TMS target comparison identifies motor network reorganization associated with behavioral improvement in writer’s cramp dystonia: A randomized, double-blind, Sham-controlled clinical trial

2.1 Study design

The study was a double-blind Sham-controlled cross-over design with data collected at Duke University Hospital between September 2018 and September 2022. The study was approved by the Duke Health Institutional Review Board (IRB # 0094131), registered on clinicaltrials.gov (NCT 06422104), and performed in accordance with the Declaration of Helsinki. All participants gave written informed consent prior to any study participation. Inclusion criteria were adults (> 18 years), diagnosed with isolated right-hand dystonia by a Movement Disorder Specialist, more than three months from the last botulinum toxin injection, more than one month from trihexyphenidyl medication, and able to sign an informed consent form. Exclusion criteria were any contraindications to receiving MRI or TMS.

2.2 MRI data acquisition and preprocessing

All study participants completed a brain imaging scan pre-TMS, and those who qualified also completed it after each TMS visit. All anatomical and functional imaging data was collected on a 3 Tesla GE scanner equipped with an 8-channel head coil. The anatomical MRI scan was acquired using a T1-weighted echo-planar sequence with the following parameters: voxel size: 256 x 256 matrix, repetition time (TR) = 7.316 ms, echo time (TE) = 3.036 ms, field of view (FOV) = 25.6 mm, 1 mm slice thickness. During fMRI sequences, participants copied holo-alphabetic sentences on an MRI-compatible writing tablet. The sentence writing was performed in a block design alternated by rest blocks (Fig. 1, task-fMRI panels). The CIGAL software presented visual writing instructions and recorded participants’ movements during the fMRI scan (6). Pre-TMS fMRI: Functional echo-planar images were acquired using the following parameters: voxel size: 3.5 x 3.5 x 4.0mm³, flip angle 90°, TR = 2 s, TE = 30 ms, FOV = 22 mm for 37 interleaved slices in ascending order, writing block: 20 s, rest block: 16 s, total: 12 blocks per fMRI. Post-TMS fMRI: Functional echo-planar images were acquired with the following parameters: voxel size: 2.0 x 2.0 x 3.0mm³, flip angle 90°, TR = 2.826 s, TE = 25 ms, FOV = 25.6 mm for 46 slices. Writing block: 20 s, rest block: 20 s, total: 5 blocks x 6 runs = 30 blocks per fMRI. fMRI images were preprocessed using fMRIPrep (7), as detailed in the supplementary methods. FMRIs with excessive head movements (defined as mean frame-wise displacement > 0.5 mm) were excluded from the study.

• Download figure
• Open in new tab

Figure 1: TMS target selection and TMS delivery.

To select a scalp target for TMS delivery, each subject completed a task-based fMRI at baseline. An individualized scalp target to the motor/premotor cortex (PMC) and primary somatosensory cortex (PSC) was then prepared using fMRI and electric field modeling. Subjects received three TMS conditions: 10 Hz rTMS to PSC, 10 Hz rTMS to PMC, and Sham rTMS to PMC (total 4,000 pulses). Each TMS condition was delivered over four stimulation blocks during a single visit. To synchronize the brain state and activate the motor network during TMS delivery, stimulation blocks were interleaved with writing blocks. TMS effect was measured using writing behavior and task-based fMRI.

2.3 MRI signal analysis

Pre-processed pre-TMS fMRIs from were input into FEAT analysis in FSL software version 6.0 (FMRIB, Oxford, UK) to generate a subject-level statistical map. A general linear model in which writing task timing (block design) was convolved with a double-gamma hemodynamic response function to generate the statistical brain map. To account for head motions, fMRIprep reported regressors (CSF, white matter, framewise displacement, and motion outliers) were regressed out from each statistical map. Spatial smoothing with a Gaussian kernel of full-width half-maximum of 5 mm and temporal high pass filter cutoff of 100 seconds was applied during the FEAT analysis. A WC group-level statistical map was then computed by importing all WC participants’ statistical brain maps from their baseline task-fMRI into a mixed-effects FLAME1 model.

