Detection of the epileptiform signals by Independent Component Analysis methods

Decompostion of epileptic EEG signals by Independent Component Analysis (ICA) methods in order to detect the epileptiform transients burried in noise and artifacts is the main motivation of this study. We used FastICA, Infomax and JADE as conventional ICA methods as well as Topographic ICA (TICA), a newer method relaxing the statistically independence assumption of former methods. Simulated and real EEG data are used to compare the performance of these ICA methods. We also showed that post-processing by Wavelet Denoising subsequent to ICA decomposition offers a better detection of the epileptic activity.