Distributed Statistical Source Localization

Advances in dense sensor array measurement of brain electrical activity promise new insights into tthe human brain. This project develops Sourcerer(R), an innovative software package for distributedanalysis of Electroencephalographic (BEG) and Event-Related Potential (BRP) data. Whereas conventionmodels approximate gross regions of the brain, Sourcerer will characterize distributed source configareas of cortex. Drawing from other Electrical Geodesics Incorporated (EGI) software, anatomic constmagnetic resonance imaging (MRI) will be available for the source modeling. Three optimization technevaluated in Phase I: polytope algorithm, simulated annealing, and Bayesian estimation. As its namerequires a skilled user for accurate results. Like any other electrical source analysis software, itneural sources rather than exact measurements. Unlike any existing approach, however, Sourcerer willcheck on its results: a Bayesian analysis of the statistical reliability of each source solution. Aerrors due to model misspecification and finite sampling of head electric fields will guide this vallocalization solutions and interpretations will be guided by the high spatial resolution yielded byLaplacian) estimates.