Human and animal EEG data suggest that epileptic seizures are not stationary events, but rather evolve dynamically over tens of seconds. We explore the processes linked to seizure dynamics by enhancing the Hodgkin-Huxley mathematical model with the physical laws governing ion movement. This enhancement enables us to replicate in silico the electrographic pattern of a typical human focal seizure, which consists of distinct phases: the onset of low-voltage fast activity, the tonic phase, the clonic phase, and the postictal suppression phase. Our study provides new insights into potential mechanisms of seizure initiation by inhibitory interneurons through K+ accumulation, as well as seizure termination and the postictal state via upregulation of the outward Na+/K+ pump current. The model also clarifies ionic mechanisms that may underlie a key feature of seizure dynamics: the progressive slowing of ictal discharges. Model predictions regarding the specific scaling of inter-burst intervals are validated in both in vitro and human seizure data, suggesting these mechanisms may be preserved across different models and species.

Neuroplasticity refers to functional and structural changes in brain regions in response to healthy and pathological activity. Activity dependent plasticity induced by epileptic activity can involve healthy brain regions into the epileptogenic network by perturbing their excitation/inhibition balance. In this article, we present a new neural mass model, which accounts for neuroplasticity, for investigating the possible mechanisms underlying the epileptogenic network expansion. Our multiple-timescale model is inspired by physiological calcium-mediated synaptic plasticity and pathological extrasynaptic N- methyl-D-aspartate (NMDA) dependent plasticity dynamics. The model highlights that synaptic plasticity at excitatory connections and structural changes in the inhibitory system can transform a healthy region into a secondary epileptic focus under recurrent seizures and interictal activity occurring in the primary focus. Our results suggest that the latent period of this transformation can provide a window of opportunity to prevent the expansion of epileptogenic networks, formation of an epileptic focus, or other comorbidities associated with epileptic activity.

Reference: Köksal-Ersöz E, Benquet P, Wendling F (2024) Expansion of epileptogenic networks via neuroplasticity in neural mass models. PLoS Comput Biol 20(12): e1012666. https://doi.org/10.1371/journal.pcbi.1012666

Dravet syndrome is a developmental and epileptic encephalopathy (DEE) that typically begins in the first year of life. This complex pathology is characterized by drug-resistant seizures, various comorbidi1es such as cognitive delay, and a risk of early death. Most cases are due to mutations of NaV1.1, a voltage-gated sodium channel expressed in fast-spiking (FS) inhibitory neurons. The pathological mechanism in the initial stage of the disease involves impaired function of those neurons, leading to network hyperexcitability. However, the details remain unclear.

Mutations of NaV1.1 may result in non-functional channels or channels with altered gating properties. We focus on the less studied case of altered gating, by investigating how it impairs neuronal ac1vity in the case of a specific mutation (A1783V). Using recordings in cell lines, Layer et al. (2021) showed that A1783V alters the voltage dependence of channel activation, as well as the voltage dependence and kinetics of slow inactivation. Slow inactivation is a mechanism dis1nct from the fast inactivation of sodium channels at each spike, developing much more slowly, during prolonged trains of depolariza1on. Implementing the three effects of the mutation in a conductance-based model, Layer et al. predict that altered activation has the largest impact on channel function, as it causes the most severe reduction in firing rate.

Using conductance-based models tailored to the dynamics of FS inhibitory neurons, we examine how the three alterations affect susceptibility to depolariza1on block, another firing deficit aside from frequency reduction. We look deeper into slow inactivation, exploiting the timescale difference with the rest of the system. We find that slow inactivation of mutant channels at lower voltage values than wild type channels favors depolarization block upon sustained stimulation. More precisely, shifting the steady-state voltage dependence of slow inactivation destroys the stable limit cycle of the full system corresponding to tonic spiking, and creates a stable equilibrium corresponding to depolarization block. The accelerated kinetics of slow inactivation in mutant channels hastens the transition from tonic spiking to depolarization block. These findings suggest that alterations of NaV1.1 slow inactivation should not be neglected as they might play an important pathological role, adding to the conclusions of Layer et al. on the consequences of altered NaV1.1 activation.

