Pith. sign in

REVIEW 3 major objections 6 minor 75 references

Across hundreds of subjects, the best EEG motor-imagery decoding pipeline changes per person; a compact portfolio of twelve recovers most of the per-subject best.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 05:45 UTC pith:ZQFM4RCD

load-bearing objection The benchmark is a valuable reusable resource; the exploitability claim needs a null model and a train-selected baseline before it's credible. the 3 major comments →

arxiv 2607.22778 v1 pith:ZQFM4RCD submitted 2026-07-24 q-bio.NC cs.LG

Subject-Level Heterogeneity in EEG Motor Imagery Decoding: A Large-Scale Benchmark and Portfolio-Based Reduction of the Search Space

Paul Barbaste , Olivier Oullier , Xavier Vasques This is my paper
classification q-bio.NC cs.LG
keywords EEGmotor imagerybrain-computer interfacepipeline benchmarkingsubject heterogeneityportfolio selectionCommon Spatial PatternsRiemannian tangent space
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper argues that EEG motor-imagery decoding has no universal best pipeline: aggregate rankings conceal that the optimal full configuration changes markedly from subject to subject. Across 52 subjects in one dataset and 109 in another, 42 and 93 distinct winning pipelines were observed, and the best feature family varied as well. The authors then use the exhaustive benchmark as an empirical landscape to build compact portfolios of pipelines. A single best global pipeline retains 94.2% of the per-subject oracle in the 52-subject dataset and 81.8% in the 109-subject dataset; a portfolio of just 12 pipelines raises retention to 96.5% and 90.0%, respectively. This matters because it suggests that personalization can be approached by choosing among a small set of strong candidate pipelines rather than by exhaustive search or a one-size-fits-all decoder.

Core claim

The decoding landscape is subject-dependent, and this heterogeneity is practically exploitable. The paper shows that on Cho2017 (52 subjects) 42 different full pipelines were each the best for some subject, and on PhysionetMI (109 subjects) 93 were; even the winning feature family was not stable. Using the benchmark itself as a performance landscape, the authors construct portfolios: the Top-K Mean heuristic, which selects the K pipelines with the highest mean training balanced accuracy, retains 96.5% of the subject-specific oracle on Cho2017 and 90.0% on PhysionetMI at K=12, compared with 94.2% and 81.8% for the single best global pipeline. The diversity that drives these gains lies mostly

What carries the argument

The central object is the subject-by-pipeline performance matrix, built from 216,714 raw benchmark evaluations across three datasets, two frequency bands, six feature-extraction families (covariance tangent-space projection, CSP, coherence-based tangent space, Hjorth, Higuchi fractal dimension, SVD entropy), multiple scalers, and several classifiers. Portfolio construction uses repeated 80/20 subject-level splits: the Top-K Mean heuristic selects the K pipelines with the highest mean balanced accuracy on training subjects, and the portfolio is evaluated on held-out subjects via the oracle-in-set metric, defined as the mean across test subjects of the best score achieved by any pipeline in th

Load-bearing premise

The results assume a selector can identify, for each new subject, the best-performing pipeline inside the portfolio; the paper states it does not yet provide such a mechanism, so without it the portfolio's 90-96.5% retention is an upper bound rather than achieved personalization.

What would settle it

Test the portfolio on new subjects with a practical selection rule, such as a two-minute calibration block evaluated on the K=12 pipelines, and compare realized accuracy against the oracle-in-set value; if the realized gain over the single best pipeline is zero or negative, the exploitable-heterogeneity claim is refuted. Alternatively, compare Top-K Mean against a random K=12 portfolio: if random selection matches its retention, the gain is not attributable to subject-level structure.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • A single fixed global pipeline leaves a measurable, dataset-dependent gap relative to the per-subject best; the gap is largest in the most heterogeneous dataset.
  • Reducing the search space to about twelve pipelines retains 90-96.5% of the subject-specific oracle, making personalization computationally feasible.
  • The useful diversity in portfolios comes mainly from within the dominant feature family, so practitioners can focus tuning effort on scalers and classifiers inside that family.
  • The results reframe so-called BCI illiteracy: poor decoding performance may reflect a pipeline-subject mismatch rather than an inherent inability of the user.
  • The natural next test is whether the same portfolio logic holds under cross-session and participant-independent evaluation protocols.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The paper defines the oracle-in-set but provides no mechanism to choose the best pipeline for a new subject in real time; the reported retention ratios are therefore upper bounds, not achieved personalization, until a practical selector is built.
  • The 21 ties observed on PhysionetMI before tie-breaking suggest that top pipelines are often statistically interchangeable; a random portfolio of the same size could capture a substantial part of the gain, which would weaken the claim that the landscape has exploitable subject-level structure.
  • Because the selected portfolios are dominated by cov-tgsp variants, a cheaper extension would be to test whether hyperparameter diversity alone within that one family saturates the oracle-retention curve.
  • A concrete next step is to use a short calibration block per new subject to pick among the portfolio members, then compare realized accuracy with the oracle-in-set; this would directly test the practical value of the portfolio.

