Brain Stimulation 14 (2021) 391e403 Contents lists available at ScienceDirect Brain Stimulation journal homepage: http://www.journals.elsevier.com/brain-stimulation Cortical responses to noninvasive perturbations enable individual brain fingerprinting Recep A. Ozdemir a, *, Ehsan Tadayon a, Pierre Boucher a, Haoqi Sun b, Davide Momi a, g, Wolfgang Ganglberger b, M. Brandon Westover b, Alvaro Pascual-Leone c, d, e, 1, Emiliano Santarnecchi a, f, 1, Mouhsin M. Shafi a, 1, ** a Berenson-Allen Center for Noninvasive Brain Stimulation, Division of Interventional Cognitive Neurology, Beth Israel Deaconess Medical Center, Boston, MA, USA b Massachusetts General Hospital, Department of Neurology, Harvard Medical School, Boston, MA, USA c Hinda and Arthur Marcus Institute for Aging Research and Deanne and Sidney Wolk Center for Memory Health, Hebrew Senior Life, Boston, MA, USA d Department of Neurology, Harvard Medical School, Boston, MA, USA e Guttmann Brain Health Institute, Institut Guttmann de Neurorehabilitacio , Universitat Autonoma de Barcelona, Badalona, Spain f Department of Medicine, Surgery and Neuroscience, University of Siena, Italy g Department of Neuroscience, Imaging and Clinical Sciences, University of Chieti-Pescara, Chieti, Italy a r t i c l e i n f o a b s t r a c t Article history: Background: In recent years, it has become increasingly apparent that characterizing individual brain Received 22 June 2020 structure, connectivity and dynamics is essential for understanding brain function in health and disease. Received in revised form However, the majority of neuroimaging and brain stimulation research has characterized human brain 18 November 2020 function by averaging measurements from groups of subjects and providing population-level inferences. Accepted 9 February 2021 External perturbations applied directly to well-defined brain regions can reveal distinctive information Available online 12 February 2021 about the state, connectivity and dynamics of the human brain at the individual level. Objectives: In a series of studies, we aimed to characterize individual brain responses to MRI-guided Keywords: Transcranial magnetic stimulation transcranial magnetic stimulation (TMS), and explore the reproducibility of the evoked effects, differ- Electroencephalography ences between brain regions, and their individual specificity. TMS evoked Potentials Methods: In the first study, we administered single pulses of TMS to both anatomically (left dorsolateral Brain fingerprinting prefrontal cortex- ‘L-DLPFC’, left Intra-parietal lobule- ‘L-IPL) and functionally (left motor cortex- ‘L-M1’, right default mode network- ‘R-DMN, right dorsal attention network- ‘R-DAN’) defined cortical nodes in the frontal, motor, and parietal regions across two identical sessions spaced one month apart in 24 healthy volunteers. In the second study, we extended our analyses to two independent data sets (n ¼ 10 in both data sets) having different sham-TMS protocols. Results: In the first study, we found that perturbation-induced cortical propagation patterns are het- erogeneous across individuals but highly reproducible within individuals, specific to the stimulated re- gion, and distinct from spontaneous activity. Most importantly, we demonstrate that by assessing the spatiotemporal characteristics of TMS-induced brain responses originating from different cortical re- gions, individual subjects can be identified with perfect accuracy. In the second study, we demonstrated that subject specificity of TEPs is generalizable across independent data sets and distinct from non- transcranial neural responses evoked by sham-TMS protocols. Conclusions: Perturbation-induced brain responses reveal unique “brain fingerprints” that reflect causal connectivity dynamics of the stimulated brain regions, and may serve as reliable biomarkers of individual brain function. © 2021 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). * Corresponding author. ** Corresponding author. Berenson-Allen Center for Non-Invasive Brain Stimulation, Beth Israel Medical Center, Harvard Medical School, Boston, MA, USA. E-mail addresses: rozdemir@bidmc.harvard.edu (R.A. Ozdemir), mshafi@bidmc.harvard.edu (M.M. Shafi). 1 contributed equally. https://doi.org/10.1016/j.brs.2021.02.005 1935-861X/© 2021 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/ ). R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 Introduction temporal characterization of TEPs at the individual level consider- ably limits our understanding of individual response dynamics to Every human brain has a unique structural and functional pro- external perturbations, and thus restricts translational utility of file, and thus likely exhibits different patterns and dynamics of TEPs both in cognitive and clinical neurosciences [19]. brain activity both at rest and when performing a given task. A critical first step for translating spatial-temporal specificity of However, the vast majority of human neuroimaging studies to date TEPs into the clinical realm is to assess whether individual-specific have characterized human brain function by averaging measure- signatures in TEPs are reliable across measurements, and more ments from groups of subjects and providing population-level in- importantly, distinct enough to identify an individual from a large ferences [1,2]. Recent studies that seek to characterize brain group of subjects. Therefore, in a series of studies, we aimed to connectivity and dynamics at the individual level have employed characterize individual brain responses to image-guided TMS, and resting-state neuroimaging techniques (e.g. resting-state func- explore the reproducibility of the evoked effects, differences be- tional Magnetic Resonance Imaging-‘fMRI’ or resting-state elec- tween brain regions, and their individual specificity. In the first troencephalography) in which direct assessments of causality are study, we focused on the spatial-temporal evolution of brain re- not possible [3,4]. External perturbations applied directly to well- sponses at the individual level and hypothesized that controlled defined brain regions can yield fundamental insights into the perturbations of distinct brain regions by TMS will reveal distinc- causal interactions and dynamics of large-scale brain networks at tive patterns of activation dynamics reflecting unique connectivity the individual level [5]. Single pulses of TMS provide such non- characteristics of the stimulated region for each individual, and invasive perturbations safely and with precise experimental con- thus enable cortical fingerprinting (Study-I). In the second study, trol over an array of stimulation parameters (e.g. stimulus location, we first tested whether our main results from Study-I generalize to timing, intensity and duration). Simultaneously recording fast other independent data sets, and then examined subject specificity evolving TMS-evoked cortical responses with electroencephalog- of TEPs against non-transcranial neural responses evoked by two raphy (TMS-EEG) can probe causal brain connectivity dynamics different sham-TMS protocols (Study-II). with a temporal resolution comparable to the timescale of neural events. Participants and methods Trans-synaptic effects of TMS have long been established. TMS- induced motor evoked potentials (MEPs) recorded from contralat- Study-I: Data collected from 24 