Decomposing EEG data into space-time-frequency components using Parallel Factor Analysis

Area of Science:

• Electrophysiology
• Neuroscience
• Signal Processing

Background:

• Summarizing electroencephalographic (EEG) data efficiently is challenging.
• Previous methods like PCA and ICA have limitations in capturing spatial, spectral, and temporal dimensions simultaneously.
• Existing frequency/time decompositions often neglect spatial information.

Purpose of the Study:

• To apply a unique three-way decomposition (Parallel Factor Analysis - PARAFAC) to multichannel EEG data.
• To identify and characterize distinct spatial, spectral, and temporal components within EEG signals.
• To introduce a novel method (Source Spectra Imaging - SSI) for localizing electrical current sources from EEG spectra.

Main Methods:

• Framing EEG data as a three-way array (channel, frequency, time) for PARAFAC decomposition.
• Analyzing EEG recordings from five subjects during resting state and mental arithmetic tasks.
• Developing and applying Source Spectra Imaging (SSI) to estimate current source locations based on EEG spectra.

Main Results:

• Identified two common atoms across subjects with spectral peaks in theta and alpha ranges.
• Observed modulation of alpha and theta signatures by physiological state (resting vs. mental arithmetic).
• Localized theta atom activity to the anterior frontal cortex and alpha atom activity to the visual cortex using SSI.

Conclusions:

• PARAFAC provides a unique and comprehensive decomposition of EEG data.
• The identified theta and alpha components reflect distinct brain states and activities.
• The developed SSI method offers a novel approach for EEG source localization and artifact detection.