Statistics and Its Interface

Volume 4 (2011)

Number 4

A spatio-temporal solution for the EEG/MEG inverse problem using group penalization methods

Pages: 521 – 533

DOI: https://dx.doi.org/10.4310/SII.2011.v4.n4.a10

Authors

Zhimin Li (Center for Clinical Neurosciences, The University of Texas Health Science Center at Houston, U.S.A.)

Tian Siva Tian (Department of Psychology, University of Houston, Houston, Texas, U.S.A.)

Abstract

The inverse problem encountered in electroencephalography (EEG) and magnetoencephalography (MEG) studies refers to estimating neural activity given limited scalprecorded data. We propose a spatio-temporal solution using group penalization approaches. This proposed method is based on the assumption that the underlying sources of EEG/MEG measurements are smooth in the temporal domain, and focal in the spatial domain. It transforms the spatio-temporal problem to a high-dimensional linear regression problem with grouped predictors using a basis expansion. Then an iterative group elastic net algorithm is utilized to localize and estimate the source time courses. The proposed approach is shown to be effective on simulations and human MEG studies.

Keywords

inverse problem, spatio-temporal data, group elastic net, EEG/MEG

Published 17 November 2011