A data-driven brain network modeling for epileptogenic spread analysis

Epileptic seizure propagation is a dynamic process that can be triggered by local abnormal discharges, leading to widespread network abnormalities in the brain. Understanding the causal relationship between the changes in brain network characteristics and the diverse propagation dynamics of epileptic seizures is crucial. We gather stereo-EEG data from 17 patients with temporal lobe epilepsy and utilize cross-channel phase amplitude coupling to extract the dynamic functional networks. Further, the patterns of brain network changes during seizure in patients with different surgeries are assessed using Hidden Markov Model. And characteristics of state transitions under different seizure periods are explored. Results show that the frequency of state transitions increases with seizures, and all epilepsy patients have a main state network with weakly connected network structure centered on the epileptogenic zone. The occupancy ratio of main state is inversely proportional to state transition frequency, where the emergence of strongly connected networks facilitates the seizure propagation. Variability in state characteristics is observed cross patients with different surgeries. The heterogeneous epileptor network model driven by the state transition is developed to simulate seizure propagation. Results show that state transition frequency and relationships affect seizure onset time and spread range. Under the main state network, seizures occur only in the epileptogenic zone and do not propagate to surrounding regions. Additionally, increasing the proportion of the main state network delays the onset of seizures. This suggests that the characteristics of the state network and its transitions may play a role in controlling the propagation of epileptic seizures.