TMS intensity and focality correlation with coil orientation at three non-motor regions

Depósito de Investigación de la Universidad de Sevilla https://idus.us.es/ This is the Accepted Manuscript version of an article accepted for publication in Physics in Medicine & Biology. IOP Publishing Ltd is not responsible for any errors or omissions in this version of the manuscript or any version derived from it. The Version of Record is available online at https://doi.org/10.1088/1361- 6560/ac4ef9 Accepted Manuscript under CC BY-NC-ND licence IOP Publishing Physics in Medicine & Biology Journal XX (XXXX) XXXXXX https://doi.org/XXXX/XXXX xxxx-xxxx/xx/xxxxxx 1 © xxxx IOP Publishing Ltd TMS Intensity and Focality Correlation with Coil Orientation at Three Non-Motor Regions Jose Gomez-Feria1,2, Mariano Fernandez-Corazza3, Juan F Martin-Rodriguez 1,2 and Pablo Mir 1,2 1 Unidad de Trastornos del Movimiento, Servicio de Neurología y Neurofisiología Clínica, Instituto de Biomedicina de Sevilla, Hospital Universitario Virgen del Rocío/CSIC/University of Seville, Seville, Spain 2 Centro de Investigacion Biomedica sobre Enfermedades Neurodegenerativas, CIBERNED, Madrid, Spain 3 LEICI-Instituto de Investigaciones en Electrónica, Control y Procesamiento de Señales, Universidad Nacional de La Plata (UNLP), CONICET, La Plata, Argentina. E-mail: [email protected]; [email protected] Received xxxxxx Accepted for publication xxxxxx Published xxxxxx Abstract Abstract—Objective: The aim of this study is to define the best coil orientations for transcranial magnetic stimulation (TMS) for three clinically relevant brain areas: pre-supplementary motor area (pre-SMA), inferior frontal gyrus (IFG), and posterior parietal cortex (PPC), by means of simulations in 12 realistic head models of the electric field (E-field). Methods: We computed the E-field generated by TMS in our three volumes of interest (VOI) that were delineated based on published atlases. We then analysed the maximum intensity and spatial focality for the normal and absolute components of the E-field considering different percentile thresholds. Lastly, we correlated these results with the different anatomical properties of our VOIs. Results: Overall, the spatial focality of the E-field for the three VOIs varied depending on the orientation of the coil. Further analysis showed that differences in individual brain anatomy were related to the amount of focality achieved. In general, a larger percentage of sulcus resulted in better spatial focality. Additionally, a higher normal E-field intensity was achieved when the coil axis was placed perpendicular to the predominant orientations of the gyri of each VOI. A positive correlation between spatial focality and E-field intensity was found for PPC and IFG but not for pre-SMA. Conclusions: For a rough approximation, better coil orientations can be based on the individual’s specific brain morphology at the VOI. Moreover, TMS computational models should be employed to obtain better coil orientations in non-motor regions of interest. Significance: Finding better coil orientations in non-motor regions is a challenge in TMS and seeks to reduce interindividual variability. Our individualized TMS simulation pipeline leads to fewer inter-individual variability in the focality, likely enhancing the efficacy of the stimulation and reducing the risk of stimulating adjacent, non-targeted areas. Keywords— Brain stimulation, coil orientation, computer simulation, finite element analysis, patient-specific modeling, transcranial magnetic stimulation (TMS). 1. Introduction Transcranial magnetic stimulation (TMS) is a noninvasive brain stimulation technique based on the Faraday electromagnetic induction principle (Birss, 1961). Electric current is applied to conventional stimulation coils placed on the subject’s scalp, generating magnetic fields that easily penetrate the brain without being attenuated by the scalp or skull. These magnetic fields induce an electric field (E- field) in neural tissue, depolarizing the cells and increasing Journal XX (XXXX) XXXXXX Author et al 2 cortical excitability (Kobayashi and Pascual-Leone, 2003). The resultant E-field highly depends on the type of coil and its position on the scalp, with the figure-8 coil the most standardized model for achieving the largest focality (Deng et al., 2013). In general, single-pulse TMS (including pairedpulse TMS) is used to explore brain function, whereas repetitive TMS (rTMS) is used to induce changes in brain activity that can last beyond the stimulation period (Fox et al., 2013; Fregni et al., 2005; George, 2010; Rossi et al., 2009; Rossini and Rossi, 2007). Regardless of how TMS is used, there is interest in finding the coil orientation that induces the highest intensity of the E-field within a volume of interest (VOI) or