Visualization of Dynamic Brain Activities Based on the Single-Trial MEG and EEG Data Analysis

Abstract

AbstractTreating an averaged evoked-fields (EFs) or event-related potentials (ERPs) data is a main approach in the topics on applying Independent Component Analysis (ICA) to neurobiological signal processing. By taking the average, the signal-noise ratio (SNR) is increased, however some important information such as the strength of an evoked response and its dynamics (trial-by-trial variations) will be lost. The single-trial data analysis, on the other hand, can avoid this problem but the poor SNR is necessary to be improved.This paper presents a robust multi-stage data analysis method for the single-trial Magnetoencephalograph (MEG) and Electroencephalograph (EEG) recorded data. In the pre-processing stage, a robust subspace method is firstly applied for reducing a high-level unique component (additive noise) in single-trial raw data. In the second stage, a parameterized t-distribution ICA method is applied for further decomposing the overlapped common components (sources). In the post-processing stage, the source localization or scalp mapping technique and post-averaging technique are applied for visualizing the dynamic brain activities. The results on single-trial MEG and EEG data analysis both illustrate the high performances not only in the visualization of the behavior and location but also in the visualization of the trial-by-trial variations of individual evoked brain response.KeywordsNoise VarianceIndependent Component AnalysisStandard Principal Component AnalysisApply Independent Component AnalysisDynamic Brain ActivityThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.