healthy, right-handed volun- eral extremity muscles following stimulation of the motor cortex teers (16 male; mean age ¼ 29.67 ± 10.60 years, ranging from 18 to demonstrate propagation of TMS-induced activations through 49 years) were analyzed for this study. Experimental protocols and structural connections [6e8]. Similarly, TMS of non-motor regions voluntary participation procedures were explained to all partici- evokes a series of electro-cortical potentials (TEPs) that spread from pants before they gave their written informed consent to the study the perturbed node to other discrete nodes of the stimulated that conformed to the Declaration of Helsinki, and had been network, thereby providing a snapshot of the connectivity profile of approved by the Institutional Review Board of the Beth Israel the stimulated node across distributed brain regions [9,10]. Over Deaconess Medical Center, Boston, MA. Participants had no self- the past decade, TEPs have been used to directly assess a broad reported history of psychiatric or neurological diseases. range of neurophysiological properties such as cortical excitability, Study-II: The first cohort (Test retest cohort-TRT) consists of 10 excitation/inhibition balance, effective connectivity, and integrity typically healthy adult control participants (5 M/5F, of the mechanisms of plasticity. However, despite their abundant age ¼ 43 ± 18.51 yrs) from an ongoing TMS-EEG study focusing on spatial and temporal neurophysiological content, the majority of reproducibility of repetitive TMS (rTMS) induced neuromodulation studies have characterized TEPs at the group level with traditional at the cortical (EEG) and cortical-spinal (EMG) level. The second event-related-potential (ERP) averaging metrics [11e15]. Such cohort (Epilepsy) consists of 10 typically healthy adult control group averaging approaches have greatly contributed to our ex- participants (9 M/1F, age ¼ 42.2 ± 18.8 yrs) from a TMS-EEG study amination of normal and pathological brain states. However, they of epilepsy. also pose limitations for understanding the neurophysiology of spatially distributed brain responses at the individual level. Spe- Data acquisition cifically, one major consideration is the temporal and spatial data loss in amplitude-based metrics extracted by ERP averaging. Ana- A T1-weighted (T1w) anatomical MRI scan was obtained in all lyses of TEPs with ERP methodology mainly focuses on the wave- participants and used for neuronavigation for each TMS visit. In form morphology, either at a single electrode or in a subset of each visit, TMS-EEG and TMS-EMG data were collected synchro- electrodes over a particular scalp region, in which selected elec- nously. Details of MRI scanning, TMS, EEG and EMG systems for trodes are first averaged at the subject-level and then further both studies are provided in the supplementary methods. averaged across subjects over a time-window of interest to generate grand ERPs at the group-level. These grand averaged ERPs Experimental procedures are then used to identify major waveform peaks to extract voltage amplitude or latency as single-time-point metrics of brain re- Study-I: We administered single pulses of TMS to both sponses to a given stimulation [14]. Such a substantial amount of anatomically (left dorsolateral prefrontal cortex- ‘L-DLPFC’, left data reduction caused by group level averaging inevitably ignores Intra-parietal lobule- ‘L-IPL) and functionally (left motor cortex- ‘L- the rich spatial-temporal content of TEPs and may discard crucial M1’, right default mode network- ‘R-DMN, right dorsal attention information that can be potentially used to characterize individual network- ‘R-DAN’) defined cortical nodes in the frontal, motor, and subjects. For example, inter-individual differences in TEP topog- parietal regions (Fig. 1A) across 2 identical sessions spaced one raphy at a given time point manifest individual variability in the month apart. All participants underwent two separate TMS-EEG spatial configuration of activated neural sources [16,17], while in- visits for the stimulation of the two anatomical targets (L-DLPFC, dividual differences in temporal evolution of such source localized and L-IPL) in one visit and the two functional targets (R-DAN and R- TEPs reflect unique propagation patterns across distinct brain re- DMN) in another visit. Identical re-test visits for each session were gions. Taken together, TEPs provide a causal connectivity profile of performed one month later (Fig. 1A). Details of functional and stimulated brain regions [18]. However, the lack of detailed spatial- anatomical target selections are explained in supplementary 392 R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 Fig. 1. Group averaged vs individual TEPs. A: Representation of individual stimulation sites for each cortical region on a template brain. B: Grand averaged TEPs (across all subjects) at the F3 electrode (black dot over the template brains) across visits in response to L-DLPFC (left), R-DMN (middle), and L-IPL (right) stimulations (shades showing variability as one unit of standard deviation). Dotted vertical black lines show TMS. Gray shaded rectangular shows the time range used (15 mse300 ms) to compute spearman correlation coefficients (r) between the two time series in each panel. C: Individual TEPs (thin colored lines) superimposed over the grand average TEP (thick blue line) in visit-1. D: Representative TEPs from two different subjects across visits. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) methods. The order of the visits and stimulation sites within each participants were asked to wear earplugs during hotspot and RMT visit was fully randomized across participants and was kept iden- trials to protect their hearing, and to minimize external noise [21]. tical across the retest sessions. TMS was administered with a thin layer of foam placed under the At the beginning of each visit, the motor hotspot was deter- coil to minimize somatosensory contamination of the TMS-evoked mined over the hand region of left motor cortex (L-M1) for eliciting EEG potentials. To minimize auditory evoked potentials related to motor evoked potentials (MEPs) in the right first-dorsal- the TMS coil click, auditory white noise masking was used interosseous (FDI) muscle. The hotspot was defined as the region throughout the TMS stimulation. where single-pulse TMS elicited larger and more consistent MEPs Following determination of RMT, a total of 120 single pulses of in the FDI muscle, as compared to abductor pollicis brevis (APB) TMS were delivered to each stimulation target at an intensity of muscle, with the minimum stimulation intensity. Resting motor 120% RMT with randomly jittered (3000e5000 ms) inter stimulus threshold (RMT) was determined on the FDI hotspot as the mini- intervals using monophasic waveforms. mum stimulation intensity eliciting at least five MEPs (50 mV) out Study II: For TRT cohort, 150 single pulses of TMS at 120% of of ten pulses in the relaxed FDI using monophasic current wave- resting motor threshold with randomly jittered 3e5 s inter- forms [8,20]. In compliance with the IFCN safety recommendations, stimulus-intervals was delivered using biphasic waveforms. For 393 R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 this cohort, active-TMS is applied to L-M1 and functionally defined Where SI XY is the cosine similarity between