Suppressing seizure via optimal electrical stimulation to the hub of epileptic brain network

TL;DR

This paper addresses seizure control in focal epilepsy by modeling seizure propagation as a networked dynamical system and learning a data-driven surrogate via Dynamic Mode Decomposition to obtain a linear model $x_{t+1}=A x_t + B u_t$. It fuses this surrogate with Model Predictive Control to optimize electrical stimulation while selecting control nodes, comparing stimulation of the seizure onset zone (SOZ) with hub nodes in the epileptic network. The method is validated on a network-coupled Jansen-Rit platform and on real iEEG data, showing that hub-based stimulation can suppress seizure dynamics with comparable effectiveness to direct SOZ stimulation and often with lower energy, demonstrating a practical, network-theoretic approach to personalized neuromodulation. The work contributes a general, theory-grounded platform for validating neural stimulation strategies, highlights the importance of control-node selection, and points to energy-efficient stimulation patterns (e.g., low-frequency inputs) as clinically relevant. Overall, it advances network-aware, data-driven strategies for seizure suppression with potential to guide patient-specific neuromodulation therapies.

Abstract

The electrical stimulation to the seizure onset zone (SOZ) serves as an efficient approach to seizure suppression. Recently, seizure dynamics have gained widespread attendance in its network propagation mechanisms. Compared with the direct stimulation to SOZ, other brain network-level approaches that can effectively suppress epileptic seizures remain under-explored. In this study, we introduce a platform equipped with a system identification module and a control strategy module, to validate the effectiveness of the hub of the epileptic brain network in suppressing seizure. The identified surrogate dynamics show high predictive performance in reconstructing neural dynamics which enables the model predictive framework to achieve accurate neural stimulation. The electrical stimulation on the hub of the epileptic brain network shows remarkable performance as the direct stimulation of SOZ in suppressing seizure dynamics. Underpinned by network control theory, our platform offers a general tool for the validation of neural stimulation.

Figures (4)

• Figure 1: The optimal control framework for seizure suppression, including a system identification modular and an optimal control strategy modular. a, Brain nodes in SOZ (red lines) and non-SOZ (black lines) of a subject are pre-determined by clinicians and the real iEEG data is from the Epilepsy-iEEG-Multicenter-Dataset. b, Identifying the system dynamics via reconstructing the neural dynamics. c, Designning the optimal control strategy, considering two factors: control node selection (left) and the optimal control policy (right). The dark nodes indicate potential control nodes. The optimal inputs $u$ to the control nodes are designed by minimizing the error between the surrogate dynamics and the reference dynamics and control energy.
• Figure 2: A simulation of the propagation process of seizure dynamics on the whole brain. a, The neural dynamics of each node in the whole brain can be simplified as Jansen-Rit dynamics. The neural dynamics between interconnected regions are propagated through the white matter fiber connections. b, An example of the whole brain structural connectivity. c, The connection that links to the hippocampus. d, The simulated neural dynamics of the whole brain. The hippocampus is predefined as an SOZ with high-frequency abnormal discharge, while other regions with abnormal discharge are considered as propagation regions. The regions with normal discharge are non-seizure regions.
• Figure 3: Control effects on the JR simulation platform. We input the optimized electrical stimulation to the pre-selected control node and illustrate the control effects. a, The original seizure dynamics (left) and reference neural dynamics (right). b, Control effects (top) and the optimal inputs (bottom). From left to right, the control node is set to the hippocampus (SOZ), thalamus (SC and FC hub), and precuneus (Top weighted degree), respectively. The results show that stimulating the hippocampus or thalamus can successfully suppress seizure propagation. c, The mean absolute error of system dynamics and reference dynamics. d, The average energy consumed to control the system. e, The spectrum of optimal inputs on each pre-selected node.
• Figure 4: Control effects on real data. We input the optimized virtual electrical stimulation to the pre-selected control node and illustrate the control effects on the identified model $x_{k+1}=Ax_{k}+Bu_k$. a, The original seizure dynamics (gray line) and the reconstructed neural dynamics (blue dot line). b, Control effects with optimal inputs to three types of control nodes (left: SOZ, middle: top degree of FC, right: FC hub). c, Distribution of control energy (left) and magnitude (right). d, Magnitude distribution of soz (blue-paired) and non-soz (green-paired) with 3 types of control nodes. ($*$ for $p$-value$<0.05$ and $***$ for $p$-value$<1e^{-4}$)