Comparative study of linear and nonlinear computational methods for EEG high frequency oscillations analysis in epilepsy

Abstract

Epilepsy is the most common serious brain disorder worldwide, and electroencephalography (EEG ) is the most common test for diagnosing it and gathering information about seizures. Over the last decade, high frequency EEG has attracted attention of researchers due to the possibility of high frequency oscillations (HFO: above 40 Hz) are used as biomarkers of epileptogenic zones. There appears to be considerable overlap in the frequency spectrum of physiological and pathological HFO, and there are currently no established criteria for distinguishing them. Also, despite several automatic methods for detecting HFO have been proposed, it remains an area of active research. Biological signals, such as EEG, have an intrinsic nonlinear and nonstationary nature, which cannot be fully characterized by traditional linear analysis methods, but which nonlinear analysis methods may deal with. This project aims to compare the application of nonlinear methods to EEG segments, within epileptic HFO context, with traditional methods used for HFO analyses, based on time-frequency spectrum and signal energy. The nonlinear analyses which will be performed are: qSDiff analysis; Detrended Fluctuations Analysis (DFA); Multiscale Entropy (MSE); Symbolic Dynamics (SD). Intracranial EEG database acquired from patients undergoing evaluation for epilepsy surgery at Mayo Clinic in Rochester, MN, USA will be analyzed. We expect to improve the means of analysis of these promising biomarkers of epilepsy, contributing to researches in areas like identification of epileptogenic regions and prediction of seizures. (AU)