EPILAB: A Matlab® Platform for Multi-Feature and Multi-Algorithm Seizure Prediction Studies

Authors

Abstract

Purpose: A reliable seizure prediction system should be based on multi-feature and multi-prediction algorithms. The EPILEPSIAE project (http://www.epilepsiae.eu) aims to develop an intelligent alarming system able to predict seizures, and allowing patient to assess his actual risk in real-time. Having in mind the search for appropriate algorithms and their efficient combinations, EPILEPSIAE includes the development of a platform that should consider a broad range of features and prediction algorithms. This is the 'EPILAB�. It offers the possibility to apply seizure predictions algorithms, together with a broader set of Electroencephalogram (EEG) and Electrocardiogram (ECG) features.

Methods: The EPILAB platform was developed in an object oriented approach, promoting the easy incorporation of other methods. Memory problems, usual when processing large data files, are prevented by performing streaming over MATFiles. EPILAB implements a four-step processing framework. First of all data can be preprocessed, i.e. several filters can be applied to the data. Then, features can be extracted from EEG/ECG data. Features are grouped in three categories: EEG Univariate; EEG Multivariate (synchronization among others), and ECG. EPILAB also allows the application of feature reduction algorithms, such as principal component analysis and multidimensional scaling. The fourth and last step encompasses the application of prediction algorithms. Two classes of prediction algorithms are considered: computational intelligence algorithms, and threshold-based algorithms.

Results, and Discussion: EPILAB was applied to process several data sets, showing to be a very useful tool for seizure prediction studies. Future EPILAB releases will implement continuous data processing from separated data files, aiming real-time performance assessment. The communication with the future European Database of Epilepsy (another objective of EPILEPSIAE) is also expected, allowing the possibility to query and process data directly from the database.
The continuous development of EPILAB will result in a public version that will be supplied at the end of EPILEPSIAE project.

Keywords

Epileptic seizure prediction; MATLAB framework

Subject

epilepsy; seizure prediction; Fp7 project

Conference

4th International Workshop on Seizure Prediction, June 2009