TEP matrices x (visit-1) L-IPL (See supplementary methods for details). Similar to our and y (visit-2) for a given stimulation site (i.e L-DLPFC), and n is the original cohort in study-I, we used auditory noise masking to number of channels with Xit and Yit are the ith vector of all channels minimize auditory evoked potentials related to the TMS coil click. at time t for visit-1 and visit-2 respectively. For joint analyses of TEP Each participant completed two identical TMS-EEG visits spaced similarity we also concatenated TEP matrices from multiple sites approximately 2 months in average (mean ¼ 64.27 ± 39.67 days). with all possible combinations of two (n ¼ 10), three (n ¼ 10), four For IPL stimulation data were collected from 7/10 participants (in 3 (n ¼ 5) and five (n ¼ 1) sites and computed the similarity matrix for participants an error in the stimulation parameters prevented each of the combinations. We normalized combined TEP matrices collection of usable data) while all participants successfully from each site with the Euclidean norm to account for the magni- completed L-M1 stimulation for both visits. In this cohort, we tude differences in TEPs across sites before concatenating. applied a sham protocol that closely resembles the active stimu- We generated a 24x24-similarity matrix to compute similarity lation conditions in Study-I (Supplementary Fig. 1A). Specifically, matrix metrics, where each entry in the matrix rows is the SI value sham-TMS was administered on the motor hot spot of the FDI between a given subject in visit-1 and all other subjects in visit-2. muscle over L-M1. An active/sham TMS coil (Cool-B65 A/P, Mag- Thus, the diagonal of the matrix corresponds to the SI for within- Venture A/S, Farum, Denmark) was flipped to the placebo side and subject values across identical visits. The following metrics were stimulation intensity was kept identical to actual TMS, but with computed from each similarity matrix: induced currents on the opposite vertical direction to targeted gyri. Within-subject similarity: The mean of the diagonal cells in each A 3D printed 3 cm spacer was attached to the placebo side (Mag- similarity matrix shows the average similarity within subjects Venture A/S, Farum, Denmark) of the coil to further ensure the across visits. elimination of residual currents on the placebo side of the coil Between-subject similarity: The mean of each row in the matrix, (Supplementary Fig. 1A). White noise masking was presented excluding the diagonal cell, represents the average between- through earplug-earbuds at the maximum volume comfortable for subject similarity for each subject. We then computed the mean each participant. Small current pulses between 2 and 4 mA and of individual averages to determine the overall between-subject proportional to the intensity of actual TMS pulse were delivered similarity of the matrix for each stimulation condition. over the left forehead, over the frontalis muscle, using surface Accuracy: We first determined the SI rank of diagonal cells in the electrodes (Ambu Neuroline 715 12/Pouch) to approximate so- matrix for each row. An individual is correctly identified if the di- matosensory sensations arising from skin mechanoreceptors and agonal cell has the highest similarity value (rank 1) at the given row scalp muscles during active-TMS condition [22]. The main goal of and that is counted as a “hit”. Accuracy is simply the ratio of this sham protocol was to minimize AEPs as much as possible and number of hits to total number of subjects. focus on SSEPs induced by electrical stimulation. Signal-to-Noise Ratio (SNR): We converted the similarity matrix For the epilepsy cohort, 100 single pulses of TMS at 120% of RMT into a z-matrix. For each row in the matrix, we calculated the with randomly jittered 3e5 s inter-stimulus-intervals was deliv- standard deviation of similarity scores across all subjects. The ered using biphasic waveforms. TMS is applied to L-M1 and cosine similarity for each cell in the row is normalized by this anatomically defined L-IPL, and L-LDLPFC (See supplementary standard deviation to obtain a z-score. The z-score for the diagonal methods for details). Each participant completed two identical cell thus represents a measure of the signal-to-noise ratio for the TMS-EEG visits spaced approximately 2.5 weeks in average self-similarity assessment. The overall SNR is simply computed as (mean ¼ 17.5 ± 14.23 days). All participants completed L-IPL, L- the mean of the diagonal in the Z-matrix. LDLPFC and L-M1 stimulation for both visits. In this cohort, we In Study II, we also computed cosine similarity at the milli- performed a basic sham control to specifically focus on the pres- second level by removing the t from summation function at the ence of AEPs. Sham-TMS was delivered to L-M1 with the TMS coil above equation. Similarly, we used voltage values across all elec- tilted 90 from the active side (Supplementary Fig. 1B). No auditory trodes at each time point as a feature vector, where each channel noise masking or electrical stimulation is used. Participants were value corresponds to a certain vector index. The result was a sim- only asked to wear earplugs during sham and active-TMS trials to ilarity time series for each comparison and allowed us to examine protect their hearing. The main goal of this sham was to minimize within-subject similarity across sessions (Visit-1 vs Visit-2) at the SSEPs and focus on the specific contributions of AEPs to active TMS highest temporal resolution possible. We note that for the primary responses. calculation of the similarity measures with active TMS, components corresponding to the auditory-evoked potential (AEPs) were removed. EEG preprocessing and metrics EEG Source Reconstruction: All TMS evoked EEG source recon- struction was performed using Brainstorm [23]. First, digitized EEG All details for preprocessing of EEG data are provided in the channel locations and anatomical landmarks of each subject were supplementary methods. Cosine similarity of TEPs within and be- extracted from Brainsight™ (nasion ‘NAS’, left pre-auricular ‘LPA’, tween subjects across sessions were computed as follows: and right pre-auricular ‘RPA’ points), and registered onto individual Similarity Index (SI): We first generated a TEP matrix for each MRI scans in brainstorm. Next, the EEG epochs, 500 mse1000 ms subject (from averaged responses) with a fixed window size with respect to TMS pulse for each TMS trial were uploaded, and the (385 ms) covering EEG responses from 15 to 400 ms following TMS average epoch time series was generated for each subject. Forward pulses (See Supplementary Fig. 2). Each TEP matrix contains modeling of neuro electric fields was performed using the open millisecond voltage values from all channels with a 63x385 matrix MEEG symmetric boundary element method [24], all with default size. We then used cosine similarity to quantify similarity index (SI) parameter settings [23]. Noise covariance was estimated from in- between matrices as follows: dividual trials using the pre TMS (500 to 0) time window as Pn baseline. The inverse modeling of the cortical sources was per- i;t¼1 ðXit *Yit Þ formed using the minimum norm estimation (MNE) method with SI XY ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Pn ffi P ð i;t¼1 X 2it Þ*ð ni;t¼1 Yit2 Þ dynamic statistical parametric mapping (dSPM) and