pith. sign in

arxiv: 2605.08184 · v1 · submitted 2026-05-05 · 📡 eess.SP · cs.AI

Improving TMS EEG Signal Quality for Closed-Loop Neuro Stimulation via Source-Domain Denoising

Zhen Tang , Ameer Hamoodi , Stevie Foglia , Aimee Nelson , Zhen Gao This is my paper

Pith reviewed 2026-05-12 01:00 UTC · model grok-4.3

classification 📡 eess.SP cs.AI
keywords TMS-EEGartifact removalsource-domain denoisingsignal preprocessingevoked potentialsbenchmark datasetclosed-loop neurostimulationbrain-computer interface
0
0 comments X

The pith

Source-domain denoising cleans TMS EEG signals while preserving evoked potentials for closed-loop use

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests a preprocessing workflow for TMS-EEG signals that uses source-domain methods to remove artifacts from muscle and other sources. It establishes a reference dataset of cleaned signals to serve as a benchmark for developing and comparing automated cleaning algorithms. The evaluation shows that these methods improve signal quality and maintain the integrity of TMS-evoked potentials. This supports more reliable data for research into cortical dynamics and for clinical closed-loop neurostimulation systems integrated with brain-computer interfaces.

Core claim

The central finding is that source-based artifact removal techniques, when applied to TMS-EEG data, produce higher quality signals suitable for analysis, with preserved TMS-evoked potentials, as demonstrated by the creation and use of a reference preprocessed dataset for validation.

What carries the argument

The source-domain denoising pipeline, which decomposes EEG signals into independent sources to selectively eliminate artifact components before signal reconstruction.

Load-bearing premise

A dataset prepared through careful preprocessing can stand in for the unknown true brain signals and serve as a valid reference for judging other methods.

What would settle it

An experiment that uses a physical head model or simultaneous invasive brain recordings to generate TMS-EEG data with known exact brain responses, then checks if the proposed denoising recovers those responses accurately.

read the original abstract

This research addresses a validated TMS EEG cleaning pipeline and a corresponding benchmark dataset. It evaluates two widely used artifact removal pipelines. A reference dataset of carefully preprocessed EEG signals was established to support future algorithm development and enable systematic comparison of automated artifact removal strategies, despite the absence of a true physiological ground truth. The study evaluates the effectiveness of two widely used source based artifact removal approaches and examines their impact on signal quality improvement and preservation of TMS-evoked potentials. The results support the robustness of the proposed preprocessing workflow and demonstrate its potential for improving data reliability in both research and clinical applications. A key goal is integrating TMS EEG and embedding it within a larger BCI framework. Ultimately, these efforts aim to enhance understanding of cortical dynamics and expand the clinical and research applications of TMS EEG.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The paper presents a source-domain denoising workflow for TMS-EEG artifact removal aimed at closed-loop neurostimulation applications. It evaluates two established source-based artifact removal pipelines against a self-created reference dataset of carefully preprocessed EEG signals (acknowledging the absence of true physiological ground truth), assesses impacts on signal quality and TMS-evoked potential (TEP) preservation, and concludes that the workflow is robust with potential to improve data reliability for research and clinical use, including BCI integration.

Significance. If the evaluation framework holds, the work could offer a useful benchmark dataset and preprocessing pipeline for the TMS-EEG community, facilitating systematic comparisons of artifact removal methods and supporting more reliable cortical dynamics studies. The explicit provision of a reference dataset for future algorithm development is a constructive contribution despite the ground-truth limitation.

