Concurrent TMS/fMRI reveals individual DLPFC dose-response pattern

Background: TMS is a valuable tool in both research and clinical settings, playing a crucial role in understanding brain-behavior relationships and providing treatment for various neurological and psychiatric conditions. Importantly, TMS over left DLPFC is an FDA approved treatment for MDD. Despite its potential, response variability to TMS remains a challenge, with stimulation parameters, particularly the stimulation intensity, being a primary contributor to these differences. Objective: The objective of this study was to establish dose-response relationships of TMS stimulation in DLPFC by means of concurrent TMS/fMRI. Methods: Here, we stimulated 15 subjects at different stimulation intensities of 80, 90, 100 and 110 % relative to the motor threshold during concurrent TMS/fMRI. The experiment comprised two sessions: one session to collect anatomical data in order to perform neuronavigation and one session dedicated to dose-response mapping. We calculated GLMs for each intensity level and each subject, as well as at a group-level per intensity. Results: On a group level, we show that the strongest BOLD-response was at 100 % stimulation. However, investigating individual dose response-relationships showed differences in response patterns across the group: subjects that responded to subthreshold stimulation, subjects that required above threshold stimulation in order to show a significant BOLD-response and atypical responders. Conclusions: We observed qualitative inter-subject variability in terms of dose-response relationship to TMS over left DLPFC, which hints towards the motor threshold not being directly transferable to the excitability of the DLPFC. Concurrent TMS/fMRI might have the potential to improve response rates to rTMS applications. As such, it may be valuable in the future to consider implementing this approach prior to clinical TMS or validating more cost-effective methods to determine dose and target with respect to changes in clinical symptoms.

Successful TMS application requires optimal selection of stimulation target, intensity and protocol.In terms of stimulation intensity, the utilization of motor evoked potentials (MEPs) as a measure for cortical excitability has gained widespread acceptance in TMS research, owing to its cost-effectiveness and the relative ease with which behavioral outputs can be quantified.This approach has been established as a reliable tool for evaluating cortical excitability (Zadey et al., 2021).
However, the use of motor responses as a proxy for cortical excitability is not without its challenges.Firstly, since different brain regions possess varying cytoarchitectonic features that influence their susceptibility to external stimuli such as transcranial magnetic stimulation (TMS) (Turi et al., 2021).Secondly, cortical areas exhibit significant differences in their remote connectivity patterns to other functional hubs of the targeted network, complicating the interpretation of results (Biswal et al., 2010;Schuler et al., 2019).Thirdly, intrinsic excitability fluctuations may differ considerably across cortical regions (Dugué et al., 2011;Gal et al., 2010;Johnson et al., 2012).
As such, behavioural studies were inconclusive about the relationship between motor threshold and phosphene threshold stimulating different areas of the visual system (Antal et al., 2004;Battelli et al., 1999;Boroojerdi et al., 2002;Deblieck et al., 2008;Gerwig et al., 2003;Stewart et al., 2001;Stokes et al., 2013;Thompson et al., 2009).In terms of studies investigating prefrontal areas there has been evidence in EEG studies that MT and excitability in the DLPFC show different patterns (Gogulski et al., 2023;Kähkönen et al., 2005;Lioumis et al., 2009;Raffin et al., 2020).However, these studies provide no precise information about the spatial extent of the dose effect.Interleaved TMS/fMRI provides the opportunity to investigate the exact spatial underpinnings of TMS effects, but has been questioned to be able to provide information about local intensity-related effects (Rafiei and Rahnev, 2022).This has been explained, among others, by the low SNR at the stimulation spot in former experiments.
In order to tackle this shortcoming of previous studies, we investigate dose-response relationship between BOLD and TMS intensity at a local level as well as in terms of whole-brain effects with a TMS/fMRI setup of increased SNR (Navarro de Lara et al., 2017).In order to establish dose-response profiles, subjects were stimulated at different intensities of their respective motor thresholds (80, 90, 100 and 110 %).This method allows for determining excitability patterns of the DLPFC relative to the individual motor threshold.We show that TMS applied to the left DLPFC at threshold level, results in the highest increase in BOLD-response on a group level.Individual dose-response profiles indicate, however, strong inter-individual variability in excitability patterns

Sample
15 healthy right-handed subjects (8 women, age 26.32 ± 3.93) participated in the experiments after giving written informed consent.Two subjects were excluded due to excessive motion (>3 mm).One subject was not included in the analysis due to problems during segmentation / normalization.Two subjects did two sessions in order to investigate reproducibility of single-subject data.An additional sample of 10 subjects underwent the same protocol during sham stimulation.None of the subjects reported any history of neuropsychiatric diseases (DSM IV) or any clinical evidence of neurological conditions.The study was approved by the ethics board of the Medical University of Vienna and was conducted according to the Declaration of Helsinki and its recent amendments.

