Pith. sign in

REVIEW 2 minor 39 references

From EEG Cleaning to Decoding: The Role of Artifact Rejection in MI-based BCIs

T0 review · 0 major / 2 minor · reviewed 2026-05-25 · grok-4.3

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Automated artifact rejection improves motor imagery BCI decoding most for low-baseline subjects and reduces performance spread across users.

desk verdict FAAR gives a practical, low-overhead way to do adaptive artifact rejection in MI-BCIs and the 13-dataset results back the claim that gains are biggest where baseline SNR is poor. read the letter →

arxiv 2605.12408 v2 pith:V2K2SURN submitted 2026-05-12 eess.SP

Davoud Hajhassani , Bruno Aristimunha , Paul-Adrien Graignic , Apolline Mellot , Lionel Kusch , Arnaud Delorme , Thomas Semah , Arnault H. Caillet This is my paper · ORCID
classification eess.SP
keywords motorimageryBCIartifactrejectionEEGsignalqualitydecodingperformanceFAAR
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes FAAR as a lightweight automated method to reject contaminated EEG epochs in motor imagery tasks by building a Signal Quality Index from a small set of artifact-sensitive features and choosing thresholds without manual input. Tests across 13 public datasets show that the benefits of rejection depend on the individual subject and the starting signal quality, delivering the clearest gains when baseline accuracy or SNR is already low. The approach also narrows the gap in results between different subjects without discarding large amounts of data, which matters for making BCIs more reliable for a wider range of users.

What carries the argument

Fast Automatic Artifact Rejection (FAAR), a method that builds an epoch-level Signal Quality Index from artifact-sensitive features and applies adaptive thresholding to identify and reject contaminated epochs.

What would settle it

A new MI dataset in which applying FAAR to low-baseline subjects produces no accuracy gain or increases the spread of performance across subjects.

Watch

Extended reading notes

Core claim

FAAR computes a compact set of artifact-sensitive features, derives an epoch-level Signal Quality Index, and adaptively selects rejection thresholds to remove contaminated epochs without prior knowledge of artifact types or manual tuning. Evaluated on 13 MI datasets against a no-rejection baseline, AutoReject, and Isolation Forest, the method produces subject- and regime-dependent effects on decoding accuracy, with the largest improvements in low-baseline or low-SNR conditions, while reducing inter-subject performance variability without aggressive data removal and maintaining consistent behavior across offline, training, and online settings.

Load-bearing premise

A compact set of artifact-sensitive features and the derived Signal Quality Index, together with adaptive threshold selection, can reliably flag contaminated epochs across many different MI datasets without knowing the artifact types ahead of time or needing manual tuning.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

0 major / 2 minor

Summary. The paper introduces Fast Automatic Artifact Rejection (FAAR), a lightweight automated method for EEG artifact rejection in motor imagery (MI) BCIs. FAAR extracts a compact set of artifact-sensitive features, computes an epoch-level Signal Quality Index, and applies adaptive threshold selection to reject contaminated epochs without prior artifact knowledge or manual tuning. It evaluates the approach on 13 public MI datasets, comparing against a no-rejection baseline, AutoReject, and Isolation Forest, and reports that rejection effects are strongly subject- and regime-dependent (largest gains in low-baseline/low-SNR conditions), that FAAR reduces inter-subject performance variability without aggressive data removal, and that the method is consistent across offline, training, and online regimes while satisfying real-time constraints.

