Search

The current pace of development and applications of large language models (LLMs) is unprecedented and will impact future medical care significantly. In this critical review, we provide the background to better understand these novel artificial intelligence (AI) models and how LLMs can be of future use in the daily care of people with epilepsy. Considering the importance of clinical history taking in diagnosing and monitoring epilepsy-combined with the established use of electronic health records-a great potential exists to integrate LLMs in epilepsy care. We present the current available LLM studies in epilepsy. Furthermore, we highlight and compare the most commonly used LLMs and elaborate on how these models can be applied in epilepsy. We further discuss important drawbacks and risks of LLMs, and we provide recommendations for overcoming these limitations., (© 2024 The Authors. Epilepsia published by Wiley Periodicals LLC on behalf of International League Against Epilepsy.)

Objective: In people with low-grade intrinsic brain tumors, an epileptic focus is often located close to the lesion. High-frequency oscillations (HFOs) in electrocorticography (ECoG) might help to delineate this focus. We investigated the relationship between HFOs and low-grade brain tumors and their potential value for tumor-related epilepsy surgery., Methods: We analyzed pre- and postresection intraoperative ECoG in 41 patients with refractory epilepsy and a low-grade lesion. Electrodes were designated as overlying the tumor, adjacent resected tissue (peritumoral), or outside the resection bed using magnetic resonance imaging (MRI) and intraoperative photographs. We then used a semiautomated approach to detect HFOs as either ripples (80-250 Hz) or fast ripples (250-500 Hz)., Results: The rate of fast ripples was higher in electrodes covering tumor and peritumoral tissue than outside the resection (p = .04). Mesiotemporal tumors showed more ripples (p = .002), but not more fast ripples (p = .07), than superficial tumors. Rates of fast ripples were higher in glioma and extraventricular neurocytoma than in ganglioglioma or dysembryoplastic neuroepithelial tumor (DNET). The rate of ripples and fast ripples in postresection ECoG was not higher in patients with residual tumor tissue on MRI than those without. The rate of ripples in postresection ECoG was higher in patients with good than bad seizure outcome (p = .03). Fast ripples outside the resection and in post-ECoG seem related to seizure recurrence., Significance: Fast ripples in intraoperative ECoG can be used to help guide resection in tumor-related epilepsy surgery. Preresection fast ripples occur predominantly in epileptogenic tumor and peritumoral tissue. Fast ripple rates are higher in glioma and extraventricular neurocytoma than in ganglioglioma and DNET., (© 2021 The Authors. Epilepsia published by Wiley Periodicals LLC on behalf of International League Against Epilepsy.)

Objective: New insights into high-frequency electroencephalographic activity and network analysis provide potential tools to improve delineation of epileptic tissue and increase the chance of postoperative seizure freedom. Based on our observation of high-frequency oscillations "spreading outward" from the epileptic source, we hypothesize that measures of directed connectivity in the high-frequency range distinguish epileptic from healthy brain tissue., Methods: We retrospectively selected refractory epilepsy patients with a malformation of cortical development or tumor World Health Organization grade I/II who underwent epilepsy surgery with intraoperative electrocorticography for tailoring the resection based on spikes. We assessed directed functional connectivity in the theta (4-8 Hz), gamma (30-80 Hz), ripple (80-250 Hz), and fast ripple (FR; 250-500 Hz) bands using the short-time direct directed transfer function, and calculated the total, incoming, and outgoing propagation strength for each electrode. We compared network measures of electrodes covering the resected and nonresected areas separately for patients with good and poor outcome, and of electrodes with and without spikes, ripples, and FRs (group level: paired t test; patient level: Mann-Whitney U test). We selected the measure that could best identify the resected area and channels with epileptic events using the area under the receiver operating characteristic curve, and calculated the positive and negative predictive value, sensitivity, and specificity., Results: We found higher total and outstrength in the ripple and gamma bands in resected tissue in patients with good outcome (ripple total : P = .01; ripple out : P = .04; gamma total : P = .01; gamma out : P = .01). Channels with events showed lower total and instrength, and higher outstrength in the FR band, and higher total and outstrength in the ripple, gamma, and theta bands (FR total : P = .05; FR in : P < .01; FR out : P = .02; gamma total : P < .01; gamma in : P = .01; gamma out : P < .01; theta total : P = .01; theta out : P = .01). The total strength in the gamma band was most distinctive at the channel level (positive predictive value [PPV] good  = 74%, PPV poor  = 43%)., Significance: Interictally, epileptic tissue is isolated in the FR band and acts as a driver up to the (fast) ripple frequency range. The gamma band total strength seems promising to delineate epileptic tissue intraoperatively., (© 2019 The Authors. Epilepsia published by Wiley Periodicals, Inc. on behalf of International League Against Epilepsy.)