reduces the stimulation in other regions of the brain beyond VOI leading to possible side effects or unwanted plastic changes (for example, in rTMS). For all applications, an important prerequisite for examining the interaction between TMS and cortex is to know the magnitude and spatial distribution of the induced E-field at the stimulation site (Fox et al., 2004; Janssen et al., 2015; Laakso et al., 2014; Antonenko et al., 2021). Numerous studies have emphasized the idea that the effectiveness of transcranial magnetic stimulation highly depends on the orientation of the coil relative to the subject’s head (Hand et al., 2020; Janssen et al., 2015; Laakso et al., 2018, 2014; Mikkonen et al., 2018; Minjoli et al., 2017; Rotem et al., 2014; Smith et al., 2017). This implies that the direction of the induced E-field has a major effect on the efficacy of TMS, an issue that has been widely investigated in the motor cortex (M1) (Bungert et al., 2017; Fox et al., 2013; Janssen et al., 2015; Krieg et al., 2015; Laakso et al., 2018, 2014; Opitz et al., 2011; Wagner et al., 2004; Windhoff et al., 2013). These studies have shown that the best coil orientation for M1 is obtained when the coil is placed 45 degrees to the medial sagittal plane of the subject’s head, a finding corroborated by computational simulations of TMS in M1 (Barker et al., 1985; Brasil-Neto et al., 1992). This orientation induces a posterioranterior (P-A) E-field that is perpendicular to the central sulcus. This coil orientation is the best for most individuals due to the similarity in anatomical morphology and orientation of the motor cortex and central sulcus (Eichert et al., 2020). However, these interindividual anatomical similarities are not usually found in other brain regions, such as non-motor regions, where sulci and gyri distributions tend to vary (Antonenko et al., 2021; Eichert et al., 2020; Syeda et al., 2017). Therefore, differences in morphological characteristics within the anatomy of the brain play an essential role in determining the orientation of the coil that induces the largest spatial focality of the E-field in the VOI (Opitz et al., 2016; Syeda et al., 2017). Various theories have been proposed to explain why the perpendicular component of the E-field plays such an important role in generating an optimal cortical response in M1, given that the direction of the normal component of the electric field along the pyramidal apical dendrites likely influences neuronal activation (Fox et al., 2004; Hiwaki and Inoue, 2009; Janssen et al., 2015; Krieg et al., 2015; Ledergerber and Larkum, 2010; Pisoni et al., 2018). However, the most accepted explanation is provided by the theoretical cortical column cosine model of TMS efficacy (C3-model) (Fox et al., 2004). According to this model, the cortical column behaves as a functional unit where the vector of the E- field induced by TMS is decomposed into two components: 1) the perpendicular component of the E-field and 2) the component parallel to the cortical column, which is the product of the absolute value of the E-field and the cosine angle between the column and the E-field orientations. The authors of the C3-model state that the electric field perpendicular to the cortical surface contributes the most to the resultant TMS-induced brain activation. Therefore, VOIs within the brain with a greater number of sulci would include regions with more cortical surfaces perpendicular to the TMS stimulation, thereby inducing stronger brain activation. This model also indicates that the E-field decreases with depth, inducing current densities below a certain activation threshold, thereby making the signal too weak to generate a cortical response. The importance of penetration into the depth of the E-field is also mentioned in the study by Sorkhabi et al. (Sorkhabi et al., 2020). They emphasized the need to determine focal and depth current densities for the correct stimulation of specific brain regions and concluded that weak current densities do not generate a sufficiently high response to properly stimulate specific neuronal populations (Deng et al., 2013). TMS studies in non-motor brain areas have found different rules regarding good coil orientations, but have been unable to find a preferred coil orientation for all individuals such as those identified for M1. The lack of a preferred coil orientation in cortical areas outside M1 can be due to a high intersubject variability in anatomical features (Benninger et al., 2011; Hamada et al., 2008; Jacobs et al., 2009; Janssen et al., 2015; Langguth et al., 2008; Mulckhuyse et al., 2011; Pobric and Hamilton, 2006) . Janssen et al. studied optimal coil orientations for a wide range of non-motor cortical regions but using only one head model (Janssen et al., 2015). The authors found the coil