Training Datasets for Epilepsy Analysis: Preprocessing and Feature Extraction from EEG Time Series

Description

Abstract

Epilepsy, a complex neurological disorder affecting millions worldwide, is characterized by seizures. Electroencephalography (EEG) is vital for epilepsy assessment, providing insights into brain electrical activity and enhancing seizure understanding. Access to tagged training sets that include all seizure phases is essential for data-driven epilepsy analysis, including the detection, prediction, and forecasting of preictal and ictal stages. Using the sliding window technique, we extracted multiple features from preprocessed EEG time series of 20 patients from the Freiburg Seizure Prediction Database, utilizing a software tool we developed named the Training Builder. We extracted 1080 univariate and bivariate features using a two-second window length and time slips of 1 and 2 seconds, also assigning a binary class to indicate the presence or absence of epileptic seizures. These features offer a comprehensive view of seizure dynamics, facilitating the creation of accurate seizure detection and prediction models. Our feature extraction methodology enhances the performance of data science models, promising advancements in epilepsy management and treatment. This highlights the significance of time-sensitive datasets in improving diagnostic and therapeutic approaches.

Notes