Epilepsy at different scales: considerations for the medical application of dynamical models

Abstract

Epilepsy is a complex and poorly understood disease characterized by a susceptibility to recurrent seizures. One in three patients does not respond to available anti-seizure drugs, raising the need for alternative treatments. Epileptic seizures result from changes in electrical patterns of activity on the scale of large neuronal networks. The underlying mechanisms that lead to these changes can be multifactorial and patient-specific, and may range from synaptic protein dysfunction to changes in the connectivity patterns of the network. Furthermore, the changes in electrical activity observed during seizures resemble dynamic phase transitions, making dynamical modeling a natural choice for the study of epilepsy. However, although epilepsy has been a subject of computational modeling for years, little progress has been made in identifying clinically relevant parameters. Seizure prediction algorithms and intervention strategies that are employed on a systemic level to interrupt seizures have had mixed results, partly because of the lack of understanding of the underlying mechanisms of action. Because of the nature of seizure prediction and prevention methods, which measure from and act upon large networks of neurons, a systems-level understanding of epilepsy based on dynamical models could be very beneficial in improving their effectiveness. Large-scale brain networks are also being integrated into dynamical models, following recent findings on the role of distributed network metrics in epileptogenesis. Nevertheless, such models still face a problem of scaling and have not yet produced clinically relevant results. The future of dynamical modeling in epilepsy treatment seems to lie in patient-specific models.