Significance. If the multi-dataset empirical results hold, the work provides concrete evidence that automated artifact rejection should be treated as an adaptive, regime-dependent component of MI-BCI pipelines rather than a fixed preprocessing step. The finding that gains are largest under low-SNR conditions and that inter-subject variability is reduced without heavy data loss directly addresses BCI illiteracy and reliability issues; the lightweight, fully automated design also supports deployment under real-time constraints.

minor comments (2)
  1. [Abstract] Abstract: states that evaluation results and comparisons were performed but supplies no quantitative performance numbers, statistical tests, or subject-exclusion criteria, which weakens the reader's ability to gauge the magnitude of the reported subject-dependent gains and variability reduction from the abstract alone.
  2. [Methods] The description of the compact artifact-sensitive feature set and the derived Signal Quality Index would benefit from an explicit enumeration or pseudocode in the methods section to allow exact reproduction.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary and recommendation of minor revision. The assessment correctly captures the core contributions of FAAR as a lightweight, adaptive artifact rejection method whose benefits are regime- and subject-dependent, and we appreciate the recognition that these results speak to BCI reliability and illiteracy issues.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; empirical evaluation on public datasets

full rationale

The paper proposes FAAR as a lightweight automated rejection method and evaluates its impact via direct comparison to baselines (no-rejection, AutoReject, Isolation Forest) across 13 public MI datasets. No mathematical derivations, parameter fits presented as predictions, or self-citation chains appear in the abstract or described methodology. Claims about subject- and regime-dependent effects rest on the multi-dataset empirical results rather than reducing to inputs by construction. The work is self-contained as an empirical comparison.

Assumptions & free parameters 0 free parameters · 0 assumptions · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; method is presented as parameter-free and automated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of From EEG Cleaning to Decoding: The Role of Artifact Rejection in MI-based BCIs." pith.science (2026). https://pith.science/paper/V2K2SURN

@misc{pith2026260512408,
  author       = {Pith},
  title        = {Pith review of: From EEG Cleaning to Decoding: The Role of Artifact Rejection in MI-based BCIs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/V2K2SURN}},
  note         = {Machine review of arXiv:2605.12408}
}
read the original abstract

Motor imagery (MI) BCIs are sensitive to EEG artifacts, yet the practical impact of automated artifact rejection on downstream MI decoding performance remains unclear. While most work focuses on decoder design, the contribution of data curation, particularly automated rejection policies, has received comparatively less attention, despite its importance for robust machine learning pipelines. Here, we propose Fast Automatic Artifact Rejection (FAAR), a lightweight method that computes a compact set of artifact-sensitive features, derives an epoch-level Signal Quality Index, adaptively selects rejection thresholds, and automatically rejects contaminated epochs without requiring prior knowledge of artifact types or manual threshold tuning. We evaluate FAAR on 13 publicly available MI datasets and compare it to a no-rejection baseline, AutoReject, and Isolation Forest. We show rejection effects are strongly subject- and regime-dependent, with the largest gains in low-baseline/low-SNR conditions, so it should be used adaptively. FAAR reduces inter-subject performance variability, an important property for MI-BCI reliability and BCI-illiteracy, without aggressive data removal. Finally, FAAR's lightweight and fully automated thresholding yields consistent rejection behavior across offline curation, training, and online filtering, and supports real-time BCI constraints.

Figures

Figures reproduced from arXiv: 2605.12408 by the authors.

Figure 1
Figure 1. A visual overview of the MI datasets included in this study. Hexagon [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Comparison of artifact rejection methods (AR, IF, and FAAR) against the baseline on the Lee2019 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Win rate for Left-Right Imagery, percentage of subjects for whom cleaning improves balanced accuracy (BA) as a function of baseline performance. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Effect of cleaning on inter-subject variability. Mean inter-subject [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Mean percentage of epochs rejected for LR imagery, motor imagery. [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages

  1. [1]

    An EEG- based brain-computer interface for cursor control,

    J. R. Wolpaw, D. J. McFarland, G. W. Neat, and C. A. Forneris, “An EEG- based brain-computer interface for cursor control,”Electroencephalogra- phy and Clinical Neurophysiology, 1991

    work page 1991

  2. [2]

    Review on motor imagery based BCI systems for upper limb post-stroke neurore- habilitation: From designing to application,

    M. A. Khan, R. Das, H. K. Iversen, and S. Puthusserypady, “Review on motor imagery based BCI systems for upper limb post-stroke neurore- habilitation: From designing to application,”Computers in Biology and Medicine, 2020

    work page 2020

  3. [3]

