Electroencephalography (EEG), the first function brain activity imaging modality, has several natural advantages over metabolic brain imaging modalities. EEG is noninvasive, low cost, and lightweight enough to be highly mobile. Two major shifts in scientific perspective on the nature and use of human electrophysiological data are now ongoing. The first is a shift to using EEG data as a source-resolved, relatively high-resolution cortical source imaging modality. The EEGLAB signal processing environment, an open source software project of the Swartz Center for Computational Neuroscience (SCCN) of the University of California, San Diego (UCSD), began as a set of EEG data analysis running on Matlab (The Mathworks, Inc.) released by Makeig on the World Wide Web in 1997. EEGLAB was first released from SCCN in 2001. Now nearly twenty years later, the EEGLAB reference paper [4] has over 6,750 citations (now increasing by over 4 per day), the opt-in EEGLAB discussion email list links 6,000 researchers, the EEGLAB news list over 15,000 researchers, and an independent 2011 survey of 687 research respondents reported EEGLAB to be the software environment most widely used for electrophysiological data analysis in cognitive neuroscience. Our statistics show that after over the past four years, EEGLAB adoption is still growing steadily. Here, we will develop a framework for thorough comparison of preprocessing methods, and will apply machine learning methods on the large body of data collected by our laboratory to build optimized, automated data processing pipelines. We will greatly augment the power of the EEGLAB environment by providing a cross-study meta-analysis capability and will revise the software architecture to use a file and metadata organization compatible with the Brain Imaging Data Structure (BIDS) framework first developed for fMRI/MRI data archiving. These tools will integrate the HED annotating system allowing for meta-analysis across large corpus of studies. We will implement beamforming within EEGLAB. We will develop a hierarchical Bayesian framework for clustering effective sources on multiple measures across subjects and studies, and will develop tools to perform statistical testing on information flow measures at these scales. Although EEG and MEG recording have co- existed for four decades, little available software can combine both data types, recorded concurrently (`MEEG' data), to enhance source separation. We recently showed that ICA decomposition also allows joint MEEG effective source decomposition and will integrate MEG and joint MEEG data decomposition and imaging into the EEGLAB tool set. We will build tools to use MRI- and fMRI-derived anatomical atlases to inform the interpretation of EEG and MEG brain source dynamics. These radical improvements will further the use of non-invasive human electrophysiology for 3-D functional cortical brain imaging in the U.S. and worldwide, thereby accelerating progress in noninvasive basic and clinical human brain research using highly time- and space-resolved measures of brain electromagnetic dynamics.
The EEGLAB signal processing environment is now used in many electrophysiological research and teaching laboratories worldwide. To accelerate progress in basic and clinical cognitive neuroscience, we will continue maintenance and development of the EEGLAB environment, introducing new tools for source separation and localization, source clustering, automatic artifact management, and across-studies meta-analysis, and will extend its scope to process magnetoencephalographic (MEG) and joint EEG/MEG data and to highlight parallels between EEG/MEG source dynamics and results of existing research using fMRI and other brain imaging.
Showing the most recent 10 out of 49 publications