Objective: The most common reported seizure-precipitant is stress. We recently showed a biologic basis for stress sensitivity of seizures: cortisol levels in people with stress-sensitive epilepsy correlated with focal interictal epileptiform discharges (IEDs) on electroencephalography (EEG). Here we aimed to determine whether the effect of cortisol on the epileptic brain is global or focal, and whether cortisol affects all brains or just those of stress-sensitive people. Because epilepsy is associated with changes in functional brain connectivity, we studied the relationship between cortisol and changes in global and focal (node-centered) functional connectivity measures for individuals with stress-sensitive and non-stress-sensitive epilepsy., Methods: Seventeen people with epilepsy underwent long-term (>24 h) EEG recording. During the first 5 h after waking, saliva was collected every 15 min for cortisol measurements. Theta-band functional connectivity was assessed for every 15 min of the recording. We calculated the average phase-lag index (PLI) between all channels as a measure of global functional connectivity. We used network Strength, the averaged PLI per channel, as focal functional connectivity measure. We correlated cortisol, global, and focal functional connectivity (Strength) with IED frequency using linear mixed models. Analyses were split for people with and without stress-sensitivity of seizures., Results: Cortisol was negatively correlated with global functional connectivity in people with stress-sensitive seizures (estimate -0.0020; P < .01), whereas not in those without stress-sensitivity (estimate -0.0003; P = .46). This relationship occurred irrespective of the presence of IEDs on a channel (channels without IEDs and stress-sensitivity: estimate -0.0019; P < .01, non-stress-sensitive -0.0003; P = .41). Global and focal functional connectivity were negatively correlated with IED frequency, irrespective of stress sensitivity of seizures or channel type., Significance: People with stress-sensitive epilepsy have a whole-brain neuronal response to cortisol that is different from that of people with non-stress-sensitive epilepsy. This offers a basis for understanding seizure genesis in stress-sensitive epilepsy, which might require a different treatment approach., (Wiley Periodicals, Inc. © 2017 International League Against Epilepsy.)

Objective: Technology for localizing epileptogenic brain regions plays a central role in surgical planning. Recent improvements in acquisition and electrode technology have revealed that high-frequency oscillations (HFOs) within the 80-500 Hz frequency range provide the neurophysiologist with new information about the extent of the epileptogenic tissue in addition to ictal and interictal lower frequency events. Nevertheless, two decades after their discovery there remain questions about HFOs as biomarkers of epileptogenic brain and there use in clinical practice., Methods: In this review, we provide practical, technical guidance for epileptologists and clinical researchers on recording, evaluation, and interpretation of ripples, fast ripples, and very high-frequency oscillations., Results: We emphasize the importance of low noise recording to minimize artifacts. HFO analysis, either visual or with automatic detection methods, of high fidelity recordings can still be challenging because of various artifacts including muscle, movement, and filtering. Magnetoencephalography and intracranial electroencephalography (iEEG) recordings are subject to the same artifacts., Significance: High-frequency oscillations are promising new biomarkers in epilepsy. This review provides interested researchers and clinicians with a review of current state of the art of recording and identification and potential challenges to clinical translation., (Wiley Periodicals, Inc. © 2017 International League Against Epilepsy.)