orientation that would induce the largest effective E-field in these regions without analyzing the field focality. These simulations were performed on a general but highly realistic head model, and the selected target regions were spheres measuring 3 mm in radius, with their centers located on the cortical surface. However, this VOI selection method might be inaccurate because a sphere does not represent the actual volume that contains the region of interest of a specific individual. Although this work concluded that there are better coil orientations for different non-motor regions, the results might be inconclusive because the authors only used one head model, and the potential interindividual Journal XX (XXXX) XXXXXX Author et al 3 variations in brain anatomy could potentially vary the optimal coil orientations. The authors also did not use an intensity threshold to select only those current densities that could cause a neural effect (Konakanchi et al., 2020). Several studies have indicated a threshold of approximately 50 V/m for an electric field capable of stimulating neuronal populations in TMS (Lee et al., 2015; Sorkhabi et al., 2020). As in these studies, we focused on analyzing the spatial focality of TMS on the various VOIs that would minimize the potential undesirable effects of repetitive TMS. The purpose of our study was to analyze, in 12 subjects, whether there is a better coil orientation in terms of spatial focality for the treatment of three non-motor VOIs that play an important role in various conditions: 1) the presupplementary motor area (pre-SMA), which is likely related to Parkinson’s disease, Tourette syndrome and central fatigue (Arai et al., 2012; MacDonald and Halliday, 2002; Potgieser et al., 2014); 2) the inferior frontal gyrus (IFG), which is involved in Tourette syndrome (Aron et al., 2003; Hampshire et al., 2010); and 3) the posterior parietal cortex (PPC), which is correlated with dystonia (Porcacchia et al., 2014). We also analyzed the importance of the perpendicular to the cortical surface component of the E-field to achieve the largest average field within the VOI and its correlation with the position of the coil handle. An important goal of this study is to correlate the best coil orientation established by spatial focality with the individual geometrical characteristics of the VOIs. We hypothesize that, due to inter-individual changes in anatomical brain properties, there is no globally preferred coil orientation for maximizing spatial focality, but there are optimal orientations for each individual. These best orientations might be related to the distribution of sulci and gyri within each individual’s VOI. We also hypothesize that the largest effective E-field (the average of the normal-to- cortex component of the E-field) is obtained when the coil handle is oriented primarily perpendicular to the sulci walls of the VOI. 2. Methods 2.1. Subjects Twelve healthy individuals (5 women, 7 men, mean age 30 ± 4 years) participated in the study. Magnetic resonance images (MRI) were taken at the Virgen del Rocio University Hospital (Seville, Spain) and inspected by doctors and technicians to corroborate the absence of any possible conditions and ensure that the selected subjects were healthy. The subjects were all informed about the MRI procedure and signed informed consent. The study was approved by the Ethics Committee of our institution and followed the standards of the Declaration of Helsinki. 2.2. Imaging We acquired structural T1 and T2-weighted magnetic resonance images using a 3T magnetic resonance imaging scanner (Ingenia, Philips Healthcare) and a 16 channel SENSE head. T1-weighted scans were acquired using the following parameters: repetition time (TR)/echo time (TE), 8.2/3.8 ms; flip angle 8; field of view (FOV) 240 mm; section thickness 1 mm; matrix 240 × 222; and voxel size 0.94 mm. The T2- weighted scans were acquired using the following parameters: TR/TE 3000/80 ms; FOV 245 mm; section thickness 4 mm; and matrix 444 × 315. 2.3. Geometrically realistic head models Before computing the current densities induced by TMS in the various subjects, the first step was the construction of individual and highly realistic tetrahedral head models (supplementary Fig. S1). Structural magnetic resonance images weighed T1 and T2 were segmented using the SimNIBS pipeline (Gomez-Feria et al., 2021; Saturnino et al., 2019). The segmentation contained eight different tissues with five different conductivities (Supplementary Table S1) (Opitz et al., 2011; Wagner et al., 2004), namely white matter, GM, cerebrospinal fluid, skull, eyes, sinus cavities, air cavities, and scalp (Supplementary Fig. S1). Within SimNIBS, we used the command ‘headreco’ to create the model, which relied on SPM12 and CAT12 for segmentation (Saturnino et al., 2019). The advantage of using SimNIBS is that it is a complete and fully automated process that does not require manual intervention. We used SimNIBS for surface extraction and the Iso2mesh MATLAB package (Fang and Boas, 2009) for the volumetric mesh generation. The resulting finite element head models have approximately 8 million of tetrahedra (~1.5 millions corresponding to the gray matter), with an average volume of 0.5 mm3 which can be considered as the spatial resolution of the models. We used custom MATLAB scripts for coil placement and electromagnetic simulations. 