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

3 major / 6 minor

Summary. The paper reports a large-scale, standardized within-session benchmark of EEG motor imagery decoding pipelines across three public datasets (Cho2017: 52 subjects, PhysionetMI: 109 subjects, Zhou2016: 4 subjects), using the MOABB LeftRightImagery framework, two frequency bands, six feature families, multiple preprocessing steps, and classical/MLP classifiers. The authors report that cov-tgsp and CSP are the strongest families overall, but that the identity of the best full pipeline varies strongly across subjects (42 distinct winners across 52 Cho2017 subjects; 93 across 109 PhysionetMI subjects). They then construct compact portfolios of pipelines from the benchmark and report oracle-retention ratios, claiming a single best global pipeline retains 94.2%/81.8% of the subject-specific oracle and a K=12 Top-K Mean portfolio raises this to 96.5%/90.0%. The paper concludes that the decoding landscape is subject-dependent and that this heterogeneity can be exploited through compact portfolios.

Significance. If the heterogeneity and portfolio claims are supported, the paper would be a useful contribution to BCI benchmarking: it provides a large, standardized, publicly reproducible evaluation of a broad pipeline space, checks robustness across accuracy/balanced-accuracy/AUC/F1/precision/recall, and proposes a concrete way to reduce a large search space to a compact set of candidate pipelines. The availability of code and the use of held-out subjects in portfolio evaluation are clear strengths. However, the central interpretive claims about “exploitable” subject-level heterogeneity currently rest on oracle-based quantities and lack any null-model comparison, so the practical conclusion is not yet established.

major comments (3)
  1. [§3.2, Table 2; §3.5, Fig. 4] The evidence for “true” subject-level heterogeneity and portfolio complementarity is not benchmarked against a null model. The subject-by-pipeline matrices contain roughly 864–1048 scores per subject. Under the null that each pipeline has the same expected accuracy for a given subject (no subject×pipeline interaction), the argmax is essentially random: with that many near-tied noisy scores, nearly every subject will have a different “winner”, and the max-of-K oracle-in-set mechanically increases with K. The reported distinct-winner counts (42/52, 93/109) and retention gains (94.2→96.5% Cho2017; 81.8→90.0% PhysionetMI) are therefore not sufficient to establish that the landscape is subject-dependent or that the gains reflect real complementarity. Please add a permutation or synthetic-null control (e.g., permuting subject labels across pipelines, or simulating scores from subject and pipel
  2. [§2.6, §3.5, §5] The K=1 baseline is not a feasible fixed-pipeline baseline. Section 2.6 defines the “fixed best global pipeline” as the pipeline with the highest mean test performance, i.e., it is selected after seeing the held-out test subjects. Consequently, the reported retention at K=1 (94.2%/81.8%) is an oracle-selected upper bound, not the performance of a train-selected single pipeline. In addition, the portfolio quantities oracle-in-set and oracle-retention ratio (Eq. 1) are upper bounds: no selector exists to choose the best pipeline for a new subject, as Section 4 concedes. The conclusions in Section 5, which state that heterogeneity is “practically exploitable” through compact portfolios, overstate what the current analysis supports. Please supply a train-selected K=1 baseline and/or explicitly frame all portfolio results as upper bounds pending a calibration-based or transfer-based selector.
  3. [§3.4, Table 1] The family-level statistical comparisons use, for each subject, the best-performing pipeline within each family, but the number of pipeline variants differs substantially across families and bands (e.g., Cov+TGSP 8–30 Hz in Cho2017 has only 16 pipelines, while most other family-band cells have 88). Best-of-family scores systematically favor families with more variants, so the Friedman/Wilcoxon comparisons in §3.4 do not test family quality per se. This is a load-bearing issue for the descriptive claim that cov-tgsp and CSP are the “strongest” families. Please either control for the number of variants (e.g., matched subsampling or per-family model-selection estimates) or explicitly state and justify the best-of-family aggregation, or report the sensitivity of the family ranking to this choice.
minor comments (6)
  1. [Abstract vs. §3.1] The abstract states 44,928, 109,000, and 4,192 subject-level observations, while the full-text abstract states 61,464, 132,762, and 22,488. The latter are raw-row counts; please reconcile the terminology to avoid confusion.
  2. [§2.2.2, §2.4] Several free parameters are stated without justification, including HFD maximum scale k=10, the post-cue window 0.6–2.0 s, and the two frequency bands. A short sensitivity analysis or citation for these choices would strengthen the paper.
  3. [Figure 4] The right-column panels have an axis label that appears cut off (“ balanced accuracy”), and the y-axis in panels A/C/E should be explicitly labeled as the oracle-retention ratio. Also, the shaded regions are fold-wise standard deviations; a note on how many folds are included would help.
  4. [Table 2] The “Ties” column reports tie cases before tie-breaking, but the tie-breaking rule is not described. Please state how ties were resolved for the distinct-winner counts.
  5. [§2.6] The portfolio-selection strategies (greedy, regret_greedy, submodular coverage, etc.) are described verbally; the composite objective is not given a precise formula. Please include the exact objective used for the non-Top-K strategies, or point to the code with a stable version/DOI, so the results are reproducible.
  6. [Throughout] The manuscript uses “PhysionetMI” and “PhysionetMotorImagery” interchangeably; please standardize. There are also duplicated references (e.g., [50] and [64] both list Cho2017; [57] repeats [15]).

