2.2. Human model

A realistic finite element model was generated from T1-, T2- and diffusion-weighted MR images of a 25-year-old male subject (Fig. 3a). A detailed description of the construction process of an earlier version of the model can be found in Rampersad et al. (2014). All MRI scans were acquired on a 3T scanner (Magnetom Trio, Siemens, Munich, Germany) with a 32-channel head coil, 1 mm³ voxel size and a field of view that captured the complete head. The T1- and T2-weighted images were segmented into compartments representing the skin, skull compacta and skull spongiosa using a gray-value based active contour model and thresholding techniques. Eye, muscle and vertebrae segmentations were added manually. The foramen magnum and the two optic canals were modeled as skull openings. The segmentation was then converted into triangular surface meshes and smoothed using CURRY (Compumedics Neuroscan, Charlotte, NC). Segmentation masks of cerebral gray matter (GM), cerebral white matter (WM), cerebellar GM, cerebellar WM, brainstem and ventricles were extracted from a brain parcellation created with Freesurfer (http://surfer.nmr.mgh.harvard.edu). Triangular meshes of the brain surfaces were created and corrected using MATLAB and iso2mesh. Surface meshes of seven skull cavities were created with SimNIBS (Thielscher et al., 2015) using SPM12 (https://www.fil.ion.ucl.ac.uk/spm/) and the CAT12 toolbox (http://www.neuro.uni-jena.de/cat/index.html). All surfaces were then combined into a high-quality 3D Delaunay triangulation via TetGen. This procedure resulted in a mesh consisting of 787k nodes and 4.84M linear tetrahedral elements with a maximum element size of 1.8 mm³ in the brain. The total brain volume of the model (excluding cerebellum and brainstem) was 1070 cm³. The diffusion-weighted images were processed following previously described methods (Rampersad et al., 2014); conductivity tensors were calculated using the volume-normalized approach (Opitz et al., 2011) and multiplied with the effective conductivity values listed in Table 1 for GM and WM. All other compartments were assigned isotropic conductivities (Table 1)². Segmentation of deep brain structures used as anatomical targets for analysis of our results, the left hippocampus and right pallidum, were produced by Freesurfer and then mapped onto the tetrahedral mesh (Fig. 3b). To construct a target region for the left motor cortex (M1), the location of the cerebral representation of the first dorsal interosseus (FDI) muscle of the right hand was experimentally determined in the volunteer on which the model was based, using single-pulse transcranial magnetic stimulation and electromyography (Rampersad et al., 2014).

a) Geometry of the model. b) Location of the three selected brain structures: left hippocampus (blue), right pallidum (green) and FDI area of left motor cortex (red). c) Section of gray and white matter with arrows representing the preferred direction vectors n→pref. See Fig. S2b for an image of n→pref in the whole brain. d) Electrodes on the skin surface of the model. Each simulated configuration in Studies 1 and 2 consisted of two electrodes on the left side of the head (blue) that supplied current I₁ and two on the right (red) that supplied I₂. The configuration shown here is the standardized configuration used in Study 1 with electrodes at the C1, C2, C5 and C6 locations of the 10-10 system.

Conductivities denoted with an asterisk were modeled anisotropically based on diffusion-weighted images.