Spatio-temporal independent component classification for localization of seizure onset zone

Abstract

Localization of the epileptic seizure onset zone (SOZ) as a step in presurgical planning leads to higher efficiency in surgical and stimulation treatments. However, the clinical localization procedure is a difficult, long procedure with increasing challenges in patients with complex epileptic foci. The interictal methods have been proposed to assist in presurgical planning with simpler procedures for data acquisition and higher speeds. In this study, spatio-temporal component classification (STCC) is presented for the localization of epileptic foci using resting-state functional magnetic resonance imaging (rs-fMRI) data. This method is based on spatio-temporal independent component analysis (ST-ICA) on rs-fMRI data with a component-sorting procedure based on the dominant power frequency, biophysical constraints, spatial lateralization, local connectivity, temporal energy, and functional non-Gaussianity. STCC was evaluated in 13 patients with temporal lobe epilepsy (TLE) who underwent surgical resection and had seizure-free surgical outcomes after a 12-month follow-up. The results showed promising accuracy, highlighting valuable features that serve as SOZ functional biomarkers. Unlike most presented methods, which depend on simultaneous EEG information, the occurrence of epileptic spikes, and the depth of the epileptic foci, the presented method is entirely based on fMRI data making it independent of such information, simpler to use in terms of data acquisition and artifact removal, and considerably easier to implement.