A1534
Title: Multivariate Hawkes process modeling of dynamic functional connectivity network in MEG for epilepsy
Authors: Jaehee Kim - Duksung Womens University (Korea, South) [presenting]
Abstract: Epilepsy is increasingly recognized as a disorder of large scale brain networks, in which seizures and interictal dysfunction arise from abnormal interactions among distributed regions on fast temporal scales. Conventional functional connectivity approaches, including static and dynamic methods, rely on temporal averaging or sliding windows, limiting sensitivity to transient, directed, and history dependent neural interactions central to epileptic dynamics. Magnetoencephalography provides millisecond temporal resolution and direct sensitivity to neuronal activity. However, most connectivity analyses rely on correlation or coherence measures that assume continuous signals and symmetric interactions, while dynamic approaches impose temporal smoothness and stationarity, restricting detection of brief and asymmetric excitation. An event driven framework using the multivariate Hawkes process is proposed to model directed and history dependent functional connectivity. This approach captures self and cross excitation, enabling inference of causal and temporally asymmetric interactions without arbitrary windowing. Applied to resting state MEG data from 44 patients and 46 controls, the model revealed increased recurrent excitation and lateralized propagation in patients, whereas controls showed weaker and more balanced dynamics.
