Delineating epileptogenic networks using brain imaging data and personalized modeling in drug-resistant epilepsy

Virtual brain twins: from basic neuroscience to clinical use

Virtual brain twins are personalized, generative and adaptive brain models based on data from an individual's brain for scientific and clinical use. After a description of the key elements of virtual brain twins, we present the standard model for personalized whole-brain network models. The personalization is accomplished using a subject's brain imaging data by three means: (1) assemble cortical and subcortical areas in the subject-specific brain space; (2) directly map connectivity into the brain models, which can be generalized to other parameters; and (3) estimate relevant parameters through model inversion, typically using probabilistic machine learning. We present the use of personalized whole-brain network models in healthy ageing and five clinical diseases: epilepsy, Alzheimer's disease, multiple sclerosis, Parkinson's disease and psychiatric disorders. Specifically, we introduce spatial masks for relevant parameters and demonstrate their use based on the physiological and pathophysiological hypotheses. Finally, we pinpoint the key challenges and future directions.

Multi-scale modelling of the epileptic brain: advantages of computational therapy exploration

Objective: Epilepsy is a complex disease spanning across multiple scales, from ion channels in neurons to neuronal circuits across the entire brain. Over the past decades, computational models have been used to describe the pathophysiological activity of the epileptic brain from different aspects. Traditionally, each computational model can aid in optimizing therapeutic interventions, therefore, providing a particular view to design strategies for treating epilepsy. As a result, most studies are concerned with generating specific models of the epileptic brain that can help us understand the certain machinery of the pathological state. Those specific models vary in complexity and biological accuracy, with system-level models often lacking biological details.Approach: Here, we review various types of computational model of epilepsy and discuss their potential for different therapeutic approaches and scenarios, including drug discovery, surgical strategies, brain stimulation, and seizure prediction. We propose that we need to consider an integrated approach with a unified modelling framework across multiple scales to understand the epileptic brain. Our proposal is based on the recent increase in computational power, which has opened up the possibility of unifying those specific epileptic models into simulations with an unprecedented level of detail.Main results: A multi-scale epilepsy model can bridge the gap between biologically detailed models, used to address molecular and cellular questions, and brain-wide models based on abstract models which can account for complex neurological and behavioural observations.Significance: With these efforts, we move toward the next generation of epileptic brain models capable of connecting cellular features, such as ion channel properties, with standard clinical measures such as seizure severity.

Bifurcations and bursting in the Epileptor

The Epileptor is a phenomenological model for seizure activity that is used in a personalized large-scale brain modeling framework, the Virtual Epileptic Patient, with the aim of improving surgery outcomes for drug-resistant epileptic patients. Transitions between interictal and ictal states are modeled as bifurcations, enabling the definition of seizure classes in terms of onset/offset bifurcations. This establishes a taxonomy of seizures grounded in their essential underlying dynamics and the Epileptor replicates the activity of the most common class, as observed in patients with focal epilepsy, which is characterized by square-wave bursting properties. The Epileptor also encodes an additional mechanism to account for interictal spikes and spike and wave discharges. Here we use insights from a more generic model for square-wave bursting, based on the Unfolding Theory approach, to guide the bifurcation analysis of the Epileptor and gain a deeper understanding of the model and the role of its parameters. We show how the Epileptor's parameters can be modified to produce activities for other seizures classes of the taxonomy, as observed in patients, so that the large-scale brain models could be further personalized. Some of these classes have already been described in the literature in the Epileptor, others, predicted by the generic model, are new. Finally, we unveil how the interaction with the additional mechanism for spike and wave discharges alters the bifurcation structure of the main burster.

Personalized modeling of Alzheimer's disease progression estimates neurodegeneration severity from EEG recordings

INTRODUCTION

Early identification of Alzheimer's disease (AD) is necessary for a timely onset of therapeutic care. However, cortical structural alterations associated with AD are difficult to discern.

METHODS

We developed a cortical model of AD-related neurodegeneration accounting for slowing of local dynamics and global connectivity degradation. In a monocentric study we collected electroencephalography (EEG) recordings at rest from participants in healthy (HC, n = 17), subjective cognitive decline (SCD, n = 58), and mild cognitive impairment (MCI, n = 44) conditions. For each patient, we estimated neurodegeneration model parameters based on individual EEG recordings.

RESULTS

Our model outperformed standard EEG analysis not only in discriminating between HC and MCI conditions (F1 score 0.95 vs 0.75) but also in identifying SCD patients with biological hallmarks of AD in the cerebrospinal fluid (recall 0.87 vs 0.50).

DISCUSSION

Personalized models could (1) support classification of MCI, (2) assess the presence of AD pathology, and (3) estimate the risk of cognitive decline progression, based only on economical and non-invasive EEG recordings.