    Five-class motor imagery BCI classification and its application to brain-controlled wheelchairs,

    H. Pan, B. Teng, Z. Liu, S. Tong, X. Yu, and Z. Li, “Five-class motor imagery BCI classification and its application to brain-controlled wheelchairs,”Cognitive Neurodynamics, 2026

    work page 2026

  4. [4]

    Adaptable Neuro- physiological Biometric Authentication System and Method,

    A. H. Caillet, A. Shtarbanov, and T. Semah, “Adaptable Neuro- physiological Biometric Authentication System and Method,” Patent WO2025219294, published Oct. 22, 2025; filing no. WO2025EP60157 (Apr. 11, 2025)

    work page 2025

  5. [5]

    A comprehen- sive review of EEG-based brain-computer interface paradigms,

    R. Abiri, S. Borhani, E. W. Sellers, Y . Jiang, and X. Zhao, “A comprehen- sive review of EEG-based brain-computer interface paradigms,”J. Neural Eng., 2019

    work page 2019

  6. [6]

    A survey of signal processing algorithms in brain-computer interfaces based on electrical brain signals,

    A. Bashashati, M. Fatourechi, R. K. Ward, and G. E. Birch, “A survey of signal processing algorithms in brain-computer interfaces based on electrical brain signals,”J. Neural Eng., 2007

    work page 2007

  7. [7]

    Review of challenges associated with the EEG artifact removal methods,

    W. Mumtaz, S. Rasheed, and A. Irfan, “Review of challenges associated with the EEG artifact removal methods,”Biomedical Signal Processing and Control, 2021

    work page 2021

  8. [8]

    Improved Riemannian potato field: An automatic artifact rejection method for EEG,

    D. Hajhassani, Q. Barth ´elemy, J. Mattout, and M. Congedo, “Improved Riemannian potato field: An automatic artifact rejection method for EEG,” Biomedical Signal Processing and Control, 2026

    work page 2026

Show all 39 references
  1. [9]

    Methods for artifact detection and removal from scalp EEG: A review,

    M. K. Islam, A. Rastegarnia, and Z. Yang, “Methods for artifact detection and removal from scalp EEG: A review,”Neurophysiologie Clinique/Clinical Neurophysiology, 2016

    2016

  2. [10]

    Removal of artifacts from EEG signals: a review,

    X. Jiang, B. G. Bin, and Z. Tian, “Removal of artifacts from EEG signals: a review,”Sensors (Basel), 2019

    2019

  3. [11]

    Investigating effects of different artefact types on motor imagery BCI,

    L. Frølich, I. Winkler, K.-R. M ¨uller, and W. Samek, “Investigating effects of different artefact types on motor imagery BCI,” inProc. 2015 37th Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. (EMBC), 2015

    2015

  4. [12]

    The effect of processing techniques on the classification accuracy of brain-computer interface systems,

    A. Adolf, C. M. K ¨oll˝od, G. M ´arton, W. Fadel, and I. Ulbert, “The effect of processing techniques on the classification accuracy of brain-computer interface systems,”Brain Sciences, 2024

    2024

  5. [13]

    An improved feature extraction algorithms of EEG signals based on motor imagery brain-computer interface,

    X. Geng, D. Li, H. Chen, P. Yu, H. Yan, and M. Yue, “An improved feature extraction algorithms of EEG signals based on motor imagery brain-computer interface,”Alexandria Engineering Journal, 2022

    2022

  6. [14]

    Comparison of two methods of removing EOG artifacts for use in a motor imagery-based brain computer interface,

    M. Mohammadi and M. R. Mosavi, “Comparison of two methods of removing EOG artifacts for use in a motor imagery-based brain computer interface,”Evolving Systems, 2021

    2021

  7. [15]

    Subject-dependent artifact removal for enhancing motor imagery classi- fier performance under poor skills,