Epileptic seizures are transient occurrences of abnormal, excessive, or synchronous neuronal activity in the brain, representing a core manifestation of epilepsy. Moreover, seizures are highly specific to the individual in terms of where they occur, what brain networks they involve and how they propagate through the brain. To further understand the individualised mechanisms of seizures we have developed a simulation-based inference technique based on long-short-term-memory neural networks and neural mass models to infer neurophysiological population-level variables from intracranial electroencephalography (iEEG) recordings. Using long-term iEEG recordings from 15 patients and involving 2000+ seizures, the inference framework has been applied to reveal how stability and chaoticity of the brain networks during seizures varies across individuals and their seizures. While there were clear individual differences, many patients showed a consistent increase in stability based on the eigenspectrum just after seizure onset. On the other hand the chaoticity dynamics based on the Lyapunov spectrum were less consistent. The neurophysiological population-level variable estimates involved in this analysis were also employed to predict if a seizure will be short or long just based on the estimates during the first few seconds of a seizure. Moreover, the estimates were used to predict the occurrence of seizures through a seizure forecasting analysis applied to the long term recordings. Results and implications for the field will be discussed.

In the field of computational epilepsy, neural field models helped to understand some large-scale features of seizure dynamics. These insights however remain on general levels, without translation to the clinical settings via personalization of the model with the patient-specific structure. In particular, a link was suggested between epileptic seizures spreading across the cortical surface and the so-called theta-alpha activity (TAA) pattern seen on intracranial electrographic signals, yet this link was not demonstrated on a patient-specific level. I will present a single patient computational study linking the seizure spreading across the patient-specific cortical surface with a specific instance of the TAA pattern recorded in the patient. Using the realistic geometry of the cortical surface we performed the simulations of seizure dynamics in The Virtual Brain platform, and we show that the simulated electrographic signals qualitatively agree with the recorded signals. Furthermore, the comparison with the simulations performed on surrogate surfaces reveals that the best quantitative fit is obtained for the real surface. The work illustrates how the patient-specific cortical geometry can be utilized in The Virtual Brain for personalized model building, and the importance of such approach.

The biological mechanisms underlying the spontaneous and recurrent transition to seizures in the epileptic brain are still poorly understood. As a result, seizures remain uncontrolled in a substantial proportion of patients. Brain stimulation is an emerging and promising method to treat various brain disorders, including drug-refractory epilepsy. Selected stimulation protocols previously demonstrated therapeutic efficacy in reducing the seizure rate. The stimulation efficacy critically depends on chosen stimulation parameters, such as the time point, amplitude, and frequency of stimulation. This study aims to explore the neurobiological impact of 1Hz stimulation and provide the mechanistic explanation behind its seizure-delaying and seizure-suppressing effects. We study this effect using a computational model, a modified version of the Epileptor-2 model, in close comparison with such stimulation effects on spontaneous seizures recorded \textit{in vitro} in a high-potassium model of ictogenesis in rat hippocampal slices. In particular, we investigate the mechanisms and dynamics of spontaneous seizure emergence, the seizure-delaying effect of the stimulation, and the optimal stimulation parameters to achieve the maximal anti-seizure effect. We show that the modified Epileptor-2 model replicates key experimental observations, and captures seizure dynamics and the anti-seizure effects of low-frequency electrical stimulation (LFES) observed in hippocampal slices. We identify the critical thresholds in the model for seizure onset and determine the optimal stimulation parameters -- timing, amplitude, and duration -- that exceed specific thresholds to delay seizures without triggering premature seizures. Our study highlights the central role of sodium-potassium pump dynamics in terminating seizures and mediating the LFES effect.

