Finite element method (FEM) head model
The finite element method (FEM) solves the forward problem by discretizing the entire head volume into small elements (tetrahedra or hexahedra) and numerically solving Poisson's equation across them. Unlike BEM, FEM can represent arbitrary, spatially varying and anisotropic conductivities — skull layering and anisotropy, white-matter anisotropy from diffusion MRI, cerebrospinal-fluid pathways, surgical defects — at the cost of building a full volumetric mesh and greater computation.
FEM is the most flexible and, with an accurate segmentation, the most accurate forward approach, but its accuracy is only as good as the tissue segmentation and assigned conductivities it is given. Modelling the dipole source singularity inside the mesh (the source-modelling problem) also requires care, using approaches such as the subtraction, partial-integration or Saint-Venant methods.
FEM is not automatically better than BEM: a beautiful FEM mesh with wrong conductivities is worse than a simple three-shell BEM with good ones. Match model complexity to the reliability of the underlying tissue data.