Objective: Children with rolandic spikes may or may not have seizures, ranging from benign rolandic epilepsy to severe atypical rolandic epilepsy. We investigated whether ripples (80-250 Hz), superimposed on rolandic spikes in surface electroencephalography (EEG), can differentiate between different entities., Methods: In this cohort study we analyzed the EEG studies of children with rolandic spikes without other EEG or magnetic resonance imaging (MRI) abnormalities. They were divided into the following three groups: (1) rolandic spikes but no epilepsy, (2) typical rolandic epilepsy, and (3) atypical and symptomatic rolandic epilepsy. Ripples superimposed on rolandic spikes were marked in 10 minutes of EEG, and compared to the number of seizures before the EEG. Receiver operating characteristic (ROC) curves were constructed to determine the predictive value of ripples and spikes for having epilepsy (groups 2 and 3) and for differentiating benign courses (groups 1 or 2) from atypical and symptomatic epilepsy (group 3). Ripples were also marked in the time frequency spectrum of averaged rolandic spikes., Results: Ripples were found in 13 of 22 children. Children without epilepsy showed no ripples, except for a single child with only one ripple. The number of ripples showed a significant positive correlation with the number of seizures (ρ = 0.70, p = 0.001), whereas spikes had a borderline significant correlation (ρ = 0.43, p = 0.05). Presence of more than two ripples was a predictor for having seizures (area under the curve [AUC] 0.84), whereas spikes could not predict having seizures (AUC 0.53). More than five ripples predicted the difference between benign courses and atypical and symptomatic epilepsy (AUC 0.91, sensitivity 63%, specificity 100%). Ripples in the time frequency spectra appeared in all children and were not related to seizures., Significance: Absence of ripples on top of rolandic spikes predicts a relative benign clinical entity, whereas in the presence of several ripples, the child is likely to have more seizures than classical rolandic epilepsy, and pharmacologic treatment might be needed., (Wiley Periodicals, Inc. © 2016 International League Against Epilepsy.)

Purpose: Epileptic high-frequency oscillations (HFOs; 80-500 Hz) may be used to guide neurosurgeons during epilepsy surgery to identify epileptogenic tissue. We studied the effect of the anesthetic agent propofol on the occurrence of HFOs in intraoperative electrocorticography (ECoG)., Methods: We selected patients who were undergoing surgery for temporal lobe epilepsy with a standardized electrode grid placement. Intraoperative ECoG was recorded at 2,048 Hz following cessation of propofol. The number and distribution of interictal spikes, ripples (R [80-250 Hz]), and fast ripples (FRs; 250-500 Hz) were analyzed. The amount of events on mesiotemporal channels and lateral neocortical channels were compared between patients with a suspected mesiotemporal and lateral epileptogenic area (Student's t-test), and HFOs were compared with the irritative zone, using correlation between amounts of events per channel, to provide evidence for the epileptic nature of the HFOs. Next, the amount of events within the first minute and the last minute were compared to each other and the change in events over the entire epochs was analyzed using correlation analyses of 10 epochs during the emergence periods (Spearman rank test). We studied whether the duration of HFOs changed over time. The change in events within presumed epileptogenic area was compared to the change outside this area (Student's t-test). Periods of burst suppression and continuous background activity were compared between and within patients (t-test)., Key Findings: Twelve patients were included: five with suspected mesiotemporal epileptogenic area and three with suspected lateral epileptogenic area (and four were "other"). Spikes, ripples, and FRs were related to the suspected epileptogenic areas, and HFO zones were related to the irritative zones. Ripples and FRs increased during emergence from propofol anesthesia (mean number of ripples from first minute-last minute: 61.5-73.0, R = 0.46, p < 0.01; FRs: 3.1-5.7, R = 0.30, p < 0.01) and spikes remained unchanged (80.1-79.9, R = -0.05, p = 0.59). There was a decrease in number of channels with spikes (R = -0.18, p = 0.05), but no change in ripples (R = -0.13, p = 0.16) or FRs (R = 0.11, p = 0.45). There was no change in the durations of HFOs. The amount of HFOs in the presumed epileptogenic areas did not change more than the amount outside the presumed epileptogenic area, whereas spikes paradoxically decreased more within the suspected epileptogenic area. Six patients showing burst-suppression had lower rates of ripples than six other patients with continuous background activity (p = 0.02). No significant difference was found between burst suppression and continuous background activity in four patients, but there was a trend toward showing more ripples during continuous background activity (p = 0.16)., Significance: Propofol, known for its antiepileptic effects, reduces the number of epileptic HFOs, but has no effect on spikes. This enforces the hypothesis that, in epilepsy, HFOs mirror the disease activity and HFOs might be useful for monitoring antiepileptic drug treatment. It is feasible to record HFOs during surgery, but propofol infusion should be interrupted for some minutes to improve detection., (Wiley Periodicals, Inc. © 2012 International League Against Epilepsy.)