2.4 Personalized TMS target selection

For each TMS study participant, a two-voxel cortical brain mask was generated for TMS targeting to the premotor and motor cortex (PMC) and for TMS targeting to the primary somatosensory cortex (PSC). To constrain the stimulation target to the PMC and PSC regions, each participant’s statistical brain map was overlayed on the WC group statistical brain map, and anatomical masks for precentral (for PMC) and postcentral gyrus (for PSC) from the Harvard-Oxford MNI atlas and the participant’s anatomical scan. Two consecutive voxels in the anatomic region of the left precentral/postcentral gyrus, with peak activation in the participant’s statistical map and the WC group statistical map and within 1 cm from the scalp surface were selected as the fMRI-guided PMC and PSC target for TMS delivery (Fig. 1, target selection panel). The fMRI-guided cortical brain masks were then used to perform prospective electric field (E-field) modeling on each participant’s scalp, as detailed in the supplementary methods. The purpose of the modeling was to identify the optimal position and orientation of the TMS coil on the participant’s scalp. This optimal coil setup was then used for the TMS intervention in the neuronavigation system (Brainsight, Rogue Research, Canada, version 2.4.9). The final personalized TMS targets at left PMC and left PSC for the 12 WC participants are shown overlayed on the MNI brain (Fig. 2).

• Download figure
• Open in new tab

Figure 2: Individualized targets for rTMS to PMC and PSC in WC.

The cortical target for rTMS delivery to the left PMC (blue) and left PSC (red) was developed using fMRI and electric field modeling. The final target for PMC-TMS and PSC-TMS for each WC participant is shown overlayed on a standard MNI brain.

2.5 TMS stimulation

Eligible participants received three TMS visits. The three TMS visits consisted of 10 Hz rTMS to PSC, 10 Hz rTMS to PMC, and Sham rTMS to PMC (Fig. 1, rTMS panel). Each rTMS visit was separated by a minimum of one week to allow signal washout. To negate any order effect, each participant was randomized to one of six possible orders for the three TMS conditions. All TMS was performed using an A/P Cool B65 coil attached to a MagVenture R30 device (MagVenture, Farum, Denmark). During each TMS visit, participants first completed an optimal motor cortex localization and motor threshold calculation as detailed in the supplementary methods (8). The active TMS paradigm delivered to each cortical target was 25 trains applied at 10 Hz rTMS with biphasic pulses and an inter-train interval of 10 seconds at 90% resting motor threshold (RMT) for a total of 1000 pulses delivered in a single block (∼five minutes), while participants were sitting in a recliner. To ensure a consistent brain state and specifically target the stimulation to the motor network, each stimulation block was preceded by a writing block (∼five minutes) in which participants performed a sentence copying task. A total of four blocks of TMS alternated with four writing blocks were performed (a total of 4000 pulses per TMS visit) (Fig. 1, rTMS panel). To perform Sham stimulation, the same AP coil was used in placebo mode, which produced clicking sounds and somatosensory sensation from scalp electrodes similar to the active mode but without a significant electric field induced in the brain (9). As previously reported, this type of stimulation allows participants to stay blinded during the experiment.

2.6 Retrospective TMS coil deviations

During each TMS block, data on the experimental TMS coil location and orientation was recorded every 500 ms in the neuronavigation system and snapped to scalp reconstruction mode prior to exporting it. Data was then imported into SimNIBS software (version 2.0.1 / 3.2.6) to calculate the deviations from the intended TMS coil position and orientation using the retrospective Targeting and Analysis Pipeline (TAP) (10). The TMS coil placement (position and orientation) data were first extruded outwards along the scalp normal by adjusting it for the participant’s recorded hair thickness for each TMS visit. These TMS coil placement data were then used to compute the coil placement deviation from the optimized coil setup constructed during the prospective E-field modeling reported in the supplementary methods. The deviations in the TMS coil placement were calculated in the normal and tangential planes and reported as changes in distance (mm) and angle (degrees). The direct distance between the actual and optimized target was also calculated, as previously reported (10, 11). Due to technical issues with the neuronavigation software, the experimental TMS coil location and orientation during one TMS block was excluded from one participant’s Sham TMS visit.

