Head Modeling for E/MEG Source Analysis: Image Segmentation, Domain Decomposition and Solution of the Forward Problem

Issues in head modeling for E/MEG source analysis are discussed and some of the methods developed by our group are outlined. For image segmentation, a Kalman filter based edge tracing algorithm is described that performs well on noisy images and where intensity nonuniformity and partial volume effects are factors. For domain decomposition, a high resolution finite volume method is described that simplifies grid generation, meshing and source modeling. For the solution of the forward problem, a polynomial preconditioned conjugate gradient method is described that is efficient in terms of both computational speed and memory utilization.