Feature Extraction from Distributions of Phase Synchronization Values of EEG Recordings

ABSTRACT

ABSTRACT Epilepsy is the most common neurological pathology. Despite treatments available to patients, only 58% to 73% will be free of seizures. This uncertainty in treatment outcomes can lead to other psychiatric affectations in cases where treatment success may be in doubt. Seizure prediction models (SPMs) emerged as a measure to help determine when patients may be susceptible to an imminent crisis. These models are based on the continuous monitoring of patient's EEG signals and subsequent continuous analysis to identify features that differentiate ictal from interictal states. This is an ongoing field of research whose aim is to establish a robust set of features to feed the SPM and obtain a high degree of certainty regarding when the next seizure will occur. In this work we propose the analysis of phase differences of EEG as a method to extract features capable of discriminating ictal and preictal states in patients; specifically, the numeric distance between Q1 and Q3 of the distribution of phase differences. We compared this values to other phase synchronization methods and tested our hypothesis getting a p < 0.0009 with our proposed method.