Electrical stimulation is an increasingly popular method to terminate epileptic seizures; yet it is not always successful. One of the potential reasons for inconsistent efficacy is that stimuli are applied empirically, without considering the underlying dynamical properties of a given seizure. In this work, we find that different seizure types have vastly different responses to controlling stimuli. We use the Taxonomy of Seizure Dynamics to model different onset dynamotypes, then determine the ability of ictal stimulation to abort seizures after they have started. Within the model, the aborting input is realized as an applied stimulus trying to force the system from a bursting state to a quiescent or resting state. This transition requires bistability, which is not present in all onset dynamotypes. We examine how topological and geometric differences in bistable phase spaces affect the probability of termination as the burster progresses from onset to offset. We find that the most significant determining factors are (1) the presence or absence of a baseline (DC) shift and (2) the dynamotype (onset/offset bifurcations) of the burster. Generally, we find that bursters that have a DC shift are far more likely to be terminated than those without because they are not as sensitive to the phase at which stimulation occurs. Furthermore, we observe that the probability of termination varies throughout the burster’s duration and is highly correlated to its dynamotype. Our model provides a method to predict the optimal method of termination for each dynamotype. We conclude that strategies for aborting seizures with ictal stimulation must account for seizure dynamotype to optimize efficacy. 

Optogenetic inhibition of excitatory neuronal populations has emerged as a promising strategy for treating refractory epilepsy. However, achieving consistent seizure suppression in animal models using optogenetics has proven difficult. These challenges are largely due to suboptimal stimulation protocols, which involve numerous complex and interacting variables.

To better understand the impact of these parameters, a mathematical model was developed by fitting data from the chloride-conducting opsin GtACR2 and integrating it with a conductance-based model of a CA1 pyramidal neuron. This framework was further extended to include potassium-conducting opsins, allowing for comparative analysis of different inhibitory mechanisms. Simulations using this model show that the effectiveness of optogenetic modulation is highly dependent on stimulation parameters, the neuronal environment, and the opsin ion type.

Additionally, a simplified opsin model was introduced to facilitate the incorporation of experimentally derived parameters describing opsin dynamics. This model enables systematic studies of opsin properties, revealing that variations in kinetics have minimal effect on the overall outcome of optogenetic inhibition, provided that the factor of change to opening and closing time constants remains between 0.5 to 2. Moreover, the previously determined impact of the stimulation paradigm is largely unaffected by changes in opsin characteristics.

In summary, computational modeling of optogenetic inhibition in CA1 pyramidal neurons provides valuable insights into how stimulation settings, opsin properties, and neuronal physiology interact. The proposed simplified opsin model offers a data-efficient and accessible tool to support future research aimed at optimizing optogenetic strategies for seizure suppression.

Objective: Responsive neurostimulation (RNS) for epilepsy applies therapeutic stimulation in response to detected epileptic activity using intracranial electrodes. Prior studies have found that people with successful seizure reduction using RNS devices undergo functional network changes [1]. We hypothesise that the timing of stimulation, specifically stimulation during non-epileptic brain states, is critical in driving long-term changes to restore healthy brain networks.

Methods: Simulation episodes occurring in low-risk epochs was characterized in a cohort of 40 people with epilepsy treated with the RNS System at the University of Utah, USA. People were categorized into three groups based on their seizure reduction values: super responders (>90% reduction, n=10), intermediate responders (≥ 50% reduction and ≤ 90% reduction), n=19, and poor responders (<50% reduction, n=11). Low-risk periods were defined as days which had fewer long epileptic episode counts than the weekly average from the prior week.

Results: Super-responders spent more time in and experienced more frequent stimulation during low-risk periods (Pearson’s correlation: p=1.88e-3 and p=1.59e-3). Additionally, circadian rhythmicity was an indicator of improved seizure reduction (Pearson’s correlation: p=0.045 and 1.55e-3), and stimulation of super-responders was less phase- locked to long epileptic episodes (two-sample Kuiper test; p<0.01). Baseline seizure frequency, epilepsy duration, and age was not a statistical predictor of outcome.

Conclusions: Individuals with greater seizure reduction experienced more stimulation episodes during low-risk periods and spent more time in low-risk periods from stimulation onset. Targeting low-risk brain states may improve seizure reduction though neuromodulation via network restructuring driven by neuromodulation-induced plasticity.

[1] Khambhati, A.N., Shafi, A., Rao, V.R. and Chang, E.F., 2021. Long-term brain network reorganization predicts responsive neurostimulation outcomes for focal epilepsy. Science Translational Medicine, 13(608), p.eabf6588.

