Probing dlPFC excitability with TMS-EEG across coil angle, location, and intensity reveals a distinct sensitivity profile for early local TEPs
DOI: Pending DOI
Synopsis
The early local TMS-Evoked Potential (EL-TEP) is a promising biomarker of cortical excitability linked to depression severity and treatment response, yet how stimulation parameters shape it remains unknown. We delivered robotic neuronavigated TMS-EEG to the left dlPFC in ten healthy participants, varying intensity, coil angle, and cortical target. The EL-TEP was primarily driven by intensity, with larger amplitudes at more posterior dlPFC sites. Coil angle and anterior-posterior location did not consistently modulate the EL-TEP, contrasting sharply with the artifact (12–20 ms) and vertex potential (80–250 ms), which were strongly shaped by both. This dissociation suggests the EL-TEP reflects genuine cortical excitability rather than peripheral contamination, and that maximizing it requires artifact-minimizing targets and intensity titration to inform personalized TMS therapy.
Background
TMS to the dlPFC is FDA-cleared for treatment-resistant major depressive disorder1, and single-pulse TMS-EEG has emerged as a tool to probe the neural effects of stimulation2. The early local TEP (EL-TEP, 20–60 ms) is a short-latency response near the stimulation site reflecting local cortical excitability, with emerging evidence linking it to depression severity and treatment response, making it a promising clinical biomarker3,4. Interpreting the EL-TEP requires separating it from the early TMS artifact (12–20 ms), driven by scalp muscle activation and electromagnetic noise5, and from the vertex potential (N100-P200, 80–250 ms), reflecting multisensory responses to the auditory click and scalp sensation of TMS6. How parameter adjustments shape these responses remains poorly understood, which is why clinical TMS defaults to fixed protocols, yet these neural signatures offer a rich window into stimulation effects7,8. Recent advances in neuronavigation and real-time EEG monitoring enable systematic within-session exploration of TMS parameter spaces, opening the door to individualized optimization3. A critical gap remains in understanding how stimulation intensity, coil angle, and cortical target location shape these signatures within the same individuals. Here we varied these parameters across anterior and posterior dlPFC in healthy participants while recording 64-channel EEG, asking how parameter selection affects prefrontal excitability as indexed by the EL-TEP and how it can minimize artifact and sensory confounds to maximize cortical signal-to-noise.
Methods
Single-pulse TMS was delivered to the left dlPFC of ten healthy participants using a MagVenture B65-AP figure-of-eight coil and MagPro X100 stimulator. Targets included anterior dlPFC (MNI: -40.6, 41.7, 34.3) and posterior dlPFC (MNI: -37, 23, 47)9,10, transformed to individual space via nonlinear MNI registration. For each target region, 21 locations were tested across a coarse and fine grid, distributed along anteroposterior and mediolateral axes and constrained to the same gyrus where possible. Thirteen coil angles were tested in 15-degree steps spanning 180 degrees, then mirrored to cover the full 360 degrees under the assumption that the cortical response to biphasic stimulation is not sensitive to whether the initial induced current is positive or negative. Neuronavigation was performed with NaviNIBS11 with real-time coil tracking. Scalp EEG was recorded with a 64-channel TMS-compatible ActiCap system (Brain Products) at 25 kHz with auditory masking12. TMS-EEG data were preprocessed using an automated artifact rejection pipeline for TMS-evoked potentials (AARATEP)13, which includes interpolation of the TMS pulse artifact, independent component analysis for removal of decay and muscle artifacts, and trial-level rejection of residual contamination14. Three measures were extracted: left prefrontal EL-TEP peak-to-peak amplitude (20-60 ms) from a six-channel ROI over left dlPFC, artifact-window GMFA (12-20 ms), and vertex potential amplitude (80-250 ms) from a six-channel ROI around Cz. Per-block measures were modelled with linear mixed models including a subject random intercept and stimulation intensity as a covariate, with coil angle or cortical target entered as a categorical factor. Estimated marginal means and comparison intervals were obtained with emmeans15.
