Computational NeuroscienceReviewComputational models of epileptiform activity
Introduction
Epileptiform activity refers to brain activity recorded during epileptic phenomena that cover not only the seizure episodes but also the plethora of abnormal transient events occurring outside seizures like, for instance, interictal spikes or high-frequency oscillations in partial epilepsies.
To date, significant progress has been achieved regarding the numerous techniques aimed at recording brain activity. Epilepsy research has always benefited from these advances. Epileptiform activity can be recorded either globally using EEG/MEG, fMRI or SPECT instruments as well as very locally using intracerebral micro- and macro-electrodes (intracellular, MUAs, LFPs, iEEG) or microscopic imaging. The multiplicity of recording techniques, each characterized by its own time and space resolution, has also led to a need (i) for advanced information processing techniques aimed at extracting/describing the multimodal information conveyed by the observations, on the one side and (ii) for computational models aimed at decoding/explaining the mechanisms underlying their generation, on the other side.
This demand explains why the field of computational modeling in epilepsy has grown rapidly over the past decades. Numerous models were proposed to (i) investigate the complex pathophysiological factors leading to ictogenesis and/or epileptogenesis and often resulting from multiple causes and (ii) help the interpretation of epileptiform activity, whatever the recording technique. These models have progressively been more widely accepted and are now considered as an efficient way of integrating structural, functional and dynamical data about neural systems (coming from neurobiology, neurophysiology and neurology research) into “coherent and interpretable views”. Models have the unique ability to identify key – possibly hidden – variables and relate these variables across multiple levels of descriptions. For instance, computational models of fast ripples observed in electrophysiological signals recorded from epileptogenic zones allowed for connecting pyramidal cell abnormal firing patterns to pathological oscillations in small-scale neural networks. Another recognized virtue of models is the capacity to generate hypotheses that can be tested experimentally. For example, some models have successfully predicted the alteration of GABAb receptor-mediated responses on thalamocortical cells in the genesis of spike-wave patterns in absence seizures.
In this article we review computer models that have been developed to reproduce and explain epileptiform activity. Conversely to reviews already published (Lytton, 2008, Wendling, 2008) or available online (http://www.scholarpedia.org/article/Models_of_epilepsy) which are based on the type of modeling approach (typically, from microscopic to mesoscopic or macroscopic level) or on the mechanisms (typically excitability and synchronization), this review follows a less traditional outline in the sense that its starts from the type of epileptiform activity under study. For both interictal and ictal activity, in the context of partial and generalized epilepsies, we reviewed most of the models proposed so far and classified them according to the following taxonomy: neural mass models (NMMs), neural field models (NFMs), detailed networks and formal mathematical models. This review shows the richness and complementarity of these various modeling approaches as well as the extreme productivity of the computational neuroscience community in the field of epilepsy research. This review ends with a discussion about the advantages, limitations and future of modeling approaches along with some perspectives in epilepsy research and clinical epileptology.
Section snippets
General principles of detailed network, neural mass, neural field and formal mathematical models
The challenge faced by any modeling approach is to capture the essential features of the system under study in a simple – but not too simple – description of its behavior (defined by the model variables) under controlled conditions (defined by the model parameters). In the brain, the level of modeling, i.e. the level at which this description is elaborated, is a crucial choice and strongly depends on (i) the type of behavior to be reproduced in the model and (ii) the nature of the data it
Interictal epileptic spikes and bursts
Interictal epileptic spikes (IESs) are very often observed in human partial epilepsies as well as in most experimental models of focal epilepsy (Schwartzkroin and Wheal, 1984). During epileptogenesis, a number of experimental studies also reported the appearance of isolated epileptic spikes during the latent period (Avoli et al., 2006, Staley and Dudek, 2006). It is commonly admitted that IESs are polymorphic events (Alarcon et al., 1994). Nevertheless, to a large extent, IESs correspond to
Fast oscillations at seizure onset (gamma, high gamma, chirps)
In partial epilepsies, seizures are often characterized by the appearance of fast oscillations in LFPs or EEG signals, sometimes referred to as “chirps” (Schiff et al., 2000). Typically, in human neocortical epilepsies, rapid discharges observed at seizure onset range from 70 to 120 Hz. They constitute a characteristic electrophysiological pattern in focal seizures characterized by a noticeable increase of intracerebral EEG signal frequency (Lee et al., 2000, Wendling et al., 2003). In mesial
Discussion and perspectives
The need for integrative approaches. Epilepsy is a complex dynamical disease (Lopes da Silva et al., 2003). In fact, the term “epilepsy” refers to a wide variety of neurological syndromes and disorders. Any brain damage can potentially lead to a disruption of its activity which, in some cases, can give rise to epileptic seizures. Over the past decades, basic and clinical research has advanced our understanding of the pathophysiology of epilepsy. The general and commonly-accepted picture is that
Acknowledgement
The authors acknowledge financial support from the French “Programme de Recherche Translationnelle en Santé” (Call PRTS 2013, Project # ANR-13-PRTS-0011, title: “VIBRATIONS”). They thank Prof. O. Raineteau for providing the hippocampal slice confocal image in Fig. 1.
Glossary
- EEG
- electroencephalography
- iEEG
- intracranial or intracerebral EEG
- sEEG
- stereotactic EEG
- ECoG
- electrocorticography
- LFP
- local field potential
- MUA
- multiunit-activity
- SPECT
- single-photon emission computed tomography
- IES
- interictal epileptic spike
- HFO
- high-frequency oscillation
- FR
- fast ripple
- SW
- spike-wave
- SE
- status epilepticus
- RSE
- refractory status epilepticus
- NMM
- neural mass model
- NFM
- neural field model
- TLE
- temporal lobe epilepsy
- EZ
- epileptogenic zone
- PSP
- post-synaptic potential
- EPSP
- excitatory PSP
- IPSP
- inhibitory PSP
- CA
- Cornu Ammonis
References (173)
- et al.
Seizure prediction in hippocampal and neocortical epilepsy using a model-based approach
Clin Neurophysiol
(2014) - et al.
A model of the coupling between brain electrical activity, metabolism, and hemodynamics: application to the interpretation of functional neuroimaging
Neuroimage
(2002) - et al.
Neural networks involving the medial temporal structures in temporal lobe epilepsy
Clin Neurophysiol
(2001) - et al.
Pitfalls of high-pass filtering for detecting epileptic oscillations: a technical note on false ripples
Clin Neurophysiol
(2010) - et al.
Modern concepts of seizure modeling
Int Rev Neurobiol
(2014) - et al.
Localization of extended brain sources from EEG/MEG: the ExSo-MUSIC approach
Neuroimage
(2011) - et al.
Automatic detection of fast ripples
J Neurosci Methods
(2013) Status epilepticus in adults
Lancet
(1990)- et al.
Description and computational modeling of the whole course of status epilepticus induced by low dose lithium-pilocarpine in rats
Brain Res
(2011) - et al.
Recording of fast activity at the onset of partial seizures: depth EEG vs. scalp EEG
Neuroimage
(2012)