Characterization of Spontaneous Neuromagnetic Brain Rhythms Using Independent Component Analysis of Short-Time Fourier Transforms

Abstract

Unsupervised spatiotemporal and spectral characterization of spontaneous neuromagnetic brain rhythms over long time scales (minutes) would be useful for basic and clinical neuroscience. We recently showed that after applying a sparsifying transform (the short-time Fourier transform or STFT) to MEG data, independent component analysis (ICA) identified sources of oscillatory activity [1]. STFT on MEG data results in a 3-way data structure with temporal (time points), spatial (channels or source signals) and spectral (frequency bins) dimensions. Here, we propose to treat the 3-way data by using ICA to impose sparseness in the space–frequency dimension, resulting in a “spatial Fourier-ICA” (sFICA). Results of sFICA applied to STFTs of source-level MEG data from subjects who received natural stimulation or were resting suggest that sFICA is an efficient technique to identify both stimulus-related and intrinsic dynamics of neuromagnetic brain rhythms.