constraining source dipoles to the cortical surface. The resulting output of EEG 394 R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 source reconstruction was the MNE current density time series for permutation t-tests (n ¼ 1000) by randomly shuffling 50% of sub- each cortical vertex. jects across compared TMS conditions (i.e., 50% of subjects shifted Additionally, we also computed global-mean-field-power from sham-TBS to active-TMS or vice versa) and determined cluster (GMFP) and spectral power of resting state EEG (Pre-TMS period) size and magnitude of significant adjacent time points at each to compare fingerprinting performance of SI with these well iteration. Finally, we re-compute “p” values of significant clusters in establish electro-cortical metrics (see supplementary methods for the main analyses by calculating the probability of their size and further details). magnitude in the permutation analyses. A cluster in the main an- alyses is considered to survive permutation, and thus significant, only if both the size and magnitude of a given cluster is above 95% Statistical analysis of all cluster sizes and magnitudes derived from permutation tests. All statistical analyses were performed using custom scripts utilizing Matlab statistical toolbox (Version 17A, The MathWorks Results Inc., Natick, MA). We ran nonparametric Wilcoxon rank sum test to compare within-subject similarity, between-subject similarity, and Study I SNR metrics across stimulation sites and conditions. The minimum significance level was set as (p < 0.05) and corrected for multiple Individual TEPs are unique and different from grand averaged TEPs comparisons for all statistical analyses. We first examined the reproducibility of TEPs at the group level. Cluster-based permutation paired sample t-test statistics were Consistent with prior studies reporting high test-retest reliability of performed to compare similarity time series at each time point TEPs [ [14]], TEP waveforms were highly reproducible across ses- across TMS conditions. First, we ran paired sample t-tests at each sions (L-DLPFC visit1 vs visit2: r ¼ 0.95, L-M1 visit1 vs visit2: sample to determine significant time points between each com- r ¼ 0.93, L-IPL visit1 vs visit2: r ¼ 0.98, R-DMN visit1 vs visit2: parison separately. We then computed the length of adjacent sig- r ¼ 0.94, and R-DAN visit1 vs visit2: r ¼ 0.89) at the group level nificant time points and sum of t-scores for significant time points (Fig. 1B). However, individual TEPs differed markedly across sub- to determine (1) cluster size and (2) cluster magnitude in the main jects and were substantially distinct from the group-average TEP analyses, respectively. Following main analyses, we performed (Fig. 1C), clearly demonstrating that averaged TEPs at the group Fig. 2. Source reconstruction of TEPs across subjects with high and low similarity. A: TEPs from LDLPFC stimulation for three representative subjects (S1, S2 and S3) with selected peaks (colored vertical lines). B and C: topography and corresponding source reconstructions of selected time points in A. Note that subject S1 has a similar topography compared to S2, but different topography from S3; the corresponding source reconstructions of these topographies reflect similar localization of cortical sources and propagation patterns for S1 and S2 but not for S3. (see Supplementary Methods for details of source reconstruction). 395 R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 level do not represent individual response characteristics to per- SNR values (Fig. 3B). On the other hand, the resulting similarity turbations. Nonetheless, TEPs for a given subject were highly matrix for TMS-EEG (Fig. 3A, right-panel) demonstrated that TEPs correlated across repeated sessions, suggesting that brain re- from different visits were substantially more similar within than sponses to perturbations were reproducible within the individual between individuals. On average, similarity for within-subject TEP (Fig. 1D). Source reconstruction of TEPs from individual subjects topographies (0.61) was significantly higher than between-subject confirmed that high topographic similarity between subjects re- similarity indices (0.24; p < 0.001), thereby allowing an individual flects consistent activation of similar brain regions, whereas topo- subject to be identified with 74% accuracy. graphical differences between subjects reflect distinct propagation patterns of evoked brain activity across distributed brain regions (Fig. 2). Interregional differences in test-retest similarity of TEPs We examined whether similarity of TEPs changes as a function Spatial temporal evolution TMS evoked potentials are stable and of the stimulation site (Fig. 4A). Highest accuracy for identifying unique subjects across repeat sessions was achieved for L-DLPFC, followed Here we tested our main hypothesis and examined whether by L-M1 and L-IPL stimulation (Fig. 4B), indicating that L-DLPFC TMS-evoked propagation patterns are stable across sessions and stimulation generates more unique spatial-temporal pattern across unique between individuals, thereby enabling brain fingerprinting. individuals. Although R-DAN stimulation has the lowest accuracy We also examined whether these properties are a result of the (47%), it was considerably above the chance level (4.3%). Signal-to- perturbation rather than an invariant feature of the ongoing noise ratios (SNR) for L-DLPFC stimulation was significantly higher background EEG activity. Thus, we computed the cosine similarity than R-DMN (LDLPFC vs DMN: Z¼2.88, p ¼ 0.003), while no other of 1) rs-EEG before TMS and 2) the spatial-temporal evolution of comparison was significant after correcting for multiple compari- TEPs for a single stimulation site (LDLPFC), both within- and sons (p > 0.05). Between-subject similarity was highest in L-M1 between-subjects, across sessions (Fig. 3). The resulting similarity stimulation (0.30) and it was significantly different from DMN matrix for rs-EEG (Fig. 3A, left-panel) revealed poor fingerprinting (M1vs DMN: Z¼3.94, p ¼ 0.0007), suggesting that TMS of M1 pro- performance with only 12.5% identification accuracy and with low duces more generic brain responses across subjects. Fig. 3. Fingerprinting with rs-EEG vs TEPs: A: Similarity matrix for Resting EEG (left) and L-DLPFC stimulation (right). Each color-coded cell represent the magnitude of cosine similarity across repeat sessions. Diagonal cells show each subjects’ own cosine similarity while cells in each row (excluding the diagonal) show each subject’s cosine similarity with every other subject in the group across repeat sessions. B: Accuracy (left), SNR (middle), and average within-subject similarity (left) metrics computed from the matrices in A. Error bars in B denote one unit of standard error of measurement (SEM), and stars denote significant comparisons across conditions after controlling for multiple corrections. