Evolving functional network properties and synchronizability during human epileptic seizures

We assess electrical brain dynamics before, during, and after one-hundred human epileptic seizures with different anatomical onset locations by statistical and spectral properties of functionally defined networks. We observe a concave-like temporal evolution of characteristic path length and cluster coefficient indicative of a movement from a more random toward a more regular and then back toward a more random functional topology. Surprisingly, synchronizability was significantly decreased during the seizure state but increased already prior to seizure end. Our findings underline the high relevance of studying complex systems from the view point of complex networks, which may help to gain deeper insights into the complicated dynamics underlying epileptic seizures.

💡 Research Summary

This paper investigates the temporal evolution of functional brain network topology and synchronizability during human epileptic seizures using graph‑theoretical methods applied to intracranial EEG recordings. Data were collected from 60 patients undergoing presurgical evaluation for drug‑resistant epilepsy, yielding 100 focal‑onset seizures recorded with an average of 53 bipolar channels at 200 Hz. After bipolar re‑referencing and normalization, the authors computed pairwise maximum lag‑cross‑correlation coefficients (ρmax ij) within non‑overlapping 2.5‑second sliding windows. For each window, an adaptive threshold was applied to the correlation matrix to produce a binary adjacency matrix A(w) that guaranteed graph connectivity; the threshold was lowered until the second smallest Laplacian eigenvalue λmin became positive, ensuring a single connected component.

From each time‑resolved graph the authors extracted three classical topological metrics: average shortest path length L(w), clustering coefficient C(w), and edge density ε(w). To assess deviations from randomness, they compared C and L to those of degree‑preserving random graphs (C_r, L_r) and examined the ratios C/C_r and L/L_r. Synchronizability was quantified via the eigenratio S(w)=λmax/λmin of the Laplacian, where a smaller S indicates a more stable globally synchronized state.

The results reveal a characteristic “concave‑shaped” trajectory for both structural and spectral measures. At seizure onset, both C/C_r and L/L_r are slightly above unity, indicating a modest shift toward a more regular (small‑world‑like) topology. Mid‑seizure, these ratios peak, reflecting maximal regularity, while edge density ε modestly rises. Simultaneously, the eigenratio S reaches its maximum, signifying the lowest synchronizability. Prior to seizure termination, C/C_r and L/L_r decline back toward random‑network values, ε continues to increase, and S drops sharply, indicating a rapid restoration of synchronizability. The dynamics of S are driven primarily by changes in λmin; reductions in λmin during the seizure middle phase suggest transient fragmentation into local sub‑structures that impede global synchronization. λmax, by contrast, follows ε, consistent with theoretical bounds linking edge density to the largest Laplacian eigenvalue.

Importantly, these patterns are consistent across seizures regardless of the anatomical origin of the ictal focus, suggesting a universal network‑level mechanism. The authors interpret the mid‑seizure decrease in synchronizability as a protective self‑regulatory process that prevents runaway hyper‑synchrony, while the subsequent increase in synchronizability may facilitate seizure termination by allowing the network to re‑establish a globally coherent state.

The study demonstrates that functional brain networks undergo systematic topological reconfiguration during seizures, moving from a relatively random configuration to a more regular one and back again, while synchronizability follows an opposite trajectory. These findings support the view that seizure dynamics are governed not only by local neuronal excitability but also by global network architecture. The authors propose that monitoring λmin (or the eigenratio S) in real time could serve as a biomarker for imminent seizure termination and might inform closed‑loop neuromodulation strategies aimed at promoting network configurations that favor termination. Future work should integrate structural connectivity data, explore higher‑order network measures, and test whether targeted perturbations of network topology can reliably abort seizures.