Recursive independent component analysis (ICA)-decomposition of ictal EEG to select the best ictal component for EEG source imaging

Objective: This study aimed to present a new ictal component selection technique, named as recursive ICA-decomposition for ictal component selection (RIDICS), for potential application in epileptogenic zone localization. Methods: The proposed technique decomposes ictal EEG recursively, eliminates a few unwanted components in every recursive cycle, and finally selects the most significant ictal component. Back-projected EEG, regenerated from that component, was used for source estimation. Fifty sets of simulated EEGs and 24 seizures in 8 patients were analyzed. Dipole sources of simulated-EEGs were compared with a known dipole location whereas epileptogenic zones of the seizures were compared with their corresponding sites of successful surgery. The RIDICS technique was compared with a conventional technique. Results: The RIDICS technique estimated the dipole sources at an average distance of 12.86 mm from the original dipole location, shorter than the distances obtained using the conventional technique. Epileptogenic zones of the patients, determined by the RIDICS technique, were highly concordant with the sites of surgery with a concordance rate of 83.33%. Conclusions: Results show that the RIDICS technique can be a promising quantitative technique for ictal component selection. Significance: Properly selected ictal component gives good approximation of epileptogenic zone, which eventually leads to successful epilepsy surgery. (C) 2019 International Federation of Clinical Neurophysiology. Published by Elsevier B.V. All rights reserved.