Computational modelling is a powerful tool for exploring brain mechanisms underlying neuroimaging data. In the future, it could be used as a diagnostic tool in the context of individual treatment planning, for example when applied to the intracranial EEG data recorded during presurgical evaluation. In this contribution, we use a computational model to identify different epileptic brain states and the synaptic mechanisms involved in epileptic seizure transitions from EEG data. Building on the modelling work of Wendling et al., we replicate rat slice (n=4, high potassium model of epilepsy) and human intracranial EEG data (4 patients from Motol hospital, 15 patients from the European Epilepsy Database). In a data-driven manner, we fit five-second segments of simulated and real data, providing us with a) an automatized classification of the observed type of epileptic brain activity, and b) estimates of the underlying levels of excitatory and inhibitory synaptic gains across time. The modelling environment could correctly classify epilepsy brain states (interictal, preonset, onset, ictal) with a mean sensitivity of 0.99 on model, 0.77 on rat, and 0.56 on human data (4-way classification task). We then study excitation/ inhibition changes in 205 seizure transitions, across seven different types of seizure onset patterns, in brain locations inside and outside of the seizure onset zone (SOZ). The transition to seizure was linked to significant increases in excitation and inhibition. Seizure types clearly differed in their associated changes (classification of seizure type: p < .01), already in the interictal interval and also in non-onset channels. Thus, different seizure types are linked to distinct synaptic mechanisms that are pre-determined at least tens of seconds before seizure onset and not (only) localized in SOZ.

Abnormal dynamics can manifest on EEG during resting states in individuals with epilepsy. The underlying mechanisms that contribute to this pathological activity—and how these mechanisms relate to the brain’s propensity to generate seizures—are poorly understood. Neural mass model calibration offers a powerful framework for interpreting EEG data by inferring the physiological mechanisms underlying the observed signals. Here, we will discuss key technical challenges involved in performing model calibration and parameter inference on EEG recordings during both resting and seizure states. We will describe algorithms used to attempt to overcome these challenges, including Bayesian inference methods and global search heuristics. Finally, we will present an application of parameter inference that demonstrates potential for identifying novel biomarkers of seizure susceptibility, and for developing hypotheses on the most effective interventions to rectify abnormal EEG dynamics.

Cognitive function emerges from cortical network dynamics and is often impaired in neuropsychiatric disorders like epilepsy. Physics and information theory suggest that brain networks operate optimally at a critical state between order and disorder. In this talk I will review work using comprehensive cognitive testing and multiday intracranial EEG from persons with epilepsy (PwE) to demonstrate that proximity to critical dynamics predicts cognitive performance across multiple domains. I will show that heterogeneous factors known to impact cognition— including interictal epileptiform activity, antiseizure medications, and sleep-like episodes— all act on one common endpoint: to perturb the critical state. While these findings suggest potential therapeutic targets, they also highlight a fundamental challenge in epilepsy: interventions aimed at optimizing critical dynamics for cognitive function must carefully consider the risk of increasing seizure susceptibility, as proximity to criticality may also lower the threshold for seizure generation. In other words, seizures may be the price for a brain being tuned to optimal computation. In conclusion, these results suggest critical dynamics to be the setpoint to measure optimal network function, thereby providing a unifying framework for the heterogeneous mechanisms impacting cognition in conditions like epilepsy.

References:

• Müller P, Miron G, Holtkamp M, Meisel C. Critical dynamics predicts cognitive performance and provides a common framework for heterogeneous mechanisms impacting cognition. Proceedings of the National Academy of Sciences 2025. DOI: 10.1073/pnas.2417117122.

• Müller P, Meisel C. Spatial and temporal correlations in human cortex are inherently linked, predicted by functional hierarchy, vigilance state and antiepileptic drug load. PLoS Comput Biol. 2023 Mar 3;19(3):e1010919. doi: 10.1371/journal.pcbi.1010919.

As a translational neuroengineering lab, getting advancements into patient care governs our work. The field has produced numerous implantable, sub-scalp, and wearable devices that detect and forecast seizures, but what problems do people with epilepsy actually want solved? I’ll highlight practical ways that continuous data can inform epilepsy management and introduce a novel platform that allows us to measure and act on these data in real-time.