Using ICA and realistic BOLD models to obtain joint EEG/fMRI solutions to the problem of source localization

We develop two techniques to solve for the spatio-temporal neural activity patterns using Electroencephalogram (EEG) and Functional Magnetic Resonance Imaging (fMRI) data. EEG-only source localization is an inherently underconstrained problem, whereas fMRI by itself suffers from poor temporal resolu...

Full description

Saved in:
Bibliographic Details
Published in:NeuroImage (Orlando, Fla.) Fla.), 2009-01, Vol.44 (2), p.411-420
Main Authors: Brookings, Ted, Ortigue, Stephanie, Grafton, Scott, Carlson, Jean
Format: Article
Language:English
Subjects:
Algorithms
Brain
Brain - physiology
Brain Mapping - methods
Brain research
Computer Simulation
Data Interpretation, Statistical
Electrodes
Electroencephalography
Electroencephalography - methods
Humans
Image Interpretation, Computer-Assisted - methods
Magnetic Resonance Imaging - methods
Models, Neurological
Models, Statistical
NMR
Nuclear magnetic resonance
Principal Component Analysis
Reproducibility of Results
Sensitivity and Specificity
Citations: Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We develop two techniques to solve for the spatio-temporal neural activity patterns using Electroencephalogram (EEG) and Functional Magnetic Resonance Imaging (fMRI) data. EEG-only source localization is an inherently underconstrained problem, whereas fMRI by itself suffers from poor temporal resolution. Combining the two modalities transforms source localization into an overconstrained problem, and produces a solution with the high temporal resolution of EEG and the high spatial resolution of fMRI. Our first method uses fMRI to regularize the EEG solution, while our second method uses Independent Components Analysis (ICA) and realistic models of Blood Oxygen-Level Dependent (BOLD) signal to relate the EEG and fMRI data. The second method allows us to treat the fMRI and EEG data on equal footing by fitting simultaneously a solution to both data types. Both techniques avoid the need for ad hoc assumptions about the distribution of neural activity, although ultimately the second method provides more accurate inverse solutions.
ISSN:1053-8119
1095-9572
DOI:10.1016/j.neuroimage.2008.08.043