seizures (ES) and healthy controls (HC). We hypothesised that people with ES would display deficits in interoception compared to HC, given the neuroanatomical correlates of interoception are frequently implicated in epilepsy. Methods We included 37 participants with FS, 48 with ES and 34 HC. All participants completed measures of (1) interoceptive accuracy (IA) – objective performance on a heartbeat tracking task (HTT) and heartbeat discrimination task (HDT); (2) trait interoceptive sensibility – subjective sensitivity to internal sensations (measured using the Porges Body Perception Questionnaire) and (3) state interoceptive sensibility – subjective trial-by-trial measures of confidence in heartbeat detection. These were used to calculate interoceptive awareness (metacognitive measure). All completed self-report ratings of dissociation (MDI, SDQ), depression (BDI), and anxiety (BAI). Results Across the groups, there was a statistically significant difference in IA measured through the HTT (χ 2 (2)=12.474, p =0.002) with the highest mean score demonstrated in HC (&xbar;=76.76), followed by ES (&xbar;=54.29) and then FS groups (&xbar;=50.14). Pairwise comparisons revealed significant differences in IA when comparing HC to both ES ( p =0.010) and FS ( p =0.003), but not between ES and FS groups ( p =0.582). Interoceptive sensibility also differed significantly between groups (F (2, 116)=11.769, p <0.001) and was highest in FS, followed by ES, then HC but did not differ between ES and FS groups ( p =0.969). Significant differences in HDT awareness were found between groups (F=5.182, p =0.007). Pairwise comparisons revealed significant differences between ES and FS ( p =0.005). Mean HDT awareness for FS (&xbar;=0.495) was lower than for ES (&xbar;=0.598) and HC (&xbar;=0.548). People with FS displayed the highest levels of dissociation, depression, and anxiety with dissociation and depression scores being significantly higher than those with ES ( p <0.001 and p =0.030 respectively). Conclusions Both ES and FS groups demonstrated deficits in interoception (reduced IA and increased interoceptive sensibility) compared to HC. Interoceptive awareness on the HDT significantly differed between ES and FS groups. Whilst interoceptive deficits in people with ES could be related to the effects of epilepsy on interoceptive networks, interoceptive awareness in ES appears to be higher compared to those with FS. The role of interoception merits further study in these patient populations.

A lot of mileage has been made recently on the long and winding road toward seizure forecasting. Here we briefly review some selected milestones passed along the way, which were discussed at the International Conference for Technology and Analysis of Seizures-ICTALS 2022-convened at the University of Bern, Switzerland. Major impetus was gained recently from wearable and implantable devices that record not only electroencephalography, but also data on motor behavior, acoustic signals, and various signals of the autonomic nervous system. This multimodal monitoring can be performed for ultralong timescales covering months or years. Accordingly, features and metrics extracted from these data now assess seizure dynamics with a greater degree of completeness. Most prominently, this has allowed the confirmation of the long-suspected cyclical nature of interictal epileptiform activity, seizure risk, and seizures. The timescales cover daily, multi-day, and yearly cycles. Progress has also been fueled by approaches originating from the interdisciplinary field of network science. Considering epilepsy as a large-scale network disorder yielded novel perspectives on the pre-ictal dynamics of the evolving epileptic brain. In addition to discrete predictions that a seizure will take place in a specified prediction horizon, the community broadened the scope to probabilistic forecasts of a seizure risk evolving continuously in time. This shift of gears triggered the incorporation of additional metrics to quantify the performance of forecasting algorithms, which should be compared to the chance performance of constrained stochastic null models. An imminent task of utmost importance is to find optimal ways to communicate the output of seizure-forecasting algorithms to patients, caretakers, and clinicians, so that they can have socioeconomic impact and improve patients' well-being.

Abnormal EEG features are a hallmark of epilepsy, and abnormal frequency and network features are apparent in EEGs from people with idiopathic generalised epilepsy in both ictal and interictal states. Here, we characterise differences in the resting-state EEG of individuals with juvenile myoclonic epilepsy (JME) and assess factors influencing the heterogeneity of these EEG features. We collected EEG data from 147 participants with JME through the Biology of Juvenile Myoclonic Epilepsy (BIOJUME) study. 95 control EEGs were acquired from two independent studies (Chowdhury et al. (2014) and EU-AIMS Longitudinal European Autism Project). We extracted frequency and functional network-based features from 10-20s epochs of resting-state EEG, including relative power spectral density (PSD), peak alpha frequency, network topology measures and Brain Network Ictogenicity (BNI): a computational measure of the propensity of networks to generate seizure dynamics. The influence of covariates such as age, sex, antiseizure medication, EEG time and epoch length were investigated for each EEG feature prior to testing for differences between JME and control EEGs using univariate, multivariable and receiver operating curve (ROC) analysis. Additionally, associations of clinical phenotypes (seizure type, seizure control) with EEG features were investigated in the JME cohort. P-values were corrected for multiple comparisons. Univariate analysis showed significant differences in PSD in delta (2-5Hz) (p=0.0007, hedges' g=0.55) and low-alpha (6-9Hz) (p=2.9x10-8, g=0.80) frequency bands, peak alpha frequency (p=0.000007, g=0.66), functional network mean degree (p=0.0006, g=0.48) and BNI (p=0.00006, g=0.56) between JME and controls. Since age (p=0.009) and epoch length (p=1.7x10-8) differed between the two groups and were potential confounders, we controlled for these covariates in multivariable analysis where disparities in EEG features between JME and controls remained. ROC analysis showed low-alpha PSD was optimal at distinguishing JME from controls, with an area under the curve of 0.72. Lower average normalized clustering coefficient and shorter average normalized path length were associated with poorer seizure control in JME patients. To conclude, individuals with JME have increased power of neural oscillatory activity at low-alpha frequencies, along with increased BNI compared to controls, supporting evidence from studies in other epilepsies with considerable external validity. In addition, the impact of confounders on different frequency-based and network-based EEG features observed in this study highlights the need for careful consideration and control of these factors in future EEG research in IGE particularly for their use as biomarkers.