Item

Abstract

Electroencephalography (EEG) allows noninvasive assessment of neuronal brain activity by means of source reconstruction (inverse modeling). A state of the art modeling of the field distribution with high spatial resolution (~ 1 mm) is performed with the finite element (FE) method. Realistic FE models of the human head depend on the segmentation of the different tissues inside the head. A variety of segmentation algorithms—like adaptive fuzzy c-means (AFCM) or SPM’s tissue probability map (TPM) algorithm—estimate for each element the probabilities of belonging to certain tissue classes. In the classical way the most likely tissue determines the element conductivity. We refer to this as the ordinary model. Here, we tested alternative strategies for assigning conductivities to elements. We estimated a weighted average of conductivities of all tissues which had a probability higher than zero and assigned this to the elements. We tried geometric averaging and an anisotropic conductivity model, but in general we did not observe improvements over the ordinary model.