Anticipating epileptic seizures through the analysis of EEG synchronization as a data classification problem

TL;DR

This paper presents a graph-based EEG analysis method to predict epileptic seizures by detecting synchronization pattern changes, aiming for real-time, patient-specific seizure prediction.

Contribution

It introduces a novel graph model and synchronization measures for EEG analysis, enabling real-time, patient-specific seizure prediction with simple, computationally efficient methods.

Findings

EEG synchronization patterns change before seizures

The method can identify preictal states in real-time

Simple algorithms effectively highlight synchronization variations

Abstract

Epilepsy is a neurological disorder arising from anomalies of the electrical activity in the brain, affecting about 0.5--0.8\% of the world population. Several studies investigated the relationship between seizures and brainwave synchronization patterns, pursuing the possibility of identifying interictal, preictal, ictal and postictal states. In this work, we introduce a graph-based model of the brain interactions developed to study synchronization patterns in the electroencephalogram (EEG) signals. The aim is to develop a patient-specific approach, also for a real-time use, for the prediction of epileptic seizures' occurrences. Different synchronization measures of the EEG signals and easily computable functions able to capture in real-time the variations of EEG synchronization have been considered. Both standard and ad-hoc classification algorithms have been developed and used. Results on scalp EEG signals show that this simple and computationally viable processing is able to highlight the changes in the synchronization corresponding to the preictal state.