2.4. Definition of target region The brain regions in the right hemisphere were selected based on published atlases, selecting the pre-SMA from Sallet’s dorsal frontal connectivity-based parcellation atlas (Sallet et al., 2013), IFG from the Harvard-Oxford cortical structural atlas (Desikan et al., 2006), and PPC from the Talairach atlas (Lancaster et al., 2007; Talairach and Tournoux, 1988). Using the FSL command flirt linear transformation tool (Jenkinson et al., 2012; Smith et al., 2004; Woolrich et al., 2009) and the standard2example matrix, we registered each subject’s brain in the Montreal Neurosciences Institute (MNI) 152 space and extracted the selected VOI based on the corresponding atlas. We then performed a reverse normalization using an inverse matrix transformation to convert these regions back to each subject’s native space. Journal XX (XXXX) XXXXXX Author et al 4 Following this procedure, we were able to minimize interindividual errors in the VOI selection (Fig.1). 2.5. Induced E-field modeling The electrical forward problem in TMS can be stated as follows: given the head model, a coil model and location, the electrical conductivities of the various tissues, and a stimulator intensity, calculate the electrical field in the head, more precisely in the GM volume. This problem is governed by Maxwell’s equations, where the resulting electric field 𝐸𝐸 �  in the conductor volume can be defined as follows (Wang and Eisenberg, 1994): 𝐸𝐸 �  =−∇ � �  𝜙𝜙−𝜕𝜕𝐴𝐴 𝜕𝜕𝜕𝜕 (1) where ϕ is the electric potential and A is the magnetic vector potential. The second term on the right side of Equation 1 (𝜕𝜕𝐴𝐴/𝜕𝜕𝜕𝜕) is the primary field, which depends only on the location of the coil and can be calculated analytically. The first term (−∇ � �  𝜙𝜙) is the secondary electric field that is produced by the induced currents and depends on the geometry of the volume and the conductivities. This secondary field requires solving the following Poisson equation (Wang and Eisenberg, 1994): ∇ � �  ⋅�𝜎𝜎∇ � �  𝜙𝜙�=−∇ � �  ⋅�σ𝜕𝜕𝐴𝐴 𝜕𝜕𝜕𝜕 � (2) where σ represents the electrical conductivity tensor with the Neumann boundary condition that the electric current does not leave the volume (i.e., �𝜎𝜎∇ � �  𝜙𝜙�⋅𝑛𝑛 �  = 0, where 𝑛𝑛 �  is the normal vector of the boundary). The left side of Equation 2 �∇ � �  ⋅�𝜎𝜎∇ � �  𝜙𝜙�� is equivalent to the Poisson equations that govern the forward problems of electroencephalography, transcranial electrical stimulation, and electrical impedance tomography (Fernández-Corazza et al., 2020, 2013). The right side of Equation 2 is the source term, which changes for the different techniques. In TMS, the sources of the scalar potential ϕ are the components of the primary electric field. We modelled two 8-shaped commercial coils: the Magstim D70 Alpha flat coil (70 mm in diameter) (Magstim, Wales, UK) and the Magstim D40 Alpha flat coil (40 mm in diameter) and set a power stimulator output of 50%. The characteristics of these coils, such as the number of turns and the internal/external radius, were required to run the simulations and were extracted from other studies (Laakso et al., 2014; Salinas et al., 2007; Thielscher and Kammer, 2004), or asked directly to the manufacturer Magstim. The characteristics we employed to model the D70 coil are as follows: distance between centers, 47 mm; minimum radius, 28 mm; maximum radius, 45 mm; and 9 turns. For the D40 coil, the characteristics where as follows: distance between centers, 38 mm; minimum radius, 10 mm; maximum radius, 36 mm; and 15 turns. Both coil models (40 mm and 70 mm) presented the coil handle tangentially and not perpendicular to the coil. We placed each coil in the plane tangential to the scalp, with the center of the coil placed at the closest point of the scalp to the center of the VOI and separated from the scalp by 5 mm, following the study by Laakso et al. (Laakso et al., 2014). Supplementary Fig. S2 shows the models of the two coils employed (40 and 70 mm). We solved the primary field distribution in Equation 1 analytically in terms of elliptical integrals, following the study by Wang et al. (Wang and Eisenberg, 1994) in Equation 34, assuming a source current derivative in the D70 coil of 165 A/µs, as described in the study by