Circularity Check

0 steps flagged

No circularity: the benchmark and portfolio results are empirical, externally grounded evaluations, not derivations that reduce to their inputs.

full rationale

The paper's central claims are empirical benchmark results on three public datasets (Cho2017, PhysionetMI, Zhou2016) under a standardized MOABB within-session protocol. Pipeline scores are generated by training classifiers on held-out trial splits and evaluating on unseen test trials, so the performance landscape is not derived from the portfolio objective or from a fitted parameter. The portfolio analysis is also structured to avoid circularity: portfolios are selected on training subjects and evaluated on held-out test subjects (Section 2.6: 'The portfolio was always learned on the training subjects only and then evaluated on the held-out test subjects'). The oracle metrics ('global oracle', 'oracle-in-set') are explicitly defined as retrospective upper bounds: 'This quantity represents the best performance that the portfolio could offer if, for each test subject, the most suitable pipeline within that portfolio could be chosen.' The paper openly acknowledges that no operational selector is provided (Section 4: 'it does not yet provide an operational mechanism for selecting the best pipeline for a truly new user in real time'), which is a practical limitation rather than a circularity. The K=1 baseline is defined as the single pipeline with the highest mean test performance, which is an oracle-selected reference and may understate portfolio gains, but this is a benchmarking/statistical choice, not a self-definitional or fitted-input-called-prediction reduction. The self-citations present (e.g., references [9], [16], [82]) are used as background or domain context and are not load-bearing for the benchmark derivation or the portfolio conclusion. There is no uniqueness theorem imported from the authors' prior work, no ansatz smuggled in via citation, and no renaming of a known result presented as derivation. Therefore no specific circular step can be exhibited.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 0 invented entities

No new entities or fitted constants were introduced. The central claims rest on MOABB data loading, leakage-free within-session evaluation, and the oracle-in-set upper-bound construction. The main unexamined assumption is that oracle-in-set can be realized by a selector, which the paper itself acknowledges is missing.