    M. Tob ´on-Henao, A. ´Alvarez-Meza, and G. Castellanos-Dom ´ınguez, “Subject-dependent artifact removal for enhancing motor imagery classi- fier performance under poor skills,”Sensors, 2022

    2022

  8. [16]

    Removing electroencephalographic ar- tifacts by blind source separation,

    T.-P. Jung, S. Makeig, C. Humphries, T.-W. Lee, M. J. McKeown, V . Iragui, and T. J. Sejnowski, “Removing electroencephalographic ar- tifacts by blind source separation,”Psychophysiology, 2000

    2000

  9. [17]

    EEG is better left alone,

    A. Delorme, “EEG is better left alone,”Sci. Rep., 2023

    2023

  10. [18]

    EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis,

    A. Delorme and S. Makeig, “EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis,”J. Neurosci. Methods, 2004

    2004

  11. [19]

    Real-time neuroimaging and cognitive monitoring using wearable dry EEG,

    T. R. Mullen, C. A. Kothe, Y . M. Chi, A. Ojeda, T. Kerth, S. Makeig, T.-P. Jung, and G. Cauwenberghs, “Real-time neuroimaging and cognitive monitoring using wearable dry EEG,”IEEE Transactions on Biomedical Engineering, vol. 62, no. 11, pp. 2553–2567, 2015

    2015

  12. [20]

    MEG and EEG data analysis with MNE-Python,

    A. Gramfort, M. Luessi, E. Larson, D. A. Engemann, D. Strohmeier, C. Brodbeck,et al., “MEG and EEG data analysis with MNE-Python,” Front. Neurosci., 2013

    2013

  13. [21]

    FASTER: fully automated statistical thresholding for EEG artifact rejection,

    H. Nolan, R. Whelan, and R. B. Reilly, “FASTER: fully automated statistical thresholding for EEG artifact rejection,”J. Neurosci. Methods, 2010

    2010

  14. [22]

    Autoreject: automated artifact rejection for MEG and EEG data,

    M. Jas, D. A. Engemann, Y . Bekhti, F. Raimondo, and A. Gramfort, “Autoreject: automated artifact rejection for MEG and EEG data,”Neu- roImage, 2017

    2017

  15. [23]

    Reliable and fast automatic artifact rejection of long-term EEG recordings based on Isolation Forest,

    R. Zhang, R. Rong, J. Q. Gan, Y . Xu, H. Wang, and X. Wang, “Reliable and fast automatic artifact rejection of long-term EEG recordings based on Isolation Forest,”Med. Biol. Eng. Comput., 2024

    2024

  16. [24]

    The Riemannian Potato: an automatic and adaptive artifact detection method for online experiments using Riemannian geometry,

    A. Barachant, A. Andreev, and M. Congedo, “The Riemannian Potato: an automatic and adaptive artifact detection method for online experiments using Riemannian geometry,” inProc. TOBI Workshop IV, 2013

    2013

  17. [25]

    The Rieman- nian potato field: a tool for online signal quality index of EEG,

    Q. Barth ´elemy, L. Mayaud, D. Ojeda, and M. Congedo, “The Rieman- nian potato field: a tool for online signal quality index of EEG,”IEEE Trans. Neural Syst. Rehabil. Eng., 2019

    2019

  18. [26]

    An automatic Riemannian artifact rejection method for P300-based BCIs,

    D. Hajhassani, J. Mattout, and M. Congedo, “An automatic Riemannian artifact rejection method for P300-based BCIs,” inProc. 2024 32nd Eur . Signal Process. Conf. (EUSIPCO), 2024

    2024

  19. [27]

    Good data? The EEG Quality Index for automated assessment of signal quality,

    S. D. Fickling, C. C. Liu, R. C. N. D’Arcy, S. Ghosh Hajra, and X. Song, “Good data? The EEG Quality Index for automated assessment of signal quality,” inProc. 2019 IEEE 10th Annu. Inf. Technol., Electron. Mobile Commun. Conf. (IEMCON), 2019