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) 396 R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 Fig. 4. Similarity differences across stimulation sites. A: Similarity matrix for each stimulation site across repeat sessions. B: Comparison of similarity matrix metrics. Error bars in B denotes one unit of standard error of measurement SEM, and stars denotes significant comparisons across conditions after controlling for multiple corrections. Combined TEPs from multiple sites reveals unique cortical differences from L-M1 (L-DLPFC vs L-M1: Z¼3.21, p ¼ 0.022), sug- neurophysiology gesting that spatial-temporal characteristic of TEPs originating We next combined and jointly analyzed TEPs induced by TMS to from non-motor cortices are considerably different between in- multiple sites to characterize individual evoked brain dynamics dividuals but highly reproducible within the individual across across different brain regions (Fig. 5A). Signal-to-noise ratio (SNR) sessions (Fig. 6A and B). gradually increased from a single site (1.90) to five site (4.05) On the other hand, fingerprinting results from sham-TMS are combinations, with significant differences between averages of markedly different across data sets and provide interesting insights single-site and two-site (Z¼3.74, p ¼ 0.001), and between two-site regarding the spatial-temporal propagation patterns of non- and three-site (Z¼3.13, p ¼ 0.001) combinations, primarily due to a transcranially evoked responses (non-TEPs). For the TRT cohort, decrease in between-subject similarity. Perfect (100%) accuracy in sham-TMS (Fig. 6A, left upper panel) produced 50% accuracy fingerprinting individual subjects was achieved by considering the (Fig. 6B, right upper panel), comparable between-subject similarity response to perturbation across five sites (Fig. 5B), suggesting that and SNR to L-M1 (p > 0.05), but significantly higher between- the patterns of perturbation-evoked dynamics are a defining and subject similarity (Sham-TMS vs L-IPL: Z¼2.87, p ¼ 0.029) and unique feature of each individual human brain. This is also lower SNR (Sham-TMS vs L-IPL: Z¼-4.21, p ¼ 0.015) than L-IPL consistent with our control analyses showing that TEPs are specific (Fig. 2B, right middle and lower panels). For the epilepsy cohort, to the stimulated region (Supplementary Fig. 3), and provide a sham-TMS resulted in 80% accuracy (Fig. 6B, right upper panel). more unique characterization of individual brain dynamics as However, sham-TMS had substantially higher between-subject compared to other widely used electrophysiological metrics such as similarity (Fig. 2B, right middle panel) than all active-TMS condi- resting state power spectra or the Global Mean Field activation tions (Sham-TMS vs L-M1: Z¼4.68, p ¼ 0.011; Sham-TMS vs L-IPL: produced by the TMS pulse (Supplementary Figs. 4 and 5). Z¼4.93, p ¼ 0.010; Sham-TMS vs L-DLPFC: Z¼5.27, p ¼ 0.009). Accordingly, sham-TMS had significantly lower SNR (Fig. 6B, right lower panel) than all the active-TMS conditions (Sham-TMS vs L- Study II M1: Z¼-2.24, p ¼ 0.045; Sham-TMS vs L-IPL: Z¼-2.98, p ¼ 0.027; Sham-TMS vs L-DLPFC: Z¼-4.31, p ¼ 0.010). Additionally, between- Subject specificity of TEPs is generalizable across independent data subject similarity in sham-TMS from epilepsy cohort was also sets significantly higher than sham-TMS from TRT cohort (Z¼2.91, Main fingerprinting results from both cohorts were provided in p ¼ 0.028). Fig. 6. We first replicated our original findings for active-TMS We further examined the role of AEPs in fingerprinting perfor- conditions in both data sets with high fingerprinting accuracies mance in both sham-TMS data sets and computed cosine similarity ranging from 80 to 100%. Similar to our original results, L-M1 had across visits (sham-TMS in visit-1 vs sham-TMS in visit-2) at the highest between-subject similarity among the active-TMS condi- millisecond resolution with and without keeping AEP components. tions in both data sets with significant differences from L-IPL (L-M1 We first compared within-subject similarity between the two vs L-IPL: Z¼2.65, p ¼ 0.041) in the TRT data set (Fig. 6A, right middle sham-TMS conditions in the different cohorts, and found that panel). Accordingly, L-IPL had significantly higher SNR in the TRT sham-TMS from epilepsy cohort has significantly higher within- cohort than M1 (L-M1 vs L-IPL: Z¼ 2.87, p ¼ 0.043). Similarly, L- subject similarity in the 80e120 ms and 180e250 ms time DLPFC had the highest SNR in the epilepsy cohort with significant 397 R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 Fig. 5. Combining TEPs from multiple sites. A: Representative similarity matrices were shown for single site (LDLPFC: left upper panel), combination of three different sites (LDLPFC- DAN-DMN), and combination of five sites. B: Comparison of similarity matrix metrics between single-site, two-site, three-site, four-site and five-site combinations. Bars represent average metric values for all the possible combinations. Error bars represent one unit of SEM. windows, suggesting that presence of stronger AEP components in done in our original analysis of active-TMS presented in Figs. 4e5 the epilepsy cohort increases within-subject similarity at time above). Our analyses showed very poor fingerprinting perfor- windows consistent with the temporal peaks of AEPs (Fig. 7). mance for sham-TMS fingerprinting all active-TMS conditions both Next, we computed within-subject similarity time series for in TRT and Epilepsy cohorts (Fig. 9), with accuracies ranging only active-TMS conditions across sessions (See supplementary 20%e40%, and low SNR values, altogether suggesting that individ- methods for details). We used these time series as our reference ual response specificity in active-TMS is distinct from the residual points and compared them with similarity time series obtained by sham-TMS responses after removing AEPs. computing cosine similarity between sham-TMS in visit-1 and active-TMS conditions in visit-2. Our goal was to examine the Discussion extent to which the spatial-temporal characteristics of sham- evoked non-TEPs are shared by active-TMS evoked TEPs. Thus, we The characterization of the individual connectome and resulting combined data sets from both cohorts and computed the similarity connectivity dynamics is critical to understanding brain function in between sham and active-TMS conditions for each site (Fig. 8). As both health and disease. Here, we used single pulse TMS to perturb expected, we found high similarity between sham-TMS and active- different cortical nodes in frontal, motor, and parietal regions in TMS conditions when AEP components were present in both data both hemispheres, and evaluated the resulting evoked spatiotem- sets (Fig. 8A, blue colored time series). As such, the similarity time poral patterns of brain activation at the individual level using EEG. series for sham-TMS in visit-1 and active-TMS in visit-2 were not In the first study, in a primary cohort of 24 subjects, we find that significantly different from the similarity time series for active-TMS such direct, controlled external perturbations of discrete cortical conditions across visits (Fig. 8A, red colored time series). Accord- regions generate a sequence of individually distinct yet highly ingly, removing AEP components substantially reduces similarity reproducible brain responses, revealing a unique “fingerprint” of between sham-TMS and active-TMS across visits, while within- dynamical connectivity. In the second study, we confirmed our subject similarity in active-TMS conditions remained high even original findings in two independent data sets each with different after removing AEPs (Fig. 8B). sham-TMS controls, suggesting that subject specificity of brain re- Finally, we examined whether the remaining