Localization of the epileptogenic network from scalp EEG using a patient-specific whole-brain model

Abstract

Computational modeling is a key tool for elucidating the neuronal mechanisms underlying epileptic activity. Despite considerable progress, existing models often lack realistic accuracy in representing electrophysiological epileptic activity. In this study, we used a comprehensive human brain model based on a neural mass model, which is tailored to the layered structure of the neocortex and incorporates patient-specific imaging data. This approach allowed the simulation of scalp EEGs in an epileptic patient suffering from type 2 focal cortical dysplasia (FCD). The simulation specifically addressed epileptic activity induced by FCD, faithfully reproducing intracranial interictal epileptiform discharges (IEDs) recorded with electrocorticography. For constructing the patient-specific scalp EEG, we carefully defined a clear delineation of the epileptogenic zone by numerical simulations to ensure fidelity to the topography, polarity, and diffusion characteristics of IEDs. This nuanced approach improves the accuracy of the simulated EEG signal, provides a more accurate representation of epileptic activity, and enhances our understanding of the mechanism behind the epileptogenic networks. The accuracy of the model was confirmed by a postoperative reevaluation with a secondary EEG simulation that was consistent with the lesion's removal. Ultimately, this personalized approach may prove instrumental in optimizing and tailoring epilepsy treatment strategies.

Keywords: Digital brain; ECoG; EEG modeling; Epilepsy surgery; Focal cortical dysplasia (FCD); Interictal epileptiform discharges (IED).

Figure 1.

Presurgical investigations. (A) Illustration of the MRI and the 3D reconstruction of the cortex after the first surgery. (B) Illustration of the patient’s scalp EEG recording revealing prominent interictal spikes within the left frontal cortex, especially on FP1-F3 and F3-C3 electrode derivations. (C) An example of an ECoG recording showing interictal SW discharges on the borders of the supposed lesion on MRI.

Figure 2.

Epileptic COALIA (eCOALIA). (A) The neural mass model of a cortical column comprises two subpopulations of pyramidal neurons and four main types of GABAergic interneurons (PV, SST, VIP, and NGFC). This NMM is considered for modeling each cortical area of the atlas. (B) The connectivity between the cortical NMMs and subcortical regions, such as the thalamus, is established via long-range excitatory glutamatergic connections that target all cell types. The thalamic input to a cortical column is received by PYR, PV+, SST+, VIP+, and NGFC subpopulations. The thalamus receives excitatory input from the cortex. (C) The patient’s MRI is used to construct the head model. (D) Desikan-Killiany brain atlas is used to parcel the cortical surface. (E) Cortical and subcortical regions are connected via a DTI-based structural connectivity matrix and a delay matrix representing the communication delay between them. (F) The epileptogenic zone, which undergoes interictal epileptiform discharges, is defined after solving the inverse problem applied to scalp EEG signals. This region is associated with a NMM, which models the observed epileptic activity recorded with an ECoG, while the rest of the cortical regions are set in the background state. The source activity is projected on surface EEG electrodes (G) and simulated EEG signals are computed (H).

Figure 3.

Simulated intracranial and scalp EEG. (A) Real intracranial ECoG recording versus simulated LFP. The signal was simulated in one population of NMM. After optimization, we found a similarity index of 0.787 between two random spike-wave discharges. The schema represents the variation of coupling coefficient compared with the generic parameters. (B) Layered NMM with synaptic modifications to render the model similar to the real FCD’s activity. (C) Real scalp EEG of the patient versus simulated EEG with a longitudinal montage. The signal was simulated with the COALIA model, which contains 66 interconnected populations of NMM, based on the Desikan Atlas. The interictal activity parameters were inserted in the NMM of the entire left F-RMG. The EEG-IES index is a low-to-moderate 43% score.

Figure 4.

Anatomical model and scalp EEG variability. (A) Representation of a template cortex and the patient-specific cortex. The patient’s cortex was reconstructed from her MRI. The epileptogenic zone is represented in red. This region corresponds to a manual segmentation of the left rostral middle frontal gyrus. (B) Real scalp EEG of the patient versus simulated EEG with template cortex or with patient-specific cortex with longitudinal montage. In the simulations, the interictal activity was generated in the epileptogenic zone. The EEG-IES score was 92% for the atlas generated EEG and a 73% for the EEG generated with the patient’s mesh.

Figure 5.

Simulation of scalp EEG after the second surgery. (A) Brain anatomy of the patient showing the first unsuccessful (blue line) and the second (red line) successful resection of the epileptogenic zone. (B) Illustration of the location (top, red area) of the epileptogenic zone according to our model simulations based on interictal activity. A simulated surgery was made by removing the epileptic NMM from the leadfield calculation (bottom). On the brain model situated in the right, we depicted the superposition between the real surgery and our predicted epileptogenic zone. (C) The add-on of the alpha, beta, and delta rhythms in the digital brain to improve the realistic appearance of the resting-state scalp EEG. (D) Comparison between the resting-state scalp EEG (eyes closed) recorded after the second surgery of the patient and the simulated scalp EEG after the in silico surgery (right panel).