    2019

  20. [28]

    The effects of electrode impedance on data quality and statistical significance in ERP recordings,

    E. S. Kappenman and S. J. Luck, “The effects of electrode impedance on data quality and statistical significance in ERP recordings,”Psychophys- iology, 2010

    2010

  21. [29]

    Complexity analysis of EEG, MEG, and fMRI in mild cognitive impairment and Alzheimer’s disease: A review,

    J. Sun, B. Wang, Y . Niu, Y . Tan, C. Fan, N. Zhang, J. Xue, J. Wei, J. Xiang, “Complexity analysis of EEG, MEG, and fMRI in mild cognitive impairment and Alzheimer’s disease: A review,”Entropy (Basel), 2020

    2020

  22. [30]

    Najimet al.,Stochastic Processes: Estimation, Optimisation and Analysis

    K. Najimet al.,Stochastic Processes: Estimation, Optimisation and Analysis. Elsevier Ltd., 2004

    2004

  23. [31]

    Enhanced detection of arti- facts in EEG data using higher-order statistics and independent component analysis,

    A. Delorme, T. Sejnowski, and S. Makeig, “Enhanced detection of arti- facts in EEG data using higher-order statistics and independent component analysis,”NeuroImage, 2007

    2007

  24. [32]

    Evaluation of artifact subspace reconstruction for automatic artifact components removal in multi-channel EEG recordings,

    C.-Y . Chang, S.-H. Hsu, L. Pion-Tonachini, and T.-P. Jung, “Evaluation of artifact subspace reconstruction for automatic artifact components removal in multi-channel EEG recordings,”IEEE Trans. Biomed. Eng., 2020

    2020

  25. [33]

    Developing brain vital signs: Initial framework for monitoring brain function changes over time,

    S. Ghosh Hajraet al., “Developing brain vital signs: Initial framework for monitoring brain function changes over time,”Front. Neurosci., 2016

    2016

  26. [34]

    Finding a ‘Kneedle’ in a haystack: Detecting knee points in system behavior,

    V . Satopaa, J. Albrecht, D. Irwin, and B. Raghavan, “Finding a ‘Kneedle’ in a haystack: Detecting knee points in system behavior,” inProc. 2011 31st Int. Conf. ICDCSW, 2011

    2011

  27. [35]

    Kneeliverse: A universal knee- detection library for performance curves,

    M. Antunes, T. Estro, P. Bhandari, A. Gandhi, G. Kuenning, Y . Liu, C. Waldspurger, A. Wildani, and E. Zadok, “Kneeliverse: A universal knee- detection library for performance curves,” SoftwareX, vol. 30, Art. no. 102161, 2025

    2025

  28. [36]

    Aristimunhaet al.,Mother of All BCI Benchmarks (MOABB), ver

    B. Aristimunhaet al.,Mother of All BCI Benchmarks (MOABB), ver. 1.4.3. Zenodo, 2025

    2025

  29. [37]

    Multiclass brain- computer interface classification by Riemannian geometry,

    A. Barachant, S. Bonnet, M. Congedo, and C. Jutten, “Multiclass brain- computer interface classification by Riemannian geometry,”IEEE Trans. Biomed. Eng., 2012

    2012

  30. [38]

    The largest EEG-based BCI reproducibility study for open science: the MOABB benchmark,

    S. Chevallieret al., “The largest EEG-based BCI reproducibility study for open science: the MOABB benchmark,”HAL, working paper, hal- 04537061, 2024

    2024

  31. [39]

    Return of the GEDAI: Unsupervised EEG denoising based on leadfield filtering,

    T. Ros, V . F ´erat, Y . Huang, C. Colangelo, S. M. Kia, T. Wolfers, S. Vulliemoz, and M. Abele, “Return of the GEDAI: Unsupervised EEG denoising based on leadfield filtering,” bioRxiv, 2025

    2025

Pith tools

Reviewed May 25, 2026 · model on record in the stance chip above.