Signal Trajectory Analysis Reveals the Spatiotemporal Manifold of Epileptic Networks

Abstract

Abstract Objective: We aim to design a method framework for data analysis and visualization in order to reveal the spatiotemporal manifold of the epileptic networks and differentiate between the seizure-onset regions and the propagation regions. Methods: We hypothesize that signal motion is the functional substrate of epileptic networks and that signal trajectories reflect the spatiotemporal manifold of epileptic networks. This spatiotemporal manifold is visualized using a series of quantitative and interpretable methodologies. Results: A total of 454 aberrant nodes (109 seizure-onset nodes and 345 propagation nodes) were identified among the 1033 electrode nodes (606 SEEG electrodes and 427 ECoG electrodes) of 9 cases of intracranial EEG data. Significance: The multidimensional joint analysis of signal amplitude describes the signal trajectories of various frequency bands, thereby disclosing the spatiotemporal manifold of the epileptic networks. This is useful for distinguishing the seizure-onset regions from the propagation regions in order to direct epilepsy treatment.