non-TEPs in sham- sponses to perturbations of discrete brain regions is robust across TMS datasets fingerprint TEPs in active-TMS after removing AEP (as data sets, and is due primarily to the transcranial-evoked 398 R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 Fig. 6. Fingerprinting with active- and sham-TMS. A: Results for test retest cohort. Similarity matrices (left panels) and fingerprinting metrics (right panel) for sham-TMS and active- TMS from L-M1 and L-IPL, B: Results for epilepsy cohort. Similarity matrices (left panels) and fingerprinting metrics (right panel) for sham-TMS and active-TMS from L-M1, L-IPL, L- LDLPFC. Error bars in metric panels denotes one unit of standard error of measurement SEM, and stars denotes significant comparisons across conditions after controlling for multiple corrections. stimulation rather than non-transcranial sensory features. Overall, while brain responses to TMS are markedly different across in- these results thus illustrate that the patterns of perturbation- dividuals and distinct from the group mean. This similarity in TEPs evoked dynamics are defining features of each individual human across sessions within individuals reflects consistent activation of brain, and that TMS-EEG provides a reliable means of character- distributed brain regions over time, whereas differences in TEPs izing these individual-specific causal propagation patterns. across subjects reflect distinct sequences of brain activity. We specifically focused on capturing propagation dynamics of TMS-induced activations at the individual level, without employing Grand averaging ignores individual information in TMS evoked any data reduction approach either in the spatial (i.e., averaging potentials electrodes) or temporal (i.e., selecting only waveform peaks) domain. Our results demonstrate that characterizing the whole- Recent studies examining the reliability of TMS-EEG responses brain spatial-temporal response dynamics provoked by TMS gen- reported that amplitude-based TEP metrics at mid-latencies are erates individually distinct information about the causal connec- reproducible at the group level, but highly variable across subjects tivity dynamics of the stimulated networks, and that this [14,25]. Our initial TEP analyses confirmed these prior findings that information is otherwise not observable through traditional group- averaging TEPs across subjects generates reproducible waveforms level averaging metrics and resting-state EEG recordings. Notably, across sessions at the group level. However, we find that this group- the spatial-temporal response patterns at the individual level are level response poorly overlaps with individual response dynamics, also specific to the stimulated brain regions and highly reproduc- and thus does not represent the actual TMS-induced activation ible between sessions. Combining such site-specific information profile for most individuals. Importantly, however, we also find that 399 R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 Fig. 7. Similarity time series of Sham-TMS across visits for each Cohort. Colored lines (Red: Epilepsy cohort, Blue: Test retest “TRT” cohort) represent group averaged within-subject similarity time series across sessions (Visit-1 vs Visit-2) for each cohort, and shaded colored regions represent one unit of standard error of measurement (SEM). Colored blocks at the bottom of each panel show significant cluster of time-points between comparisons (Gray blocks: Significant clusters for bivariate comparisons that did not survive permutation tests and Red Blocks: Significant clusters survived permutation tests p < 0.05).). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Fig. 8. Comparison of similarity time series for active-TMS to active-TMS with sham-TMS to active-TMS across visits. A: Comparison of active-TMS to active-TMS with AEPs (red) to sham-TMS to active-TMS (blue) with AEPs. B: Comparison of active-TMS to active-TMS without AEPs (red) to sham-TMS to active-TMS (green) without AEPs. Colored blocks at the bottom of each panel show significant cluster of time-points between comparisons (Gray blocks: Significant clusters for bivariate comparisons that did not survive permutation tests and Red Blocks: Significant clusters survived permutation tests p < 0.05). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) across multiple brain regions reveals an individually unique brain outperformed widely used electrophysiological metrics, suggesting “fingerprint”, capable of identifying each participant with perfect that cortical fingerprinting is a specific function of individual dif- accuracy. Additionally, we also showed that fingerprinting sensi- ferences in spatial propagation characteristics of TEPs. tivity and specificity of TMS perturbations substantially 400 R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 Fig. 9. Fingerprinting active-TMS responses with sham-TMS. A: Test retest cohort similarity matrices (left panels) and metrics (right panels) for sham-TMS in visit-1 finger- printing active-TMS from L-M1 (Left upper panel) and L-IPL (Left lower panel) in visit-2. B: Epilepsy cohort similarity matrices (left panels) and metrics (right panels) for sham-TMS in visit-1 fingerprinting active-TMS from L-M1 (Left upper panel), L-IPL (Left middle panel) and L-DLPFC (Left lower panel) in visit-2. Error bars in metric panels denotes one unit of standard error of measurement SEM, and stars denotes significant comparisons across conditions after controlling for multiple corrections. Utility of perturbation responses in characterizing individual fundamental insights into the causal interactions and dynamics of connectivity large-scale brain networks at the individual level. In particular, TMS of cortical nodes in parietal association cortex has been shown to Given the high translational and clinical potential of establishing induce network specific propagation patterns [5] compatible with brain-behavior relationships at the individual level, recent neuro- the functional connectivity profile of the stimulated nodes, imaging studies have focused on characterizing inter-individual endorsing targeted network perturbation as a promising approach variability in brain connectivity with the goal of identifying bio- to evaluate the causal interactions and dynamics of large-scale markers of individual brain function. So far, blood-oxygenation brain networks at the individual level. As such, TMS-based as- level-dependent (BOLD) functional magnetic resonance imaging sessments of individual connectivity dynamics can complement (fMRI) during “unconstrained resting” has been the primary neu- and extend more traditional resting-state or task-based measures. roimaging tool for characterizing the individual variation in func- As an example, a large body of recent human fMRI studies have tional brain organization. However, the slow time-course of the reported extensive between-subject variance in functional and hemodynamic response function underlying the BOLD signal, as structural connectivity profile at multimodal association cortices well as the purely correlational nature of resting-state fMRI con- (i.e., frontoparietal or ventral attention) as compared to unimodal nectivity, are two fundamental limitations of this approach for sensori-motor cortices [26e28]. Consistent with this, we showed capturing causal brain interactions