Experimental setup
The experiment comprised two sessions.In the first session, functional images were collected during a standard finger-tapping fMRI task (nine 10 s blocks each separated by 10 s rest) in order to localize the representative cortical area of the right first dorsal interosseous (FDI) muscle on M1.An anatomical T1-weighted scan was acquired in order to reconstruct skin and brain models to be used for neuronavigation.The second session consisted of a concurrent TMS/fMRI run and an additional anatomical scan performed with the dedicated TMS/fMRI coil setup.A schematic flowchart of the study protocol is shown in Fig. 1.

Magnetic resonance imaging
The study was performed on a 3T Prismafit scanner (Siemens, Erlangen, Germany).Images in the first session were acquired using the Fig. 1.TMS/fMRI Procedure.For concurrent TMS/fMRI, we used our tailor-made coil setup and an optimized multiband accelerated EPI sequence for highresolution, artefact-free imaging of acute TMS effects.After assessment of functional and structural markers, we used neuronavigation for defining motor thresholds and DLPFC targets.We used two receive coil arrays, one in combination with the TMS coil and one on the contralateral side to enable full brain coverage.EPI (TR=1 s) recording was conducted continuously intermitted by three pulses of 10 Hz TMS each 30 s (320 ms acquisition pause).Intensities were applied at doses of 80, 90, 100 and 110 % of the subjects' motor threshold in a randomized order and in sum 20 TMS triplets were applied.manufacturer's default 64-channel head coil.During the finger-tapping task, functional data was acquired using an EPI sequence with TR/TE = 1800/38 ms, 23 slices, 3 mm slice thickness.In addition, anatomical MPRAGE images were acquired (TR / TE / TI = 2300 ms / 4.21 ms / 900 ms, flip angle 9 • , 160 slices, 1.25 mm slice thickness).

Neuronavigation
In order to target the M1 functional hotspot for active motor threshold (AMT) definition and the stimulation target region in the DLPFC, we used Brainsight neuronavigation software (Brainsight, Rogue Research, Canada) in combination with a Polaris Spectra infrared camera (NDI, Waterloo, Canada).MR-compatible coil and subject trackers for online positioning and tracking as well as custom mouthpieces were 3D-printed (Fig. 2).

TMS
Stimulation was performed using an MRi-B91 MR-compatible TMScoil and a MagProX100 stimulator (Magventure, Farum, Denmark).For the stimulation sessions, biphasic pulses were used with a duration of approximately 290 μs.AMT was defined for each subject as the intensity at which five twitches in the right FDI muscle out of ten stimulations were observed.

Online TMS/fMRI
The TMS target region was imported into the neuronavigation software and optimized for each subject towards the center of BA46 (MNI -42, 28, 21, Tik et al. 2017) at an 45 • angle rotation from midline, perpendicular to scalp.For this purpose, subjects' brains were normalized using SPM12 and the target coordinate was back-projected from MNI space into subject space using inverse normalization parameters.We used our custom-built concurrent TMS/fMRI setup comprising two 7-channel surface RF-coils for data acquisition.One of the 7-channel surface coils was mounted underneath the TMS coil, while the other was placed on the contralateral side to guarantee whole brain coverage.The TMS coil was positioned over the left DLPFC using the custom-built MRI adapted neuronavigation trackers (Fig. 2).The stimulation protocol was based on triplets of 10 Hz rTMS being applied every 30 s at different intensities (80, 90, 100 and 110 % of the AMT), 4 repetitions each, in a randomised order.An in-house written TMS control class (http://www.tmsfmri.com/TMSfMRITools.zip) and paradigm script in Matlab (2010) version 7.10.0(R2010a) ensured precise timing of TMS pulses with respect to the MR scanner trigger signal to avoid artefacts.The total scanning time for the stimulation paradigm was 10 min 38 s.