Results
Figure 1. Stimulation parameters and noninvasive neurophysiology across the dlPFC parameter space. (A) TMS-EEG setup and EL-TEPs. A neuronavigated robotic arm positions the TMS coil over the EEG cap and stimulation target. EL-TEPs for the anterior (red) and posterior (blue) dlPFC targets are shown across stimulation intensities. (B) Coil angle analysis. Data from a 180-degree angle sweep around the anterior (red) and posterior (blue) stimulation sites. The top row displays Cartesian tuning curves for the EL-TEP, artifact, and vertex potential (columns 1–3). The bottom row displays the same data in polar coordinates. An angle of 0 degree corresponds to the midline; negative angles point toward the left ear and positive angles toward the right ear. Error bars represent comparison intervals (alpha = 0.05). (C) Location grid analysis. Each target (anterior, red; posterior, blue) is surrounded by a coarse grid (4 locations, 7.5 mm from center) and a fine grid (16 locations, 5 mm spacing). The top row shows estimated marginal means from the coarse grid for the EL-TEP, artifact, and vertex potential (columns 1–3). The bottom row shows the corresponding fine grid results.
Dose-response. All three measures increased with stimulation intensity, suggesting a clear dose-response relationship. The EL-TEP showed larger peak-to-peak amplitudes and notably better signal-to-noise ratio (defined as the ratio of EL-TEP amplitude to pre-stimulus baseline variability) at the posterior target compared to the anterior site (Fig 1A).
Coil angle. Coil orientation was the strongest modulator of the artifact, producing approximately a 10–12 dBuV swing across the 180-degree sweep, peaking around +20 to +45 degrees and reaching its minimum near −45 to −60 degrees (Fig 1B). The vertex potential followed a similar angular profile with moderate tuning, consistent with angle-dependent changes in scalp sensation driving the somatosensory evoked response. The EL-TEP showed a subtle shift in its angular peak toward 0 to −45 degrees, offset from the artifact and vertex maxima, suggesting a dissociation between the cortical response and the peripheral and sensory components. This relative insensitivity is notable given the large artifact modulation across the same angles, arguing against the EL-TEP being a residual of incomplete artifact removal.
Location. The artifact and vertex potential were consistently larger for the anterior than the posterior dlPFC target across all analyses, including angle sweeps, coarse-grid, and fine-grid mapping (Fig 1C). This anterior dominance likely reflects greater temporalis muscle activation and stronger scalp sensation at the more lateral anterior site. The EL-TEP showed no consistent anterior-greater-than-posterior pattern, further distinguishing it from the artifact and sensory-driven components.
Fine and coarse grid. Within-region spatial variation across fine-grid (FG1–FG16) and coarse-grid (Ctr, CG1–CGn) locations did not produce individually significant differences for any measure, with estimated marginal means and comparison intervals overlapping substantially across grid positions. This spatial stability within each region suggests that small displacements along the cortical surface, within the range tested, do not drastically alter the group-aggregate evoked response, and that the between-region differences reflect genuine regional variation rather than sensitivity to precise coil placement.
Discussion
These results reveal a preliminary dissociation between the EL-TEP and the artifact and vertex components in their sensitivity to TMS parameters. The artifact and vertex potential behaved similarly, both modulated by coil angle and tending to be larger at the anterior dlPFC target, possibly reflecting a common peripheral origin, whereby anterior stimulation may activate more temporalis muscle, producing larger artifacts and a stronger scalp sensation that could drive a larger vertex potential16.
The EL-TEP showed a somewhat different pattern. Its relative insensitivity to coil angle and lack of a consistent anterior-posterior offset suggest it may not simply reflect residual artifact or sensory processing. Stimulation intensity appeared as a primary driver, consistent with recent work on prefrontal input-output excitability curves14, though individual-level analyses and more comprehensive parameter sweeps may reveal additional effects not captured here.