2.7 TMS-induced behavior changes

During each TMS visit, participants performed a behavioral writing assay before and after each four-block TMS session (Fig. 1, behavior assessment panels). The behavioral assay consisted of participants using a sensor-based pen on a digital tablet (MobileStudio Pro13; Wacom Co, Ltd, Kazo, Japan). Participants copied a holo-alphabetic sentence ten times in a writing software (MovAlyzer, Tempe, AZ, USA). The sensor-based pen recorded the x, y, and z positions and the time function of the participants’ writings. The writing software then automatically transformed the writing samples’ position parameters and time functions using a Fast Fourier transform algorithm to calculate the kinematic features. A previously detailed analysis of these kinematic writing measures showed that the sum of acceleration peaks in a single sentence (henceforth peak accelerations) (Fig. 1, behavior assessment panel, black circles) demonstrates high diagnostic potential (sensitivity, specificity, and intra-participant reliability), and associates with patient-reported dystonia and disability scales (12). In this study, the peak accelerations measure was used as the primary behavioral outcome measure. Participants performed the behavioral assessment before and after each TMS session. To minimize learning across the three TMS visits, a different holo-alphabetic sentence was used for each sequential visit. The three sentences were: Pack my box with five dozen liquor jugs; The quick brown fox jumps over the lazy dog; Jinxed wizards pluck blue ivy from the big quilt. To measure the change in peak accelerations, each participant’s post-TMS measure was normalized by the mean of their pre-TMS measure using the following equation: [(Post-TMS peak accelerations per sentence)/(mean peak accelerations for all ten sentences pre-TMS)]*100. Higher measures of peak accelerations represent greater writing dysfluency and worsening dystonia. The standard TMS adverse events survey and secondary outcome measures of clinician-rated and participant-reported dystonia scales were also collected before and after each TMS session as detailed in the supplementary methods.

2.8 TMS-induced fMRI changes

After each TMS visit, participants completed a task-based fMRI. The fMRI task design, acquisition settings, preprocessing, and run-level analyses are detailed in methods section 2.2 and 2.3 above. To compare changes in BOLD activity across the three TMS conditions, 4D time-series data were extracted from each fMRI run level FEAT directory using brain masks for regions of interest. The time-series data for each region of interest was z-scored across runs and analyzed as on-block and off-block writing epochs. To perform functional connectivity analysis, the extracted 4D time-series data for each region of interest was correlated pairwise using Pearson’s correlation (R) and Fisher z-transformed.

2.9 Brain parcellation and ROI extraction

To compare BOLD activity and functional connectivity analysis across TMS conditions, brain masks corresponding to regions of the motor network previously identified as abnormal in human dystonia studies were prepared (13–21). Anatomical brain masks were prepared using the Harvard-Oxford MNI brain atlas for the following subcortical regions: left caudate (CAU), left putamen (PUT), left globus pallidus (PAL), left thalamus (THL), left subthalamic nucleus (STN) and left substantia nigra (SN). Since the cortex and cerebellum are large brain regions, brain masks for these regions were made using the publicly available Dictionaries of Functional Modes (DiFuMo) brain atlas (22). DiFuMo is a fine-grain atlas that parcellated the brain into functional regions of 1024 components based on data from 15,000 statistical brain maps spanning 27 studies (22). To select DiFuMo brain masks for the left superior parietal cortex (SPC), left inferior parietal cortex (IPC), left supplementary motor area (SMA), and right cerebellar lobules VI and VIII (CBL), MNI coordinates from prior neuroimaging studies in dystonia were used (15, 23). For PSC and PMC, each participant’s personalized TMS target (section 2.4) was used as the center to prepare a custom 5 mm spherical mask.

2.9 Statistical Analyses

Because the study was a cross-over design with multiple visits and measures, a Mixed Model for Repeated Measures (MMRM) statistical analysis was used to compare differences within participants across the three TMS conditions. For the measure of TMS coil deviations, the MMRM covariate was the TMS condition. For examiner rating scales, the covariates were TMS condition, TMS visit, and clinician rater. For patient rating scales, the covariates were TMS condition and TMS visit. For the measures of peak accelerations, BOLD activity, and functional connectivity analyses, the covariates were TMS condition, visit, and interaction of TMS condition*visit. Differences across participants were accounted for by including participants as a random effect variable. To focus on clinically meaningful differences induced by TMS, functional connectivities that showed small effect sizes (defined as Cohen’s D ≤ |0.2|) between active and Sham conditions were excluded from statistical analyses. Changes in BOLD activity during the on-block and off-block writing were analyzed separately using MMRM. To perform brain-behavior correlation, BOLD activity or functional connectivity was correlated within participants with their measure of normalized peak acceleration behavior using Pearson’s (R). To compare differences in brain-behavior correlation across TMS conditions, a generalized regression analysis was performed. For the generalized regression, the dependent variable was peak accelerations, covariates were TMS conditions, BOLD activity or functional connectivity, and the interaction term (TMS condition*BOLD activity/functional connectivity). All MMRM were fit using restricted maximum likelihood. Covariance structures of the unstructured, autoregressive first order, and compound symmetry were tested and evaluated for stability of the model fit and ability to converge. Compound symmetry reached convergence for all of the models and, therefore, was utilized. All p values were corrected for multiple comparisons using the method of Benjamini-Hochberg, with an alpha of 0.05 accepted as statistically significant. For the exploratory functional connectivity to behavior correlational analysis, a less stringent alpha threshold of 0.1 was accepted as statistically significant.