Data preprocessing and statistical analyses
Analyses of fMRI data were performed using Matlab, SPM12 and cat12 toolbox as well as AFNI and ANTS preprocessing tools.Preprocessing comprised despiking (AFNI), bias-field correction (ANTS), segmentation and normalization of anatomical images (cat12), transformation of EPIs into MNI space using anatomical deformation fields and smoothing with a 6 mm FWHM Gaussian kernel (SPM12).For single-subject (first-level) analysis, linear regression was performed at each voxel, using generalized least squares with a global approximate AR(1) autocorrelation model, drift-fit with Discrete Cosine Transform basis (128 s cutoff).Stimulation events were convolved with SPM's canonical HRF and used as regressors of interest resulting in four regressors representing the different stimulation amplitudes (80, 90, 100 and 110 % of the individual's AMT) of 10 Hz TMS triplets over the left DLPFC.Realignment parameters obtained from the previous preprocessing steps were included in the model as nuisance regressors.
Resulting single-subject beta map estimates of BOLD responses were used for group analyses.Linear regression was performed at each voxel, using generalized least squares with a global repeated measures correlation model.For effects-of-interest analysis, an F-contrast was calculated.Statistical images were calculated based on the contrasts between each stimulation intensity and rest (p>0.05,FWE cluster-level corrected).
Hotspots at the stimulation sites were determined for each stimulation intensity by selecting the maximum closest to the stimulation target.The contralateral homologue site was chosen by the nearest maximum after inverting the x-coordinate sign of the left DLPFC maximum.Note that slight changes in the peak coordinates might be explained by changes in spatial distribution of the induced E-fields Fig. 2. Concurrent TMS/fMRI-setup.In-scanner tracking and neuronavigation setup with the Polaris infrared camera mounted on a movable stand.One 3D-printed subject tracker is placed on a subject customized mouth mount for optimal tracking during concurrent stimulation.A second 3D-printed coil tracker was placed at the frontal edge of the 7-channel TMS/fMRI coil array mounted underneath the TMS coil.
depending on stimulation intensities.In order to define individual doseresponse at the stimulated target region, we calculated the mean position of the center of the coil using a fully automatic script based on the MRI fiducial markers (Woletz et al., 2019) and selected the nearest maximum to this point in the individual t-contrast for stimulation over all intensities (80 %/90 %/100 %/110 %).We further selected the maximum in the right DLPFC in order to calculate remote dose-response relationship over the four regressors.Betas were extracted from individual beta maps.Standard-error-of-the-mean was calculated based on the SPM results (residual variance estimate, ResMS; variance-covariance matrix of parameter estimates, Bcov).

Results
Group-wise mean AMT was 80.60 % (sd=7.65) of maximum stimulator output, corresponding to a dI/dt of 138.33 A/μs, (sd=13.47).Note that the effective stimulation intensity is decreased in the TMS/fMRI setup by (a) cable length, (b) hardware to suppress leakages, and (c) increased distance due to imaging coil.

TMS at each intensity leads to statistically significant activation changes at group-level
As a first step, we investigated group-wise TMS-related brain activation from concurrent TMS/fMRI for each stimulation intensity (Tcontrasts for 80, 90, 100 and 110 %).In Fig. 3, the resulting activation maps for each stimulation intensity are shown overlaid onto a brain template of the averaged normalised (cat12) anatomies of all subjects.Statistically significant increases in activation were found in the targeted left DLPFC and the right DLPFC for each intensity setting.Remote effects following stimulation were found in several cortical regions including right premotor area (PMA), supplementary motor area (SMA), supramarginal gyrus (SMG), anterior cingulate cortex (ACC), middle to posterior cingulate cortex (MCC, PCC), anterior insula (AI) and posterior insula (PI).Subcortical activation was found in thalamus and basal ganglia including striatum and substantia nigra (SN).While on a grouplevel there is a trend towards sub-threshold intensities (80 and 90 % of AMT) leading to lower changes in neural activity at the left and right DLPFC as compared to supra-threshold intensities (100 and 110 % of AMT), at the whole-brain level no significant changes (p<0.05FWE cluster-level ) were found for t-contrasts between intensities sub-> supra-threshold.
Fig. 3 indicates that across-brain activation is strongest for a stimulation intensity of 100 % AMT.Quantitative results are shown in Table 1, where coordinates and t-values of hotspots in left and right DLPFC are given for all stimulation intensities.It can be seen that the most significant activation increase at the stimulation site in left DLPFC was indeed found for stimulation with 100 % of the AMT.Remarkably, this is also the case for the contralateral DLPFC hotspots.Additionally, we modulated linear increase in BOLD-response over all four intensities (Fig. S1).
Concerning sham stimulation there was no such response of the bilateral DLPFC.Unthresholded maps for cortical activation changes are depicted in Fig. 4.