major comments (3)
  1. [Abstract] Abstract: The claim that 'the results support the robustness of the proposed preprocessing workflow' lacks any quantitative support (no metrics, error bars, statistical tests, or effect sizes are mentioned), making it impossible to assess whether the source-domain approaches outperform baselines or preserve TEPs without circular dependence on the reference.
  2. [Reference dataset] Reference dataset section: The reference is constructed via 'careful preprocessing' of EEG signals to serve as ground truth for evaluating source-separation methods, yet the manuscript does not detail how these preprocessing steps differ from the two evaluated pipelines; this creates a risk that performance metrics simply reward reproduction of similar source-separation assumptions rather than independent physiological fidelity.
  3. [Evaluation] Evaluation of artifact removal and TEP preservation: Without an independent validation strategy (e.g., simulated data with known ground truth, cross-validation against other modalities, or blinded expert scoring), the reported improvements in signal quality remain unverifiable and the central claim of robustness for closed-loop use rests on an untestable foundation.
minor comments (1)
  1. [Abstract] Abstract contains minor repetition ('evaluates two widely used artifact removal pipelines' appears in consecutive sentences with slight rephrasing); streamlining would improve clarity.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for their detailed and constructive review. We address each major comment point by point below, providing clarifications and committing to revisions that improve the manuscript's transparency and rigor without overstating our claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that 'the results support the robustness of the proposed preprocessing workflow' lacks any quantitative support (no metrics, error bars, statistical tests, or effect sizes are mentioned), making it impossible to assess whether the source-domain approaches outperform baselines or preserve TEPs without circular dependence on the reference.

    Authors: We agree that the abstract would benefit from quantitative support to substantiate the robustness claim. The results section of the manuscript contains specific metrics on signal quality (e.g., improvements in SNR and artifact reduction ratios) and TEP preservation (e.g., amplitude and latency consistency across conditions), along with statistical comparisons. In the revised manuscript, we will update the abstract to explicitly include key quantitative findings, such as average SNR gains, TEP preservation percentages, and p-values from relevant tests, to allow direct assessment of performance. revision: yes

  2. Referee: [Reference dataset] Reference dataset section: The reference is constructed via 'careful preprocessing' of EEG signals to serve as ground truth for evaluating source-separation methods, yet the manuscript does not detail how these preprocessing steps differ from the two evaluated pipelines; this creates a risk that performance metrics simply reward reproduction of similar source-separation assumptions rather than independent physiological fidelity.

    Authors: This concern about potential circularity is well-taken. The reference dataset was constructed using a multi-stage pipeline involving manual artifact rejection via visual inspection, bandpass filtering, epoching around TMS pulses, and baseline correction—steps that rely on time-domain and expert judgment rather than the automated source-domain separation techniques (e.g., ICA or equivalent source imaging) evaluated in the study. These approaches differ fundamentally in their assumptions: the reference emphasizes preservation of known TEP features through conservative cleaning, while the evaluated methods use blind source separation. We will revise the Reference dataset section to include an explicit comparison of steps, assumptions, and potential overlaps, along with a supplementary table detailing the distinctions. revision: yes

  3. Referee: [Evaluation] Evaluation of artifact removal and TEP preservation: Without an independent validation strategy (e.g., simulated data with known ground truth, cross-validation against other modalities, or blinded expert scoring), the reported improvements in signal quality remain unverifiable and the central claim of robustness for closed-loop use rests on an untestable foundation.

    Authors: We fully acknowledge the inherent limitation that true physiological ground truth is unavailable for TMS-EEG recordings, as stated in the manuscript. The reference dataset is presented as a practical benchmark for comparative evaluation rather than absolute ground truth, with TEP preservation serving as a key physiologically interpretable metric. We agree that this does not constitute fully independent validation. In the revision, we will expand the discussion and limitations sections to explicitly address this point, including caveats for closed-loop applications and suggestions for future work (e.g., integration with other modalities where feasible). However, we cannot add new simulated datasets or blinded scoring without additional data collection, which is beyond the scope of the current study. revision: partial

standing simulated objections not resolved
  • Addition of fully independent validation data such as simulated TMS-EEG with known ground truth or concurrent multi-modal recordings (e.g., fMRI), as these were not part of the original study design and no such datasets are available for re-analysis.