Clinical characteristics

Thirty-four participants were assessed for study eligibility (Fig. 3). Of those assessed for eligibility, 24 WC participants met the inclusion/exclusion criteria to participate in the TMS study and completed a baseline fMRI before the first stimulation visit. The baseline fMRI from five participants was excluded due to other neurological disorders, structural abnormalities on the MRI brain, or excess head motion. FMRIs from 19 WC participants were then used to identify a group-targeting approach for TMS. Of these 19 participants, 14 consented to participate in the TMS research visits. Two participants who consented to TMS visits were taking a medication that increased the risk of seizures and, therefore, were excluded from undergoing TMS. Twelve WC participants (11/1 males/female; mean age 55 [SD 12.91] years) completed all three TMS visits. Due to technical issues during data collection, one participant’s TMS visit (Sham) was removed from the data analysis. Thus, data from 12 WC participants and 35 TMS visits (12 participants x 3 conditions – 1 visit = 35) were used for all analyses in this study. None of the 12 WC participants reported any adverse side effects of TMS. The mean symptom duration for the 12 WC participants was 16.4 years [SD 15.54].

• Download figure
• Open in new tab

Figure 3: Consort diagram showing the recruitment, inclusion, and exclusion numbers for participants.

A total of 34 WC dystonia participants were screened. Twenty-four participants completed fMRI, and 14 consented to the TMS study. Of these, 12 participants were randomized and completed the TMS study.

No differences in TMS technical delivery

To evaluate the technical delivery of TMS, the position and orientation of the TMS coil during the three conditions were analyzed retrospectively. There were no significant differences in the position or orientation of the TMS coil across the three conditions (Fig. 4 and Supplementary Table 1). Therefore, TMS delivery across the three conditions was technically comparable.

• Download figure
• Open in new tab

Figure 4: Minimal deviation between intended and actual TMS scalp targets.

Each graph shows the deviation in the TMS coil setup, defined as the difference between the intended and actual TMS coil parameters for the three TMS conditions. The deviation in TMS coil position is measured in the normal plane (A) and tangential plane (B). The direct distance between the intended TMS coil position and the actual TMS coil position is also measured (C). The deviation in TMS coil orientation is also presented in the normal plane (D) and tangential plane (E). Each data point on a graph represents the mean TMS coil deviation across all four TMS blocks for each subject. Data are presented for Sham-rTMS (n = 11), 10 Hz rTMS to PMC (n = 12), and 10 Hz rTMS to PSC (n = 12).

10 Hz rTMS to PSC, but not PMC, improved writing dysfluency

In a within-participant comparative design, in which all participants received stimulation to both sites and at the same frequency, we found that 10 Hz rTMS to PSC significantly decreased writing dysfluency compared to Sham-TMS (Fig. 5) [PSC vs. Sham: -1.73, SE: 0.41, t(21): -4.22, p = 0.001] and PMC-TMS [PSC vs. PMC: -1.28, SE 0.40, t(21): 3.23, p = 0.012]. TMS to PMC did not show significant differences in writing dysfluency compared to Sham [PMC vs. Sham: -0.45, SE 0.41, t(21): -1.09, p = 0.639]. These results confirm prior studies that TMS can modify behavior in WC dystonia and show the first results using 10 Hz frequency, within-participant site comparisons, and delivery under a task-activated brain state. Across the clinician rating (BFM right arm dystonia and WCRS movement score) and participant-reported scales (BFM writing score and ADDS), there were small but consistent improvements in dystonia symptoms after PSC-TMS compared to Sham (Supplementary Table 2, “Difference” column, positive values represent an improvement, and negative values represent worsening dystonia). However, the effect sizes of these coarse categorical rating scales were small and had large variability, resulting in no statistical differences across the three TMS conditions.