Individual differences in TMS/fMRI dose-response profiles in target regions
Primary aim of this study was to examine the individual doseresponse profiles across the different stimulation intensities.In Fig. 5, TMS effects (beta parameter estimates and standard error) are plotted across all four intensities for all subjects and both hemispheres.For almost all subjects, above-threshold stimulation yielded stronger neural activation in both DLPFC hotspots.More specifically, in many of the subjects a linear increase in neural activation is seen within the left and the right DLPFC (see Fig. 5A).However, a subgroup of subjects shows a stimulation intensity threshold after which the activation seems to decrease with higher intensities (see Fig. 5B).Some subjects show doseresponse profiles, which are inconclusive.For instance, subject#10 shows a dose-response profile where sub-threshold stimulation seems to result in higher activation levels at the target site compared to abovethreshold intensities.
Regardless of the subjects' individual dose-response patterns, the right and left DLPFC show a large similarity in dose-response profiles strengthening the validity of found parameters.Taken together, these results indicate large inter-subject variation in stimulation doseresponse profiles.Most importantly the results show that stimulating with higher intensities does not necessarily yield higher activation at the target site for all subjects.

Reproducibility of dose-response
Comparing additional runs from two participants revealed a qualitatively similar response pattern in the stimulated DLPFC, while a less reproducible pattern in the right DLPFC (Fig. S2).

Discussion
In this study, we applied an improved concurrent TMS/fMRI setup utilizing two multichannel RF-coils for whole-brain coverage to investigate the direct effects of different TMS intensities (80, 90, 100 and 110 % of the active motor threshold) on the neural activation levels evoked by stimulation at the left DLPFC.We showed the effects of different TMS intensities delivered to the left DLPFC on a group-level as well as an individual subject level.There are two main results.We could show here that: (1) similar network structures are activated across the different stimulation intensities studied, (2) while higher stimulation intensities generally yielded stronger responses in left and right DLPFC, there is considerable inter-subject variability in individual dose-response relationships, (3) during active stimulation contralateral hemisphere gets stronger involved at all intensity levels.
Interestingly, left DLPFC stimulation at different intensities led to strong increases in contralateral hemisphere's activation.This might be due to compensatory effects following local disturbance of the left DLPFC (Friehs et al., 2023;Hartwigsen, 2018;Hartwigsen et al., 2013).
It remains, however, somewhat speculative, if the TMS effect on a local network level was acting in an inhibitory or excitatory manner.While it has been suggested that the interaction between stimulation intensity and intrinsic excitability defines functional facilitation or inhibition (Silvanto and Cattaneo, 2017;Silvanto et al., 2008), we have not yet unraveled how this pattern is reflected in BOLD response related to stimulation (Cash et al., 2017).