Highlights

Personalized cortical model estimating structural alterations from EEG recordings.Discrimination of Mild Cognitive Impairment (MCI) and Healthy (HC) subjects (95%)Prediction of biological markers of Alzheimer's in Subjective Decline (SCD) Subjects (87%)Transition correctly predicted for 3/3 subjects that converted from SCD to MCI after 1y.

The virtual aging brain: Causal inference supports interhemispheric dedifferentiation in healthy aging

The mechanisms of cognitive decline and its variability during healthy aging are not fully understood, but have been associated with reorganization of white matter tracts and functional brain networks. Here, we built a brain network modeling framework to infer the causal link between structural connectivity and functional architecture and the consequent cognitive decline in aging. By applying in-silico interhemispheric degradation of structural connectivity, we reproduced the process of functional dedifferentiation during aging. Thereby, we found the global modulation of brain dynamics by structural connectivity to increase with age, which was steeper in older adults with poor cognitive performance. We validated our causal hypothesis via a deep-learning Bayesian approach. Our results might be the first mechanistic demonstration of dedifferentiation during aging leading to cognitive decline.

Optimization of ictal aborting stimulation using the dynamotype taxonomy

Electrical stimulation is an increasingly popular method to terminate epileptic seizures, yet it is not always successful. A potential reason for inconsistent efficacy is that stimuli are applied empirically without considering the underlying dynamical properties of a given seizure. We use a computational model of seizure dynamics to show that different bursting classes have disparate responses to aborting stimulation. This model was previously validated in a large set of human seizures and led to a description of the Taxonomy of Seizure Dynamics and the dynamotype, which is the clinical analog of the bursting class. In the model, the stimulation is realized as an applied input, which successfully aborts the burst when it forces the system from a bursting state to a quiescent state. This transition requires bistability, which is not present in all bursters. We examine how topological and geometric differences in the bistable state affect the probability of termination as the burster progresses from onset to offset. We find that the most significant determining factors are the burster class (dynamotype) and whether the burster has a DC (baseline) shift. Bursters with a baseline shift are far more likely to be terminated due to the necessary structure of their state space. Furthermore, we observe that the probability of termination varies throughout the burster's duration, is often dependent on the phase when it was applied, and is highly correlated to dynamotype. Our model provides a method to predict the optimal method of termination for each dynamotype. These results lead to the prediction that optimization of ictal aborting stimulation should account for seizure dynamotype, the presence of a DC shift, and the timing of the stimulation.

The role of quantitative markers in surgical prognostication after stereoelectroencephalography

OBJECTIVE

Stereoelectroencephalography (SEEG) is the reference method in the presurgical exploration of drug-resistant focal epilepsy. However, prognosticating surgery on an individual level is difficult. A quantified estimation of the most epileptogenic regions by searching for relevant biomarkers can be proposed for this purpose. We investigated the performances of ictal (Epileptogenicity Index, EI; Connectivity EI, cEI), interictal (spikes, high-frequency oscillations, HFO [80-300 Hz]; Spikes × HFO), and combined (Spikes × EI; Spikes × cEI) biomarkers in predicting surgical outcome and searched for prognostic factors based on SEEG-signal quantification.

METHODS

Fifty-three patients operated on following SEEG were included. We compared, using precision-recall, the epileptogenic zone quantified using different biomarkers (EZq ) against the visual analysis (EZC ). Correlations between the EZ resection rates or the EZ extent and surgical prognosis were analyzed.

RESULTS

EI and Spikes × EI showed the best precision against EZc (0.74; 0.70), followed by Spikes × cEI and cEI, whereas interictal markers showed lower precision. The EZ resection rates were greater in seizure-free than in non-seizure-free patients for the EZ defined by ictal biomarkers and were correlated with the outcome for EI and Spikes × EI. No such correlation was found for interictal markers. The extent of the quantified EZ did not correlate with the prognosis.

INTERPRETATION

Ictal or combined ictal-interictal markers overperformed the interictal markers both for detecting the EZ and predicting seizure freedom. Combining ictal and interictal epileptogenicity markers improves detection accuracy. Resection rates of the quantified EZ using ictal markers were the only statistically significant determinants for surgical prognosis.