Laakso et al. (Laakso et al., 2014) for the Magstim 2002 stimulator. For the D40 coil, we asked the manufacturers (Magstim) and they reported us 369.3 µs for the pulse width and 92.4 µs for the rise time, from which we estimated a source current derivative of 89.3 A/µs. The electric potential ϕ was calculated from Equation 2 numerically using our in-house first-order finite element method with the Galerkin approach (Kwon and Bang, 2000; Preston, 1984). We solved the forward problem for eight coil orientations (45° steps) (Fig. 2) and the two types of coils, for each subject and for each VOI. Given the lack of substantial differences in the results related to the orientations of the two types of coil, we focused the analysis only on the 40-mm coil simulations. We opted for this coil instead of the 70mm coil because the evaluated regions are more difficult to target (and the 40mm coil is expected to Fig. 1. VOI selection: pre-supplementary motor area (pre-SMA), inferior frontal gyrus (IFG), and posterior parietal cortex (PPC). The left column shows the VOIs painted in red. The second column shows an example of the TMS induced electric field at 0% threshold percentile. The right column shows the E-field module only in the VOI mask. Journal XX (XXXX) XXXXXX Author et al 5 be more focal) and because the 40mm coil is the one used in our laboratory. 2.6. Metrics of evaluation To obtain the best simulation angle, we studied the spatial focality of the module, normal, and tangential components of the E-field (Supplementary Fig. S3) to determine the accuracy of the spatial targeting and the avoidance of stimulating nontarget brain regions, which could result in negative clinical side effects (Deng et al., 2013). We defined the spatial focality as the ratio of the integral of the E-field in the VOI and above a certain threshold to the integral of the E-field induced in the whole GM (also above the same threshold) (Fernandez- Corazza et al., 2016; 2021). More intuitively, the spatial focality definition used in this work is the ratio between the desired stimulation and total gray matter stimulation (which includes both desired and undesired stimulation). If the focality value equals one (or 100%), it means highest focality, and if it approaches zero, it means poor focality. This definition was applied to a set of thresholds of the E-field determined as percentiles of the total E-field: 0%, 10%, 50%, and 65% (Fig. 3). For the analysis, we focused on the results for the 65% percentile, which was a conservative threshold for most subjects and orientations, considering the 50 V/m neuronal activation threshold (Lee et al., 2015; Sorkhabi et al., 2020). 2.7. Anatomical analysis Lastly, we studied the relationship between the coil orientations and the anatomical properties of our VOIs by analyzing the sulci and gyri geometry within each VOI. We separated the wrinkled cortical surface within each VOI into two portions: surface mostly parallel to the coil plane (mostly gyri) and surface mostly perpendicular to the coil plane (mostly sulci). This separation was done by computing the angle between the normal vector and the coil plane and setting a threshold of 45 degrees to separate these two surface components. We first calculated the percentage of sulci and gyri within a selected region (that is, pre-SMA, IFG, or PPC) and compared this percentage with the spatial focality in our 12 subjects, obtaining a linear regression that correlated the spatial focality and percentage of sulcus or gyrus. Next, we calculate the average direction of the normal vector of the sulcus within the VOI and compare it with the direction of the predominant component of the TMS-induced E-field (which, in a tangential figure-of-eight coil, is equal to the direction of the coil’s handle). With this analysis and knowing that the perpendicular to the cortical component of the E-field plays an important role in obtaining the best cortical response, our goal was to find a method to predict the best coil orientation by knowing only the direction of the normal sulcus’ vector. We further assessed the association between the E-field intensity and the alignment of normal vector sulci with the normal component of the E-field with linear mixed-effect models in the three VOIs. As fixed effects, we entered the alignment of normal vector sulci with the normal component of the E-field, and the subject was entered as a random effect. To assess a potential association between intensity and focality repeated measures correlation analyzes were performed (Bakdash and Marusich, 2017) for the three VOIs, as implemented in the R package rmcorr. 