Transcranial Magnetic Stimulation (TMS)¹ is a non-invasive and safe method to modulate brain activity via electromagnetic induction and it is FDA approved for treatment of depression and pre-neurosurgical localization. Due to its ability to temporally precisely modulate local cortical processing, TMS has found many applications in basic neuroscience as well². Navigated TMS enables repeated stimulation of a given target both within and between sessions³. While highly useful, all of the existing commercially available navigators employ simplified models (e.g., spherical head approximations or “ball and stick” targeting) to estimate the stimulation “hot spot”. Previous work suggests that a relatively simple realistically shaped boundary element model (BEM) can improve the accuracy of the estimation of the TMS-induced electric field (E-field)⁴. Here, we employ a MRI-based BEM with four layers (skin, skull, cerebrospinal fluid (CSF), and brain) to compute the E-field distributions at the cortical gray-white matter boundary and study the influence of the CSF on the computational targeting and dosing of TMS.The tissue boundary surfaces were segmented from T1- and T2-weighted anatomical MRI data (Figure 1 (A-B)). We utilized a dataset from the Human Connectome Project (HCP) database collected using the 3T Siemens Connectom scanner located at the Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital. The MPRAGE and T2-SPACE datasets along with the acquisition parameters can be accessed at (http://www.humanconnectome.org/documentation/MGH-diffusion). The SimNIBS pre-processing pipeline⁵ that uses FreeSurfer⁶ , FSL⁷, and Meshfix⁸ software packages was adopted to produce the boundary surfaces (Figure 1 (C)) for skin, outer skull, inner skull, and brain (consisting of gray matter and cerebellum boundaries). The E-fields at the gray-white matter boundary of the left cerebral hemisphere were calculated based on the reciprocity principle between TMS and magnetoencephalography (MEG)⁹. The numerical implementation was based on the Helsinki BEM library¹⁰ core routines modified to accommodate the CSF (i.e., the volume between brain and inner skull boundaries) as an additional conductivity layer. The following values for the conductivities were used: brain = 0.33 S/m, CSF=1.79 S/m, skull=0.0066 S/m, and skin=0.33 S/m. The numbers of vertices for the final re-sampled BEM surfaces were: brain = 5207, inner skull = 3335, outer skull = 829, and skin = 9108. In addition, the computation was repeated for a case where the conductivity of the CSF was set to be equal to that of the brain to investigate the effect of omitting the CSF compartment. A TMS coil with wire winding geometry of a typical commercially available figure-of-eight configuration (C-B60, MagVenture, Farum, Denmark) was assumed. The E-field amplitude depends linearly on the current rate of change (dI/dt) so it only affects the relative scale but for consistency a typical value of 60 A/μs was assumed.First we investigated two separate stimulation locations 1) MC=motor cortex and 2) PFC=prefrontal cortex both with two coil orientations i) AP=anterior-posterior and ii) LR=left-right. The results shown in Figure 2 indicate that the omission of the CSF layer results in reduced E-field amplitudes of about 15 % depending on the coil position and orientation but the overall shape of the E-field is similar. For the frontal location, the E-field shape appears somewhat more focal for the case with CSF layer included. For the case with CSF included at PFC target, we also observe a difference of ~15% in the E-field amplitude between the AP and LR coil directions. Figure 3 shows a quantitative map of the E-field amplitude differences when coil locations were simulated across the entire left hemisphere for both AP and LR orientations. The maximum E-field amplitudes were estimated for each location and the relative difference was calculated for the models with and without CSF. The results show that the difference depends on the coil orientation and location in a rather intricate way, with relative differences up to 25%.Our results are in line with previously published finite-element method (FEM) calculations¹¹ in that the inclusion of the CSF layer appears to increase the estimated amplitude of the TMS-induced E-fields. However, the previous work has not shown a systematic analysis across the whole brain and according to our results the overall pattern seems to be relatively complex. Further work is required to determine the main factors contributing to the observed effects.MRI-based computational methods are important for more accurate targeting and dosing of TMS. The presented computationally efficient BEM method can be employed using standard T1/T2 weighted acquisitions and can be utilized to optimize TMS targeting on an individual-subject basis.