Epilepsy: Mitochondrial connections to the 'Sacred' disease

Over 65 million people suffer from recurrent, unprovoked seizures. The lack of validated biomarkers specific for myriad forms of epilepsy makes diagnosis challenging. Diagnosis and monitoring of childhood epilepsy add to the need for non-invasive biomarkers, especially when evaluating antiseizure medications. Although underlying mechanisms of epileptogenesis are not fully understood, evidence for mitochondrial involvement is substantial. Seizures affect 35%-60% of patients diagnosed with mitochondrial diseases. Mitochondrial dysfunction is pathophysiological in various epilepsies, including those of non-mitochondrial origin. Decreased ATP production caused by malfunctioning brain cell mitochondria leads to altered neuronal bioenergetics, metabolism and neurological complications, including seizures. Iron-dependent lipid peroxidation initiates ferroptosis, a cell death pathway that aligns with altered mitochondrial bioenergetics, metabolism and morphology found in neurodegenerative diseases (NDDs). Studies in mouse genetic models with seizure phenotypes where the function of an essential selenoprotein (GPX4) is targeted suggest roles for ferroptosis in epilepsy. GPX4 is pivotal in NDDs, where selenium protects interneurons from ferroptosis. Selenium is an essential central nervous system micronutrient and trace element. Low serum concentrations of selenium and other trace elements and minerals, including iron, are noted in diagnosing childhood epilepsy. Selenium supplements alleviate intractable seizures in children with reduced GPX activity. Copper and cuproptosis, like iron and ferroptosis, link to mitochondria and NDDs. Connecting these mechanistic pathways to selenoproteins provides new insights into treating seizures, pointing to using medicines including prodrugs of lipoic acid to treat epilepsy and to potential alternative therapeutic approaches including transcranial magnetic stimulation (transcranial), photobiomodulation and vagus nerve stimulation.

Functional connectomic profile correlates with effective anterior thalamic stimulation for refractory epilepsy

BACKGROUND

Deep brain stimulation of the anterior nucleus of the thalamus (ANT-DBS) is an effective treatment for refractory epilepsy; however, seizure outcome varies among individuals. Identifying a reliable noninvasive biomarker to predict good responders would be helpful.

OBJECTIVES

To test whether the functional connectivity between the ANT-DBS sites and the seizure foci correlates with effective seizure control in refractory epilepsy.

METHODS

We performed a proof-of-concept pilot study of patients with focal refractory epilepsy receiving ANT-DBS. Using normative human connectome data derived from 1000 healthy participants, we investigated whether intrinsic functional connectivity between the seizure foci and the DBS site was associated with seizure outcome. We repeated this analysis controlling for the extent of seizure foci, distance between the seizure foci and DBS site, and using functional connectivity of the ANT instead of the DBS site to test the contribution of variance in DBS sites.

RESULTS

Eighteen patients with two or more seizure foci were included. Greater functional connectivity between the seizure foci and the DBS site correlated with more favorable outcome. The degree of functional connectivity accounted for significant variance in clinical outcomes (DBS site: |r| = 0.773, p < 0.001 vs ANT-atlas: |r| = 0.715, p = 0.001), which remained significant when controlling for the extent of the seizure foci (|r| = 0.773, p < 0.001) and the distance between the seizure foci and DBS site (|r| = 0.777, p < 0.001). Significant correlations were independent of variance in the DBS sites (|r| = 0.148, p = 0.57).

CONCLUSION

These findings suggest that functional connectomic profile is a potential reliable non-invasive biomarker to predict ANT-DBS outcomes. Accordingly, the identification of ANT responders could decrease the surgical risk for patients who may not benefit and optimize the cost-effective allocation of health care resources.

Computational neurosciences and quantitative systems pharmacology: a powerful combination for supporting drug development in neurodegenerative diseases

Successful clinical development of new therapeutic interventions is notoriously difficult, especially in neurodegenerative diseases, where predictive biomarkers are scarce and functional improvement is often based on patient's perception, captured by structured interviews. As a consequence, mechanistic modeling of the processes relevant to therapeutic interventions in CNS disorders has been lagging behind other disease indications, probably because of the perceived complexity of the brain. However in this report, we develop the argument that a combination of Computational Neurosciences and Quantitative Systems Pharmacology (QSP) modeling of molecular pathways is a powerful simulation tool to enhance the probability of successful drug development for neurodegenerative diseases. Computational Neurosciences aims to predict action potential dynamics and neuronal circuit activation that are ultimately linked to behavioral changes and clinically relevant functional outcomes. These processes can not only be affected by the disease state, but also by common genotype variants on neurotransmitter-related proteins and the psycho-active medications often prescribed in these patient populations. Quantitative Systems Pharmacology (QSP) modeling of molecular pathways allows to simulate key pathological drivers of dementia, such as protein aggregation and neuroinflammatory responses. They often impact neurotransmitter homeostasis and voltage-gated ion-channels or lead to mitochondrial dysfunction, ultimately leading to changes in action potential dynamics and clinical readouts. Combining these two modeling approaches can lead to better actionable understanding of the many non-linear pharmacodynamic processes active in the human diseased brain. Practical applications include a rational selection of the optimal doses in combination therapies, identification of subjects more likely to respond to treatment, a more balanced stratification of treatment arms in terms of comedications, disease status and common genotype variants and re-analysis of small clinical trials to uncover a possible clinical signal. Ultimately this will lead to a higher success rate of bringing new therapeutics to the right patient populations.