3. Results The average computation time is approximately three hours for the creation of the individual-specific model (~2.5 hs for surface extraction in SimNIBS and ~30 minutes for the generation of volumetric meshes using iso2mesh), and ~6 minutes for each calculation of the induced E-field (~5 hs for each patient considering 8 orientations, 3 regions and 2 coils). All computation times consider a PC with Intel(R) Core(TM) i7–7700 CPU with a 3.00 GHz quad core processor and 32 GB of memory. The mean spatial focality values for the three VOIs and eight coil orientations are shown in Fig. 4. There was not same coil orientation that produced the largest focality for all subjects. However, all coil orientations showed an overall similar Fig. 2. Coil rotation. Example of the different coil rotations over the presupplementary motor area (pre-SMA) every 45º and using the 40mm coil. The same rotation angles were applied to the inferior frontal gyrus (IFG) and the posterior parietal cortex (PPC) to find the best coil orientations. The red vector represents the direction of the coil handle. Fig. 3. Working with threshold percentiles. Different threshold percentiles were applied in each stimulation. 0% shows the full stimulation without any cutoff value. 65% only draws intensities over that percentile. Only intensities above that threshold are expected to generate a direct cortical response. Journal XX (XXXX) XXXXXX Author et al 6 spatial focality close to 60% for pre-SMA and IFG, and close to 88% for PPC. The polar plots in Fig. 4 show the differences in spatial focality between the individual subjects for each coil orientation. The lower images of Fig. 4 represent an example of a random subject with two coil orientations at the maximum and minimum spatial focality. Supplementary Fig. S4 shows the anatomy of those head models with models with the largest focality difference. A two-way ANOVA introducing VOI and coil orientation as intraindividual factors revealed a significant main effect for VOI (p<0.0001) but not for coil orientation or the interaction between coil orientation and VOI. Higher focality values were obtained for PPC, regardless the coil orientation (Tukey’s honestly significant difference, p<0.0001) (Fig. 5). We also calculated the average value of the proportion of sulci of the 12 subjects for the different VOIs, demonstrating a significant difference between the percentage of sulci in PPC when compared to the pre-SMA and IFG (Fig. 6). In TMS, the current moves parallel to the coil plane, and the presence of more sulci walls in the VOI that are perpendicular to this plane implies that there is a larger area in the VOI for the electric current to penetrate these walls and enter the brain perpendicularly, thereby increasing the VOI intensity to the outside-VOI intensity ratio (i.e., the spatial focality), which might explain why the PPC in Fig. 5 achieved the largest spatial focality on average. To test this, we performed a linear regression analysis to test the association between the overall percentage of sulci and gyri and the spatial focality achieved in each VOI (Fig. 7). Significant positive correlations were found between the percentage of sulci and spatial focality in pre-SMA and IFG, but not in PPC. On the contrary, negative relationships were observed between the percentage of gyri and the focality achieved in the three VOIs were observed. Next, we analyzed whether the alignment of the coil orientation with the normal vectors of the sulci could predict the intensities of the normal component of the TMS-induced E-field within the three VOIs. Results from the linear mixed- Fig. 4. Spatial Focality at pre-supplementary motor area (pre-SMA) (A), inferior frontal gyrus (IFG) (B), and posterior parietal cortex (PPC) (C). The bar graphs represent the average value of spatial focality at the percentile of 65% for the 12 subjects, showing no significant differences between different coil orientations. Each dot represents the value of the spatial focality of an individual subject for a selected coil orientation, showing high variability between subjects when the coil is fixed at one position. Polar plots represent the spatial focality values for the 12 subjects with different colors per subject. The red gross line of the polar plots shows the average value for each coil orientation for all subjects. Looking at the bar plots, it can be seen a slight increase of spatial focality in 90º and 270º for IFG, and 0º and 180º for PPC. However, these differences were not significant. The bottom images represent examples of the E-field in V/m induced by TMS above the 65% percentile. The third column shows the stimulated volume, inside and outside the VOI, delimited by the solid light green line. For each region, a representative subject was selected to represent maximum (top row) and minimum (bottom row) spatial focality depending on coil orientation. Fig. 5. Multiple comparison between pre-SMA, PPC and IFG for each coil orientation. Using