Dose-dependent activity changes
We showed in this study that TMS over left DLPFC leads to activation increase in bilateral DLPFC and its associated network hubs.Interestingly, group-level results indicate that stimulation at 100 % of the motor threshold results in stronger neural activity compared to stimulation at 110 % of the motor threshold.This group-level finding might be explained by the higher between-subject variance at 110 %.On an individual level, there are subjects that show increased BOLD response with higher stimulation intensity, while other subjects show a peak at 90 or 100 % and a subsequent decrease at higher intensities.The latter might possibly be due to stronger inhibitory processes triggered by strong disturbances introduced into the neural system (Cameron et al., 2020;Noda et al., 2017).Differences in response patterns might further be explained by different patterns of cell compositions in individual cortices resulting in differences in excitation/inhibition circuits.A recent clinical study on theta-burst stimulation could even show that subthreshold stimulation led to improved clinical outcome (Lee et al., 2021).In two participants we could additionally show that dose-response patterns are reproducible in the left DLPFC, while the right DLPFC shows higher variability.However, additional research is needed in order to investigate reproducibility on a larger scale.
Most previous concurrent TMS/fMRI studies on stimulation intensity effects were limited to the primary motor cortex, M1.Many of these studies compared sub-threshold intensities to above-threshold intensities.Only few studies measured dose-response curves with a range of intensities (for a review see Bergmann et al. 2021).Hanakawa et al. (2009) measured neural activity changes for single-pulse TMS with intensities ranging between 30 and 100 % of the MT and reported statistically significant activation in M1 only for the highest intensity.These results are corroborated by a later study reporting increasing BOLD responses in M1 with increased TMS amplitude when varying intensities of 1 Hz TMS trains (Navarro de Lara et al., 2017).Very recent TMS studies targeting M1 have used a fine-grained stepwise increase of stimulation intensities or E-fields to reveal dose-response relationships and generally found a monotonic increase on a single-subject level (Numssen et al., 2021;Weise et al., 2020).Still, studies investigating the effect of systematically varied stimulation intensities in regions beyond M1 are sparse.In a sham-controlled experiment, Dowdle et al. (2018) demonstrated activation increases over intensities within a network.They did, however, not observe dose-response effects for left DLPFC stimulation.The absence of a significant finding at the targeted DLPFC in their study might be due to the technical SNR challenges of concurrent TMS/fMRI setups (Mizutani-Tiebel et al., 2022) or the investigation on a group-level was possibly cancelling out stimulation effects due to non-responders.
The novel setup used in this study (Navarro de Lara et al., 2015) was aimed at overcoming some of the limitations seen in the standard MRI birdcage head coil setup, such as low sensitivity at the target site, difficulties with accurate TMS coil positioning and associated problems with insufficient stimulation intensity.In addition, we used a  Matlab-based script for accurately controlling TMS timing with respect to the MR scanner trigger signal (http://www.tmsfmri.com/TMSfMRITools.zip).This allowed us to apply TMS bursts between EPI sequences without loss of functional images from TMS-MR interaction artefacts.This multi-channel setup allowed us moreover to apply an accelerated multiband sequence to acquire more data in shorter time, which leads to increased power.
Using this setup, we were able to show that TMS over the left DLPFC leads to local as well as widespread network activity changes, which are modulated by stimulation intensity.Importantly, we found interindividual differences in TMS dose-response patterns in the bilateral DLPFC in which above-threshold stimulation intensities do not always result in the highest activation levels.These results indicate that the linear increase in dose-response found in M1 stimulation (Navarro de Lara et al., 2017) is not directly transferable to DLPFC stimulation.

Individual dose-response profiles
In a previous TMS/fMRI study, we have shown that with increasing dose, TMS over M1 leads to increased BOLD response (Navarro de Lara et al., 2017).Here we show that the situation in DLPFC is different and the simplistic M1-mechanism is not directly applicable to DLPFC stimulation, as previous TMS-EEG studies already have suggested (Gogulski et al., 2023;Kähkönen et al., 2005;Lioumis et al., 2009;Raffin et al., 2020).We found that there are at least three groups of responders: below-threshold responders (strongest response below 100 %), above-threshold responders (strongest response at or above 110 %) and atypical responders.This clearly questions the standard approach in TMS applications to base DLPFC stimulation intensity purely on MT in clinical research as well as in cognitive neuroscience (Cole et al., 2021;Chen et al., 2023;Hanlon et al., 2013;Hanlon et al., 2016;Williams et al., 2018).
In any case the optimal dose for TMS treatment remains to be identified.While some studies argued that stimulation intensity has no direct effect on treatment response, an iTBS study on a healthy population could show that the effect on working memory was maximized at 75 % of MT, comparing between 50, 75 and 100 % stimulation intensities (Chung et al., 2018).Future research should investigate the effects of optimal stimulation intensities in terms of treatment response and behavioural outcomes.

