Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Error bounds in MEG (Magnetoencephalography) multipole localization

Conference ·
OSTI ID:975132

Magnetoencephalography (MEG) is a non-invasive method that enables the measurement of the magnetic field produced by neural current sources within the human brain. Unfortunately, MEG source estimation is a severely ill-posed inverse problem. The two major approaches used to tackle this problem are 'imaging' and 'model-based' methods. The first class of methods relies on a tessellation of the cortex, assigning an elemental current source to each area element and solving the linear inverse problem. Accurate tessellations lead to a highly underdetermined problem, and regularized linear methods lead to very smooth current distributions. An alternative approach widely used is a parametric representation of the neural source. Such model-based methods include the classic equivalent current dipole (ECD) and its multiple current dipole extension [1]. The definition of such models has been based on the assumption that the underlying sources are focal and small in number. An alternative approach reviewed in [4], [5] is to extend the parametric source representations within the model-based framework to allow for distributed sources. The multipolar expansion of the magnetic field about the centroid of a distributed source readily offers an elegant parametric model, which collapses to a dipole model in the limiting case and includes higher order terms in the case of a spatially extended source. While multipolar expansions have been applied to magnetocardiography (MCG) source modeling [2], their use in MEG has been restricted to simplified models [7]. The physiological interpretation of these higher-order components in non-intuitive, therefore limiting their application in this community (cf. [8]). In this study we investigate both the applicability of dipolar and multipolar models to cortical patches, and the accuracy with which we can locate these sources. We use a combination of Monte Carlo analyses and Cramer-Rao lower bounds (CRLBs), paralleling the work in [3] for the ECD. Results are presented for both point sources and cortical patches.

Research Organization:
Los Alamos National Laboratory
Sponsoring Organization:
DOE
OSTI ID:
975132
Report Number(s):
LA-UR-01-0643; LA-UR-01-643
Country of Publication:
United States
Language:
English