Independent Component Analysis of EEG-fMRI data for studying epilepsy and epileptic seizures

Abstract

Here we present a method for classifying fMRI independent components (ICs) by using an optimized algorithm for the individuation of noisy signals from sources of interest. The method was applied to estimate brain activations from combined EEG-fMRI data for the exploration of epilepsy. Spatial ICA was performed using the above-mentioned optimized algorithm and other three popular algorithms. ICs were sorted considering the value: of the coefficients of determination R2, obtained from the multiple regression analysis with morphometric maps of cerebral matter; of the kurtosis, which features the signal energy. The validation of the method was performed comparing the brain activations obtained with those resulted using the General Linear Model (GLM). The ICA-derived activations in different datasets comprised subareas of the GLM-revealed activations, even if the volume and the shape of activated areas do not correspond exactly. The method proposed also detects additional negative regions implicated in a default mode of brain activity, and not clearly identified by GLM. Compared with a traditional GLM approach, the ICA one provides a flexible way to analyze fMRI data that reduces the assumptions placed upon the hemodynamic response of the brain and the temporal constrains.