a two-way ANOVA and Tukey’s multiple comparison test, differences in spatial focality between regions were analyzed. PPC showed a significant difference (p<0.0001) with pre- SMA and IFG, showing on average 88% of spatial focality, compared to 60% of pre-SMA and IFG. Fig. 6. Comparison of sulci percentage between VOIs. PPC showed higher percentage of sulci than pre-SMA and IFG. ****p<0.0001, p<0.001; **, p<0.01. Journal XX (XXXX) XXXXXX Author et al 7 effects models showed that a better alignment of the coil handle with most of the normal vectors of the sulci surface was significantly associated with a higher intensity of the normal component of the E-field, which held true for the three VOIs. These analyzes were performed for all coil orientations and for all subjects (Fig. 8). Finally, a repeated measures correlation analysis was conducted to assess a potential association between the normal component of the intensity of the E-field and the spatial focality. This analysis yielded positive correlations for both measures in IFG (r = 0.42, p < 0.00001) and PPC (r = 0.25, p = 0.018), while no relationship was detected in pre-SMA (r = -0.01, p = 0.89). 4. Discussion We used finite-element calculations in highly realistic head models of 12 subjects over three non-motor regions of interest: the pre-SMA, IFG, and PPC. The overall results showed that due to differences in anatomical properties between the subjects, there was no specific coil orientation that induced the largest spatial focality within our selected VOIs. In their study, Janssen et al. (2015) found better coil orientations for their studied regions, including only two of the three VOIs that we studied, using one head model for the TMS simulations without accounting for inter-individual variabilities. We found that interindividual differences in the optimal coil orientation (Figs. 4 and 7) could be explained by anatomical distribution of sulci and gyri within the VOIs. In fact, when analyzing the overall percentage of sulci within a VOI versus the spatial focality achieved in the same VOI, we found a positive association, increasing the spatial focality of the analyzed VOI in subjects who presented a higher percentage of sulci. In contrast, we found that spatial focality decreased when the gyri percentage increased, which is likely related to an increase in the perpendicular components of the E-field when more sulci or walls are present in the VOI. Thus, VOIs with more and deeper walls increase the perpendicular component of the E-field, which in turn generates an increase of spatial focality over that region. An interesting finding of this study is the better spatial focality in all VOIs of the PPC compared with the spatial focality achieved in the pre-SMA and IFG. When analyzing the overall percentage of sulci for each region, we found that it was significantly higher in the PPC than in the pre-SMA and IFG. This is likely because the PPC presents more stable landmarks across subjects, given that the PPC includes the intraparietal sulcus (Swisher et al., 2007; Whitlock, 2017). Thus, we showed that the PPC contains a larger percentage of sulcus on average, which in turn increases the spatial focality. Another novel approach in this study was the use of our selected VOIs based on the atlases in the MNI 152 space, which were subsequently transformed back into the native space for each subject, making the VOI selection process more accurate than when employing a fixed-radius sphere (Janssen et al., 2015). As the C3-model and other studies have stated (Bungert et al., 2017; Deng et al., 2013; Janssen et al., 2015; Laakso et al., 2018, 2014; Minjoli et al., 2017; Richter et al., 2013), cortical stimulation decreases with depth, which is due to the decrease in power of the E-field. Therefore, only electric fields above a certain threshold of approximately 50 V/m can generate cortical changes with sufficiently high intensity to induce a neural response, which translates into a larger effect and efficiency of the technique. In their study, Jannsen et al. analyzed coil orientations without selecting a threshold for the E-field or a power percentile (being equivalent to 0% percentile in our study), considering small current densities that might be insufficient for inducing a neural response. In agreement with the authors’ results and the C3-model (Fox et al., 2004), the perpendicular component of the E-field played an important role in generating maximum current densities when selecting a certain coil orientation. Our analysis showed that when the normal vector of the sulci matches the predominant E-field direction induced by the TMS (which, in the tangential figure-8 coil, is the same direction as the coil handle), the normal-to-cortex intensity Fig. 7. Relationship between spatial focality and sulci