Abstracts

Rationale: Epilepsy surgery requires a precise localization of the individual epileptic zone (EZ). The virtual epileptic patient (VEP) framework was developed to predict the EZ in individual patients¹. This method uses brain network modelling, a combination of non-linear dynamical systems theory, patient individual brain structural connectivity data and ictal SEEG recordings. A model is built, in which brain regions are described by a phenomenological mathematical model, the Epileptor, capable of showing seizure-like dynamics. Activity propagates through the brain network to replicate empirical ictal recordings and to infer the EZ. This method is currently applied in the clinical trial “EPINOV”. The aim of this work is to extend the current model and overcome some of its limitations related to the low-resolution representation of the brain.

Methods: Instead of a coarse parcellation of the brain into 162 regions, which are represented by neural masses, we introduce neural fields to our model. Cortical regions are now modelled using the spatial extent of the cortical surface. Neural fields increase the resolution of the electromagnetic forward problem of the SEEG, using the vertex normal of the mesh as precise electric dipole locations. The surface extension is also applied to the hippocampus by following the cortical surface from the entorhinal and parahippocampal region into the subiculum and the folding of the cornu ammonis. Diffusion MRI is used to infer white matter connectivity between brain regions of our model. However, the typical resolution of this data is too low to resolve fibers in the hippocampus. Therefore we use high resolution data coming from post-mortem scans to infer connections between the hippocampal subfields. Another shortcoming of the current VEP method is that no variance in the parametrization of the brain regions is considered. We are introducing region variance coming from microscopic cytoarchitectonic studies and functional SEEG data analysis to further constrain our model.

Results: The neural field formulation allows for new propagating ictal waves across neighboring points of the surface. A phenomenon that has been observed empirically in multiarray and ECOG recordings but has been neglected by the current VEP model. The improved forward model creates more realistic SEEG data generated from simulated seizures. Since the high importance for temporal lobe seizures of the hippocampus, we expect that our model will significantly improve in seizure prediction due to its detailed representation in the model.

Conclusions: We introduce the virtual big brain in epilepsy as the next generation of large scale brain models equipped with high resolution data to improve seizure prediction in individual patients.

¹Jirsa, V. K., Proix, T., Perdikis, D., Woodman, M. M., Wang, H., Gonzalez-Martinez, J., Bernard, C., Benar, C., Guye, M., Chauvel, P., & Bartolomei, F. (2015). The Virtual Epileptic Patient: Individualized whole-brain models of epilepsy spread. NeuroImage, 145, 377–388.

Funding: Please list any funding that was received in support of this abstract.: This research was funded by the European Union’s Horizon 2020 Framework Programme for Research and Innovation under the Specific Grant Agreement No. 945539 (Human Brain Project SGA3).