Challenges and future research
One limitation of the current study is that we examined different stimulation intensities at the same stimulation frequency, i.e. 10 Hz.This approach has also been suggested in the literature since single pulse TMS effects might be in some cases too short to be picked up with BOLD imaging (De Graaf et al., 2018 Rafiei andRahnev, 2022).It is unclear whether single-pulses and bursts engage the same target networks (Bergmann et al., 2021).While studies have investigated the effects of both 1 Hz and 10 Hz bursts, the intensity might interact with the frequency of the stimulation in terms of affecting neural activity.Future studies might look into the interaction effects between stimulation intensity and stimulation frequency by systematically varying both of these parameters.
Another limitation of the study is that we used AMT, which might have led to lower intensity stimulation compared to RMT application.Furthermore, in this study we did not investigate state dependence or intrinsic connectivity patterns on an individualized level.While engaging in a task, dose-response might have resulted in different response patterns, while targeting personalized networks might have provided an even more specific dose-response pattern.
A major limitation is that while we observed noticeable doseresponse differences among subjects, they lack statistical and clinical validation.As the original study design did not include a sham condition, the study employed separate subject groups for the real and sham conditions, which could lead to increased inter-subject variability which reduces statistical power and makes direct comparisons difficult.To address these limitations, future research should consider employing a within-subject sham control design.Critically, even the absence of a statistically significant BOLD response for lower doses might not necessarily be interpreted as ineffective doses.Future studies have to measure behavioral outputs in combination with neural response to different intensities in order to validate the ideal dose for the respective anticipated outcome.Since hotspots in activation patterns also vary spatially among subjects, a direct comparison between subjects' doseresponse on a single target might probably not account for the individual functional brain circuits associated with the participants' individual responses to stimulation.
Concerning clinical application of this protocol there might be more cost-efficient E-field based approaches in the future (Numssen et al., 2021;Turi et al., 2022;Weise et al., 2020Weise et al., , 2022)).However, these are a more indirect measure of function and TMS/fMRI-studies have the potential to validate E-field-based dosing approaches for future application.
Another remaining challenge concerns subject motion.Subject motion can have a considerable effect on the TMS outcome due to increased coil-to-cortex distances and/or stimulations of non-target areas.In the current study, we were able to manage subject motion to some extent by using an MR-compatible online neuronavigation system to track both coil and subject position.However, it is unclear to what extent this monitoring method is transferable to the clinical population where subject compliance is much lower compared to healthy subjects.
Finally, we determined individual dose-response curves mainly at the target area and verified this stimulation pattern by looking at the same region on the contralateral side.It could be the case that for an optimal clinical outcome, the dose-response curves should be determined at a different brain region, or even be based on the functional connectivity changes between areas.Future studies could elaborate on these speculations by relating clinical treatment outcomes to doseresponse curves that were defined at different locations.

Conclusions
With this study, we were able to reveal intensity-related effects of TMS over the left DLPFC at the individual level.First, we were able to identify changes in the bilateral DLPFC for the different intensities in a group of healthy subjects.Second, we highlighted the individual differences in dose-response curves.Currently, clinical applications of TMS do not have access to an evidence-based method for determining stimulation intensity at the target site.Concurrent TMS/fMRI methods could provide an option for overcoming this shortcoming.TMS-evoked BOLD responses provide evidence of target engagement and thereby might allow for objective determination of individual dose-response relationships across the cortex.Future work is needed to establish robust doseresponse curves assessment, which could improve TMS basic research and treatment applications.To overcome the remaining limitations of this current study, future studies could use a within-subject sham control design and include behavioural measures in combination with a higher number of different intensities in order to validate the ideal dose for the desired outcome.

Data and code availability statement
All anonymised data and analysis codes are available upon request in accordance with the requirements of the institute, the funding body, and the institutional ethics board.

Fig. 3 .
Fig. 3. Comparison of TMS induced network activation at four different stimulation intensities.Immediate group-level brain response to stimulation with 80, 90, 100 and 110 % of the motor threshold is shown (p<0.05FWE cluster-level corrected).It can be seen that the different stimulation intensities activate a similar target network.

Fig. 4 .
Fig. 4. Comparison of TMS induced network activation for real and sham.Unthresholded maps on the cortical surface depict stronger bilateral BOLD increase due to real stimulation as compared to sham stimulation.

Fig. 5 .
Fig. 5. Individual differences in subjects DLPFC TMS dose-response relationships.This figure shows the beta estimates for stimulation at 80, 90, 100 and 110 % of the individual active motor threshold for left and right DLPFC peaks defined at the nearest local maxima to the center of the stimulation path intersection with the subject's brain.The middle column shows a slice of the actual subject's T-map (FWE clusterlevel-corrected p<0.05) overlayed on the indivdual's normalized brain.(A) While most subjects show a linear increase in BOLD between 80 and 110 %, others seem to respond only to intensities above the motor-threshold.(B) Some subjects show a peak at 90 or 100 %, with a consequent drop.

Table 1 t
-values and coordinates (MNI) of the peaks in left and right DLPFC for different stimulation intensities.