at the timescale of neuronal that TMS of motor cortex generated more similar TEP topographies activity. In contrast, controlled perturbations (e.g. via TMS) applied across subjects in three independent data sets, confirming common directly to well-defined brain regions, including specifically asso- response dynamics across subjects, as compared to TMS of associ- ciation cortices that underlie higher-order control functions and ation cortex sites. that show the greatest variability between individuals [7], can yield 401 R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 TMS fingerprinting versus non-transcranial evoked activity component does not play a major role in the fingerprinting accu- racy of the unique and reliable evoked potentials obtain from non- A potential confound in TMS-EEG responses could be the motor regions. contamination of TEPs with non-transcranial sensory-evoked re- sponses such as the auditory evoked potentials (AEP). In our pri- Limitations mary analysis cohort, we used noise masking to try to minimize the presence of the AEP component, and used an ICA-based approach Certain limitations of the study should be acknowledged. Our in postprocessing to remove residual AEP elements. However, to subjects from all data sets are young and healthy participants with determine the potential influence of the AEP on the similarity no known neurological, psychiatric or cognitive disorders. It is measures assessed here, we analyzed two additional control therefore essential to evaluate the validity of our results in different datasets, one with noise masking and one without, and including age groups (e.g. children, older adults) and in various clinical two different types of sham stimulation. We noticed strong AEPs in populations, to establish the spatial-temporal specificity of TEPs as all subjects in the control epilepsy cohort in which noise masking biomarkers of individual brain function in both health and disease. was not done, while AEPs were observed in only 3 out of 10 subjects Another important translational step would be focusing on “be- in the control test-retest cohort with noise masking, corroborating tween-subject” similarity characteristics to examine whether in- with the recent evidence [29] that the use of noise masking (as dividuals with similar cortical response signatures at the network done in our primary study cohort) effectively eliminates or removes level also share common cognitive or behavioral characteristics. AEPs for most of the subjects. Our analyses of sham-TMS data sets Moreover, the stability of cortical response dynamics across mul- showed that, when present, AEPs dominate sham-TMS responses tiple repeat sessions with long time intervals is yet to be and result in high degree of within-subject similarity across visits. determined. These results suggest that evoked responses in the brain, even when they are non-transcranial (i.e. AEP), may be stable over time. Conclusions Given the fact that all these sensory evoked responses are pro- cessed in individual brains with stable structural connectivity TMS-EEG can be used to characterize the causal propagation profiles, it is not surprising to observe high within-subject simi- patterns of brain responses to TMS perturbations at the timescale of larity in sensory responses evoked by sham-TMS across sessions. normal physiological function. While such TMS-evoked responses However, the spatial-temporal characteristics of these AEPs are also are highly reproducible at the group level, individual responses are highly common across subjects, as the uniqueness of each subject’s distinct from the group response and highly heterogeneous across cortical response profile is substantially lower than response subjects. Nevertheless, the spatiotemporal patterns of individual characteristics originating from active-TMS of associative cortices. TMS-evoked responses are stable over time and specific to the We also found that both within-subject and between-subject stimulated region, carrying information about individual connec- similarity are significantly decreased in sham-TMS data sets after tivity dynamics that is distinct from spontaneous activity. Most AEPs are removed in postprocessing, further confirming that the importantly, we demonstrate that an individually unique brain nature of high within-subject similarity, especially in the epilepsy fingerprint can be identified by combining spatial-temporal char- cohort, heavily depend on the presence of AEPs. In contrast, active- acteristics of TMS induced brain responses originating from TMS evoked potentials provide excellent fingerprinting with high different cortical regions. These results thus reveal spatial-temporal similarity within each individual and highly distinct propagation analyses of whole brain responses to neuroimaging-based pertur- patterns across individuals after removing the AEP component. bation as a highly promising approach to characterize causal brain- Importantly, when AEPs are removed from both active- and sham- connectivity dynamics at the individual level. Such perturbation- TMS data sets, the remaining sham-TMS responses did not effec- based characterization of brain responses may serve as reliable tively fingerprint active-TMS responses, indicating that individual biomarkers of individual brain function that could enable the lon- response specificity in active-TMS is unlikely to be due to the gitudinal tracking of individual brain dynamics across the lifespan, presence of residual non-TEPs in active-TMS. in pathological processes, and in response to therapeutic Somatosensory-evoked potentials (SSEPs) could represent interventions. another transcranial-evoked component that contributes to the observed effects. Specifically, high between-subject similarity CRediT author contribution statement derived from the motor cortex stimulation could be attributed in part to sensory afference evoked by the stimulated muscle. Indeed, Recep A. Ozdemir: designed the study, Conceptualization, in both our test-retest and epilepsy cohorts, sham-TMS responses conceptualized the framework, Data curation, collected the data, had the highest similarity with active-TMS responses from L-M1, preprocessed the TMS-EEG data, Formal analysis, Writing - original and the sham-M1 similarity remained significantly higher than the draft, preprocessed and analyzed TMS-EEG data and wrote the first similarity between sham-TMS and active-TMS responses from non- draft. All authors critically reviewed the manuscript for content and motor cortices after removing AEPs. Moreover, between-subject approve the final version for publication. Ehsan Tadayon: Formal similarity with M1 stimulation was comparable to sham-TMS be- analysis, performed rs-fMRI analysis to define individual TMS tar- tween-subject similarity in the test-retest cohort in which we gets. All authors critically reviewed the manuscript for content and applied electrical stimulation to induce somatosensory sensations. approve the final version for publication. Pierre Boucher: Data Taken together, these findings suggest that the presence of SSEPs curation, collected the data, preprocessed the TMS-EEG data. All may contribute to the high between-subject similarity seen with authors critically reviewed the manuscript for content and approve M1 stimulation. However, accuracy and SNR were higher with M1 the final version for