Introduction TMS is now widely used in research and holds therapeutic promise. Nevertheless there remains a high degree of variability in the effects reported across subjects, studies and treatment paradigms. Recently, computational approaches using realistic finite element models (FEM) ( Opitz et al., 2011 , Thielscher et al., 2011 , Windhoff et al., 2013 ) of the brain have enabled more accurate estimations of the electric fields generated during TMS protocols. These models help to differentiate between interindividual TMS variability due to gyral folding patterns and other confounds such as subject-specific conductivity anisotropy. Objectives Validation approaches of FEM-simulated electric fields induced by TMS by actual physiological responses, such as motor evoked potential (MEP) amplitudes are scarce. Here, we compared computational predictions with MR-guided TMS motor-mapping and fMRI. Furthermore, predictions of stimulated areas are compared between realistic finite element models and currently used projection approaches widely used in neuronavigation systems. Materials and methods Anatomical voxel based, diffusion weighted imaging (DTI) and functional MRI during voluntary finger movement was conducted on all subjects. An individual FEM model including conductivity anisotropy derived from DTI data was generated for each subject. We measured MRI-guided TMS-evoked MEP from four different muscles (FDI, ADM, ECR, FCR). The motor cortex was mapped at different locations with 1 cm spacing using 5 × 5 cm grids centered on the subject’s motor hot-spot for a given muscle. At each stimulus site, we evoked MEPs using two different TMS coil orientations (45° to midline and 90° to midline). For each coil position, FEM simulations of the TMS-induced electric field were generated and compared with TMS-evoked MEP amplitudes. In addition a theoretical comparison between different methods to determine stimulated cortical areas was conducted. Results FEM simulations reliably predicted motor areas as stimulation target independent of the given TMS orientations in a subject specific manner, although MEP amplitude maps differed significantly in spatial extent and amplitude across coil orientations. MEP amplitudes at different grid points were highly correlated with the perpendicular component of the electric field strength in regions with strong BOLD activations during voluntary finger movement. A similar relationship was found for the electric field component in direction of WM fiber bundles in M1. Realistic finite element simulations produce more robust results in indicating stimulated areas in comparison to projection approaches. Conclusion We conclude that taking subject-specific electric field distributions into account can improve our understanding of the variability of physiological measurements obtained during various TMS protocols. Furthermore, combining fMRI with FEM simulations may enhance the precision of targeting of brain circuits such as the dlPFC, which do not exhibit immediate behavioral responses to TMS.
Read full abstract