• Download figure
• Open in new tab

Figure 5: 10 Hz rTMS to PSC, but not PMC, reduced writing dysfluency in WC.

10 Hz rTMS to PSC significantly reduced a measure of writing dysfluency called peak accelerations compared to Sham-TMS in a within-subject analysis in WC subjects. PMC did not show significant differences in writing dysfluency compared to Sham-TMS. Each data point represents the mean change in peak accelerations for each TMS condition, with higher measures representing worsening writing dysfluency. **p<0.01, *p<0.05 after multiple comparisons correction.

10 Hz rTMS to both PMC and PSC decreased subcortical activity in the motor network

Considering the differential effect of the stimulation site on behavioral outcomes, we examined the extent to which TMS at these two target sites affected brain BOLD activity (Fig. 6A). Across the motor network, stimulation at each of the two TMS target sites showed similar patterns of brain activation (Fig. 6B). Specifically, 10 Hz rTMS decreased subcortical brain activity and increased BOLD activity at the superior parietal cortex. In the cerebellum, the brain activation pattern varied by TMS target site. PMC-TMS decreased BOLD activity in lobules VI and VIII, while PSC-TMS decreased BOLD activity in lobule VI only. Overall, these results suggest that cortically delivered 10 Hz rTMS decreased deep brain activity in the motor network with an overall activation pattern that was similar across the two stimulation sites. Changes in BOLD activity after TMS, therefore, do not explain the differential effect of stimulation site on behavioral outcomes.

• Download figure
• Open in new tab

Figure 6: 10 Hz rTMS to either PMC or PSC decreased subcortical activity in the motor network.

A) Graphs represent mean BOLD activity during the writing and rest blocks after each TMS condition [Sham (green), PMC-TMS (blue), and PSC-TMS (red)]. The mean BOLD activity is presented for brain regions of the primary somatosensory cortex (PSC, top graph) and caudate (bottom graph, PSC vs. Sham: -0.19, p = 1.65 x 10-5, PMC vs. Sham -0.14, p = 2.8 x 10-4, PSC vs. PMC: 0.05, p = 0.14). B) Heatmap represents mean BOLD activity during the writing block for subregions of the motor network across the three TMS conditions. Asterisks indicate significant differences in mean BOLD activity for PMC-TMS or PSC-TMS compared to Sham (MMRM and multiple comparisons corrected). SPC: superior parietal cortex, SMA: supplementary motor area, CAU: caudate, PUT: putamen, PAL: pallidum, THL: thalamus, STN: subthalamic nucleus, SN: substantia nigra, CBL-VI: cerebellum, lobule VI and CBL-VIII: cerebellum, lobule VIII.

PSC-TMS modified subcortical connectivity, distinct from PMC-TMS

We next considered if the behavioral outcome differences between the stimulation sites also corresponded to changes in functional connectivity (FC) after TMS. Figure 7 illustrates the changes in the magnitude of brain connections induced by PSC-TMS compared to Sham. In general, PSC-TMS weakened cortico-striatal connectivity (thin lines: PMC-CAU, PSC-CAU), cortico-cerebellar connectivity (SPC-CBLVI), and intra-cerebellar connectivity (CBL VI-VIII). PSC-TMS also strengthened striato-cerebellar connectivity (thick lines: PAL-CBL-VI) and striato-nigral connectivity (PUT-SN) (Table 1). There were no connections that showed significant differences between PMC-TMS compared to Sham. Overall, these findings demonstrate two important points. 10 Hz rTMS to PSC interleaved with writing task predominantly changed subcortical brain connectivity. Second, changes in brain connectivity induced by TMS may explain the differential effect of stimulation sites on behavioral outcomes.

• Download figure
• Open in new tab

Figure 7: 10 Hz rTMS to PSC but not PMC-induced changes in cortical-subcortical connectivity in the motor network.

Connectome represents TMS-induced changes in functional connectivity in the motor network. Each line represents a connection between two regions (black circles) of the motor network. Only connections that showed significant change after PSC-TMS compared to Sham were presented (p < 0.05 after multiple comparisons correction). There were no connections that showed significant change after PMC-TMS compared to Sham. The magnitude of connectivity change is represented by the line thickness (thin line: weakened, thick line: strengthened) and values reported in Table 1.