These observations tentatively suggest that coil angle and location choices may primarily serve as artifact-management decisions, while intensity more directly modulates the cortical response. Maximizing the EL-TEP may be best approached by selecting a posterior target and titrating intensity, though this warrants more rigorous testing. Future work integrating scalp EEG with intracranial recordings could provide mechanistic insight into EL-TEP variation and guide closed-loop, individualized TMS protocols for depression treatment.
Disclosures
The authors declare no financial disclosures
References
1. Cohen, S. L., Bikson, M., Badran, B. W. & George, M. S. A visual and narrative timeline of US FDA milestones for Transcranial Magnetic Stimulation (TMS) devices. Brain Stimul. Basic Transl. Clin. Res. Neuromodulation 15, 73–75 (2022).
2. Gruenwald, J. et al. Reliable and efficient transcranial magnetic stimulation–electroencephalography (TMS–EEG) using ultra-thin active electrodes. Commun. Eng. 4, 206 (2025).
3. Parmigiani, S. et al. Real-time optimization to enhance noninvasive cortical excitability assessment in the human dorsolateral prefrontal cortex. bioRxiv 2024.05.29.596317 (2024) doi:10.1101/2024.05.29.596317.
4. Gogulski, J., Cline, C. C., Ross, J. M., Parmigiani, S. & Keller, C. J. Reliability of the TMS-evoked potential in dorsolateral prefrontal cortex. Cereb. Cortex 34, bhae130 (2024).
5. Holden, M. M. et al. Characterising and Minimising Step and Filtering Artifacts in TMS-EEG Recordings. 2026.05.13.725016 Preprint at https://doi.org/10.64898/2026.05.13.725016 (2026).
6. Ross, J. M., Sarkar, M. & Keller, C. J. Experimental suppression of transcranial magnetic stimulation-electroencephalography sensory potentials. Hum. Brain Mapp. 43, 5141–5153 (2022).
7. Rossi, S. et al. Safety and recommendations for TMS use in healthy subjects and patient populations, with updates on training, ethical and regulatory issues: Expert Guidelines. Clin. Neurophysiol. 132, 269–306 (2021).
8. Caulfield, K. A. & Brown, J. C. The Problem and Potential of TMS’ Infinite Parameter Space: A Targeted Review and Road Map Forward. Front. Psychiatry 13, 867091 (2022).
9. Fox, M. D., Liu, H. & Pascual-Leone, A. Identification of reproducible individualized targets for treatment of depression with TMS based on intrinsic connectivity. NeuroImage 66, 151–160 (2013).
10. Fitzgerald, P. B. et al. A Randomized Trial of rTMS Targeted with MRI Based Neuro-Navigation in Treatment-Resistant Depression. Neuropsychopharmacology 34, 1255–1262 (2009).
11. Cline, C. C. et al. NaviNIBS: a comprehensive and open-source software toolbox for neuronavigated noninvasive brain stimulation. J. Neural Eng. 22, 056007 (2025).
12. Russo, S. et al. TAAC - TMS Adaptable Auditory Control: A universal tool to mask TMS clicks. J. Neurosci. Methods 370, 109491 (2022).
13. Cline, C. C., Lucas, M. V., Sun, Y., Menezes, M. & Etkin, A. Advanced Artifact Removal for Automated TMS-EEG Data Processing. in 2021 10th International IEEE/EMBS Conference on Neural Engineering (NER) 1039–1042 (2021). doi:10.1109/NER49283.2021.9441147.
14. Chen, N.-F. et al. Noninvasive profiling of input-output excitability curves in human prefrontal cortex. 2025.09.26.678876 Preprint at https://doi.org/10.1101/2025.09.26.678876 (2025).
15. Searle, S. R., Speed, F. M. & Milliken, G. A. Population Marginal Means in the Linear Model: An Alternative to Least Squares Means. Am. Stat. 34, 216–221 (1980).
16. Rogasch, N. C., Daskalakis, Z. J. & Fitzgerald, P. B. Cortical inhibition of distinct mechanisms in the dorsolateral prefrontal cortex is related to working memory performance: A TMS–EEG study. Cortex 64, 68–77 (2015).