Circularity Check

0 steps flagged

No circularity: empirical comparison of artifact removal methods on a reference dataset

full rationale

The paper is an empirical evaluation of two source-based artifact removal approaches for TMS-EEG signals. It creates a reference dataset via careful preprocessing and compares methods on signal quality and TEP preservation metrics, explicitly noting the lack of physiological ground truth. No mathematical derivation, equations, fitted parameters renamed as predictions, or self-citation chains are present in the provided text. The workflow is presented as a practical preprocessing pipeline rather than a first-principles derivation that reduces to its inputs by construction. This matches the default expectation of a non-circular empirical study.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract describes an empirical study without introducing new mathematical parameters, axioms, or postulated entities.

pith-pipeline@v0.9.0 · 5439 in / 1087 out tokens · 48640 ms · 2026-05-12T01:00:30.219547+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

10 extracted references · 10 canonical work pages

  1. [1]

    Removal of large muscle artifacts from transcranial magnetic stimulation - evoked EEG by independent component analysis

    Korhonen, R.J., Hernandez- Pavon, J.C., Metsomaa, J., Mäki, H., Ilmoniemi , R.J., Sarvas, J., 2011. Removal of large muscle artifacts from transcranial magnetic stimulation - evoked EEG by independent component analysis. Med. Biol. Eng. Comput. 49 (4), 397–407

    work page 2011

  2. [2]

    P., Kukkonen, M., Nieminen, J

    Mutanen, T. P., Kukkonen, M., Nieminen, J. O., Stenroos, M., Sarvas, J., & Ilmoniemi, R. J. (2018). Automatic and robust noise suppression in EEG and MEG: The SOUND algorithm. NeuroImage, 166, 135–151. https://doi.org/10.1016/j.neuroimage.2017.10.021

    work page doi:10.1016/j.neuroimage.2017.10.021 2018

  3. [3]

    P., Metsomaa, J., Makkonen, M., Varone, G., Marzetti, L., & Ilmoniemi, R

    Mutanen, T. P., Metsomaa, J., Makkonen, M., Varone, G., Marzetti, L., & Ilmoniemi, R. J. (2022). Source-based artifact- rejection techniques for TMS– EEG. Journal of Neuroscience Methods, 382, 109693. DOI: 10.1016/j.jneumeth.2022.109693

    work page doi:10.1016/j.jneumeth.2022.109693 2022

  4. [4]

    Signalspace projection suppresses the tACS artifact in EEG recordings

    Vosskuhl, J., Mutanen, T.P., Neuling, T., Ilmoniemi, R.J., Herrmann, C.S., 2020. Signalspace projection suppresses the tACS artifact in EEG recordings. Front. Human Neurosci. 14, 525

    work page 2020

  5. [5]

    ICLabel: an automated electroencephalographic independent component classifier, dataset, and website

    Pion-Tonachini, L., Kreutz -Delgado, K., Makeig, S., 2019. ICLabel: an automated electroencephalographic independent component classifier, dataset, and website. Neuroimage 198, 181–197. https://doi.org/10.1016/j.neuroimage.2019.05.026

    work page doi:10.1016/j.neuroimage.2019.05.026 2019

  6. [6]

    Altman, N., & Krzywinski, M. (2018). Statistics versus machine learning . Nature Methods, 15, 233–234

    work page 2018

  7. [7]

    Pion-Tonachini, L., Kreutz- Delgado, K., & Makeig, S. (2019). ICLabel: An automated electroencephalographic independent component classifier, dataset, and website . NeuroImage, 198, 181–197

    work page 2019

  8. [8]

    M., Lee, S., Lee, J., Börgers, C., Whittington, M

    Cannon, J., McCarthy, M. M., Lee, S., Lee, J., Börgers, C., Whittington, M. A., & Kopell, N. (2014). Neurosystems: Brain rhythms and cognitive processing . European Journal of Neuroscience, 39(5), 705–719. https://doi.org/10.1111/ejn.12453

    work page doi:10.1111/ejn.12453 2014

  9. [9]

    Scalp electrode impedance, infection risk, and EEG data quality

    Ferree, T.C., Luu, P., Russell, G.S., Tucker, D.M., 2001. Scalp electrode impedance, infection risk, and EEG data quality. Clin. Neurophysiol. 112, 536–544

    work page 2001

  10. [10]

    Cho, H., Ahn, M., Ahn, S., Kwon, M., & Jun, S. C. (2017). EEG datasets for motor imagery brain– computer interface. GigaScience , 6(7), https://doi.org/10.1093/gigascience/gix034

    work page doi:10.1093/gigascience/gix034 2017