Purpose: Electrical stimulation (ES) is used during intracranial electroencephalography (EEG) investigations to delineate epileptogenic areas and seizure-onset zones (SOZs) by provoking afterdischarges (ADs) or patients' typical seizure. High frequency oscillations (HFOs--ripples, 80-250 Hz; fast ripples, 250-500 Hz) are linked to seizure onset. This study investigates whether interictal HFOs are more frequent in areas with a low threshold to provoke ADs or seizures., Methods: Intracranial EEG studies were filtered at 500 Hz and sampled at 2,000 Hz. HFOs were visually identified. Twenty patients underwent ES, with gradually increasing currents. Results were interpreted as agreeing or disagreeing with the intracranial study (clinical-EEG seizure onset defined the SOZ). Current thresholds provoking an AD or seizure were correlated with the rate of HFOs of each channel., Results: ES provoked a seizure in 12 and ADs in 19 patients. Sixteen patients showed an ES response inside the SOZ, and 10 had additional areas with ADs. The response was more specific for mesiotemporal than for neocortical channels. HFO rates were negatively correlated with thresholds for ES responses; especially in neocortical regions; areas with low threshold and high HFO rate were colocalized even outside the SOZ., Discussion: Areas showing epileptic HFOs colocalize with those reacting to ES. HFOs may represent a pathologic correlate of regions showing an ES response; both phenomena suggest a more widespread epileptogenicity.

Purpose: To determine the prevalence of heart rate changes and ECG abnormalities during epileptic seizures and to determine the timing of heart rate changes compared to the first electrographic and clinical signs. To assess the risk factors for the occurrence of ECG abnormalities., Methods: We analyzed retrospectively 281 seizures in 81 patients with intractable epilepsy who had prolonged video-EEG and two-channel ECG. The nature and timing of heart rate changes compared to the electrographic and clinical seizure onset was determined. The ictal period (including one minute preictally and three minutes postictally) was analyzed for cardiac arrhythmias, conduction and repolarization abnormalities. Risk factors for cardiac abnormalities were investigated using parametric and non-parametric statistics., Results: There was an increase in heart rate of at least 10 beats/minute in 73% of seizures (93% of patients) and this occurred most often around seizure onset. In 23% of seizures (49% of patients) the rate increase preceded both the electrographic and the clinical onset. ECG abnormalities were found in 26% of seizures (44% of patients). One patient had an asystole for 30 seconds. Long seizure duration increased the occurrence of ECG abnormalities. No other risk factor was found., Conclusions: Heart rate changes occur frequently and occur around the time or even before the earliest electrographic or clinical change. The change can clarify the timing of seizure onset and the specific rate pattern may be useful for seizure diagnosis and for automatic seizure detection. ECG abnormalities occur often and repeatedly in several seizures of the same patient.