TUT-3: Source Localisation with EEG/MEG and EEG/fMRI

Date: Thursday, April 12
08:00 - 12:30

Presented by

Dr. Jan C. de Munck, VU medical center, Netherlands and Dr. Vince Calhoun, The MIND Institute/University of New Mexico, United States

Abstract

The purpose of this course is to give an overview of data modelling and analysis of EEG/MEG/fMRI data. Despite different spatial and temporal characteristics of fMRI on the one hand and MEG/EEG on the other hand, the computational tools required to convert the raw data into meaningful images are very much related. For instance, when a model driven point of view is adopted, source localisation based on EEG/MEG/fMRI data is based on (non-linear) regression analysis. In this course, Jan de Munck will highlight several topics, such as MEG/EEG source localisation, fMRI regression analysis and EEG/fMRI hemodynamic response estimation, from within a common maximum likelihood framework. Vince Calhoun will present a data driven approach, largely based upon independent component analysis and will discuss the analysis of EEG/MEG/fMRI data separately as well as the fusion of these modalities to address the source localization problem.

As a priori knowledge, it is assumed that the student knows:

• basic matrixvector algebra
• the multivariate Gaussian distribution
• basic aspects of MEG/EEG/fMRI technology

Course Outline

• Basics of parameter estimation applied to MEG/EEG/fMRI
  • linear/nonlinear parameter estimation
  • the Maximum Likelihood principle (underlying assumptions)
  • OLS/GLS estimation
  • nuisance parameters (regressors of no interest)
  • confidence intervals
  • (partial) correlation coefficients, t-test and F-test
  • the multiple comparison problem
  • fMRI: preprocessing: motion regressors, smoothing
  • EEG/fMRI: estimation of hemodynamic response
  • MEG/EEG: forward modelling (basics)
  • MEG/EEG: computational aspects of the moving dipole model
• Data driven methods (basics)
  • Introduction to data driven analysis
  • ICA/PCA/Clustering
  • fMRI: applications of ICA
  • EEG: applications of ICA
• Source localisation on the basis of MEG/EEG/fMRI data
  • Simultaneous EEG/MEG
  • Spatio-temporal covariance estimation
  • MEG/EEG: forward modelling (advanced)
  • The stationary dipole model
  • The MUSIC algorithm
  • How many dipoles?
  • The coupled dipole model
  • Beamformers
• Data driven methods (advanced)
  • Data driven source localization
  • Views on EEG-fMRI Integration
  • Inferring source localization with joint ICA

Speaker Biographies

Jan C. de Munck obtained his bachelors degree in experimental physics from University of Leiden (1983), his masters degree in physics (cum laude) from University of Amsterdam (1985), and his PhD from the same university in 1989 (cum laude). After several post doc appointments in oceanography and radiotherapy he became staff member at the MEG centre of the VUmc in Amsterdam in 1996. Since 2004 he is associate professor and coordinator brain imaging at the VUmc. His main research interests are in computational aspects of brain imaging. He has contributed by scientific publications on topics as forward/inverse modelling based on MEG/EEG/EIT data, multimodality image fusion and fMRI data analysis. He is associate editor of IEEE Trans. SP.

Vince D. Calhoun obtained his bachelors degree from the University of Kansas, in 1991, his MA degree in 1993 from John Hopkins University and in 1996 his MS degree in Information Systems from the same university. His PhD in electrical engineering was obtained in 2002 from University of Maryland Baltimore County. In 2002 he was appointed assistant clinical professor at Yale, Department of Psychiatry and since 2005 associate professor. In 2002 he was also appointed assistant professor department of psychiatry at Johns Hopkins University. Since 2006 he is associate professor at the department of electrical and computer engineering at University of New Mexico. Furthermore, since 2002 he is director of the Medical Image Analysis Laboratory, at Institute of Living, Hartford. He is associate editor of International Journal of Computational Intelligence and Neuroscience and also of IEEE Signal Processing Letters. Dr. Calhoun´s research interest is to develop techniques for making sense of complex brain imaging data. Because each imaging modality has limitations, the integration of these data is needed to understand the healthy and especially the disordered human brain. Dr. Calhoun has created algorithms which map dynamic networks of brain function, structure, and genetics and how these are impacted while being stimulated by various tasks or in individuals with mental illness such as schizophrenia.