Observers for Phenomenological Models of Epileptic Seizures

Progress towards effective treatment of epileptic seizures has seen much improvement in the past decade. In particular, the emergence of phenomenological models of epileptic seizures specifically designed to capture the electrical seizure dynamics in the Epileptor model is inspiring new approaches to predicting and controlling seizures. These new models present in various forms and contain important but unmeasurable variables that control the occurrence of seizures. These models have been used mostly as nodes in large networks to study the complex brain behaviour of seizures. In order to use this model for the purposes of seizure forecasting or to control seizures through deep brain stimulation, the states of the model will need to be estimated. Although devices such as EEG electrodes can be related to some of the states of the model, most remain unmeasured and would require an observer (as defined in control theory) for their estimation. Additionally, we would like to consider the case for large nodes of systems where the number of electrodes is far smaller than the number of nodes being estimated. In this paper, we provide methods towards obtaining the full states of these phenomenological models using nonlinear observers. In particular, we explore the effectiveness of the Extended Kalman Filter for small networks of nodes of a smoothed sixth order Epileptor model. We show that observer design is possible for this family of systems and identify the difficulties in doing so.Clinical relevance-The methods presented here can be applied with an individual epileptic patient's EEG to reveal previously hidden biomarkers of epilepsy for seizure forecasting.

From phenomenological to biophysical models of seizures

Epilepsy is a complex disease that requires various approaches for its study. In this short review, we discuss the contribution of theoretical and computational models. The review presents theoretical frameworks that underlie the understanding of certain seizure properties and their classification based on their dynamical properties at the onset and offset of seizures. Dynamical system tools are valuable resources in the study of seizures. By analyzing the complex, dynamic behavior of seizures, these tools can provide insights into seizure mechanisms and offer a framework for their classification. Additionally, computational models have high potential for clinical applications, as they can be used to develop more accurate diagnostic and personalized medicine tools. We discuss various modeling approaches that span different scales and levels, while also questioning the neurocentric view, and emphasize the importance of considering glial cells. Finally, we explore the epistemic value provided by this type of approach.

Personalised virtual brain models in epilepsy

Individuals with drug-resistant focal epilepsy are candidates for surgical treatment as a curative option. Before surgery can take place, the patient must have a presurgical evaluation to establish whether and how surgical treatment might stop their seizures without causing neurological deficits. Virtual brains are a new digital modelling technology that map the brain network of a person with epilepsy, using data derived from MRI. This technique produces a computer simulation of seizures and brain imaging signals, such as those that would be recorded with intracranial EEG. When combined with machine learning, virtual brains can be used to estimate the extent and organisation of the epileptogenic zone (ie, the brain regions related to seizure generation and the spatiotemporal dynamics during seizure onset). Virtual brains could, in the future, be used for clinical decision making, to improve precision in localisation of seizure activity, and for surgical planning, but at the moment these models have some limitations, such as low spatial resolution. As evidence accumulates in support of the predictive power of personalised virtual brain models, and as methods are tested in clinical trials, virtual brains might inform clinical practice in the near future.

Amortized Bayesian inference on generative dynamical network models of epilepsy using deep neural density estimators

Whole-brain modeling of epilepsy combines personalized anatomical data with dynamical models of abnormal activities to generate spatio-temporal seizure patterns as observed in brain imaging data. Such a parametric simulator is equipped with a stochastic generative process, which itself provides the basis for inference and prediction of the local and global brain dynamics affected by disorders. However, the calculation of likelihood function at whole-brain scale is often intractable. Thus, likelihood-free algorithms are required to efficiently estimate the parameters pertaining to the hypothetical areas, ideally including the uncertainty. In this study, we introduce the simulation-based inference for the virtual epileptic patient model (SBI-VEP), enabling us to amortize the approximate posterior of the generative process from a low-dimensional representation of whole-brain epileptic patterns. The state-of-the-art deep learning algorithms for conditional density estimation are used to readily retrieve the statistical relationships between parameters and observations through a sequence of invertible transformations. We show that the SBI-VEP is able to efficiently estimate the posterior distribution of parameters linked to the extent of the epileptogenic and propagation zones from sparse intracranial electroencephalography recordings. The presented Bayesian methodology can deal with non-linear latent dynamics and parameter degeneracy, paving the way for fast and reliable inference on brain disorders from neuroimaging modalities.