Specific Aims: NIH proposal for continued EEGLAB project funding
EEGLAB began as a set of Matlab (The Mathworks, Inc.) functions for electroencephalographic (EEG) data analysis and visualization released on the Web as The EEG / ICA Toolbox from The Salk Institute (La Jolla CA) by Makeig and colleagues fifteen years ago (1997). The first version of EEGLAB, incorporating a coherent software architecture, a well-organized graphic user interface, and a larger collection of signal processing and data visualization tools was released from the Swartz Center for Computational Neuroscience (SCCN) at UCSD in 2001. Now some ten years later, the EEGLAB reference paper (Delorme & Makeig, 2004) has over 2,300 citations (Google Scholar), increasing at a rate of over 1 per day, the opt-in EEGLAB discussion email list links over 4,500 researchers, the EEGLAB news list over 9,000 researchers, and a recent survey of 687 research respondents (Hanke & Helcencko, 2011) has reported EEGLAB to be the software environment most widely used for electrophysiological data analysis (by a wide margin). In addition, at least 35 EEGLAB plug-in tools have now been released by researchers from many laboratories. Since EEGLAB software is now a de facto standard supporting a wide range of EEG and other electrophysiological research studies and teaching labs, our responsibility to maintain and further enhance EEGLAB is, we feel, an important one.
A major shift in scientific perspective on the nature and use of electrophysiological data is now ongoing -- a shift from measurement and visualization of individual channel signals (in the 'recording channel space') to visualizing and interpreting the data directly within a suitable inverse model representing activity reaching the electrodes by volume conduction from a set of effective data sources in native 'brain space'. An equivalent shift, via the development and exploitation of an appropriate inverse imaging model, made possible the phenomena of structural and functional magnetic resonance imaging (fMRI) (Lauterbur, 1974; Mansfield, 1977). While the electrophysiological inverse problem is still difficult, dramatic progress is being made through combined use of multimodal imaging and modern statistical signal processing methods (Delorme et al., 2012) and the future for EEG both in basic brain research and for a wide variety of clinical and other purposes (Makeig et al., 2012) appears bright.
Recovering the considerable degree of spatial source resolution available in high-density scalp EEG and other electrophysiological data, while retaining its natural advantage over other functional imaging methods in temporal resolution, has begun to yield a steady stream of new information about patterns of distributed brain processing supporting human behavior and experience. Methods focused on modeling and measurement of data sources rather than features of the recorded channel signals themselves were first widely applied to magnetoencephalographic (MEG) data. EEG data, however, provide much of the same information as MEG (Liu, 2002) and have quite substantial and increasing cost and portability advantages, making promotion of new EEG methods for source space analysis of increasing interest and importance for brain and health research. However, applying new source signal and signal processing models to electrophysiological data is complex and increasingly involves application of modern mathematical methods whose details not within the training of most cognitive neuroscience and health research professionals. The readily extensible EEGLAB environment, therefore provides a relatively easy-to-use platform for applying new methods in basic neuroscience and clinical health research. We propose, therefore, to support both continued maintenance and further development of the EEGLAB signal-processing environment (following current project 5R01-NS047294-08).
Our proposed plans for accomplishing this may be grouped under three Specific Aims:
1. Strengthen EEGLAB core tools, by adding support at the STUDY level for and improved source localization including MR-based and electrode-position based source imaging and improved 3-D brain source graphics, modeling of effective connectivity, advanced approaches to source identification including independent component analysis (ICA) and beamforming, improved memory use and support for high-performance cluster/cloud/GPU computing, and continued support of a fully free and fully functional compiled version of EEGLAB.
2. Further support development and maintenance of externally developed tools and toolboxes that interface directly to EEGLAB. Important examples include extensive, in-house developed plug-in toolboxes BCILAB, SIFT, MPT, NFT, HeadIT, and MoBILAB, and externally developed toolboxes Fieldtrip, ERPLAB, and LIMO EEG. We will allocate resources to further support interfacing new toolboxes to work directly with EEGLAB studies, and will develop tools to export results to and from standard packages for advanced statistics, R, SPSS, and Statistica.
3. Maintain EEGLAB code and data, and further develop and educate EEGLAB user and developer communities using a new EEGLAB forum website and chatroom, a more extensive Online EEGLAB Workshop and associated online course, and the addition into the EEGLAB framework of a new open-source private laboratory database and public data sharing system, HeadIT, that interacts directly with EEGLAB.
- Scott Makeig, Arnaud Delorme, et al.
March 5, 2012