SEEG Virtual Brain Twins Mirror Biological Networks in Epileptogenic Zone and Connectivity Properties

Introduction: SEEG recordings provide direct access to epileptic networks but are inherently limited by sparse spatial sampling, typically covering only a fraction of brain regions per patient. This spatial limitation creates gaps in comprehensive network analysis, particularly for validating connectivity hypotheses in subcortical regions. Virtual brain twins address this limitation through simulated seizures that provide whole-brain coverage. We hypothesized that virtual brain twins would successfully recreate the Interictal Suppression Hypothesis (ISH), demonstrating that non-epileptogenic zones actively suppress epileptogenic zones during interictal periods, and that ictal connectivity patterns would mirror those observed in real patient data.

Methods: We utilized the Virtual Epileptic Patient (VEP) cohort of 30 drug-resistant epilepsy patients, providing patient-specific anatomical and functional data including brain regions, structural connectivity, reconstructed electrodes, and simulated brain activity. This personalized modeling generates virtual brain twins based on individual brain data. We employed partial directed coherence (PDC) analysis in the alpha band (8-13 Hz) to assess directional connectivity. We first tested the ISH on seizure onset zones, representing the first ISH testing in a virtual patient cohort. Subsequently, we analyzed basal forebrain connectivity across seven cortical networks during interictal versus ictal states.

Results: We successfully validated the Interictal Suppression Hypothesis (ISH) for the first time in a virtual patient cohort, demonstrating that virtual epileptogenic zones exhibited significantly higher inward connectivity and reduced outward connectivity compared to non-seizure onset regions (p < 0.001). This validation confirmed the utility of virtual brain twins for testing network-level seizure mechanisms. Building on this validation, basal forebrain regions showed pronounced seizure-related reorganization of attention network connectivity: decreased PDC to the Ventral Attention Network (p < 0.01) and increased PDC to the Dorsal Attention Network (p < 0.01). Notably, this pattern of disrupted attention network connectivity in the subcallosal area mirrors our previous findings of compromised BNST-attention network control in temporal lobe epilepsy patients, consistent with the anatomical proximity of these critical subcortical structures.

Conclusion: Virtual brain twins extend SEEG's precision to achieve comprehensive whole-brain network analysis. Our findings demonstrate that basal forebrain regions, specifically the subcallosal area, undergo systematic seizure-related reorganization of attention network connectivity, supporting their potential as neuromodulation targets. The observed shift from ventral to dorsal attention network connectivity reveals adaptive reconfiguration mechanisms that parallel our BNST findings, suggesting that seizure-related attention network disruptions may represent a common pathophysiological mechanism affecting this anatomically contiguous subcortical region.