TMS-induced activation of PSC is associated with a reduction in writing dysfluency

We next asked if there was a relationship between TMS-induced brain activation and behavior changes and if this relationship was dependent on the stimulation site. A brain activity-behavior correlational analysis was performed for all brain regions in Table 1 that showed significant differences in functional connectivity between PSC-TMS and Sham. Among all these regions, only PSC BOLD activity showed a significant correlation with peak acceleration behavior. Specifically, TMS-induced increase in PSC BOLD activity was associated with a significant reduction of peak accelerations in WC participants (R = -0.84, p = 0.02) (Fig. 8). In contrast, no brain-behavior correlations were observed after TMS to PMC (R = -0.39, p = 0.75) or Sham (R = 0.15, p = 0.76). When compared across TMS conditions, there was a significant difference in brain activity-behavior association between PSC-TMS and Sham (PSC vs. Sham = -3.32, p = 0.0084 generalized regression and multiple comparisons corrected). Collectively, these results suggest that TMS-induced changes in the association between brain activity and behavioral measures were dependent on the stimulation site, and TMS-induced activation of PSC was associated with a reduction in writing dysfluency.

• Download figure
• Open in new tab

Figure 8: TMS-induced increase in BOLD activity at primary somatosensory cortex correlated with reduced writing dysfluency.

Graph represents the correlation between BOLD activity at the primary somatosensory cortex (x-axis) and behavior of peak accelerations (y-axis) for the three TMS conditions. Each data point represents the correlation between a WC participant’s PSC BOLD activity and peak accelerations behavior for each TMS condition. The shaded blue region represents the confidence region for the fitted lines. PSC vs. Sham difference = -3.32, p=0.0084 generalized regression and multiple comparisons corrected.

Reorganization of the motor network after PSC-TMS was associated with reduced writing dysfluency

Lastly, we explored if there was an association between TMS-induced changes in functional connectivity and behavior in the motor network and if this association was dependent on the stimulation site. From this analysis, we made three observations (Fig. 9). First, under Sham-TMS condition (untreated dystonia), nine significant FC relationships with writing dysfluency were identified, with the majority of them in connection to the cerebellum (cortico-cerebellum, subcortical-cerebellum, intra-cerebellum). Second, compared to the Sham, correlations between writing dysfluency and FC involving the cerebellum were no longer present following 10 Hz rTMS at either PMC or PSC. The loss of FC-behavior correlations in these regions was observed to a greater extent after PSC-TMS than PMC-TMS. Third, PSC-TMS differed from the other stimulation sites in brain-behavior correlations at cortical (PSC-SPC) and subcortical (CAU-SN, PUT-PAL, PUT-SN) brain regions (Fig. 9B, indicated by the “+” sign). Of these cortical and subcortical connections, PSC-TMS significantly differed from Sham in brain-behavior associations at the cortico-cortical (PSC-SPC) connection (PSC vs. Sham: -14.6, p = 0.075, generalized regression, multiple comparisons corrected) (Fig. 9A). Collectively, these results demonstrated that reduced writing dysfluency after PSC-TMS may be mediated by loss of brain-behavior associations to the cerebellum, and/or gain of brain-behavior associations to cortical and subcortical brain regions. Furthermore, the association between writing dysfluency and intracortical connectivity (PSC-SPC) may be a key signature for brain-behavior changes after PSC-TMS.

• Download figure
• Open in new tab

Figure 9: Reorganization of the motor network after PSC-TMS correlated with reduced writing dysfluency.

A) Graph represents the correlation between functional connectivity (FC) (x-axis) and peak acceleration behavior (y-axis) for the three TMS conditions. Each data point represents a WC participant’s connectivity-behavior relationship. Strengthening of connectivity between the primary somatosensory cortex and superior parietal cortex was associated with a reduction in peak accelerations behavior after PSC-TMS compared to Sham-TMS (PSC vs. Sham: -14.6, p = 0.075, generalized regression, multiple comparisons corrected). B) A heatmap of the brain connectivity-behavior relationship in the motor network. Each box represents the mean correlation (R) between peak accelerations behavior and a functional connection for each TMS condition. Heatmap is reported only for connections with R ≥ 0.6 for at least one TMS condition (indicated by an asterisk). “+” indicates connectivity-behavior relationships that are different for PSC-TMS compared to both Sham-TMS and PMC-TMS.

Overall, TMS target comparison demonstrated that 10 Hz rTMS to the primary somatosensory cortex but not the premotor cortex significantly changed functional connectivity and markedly redistributed brain-behavior associations that spanned cortical and subcortical regions of the motor network.