publication. Davide Momi: Data curation, stimulation than with sham stimulation even in the test-retest collected the data, All authors critically reviewed the manuscript cohort, suggesting that the observed subject specificity is not just for content and approve the final version for publication. Alvaro due to the presence of SSEPs. Notably, the between-subject simi- Pascual-Leone: designed the study, Writing - review & editing, larity values were significantly lower, and the SNR significantly oversaw study conduction and edited the first draft. All authors higher, with stimulation of non-motor associative regions in com- critically reviewed the manuscript for content and approve the final parison to sham, suggesting that this somatosensory-evoked version for publication. Emiliano Santarnecchi: overviewed the 402 R.A. Ozdemir, E. Tadayon, P. Boucher et al. Brain Stimulation 14 (2021) 391e403 selection of stimulation sites. All authors critically reviewed the [5] Ozdemir RA, et al. Individualized perturbation of the human connectome reveals reproducible biomarkers of network dynamics relevant to cognition. manuscript for content and approve the final version for publica- Proc Natl Acad Sci Unit States Am 2020. https://doi.org/10.1073/ tion. Mouhsin M. Shafi: designed the study, Conceptualization, pnas.1911240117. conceptualized the framework. overviewed the selection of stim- [6] Hallett M. Transcranial magnetic stimulation and the human brain. Nature ulation sites, Writing - review & editing, oversaw study conduction 2000;406:147e50. [7] Kobayashi M, Pascual-Leone A. Transcranial magnetic stimulation in and edited the first draft. All authors critically reviewed the neurology. Lancet Neurol 2003;2:145e56. manuscript for content and approve the final version for [8] Rothwell JC, et al. Magnetic stimulation: motor evoked potentials. Electro- publication. encephalogr Clin Neurophysiol Suppl 1999;52:97e103. [9] Ilmoniemi RJ, et al. Neuronal responses to magnetic stimulation reveal cortical reactivity and connectivity. Neuroreport 1997;8:3537e40. Declaration of competing interest [10] Komssi S, et al. Ipsi-and contralateral EEG reactions to transcranial magnetic stimulation. Clin Neurophysiol 2002;113:175e84. [11] Bagattini C, et al. Predicting Alzheimer’s disease severity by means of The authors declare no competing interests. TMSeEEG coregistration. Neurobiol Aging 2019;80:38e45. [12] Bortoletto M, Veniero D, Thut G, Miniussi C. The contribution of TMSeEEG Acknowledgments coregistration in the exploration of the human cortical connectome. Neuro- sci Biobehav Rev 2015;49:114e24. [13] Darmani G, Ziemann U. Pharmacophysiology of TMS-evoked EEG potentials: a We thank all participants who took part in the study for their mini-review. Brain Stimulation 2019;12:829e31. effort. The study was supported by the Harvard-MIT BROAD Insti- [14] Kerwin LJ, Keller CJ, Wu W, Narayan M, Etkin A. Test-retest reliability of transcranial magnetic stimulation EEG evoked potentials. Brain Stimulation tute (Grant ID 6600024-5500000895). A.P.-L. is supported by the 2018;11:536e44. Berenson-Allen Foundation, the Sidney R. Baer Jr. Foundation, NIH [15] Lioumis P, Ki ci c D, Savolainen P, Ma €kela€ JP, Ka €hko€nen S. Reproducibility of Grants R01HD069776, R01NS073601, R21 MH099196, R21 TMSdevoked EEG responses. Hum Brain Mapp 2009;30:1387e96. NS082870, R21 NS085491, and R21 HD07616, and The Harvard [16] Murray MM, Brunet D, Michel CM. Topographic ERP analyses: a step-by-step tutorial review. Brain Topogr 2008;20:249e64. Clinical and Translational Science Center (National Center for [17] Michel CM, Murray MM. Towards the utilization of EEG as a brain imaging Research Resources and the National Center for Advancing Trans- tool. Neuroimage 2012;61:371e85. lational Sciences Grant UL1 RR025758). E.S. is supported by the [18] Hallett M, et al. Contribution of transcranial magnetic stimulation to assess- ment of brain connectivity and networks. Clin Neurophysiol 2017;128: Beth Israel Deaconess Medical Center via the Chief Academic Offi- 2125e39. cer Award 2017, the Defence Advanced Research Projects Agency [19] Fox MD, et al. Resting-state networks link invasive and noninvasive brain via HR001117S0030, and NIH Grants P01 AG031720-06A1, R01 stimulation across diverse psychiatric and neurological diseases. Proc Natl Acad Sci Unit States Am 2014;111:E4367e75. MH117063-01, and R01 AG060981-01. M.M.S. is supported by the [20] Rossini PM, et al. Non-invasive electrical and magnetic stimulation of the Citizens United for Research in Epilepsy foundation, the Football brain, spinal cord, roots and peripheral nerves: basic principles and proced- Players Health Study at Harvard University, and NIH Grants R01 ures for routine clinical and research application. An updated report from an IFCN Committee. Clin Neurophysiol 2015;126:1071e107. MH115949, R01AG060987, R01 NS073601, and P01 AG031720- [21] Rossi S, Hallett M, Rossini PM, Pascual-Leone A. Safety, ethical considerations, 06A1. The content of this paper is solely the responsibility of the and application guidelines for the use of transcranial magnetic stimulation in authors and does not necessarily represent the official views of clinical practice and research. Clin Neurophysiol 2009;120:2008e39. Group, S. of T. C. Harvard University and its affiliated academic health care centers, [22] Rossi S, et al. A real electro-magnetic placebo (REMP) device for sham or NIH. transcranial magnetic stimulation (TMS). Clin Neurophysiol 2007;118: 709e16. Appendix A. Supplementary data [23] Tadel F, Baillet S, Mosher JC, Pantazis D, Leahy, Brainstorm RM. A user-friendly application for MEG/EEG analysis. Comput Intell Neurosci 2011;2011:1e13. [24] Gramfort A, et al. MNE software for processing MEG and EEG data. Neuro- Supplementary data to this article can be found online at image 2014;86:446e60. https://doi.org/10.1016/j.brs.2021.02.005. [25] Rogasch N, et al. TMS-evoked EEG potentials from prefrontal and parietal cortex: reliability, site specificity, and effects of NMDA receptor blockade. bioRxiv 2019. https://doi.org/10.1101/480111. References [26] Bürgel U, et al. White matter fiber tracts of the human brain: three- dimensional mapping at microscopic resolution, topography and inter- [1] Van Essen DC, et al. The Wu-minn human connectome project: an overview. subject variability. Neuroimage 2006;29:1092e105. Neuroimage 2013;80:62e79. [27] Frost MA, Goebel R. Measuring structuralefunctional correspondence: spatial [2] Greicius MD, Krasnow B, Reiss AL, Menon V. Functional connectivity in the variability of specialised brain regions after macro-anatomical alignment. resting brain: a network analysis of the default mode hypothesis. Proc Natl Neuroimage 2012;59:1369e81. Acad Sci Unit States Am 2003;100:253e8. [28] Mueller S, et al. Individual variability in functional connectivity architecture of [3] Finn ES, et al. Functional connectome fingerprinting: identifying individuals the human brain. Neuron 2013;77:586e95. using patterns of brain connectivity. Nat Neurosci 2015;18:1664e71. [29] Rocchi L, et al. Disentangling EEG responses to TMS due to cortical and pe- [4] Tavor I, et al. Task-free MRI predicts individual differences in brain activity ripheral activations. Brain Stimulation 2020;14(1):4e18. https://doi.org/ during task performance. Science 2016;352:216e20. 10.1016/j.brs.2020.10.011. ISSN 1935-861X. 403