%0 Journal Article
%T Independent Component Analysis for Source Localization of EEG Sleep Spindle Components
%A Erricos M. Ventouras
%A Periklis Y. Ktonas
%A Hara Tsekou
%A Thomas Paparrigopoulos
%A Ioannis Kalatzis
%A Constantin R. Soldatos
%J Computational Intelligence and Neuroscience
%D 2010
%I Hindawi Publishing Corporation
%R 10.1155/2010/329436
%X Sleep spindles are bursts of sleep electroencephalogram (EEG) quasirhythmic activity within the frequency band of 11¨C16£¿Hz, characterized by progressively increasing, then gradually decreasing amplitude. The purpose of the present study was to process sleep spindles with Independent Component Analysis (ICA) in order to investigate the possibility of extracting, through visual analysis of the spindle EEG and visual selection of Independent Components (ICs), spindle ¡°components¡± (SCs) corresponding to separate EEG activity patterns during a spindle, and to investigate the intracranial current sources underlying these SCs. Current source analysis using Low-Resolution Brain Electromagnetic Tomography (LORETA) was applied to the original and the ICA-reconstructed EEGs. Results indicated that SCs can be extracted by reconstructing the EEG through back-projection of separate groups of ICs, based on a temporal and spectral analysis of ICs. The intracranial current sources related to the SCs were found to be spatially stable during the time evolution of the sleep spindles. 1. Introduction Sleep spindles are characteristic transient oscillations that appear on the electroencephalogram (EEG) during nonrapid eye movement (non-REM) sleep. They are characterized by progressively increasing, then gradually decreasing waveforms with frequencies ranging from 11 to 16£¿Hz. Sleep spindles characterize sleep onset, being one of the defining EEG waveforms of stage 2 sleep. They are affected by medication, aging, and brain pathology and may be involved in learning processes [1]. Analyses of scalp-recorded sleep spindles have demonstrated topographic distinction between two sleep spindle classes: ¡°slow¡± spindles, with spectral peak frequency at around 12£¿Hz, and ¡°fast¡± spindles, with spectral peak frequency at around 14£¿Hz. Slow spindles are more pronounced over frontal scalp electrodes, while fast spindles exhibit mainly parietal and central scalp distribution [2¨C5]. Independent Component Analysis (ICA) is a statistical technique used for solving the Blind Source or Signal Separation (BSS) problem [6, 7]. Suppose that data measured in an experiment are expressed through an n-dimensional vector , where is the number of measured data time samples. The BSS problem relates to recovering unknown ¡°source¡± signals from their mixtures, that is, the measured data, without prior knowledge about the mixing mechanism producing the measured data. Sources and measured data are related through where is the unknown ¡°mixing¡± matrix. It should be noted that, in the BSS context, the term
%U http://www.hindawi.com/journals/cin/2010/329436/