Realistically shaped volume conductor models of the head

Usually the head is described as a set of concentric spheres, each compartment with linear, isotropic, homogeneus conductivity. However, the solution of the inverse problem for MEG/EEG will be more precise if the compartments have a realistic shape obtained automatically from MRI-scans of each subject of experiment and the inverse problem solved by the boundary element method. The procedure is both expensive and time consuming. Dramatic increases in efficiency might result from use of a standard realistically shaped model, a possibility is tested here by means of simulations. Models from MRI-scans of three subjects were scaled in different ways. One model acted as our standard model. It was used to calculate a magnetic field (MF) produced by a current dipole. Another model (scaled one) was used to compute from the MF distribution the location and strength of the source. For sources in the frontal and occipital parts, two of the models could be interchanged. The error in the dipole location was smaller than 2mm, and that in the strength was a few percent. The third model gave larger errors when it was interchanged with one of the other two, possibly due to this head looking quite different from the other two. The number of points used to calculate the magnetic field cannot be too small. In our case 57 was sufficient. Using a sphere instead of a realistically shaped model in the inverse procedure increased the errors which were larger in the frontal part of the head, especially for deep sources. The 256 MRI-scans available were used to generate various models of the same subject. The errors in the location procedure were smaller than 2 mm