and gyri percentages. Linear regression analysis between the percentage of overall spatial focality and the percentage of sulci and gyri within the selected VOIs: pre-supplementary motor area (pre-SMA), inferior frontal gyrus (IFG), and posterior parietal cortex (PPC). Each dot represents the average of spatial focality for all coil directions in one subject. A clear relationship is shown for the three VOIs, except for the percentage of sulci in PPC (P=0.3551). Journal XX (XXXX) XXXXXX Author et al 8 reached maximum strength in that region. This finding suggests an ‘ad-hoc’ and practical orientation that is very close to optimal for increasing the intensity of TMS stimulation without the need to perform TMS simulations. In other words, to achieve the highest intensity of the normal component of the E-field, the handle of a figure-eight coil should be oriented perpendicularly to the orientation of the targeted gyrus. The predominant direction of the E-field will thereby match the predominant normal vector of the sulci for a specific VOI, and the normal component of the E-field penetrating the brain will be at maximum. Therefore, knowing the direction of the gyrus to stimulate can predict an approximate coil position that will likely deliver the largest average intensity in the target or VOI. However, this estimate is not as accurate as the entire TMS simulation model, which is far superior (if available) in determining the best coil orientation for specific individuals due to the complex anatomical topography of the brain. Moreover, optimal coil orientations can be obtained using certain optimization approaches (Gomez et al., 2021). Focality and normal intensity show moderate-to-large positive correlations for both IFG and PPC. This finding is somehow expected in some way when stimulating large regions of the brain, such as IFG and PPC (average IFG volume = 8.42 cm3, PPC = 13.27 cm3), as those coil orientations that result in an increase of the normal component of the E-field (intensity) at the VOI will also increase the spatial focality, measured in our study as the ratio between the VOI intensity and the total (VOI and non-VOI) intensity. In smaller areas, such as the case of pre-SMA (average pre-SMA volume = 6.71 cm3), this correlation was not found. One possible explanation to this is that, when stimulating small volumes, the electric field could dissipate more to outer regions from the VOI, thus, lowering the ratio of what is being stimulated inside the VOI vs outside the VOI (i.e. spatial focality). 5. Conclusion We studied the spatial focality and normal-to-cortex E- field intensities for three non-motor TMS targets (the pre- SMA, IFG and PPC) in 12 detailed individualized MRI-based head models. We found no coil orientation that maximized the spatial focality across all subjects, but we did find that maximum spatial focality is obtained in the VOIs with the largest proportion of sulci. Moreover, the PPC was the target with the largest spatial focality among the three studied VOIs and the region with the largest percentage of sulcus. The highest focality is sought in repetitive TMS to reduce undesired effects due to unwanted outside-VOI stimulation. The maximum averaged directional E-field was obtained for coil orientations where the handle of the coil was more perpendicular to the predominant sulci orientations within the VOI. We found a positive correlation between spatial focality and E-field intensity for PPC and IFG but not for pre-SMA. For the stimulation of the latter region, the prioritization of focality over intensity, or vice versa, should be adapted to the objectives of each application. Future experimental studies are needed to test the predictions of our approach. One possibility would be to apply dual coil TMS to assess effective connectivity or plasticity, at different coil locations and orientations, or in combination Fig. 8. Linear association between the E-field and alignment of normal vector sulci with normal component of E-field. Images on the left represent how the E-field increases when the normal component is aligned with the normal vector of sulci. Linear mixed-effect models that includes information for the 12 head models and eight coil orientations revealed that alignment of normal vector sulci with normal component could predict E-field, in the three VOIs (all p<0.0001). X-axis is the cosine of the angle α between the predominant orientation of the VOI sulci walls and the orientation of the coil’s handle. It ranges from 0 to 1, where a larger value means more alignment between these orientations. Images on the right show examples of maximum intensities within the VOI when normal vectors of sulci and normal components of TMS are more aligned; and examples of minimum intensities when these components are less aligned.