Subspace Angles: A Metric for Comparisons in EEG and MEG

In forward head modeling, various approximations are made in order to keep the problem tractable. Simplifications can yield models ranging from simple spherical models to multi-tessellated arbitrary surfaces in a boundary element model (BEM). Spherical head models differ in the number of shells and the assumed conductivities. Other assumptions in the BEM include the choice of basis sets, such as constant, linear, or quadratic variations of the voltages across the individual areal elements, or the selection of error-weighting method, such as collocation, Galerkin, or `direct` methods. Numerical versus analytic integration can also yield numerical differences. These differences in parameters and approximations can yield models whose external fields (EEG potentials or MEG magnetic fields) differ for the same internal source configuration. Quantitative measures are needed to determine if these differences are significant.