free parameters (5)
  • HFD maximum scale k_max = 10
    Hand-selected in Section 2.2.2; affects only the nonlinear feature family, which is secondary in the results.
  • post-cue epoch window = 0.6-2.0 s
    Hand-selected in Sections 2.1 and 2.2.1; applied uniformly across pipelines and could affect absolute accuracy but not the central portfolio logic.
  • frequency bands = 8-15 Hz and 8-30 Hz
    Chosen in Section 2.2.1 as standard mu/beta bands; both are benchmarked, so neither is fitted to outcomes.
  • ShuffleSplit settings = 10 repetitions, test_fraction=0.2, seed=42
    Hand-selected in Section 2.6; determines the cross-validated oracle-retention estimates.
  • MLP architecture set = 17 predefined architectures
    MLP_1 to MLP_17 in Section 2.4; the choice of this set influences which pipelines can win, though no per-dataset tuning is performed.
axioms (5)
  • domain assumption MOABB's dataset loaders and LeftRightImagery event annotations correctly map trials to labels.
    Section 2.1; if label mapping is wrong, all accuracy numbers are invalid.
  • domain assumption Within-session accuracy on MOABB splits is leakage-free and a meaningful measure of pipeline quality.
    Section 2.5; the paper states this protocol evaluates unseen trials from the same session only.
  • ad hoc to paper Best-pipeline-within-family scores can be compared across families without correcting for unequal numbers of variants.
    Section 3.4 implicitly assumes this; family sizes differ (16 vs 88 pipelines), which biases best-of-N comparisons.
  • ad hoc to paper Oracle-in-set retention can be interpreted as practical exploitability.
    Defined in Section 2.6; this assumes a perfect per-subject selector, which the paper does not provide.
  • domain assumption The three public datasets are sufficiently representative to support subject-level heterogeneity claims.
    Section 2.1; Zhou2016 has only 4 subjects, and all results are within-session only.

pith-pipeline@v1.3.0-alltime-deepseek · 26035 in / 13397 out tokens · 137449 ms · 2026-08-01T05:45:27.508078+00:00 · methodology

0 comments
read the original abstract

Robust EEG motor imagery decoding remains limited by strong inter-individual variability, making it difficult to identify pipelines that generalize across users. We present a large-scale, standardized within-session benchmark of decoding pipelines across three public datasets: Cho2017 (52 subjects), PhysionetMI (109 subjects), and Zhou2016 (4 subjects). Using a common MOABB LeftRightImagery setting, two frequency bands (8-15 Hz and 8-30 Hz), and a broad combination of feature extraction, preprocessing, and classification steps, we analyzed 216,714 raw evaluation rows, which after structured aggregation yielded 44,928, 109,000, and 4,192 subject-level observations respectively. Covariance tangent-space projection (cov-tgsp) and Common Spatial Patterns (CSP) consistently defined the strongest methodological families, though their relative ordering was dataset-dependent. On Cho2017, the best family-level mean accuracy came from cov-tgsp in 8-30 Hz (0.712 +/- 0.140), whereas Zhou2016 favored CSP (0.832 +/- 0.121 in 8-15 Hz). These aggregate rankings concealed substantial subject-level heterogeneity: 42 distinct winning pipelines across 52 Cho2017 subjects, and 93 across 109 PhysionetMI subjects. We then used the benchmark as an empirical performance landscape for building compact portfolios of pipelines of size K. Several construction procedures were compared, including a ranking-based Top-K Mean heuristic and search-based strategies. Results were broadly consistent, with Top-K Mean giving the best trade-off. A single best global pipeline already retained 94.2% of the oracle in Cho2017 and 81.8% in PhysionetMI; at K = 12, oracle retention rose to 96.5% and 90.0%. The landscape is therefore subject-dependent, and this heterogeneity can be exploited through compact portfolios that make personalization more feasible.

Figures

Figures reproduced from arXiv: 2607.22778 by Olivier Oullier, Paul Barbaste, Xavier Vasques.

Figure 1
Figure 1. Figure 1: Global distribution of subject-level benchmark scores across methodological families, datasets, and frequency bands. Each point represents an aggregated subject-level pipeline score after collapsing repeated raw outputs at the subject-session￾pipeline level and averaging across sessions when applicable. Methods are displayed by family on the x-axis, while marker position and grouping allow visual compariso… view at source ↗
Figure 2
Figure 2. Figure 2: Comparative performance of methodological families across frequency bands and datasets. Panel A shows mean subject-level accuracy for each methodological family in the 8-15 Hz and 8-30 Hz bands after averaging across datasets. Panel B shows mean subject-level accuracy for the same families in Cho2017, PhysionetMI, and Zhou2016 after averaging across frequency bands. Error bars denote standard deviations de… view at source ↗
Figure 3
Figure 3. Figure 3: Subject- and dataset-level heterogeneity of optimal decoding configurations. Panel A displays the best-performing feature family for each subject across Cho2017, PhysionetMI, and Zhou2016. Panel B reports the distribution of winning feature families across datasets. Panel C reports the distribution of winning preprocessing scalers across datasets. Panel D reports the distribution of winning classifiers acr… view at source ↗
Figure 4
Figure 4. Figure 4: Portfolio optimization performance across datasets and portfolio sizes. Multi-panel summary of portfolio optimiza￾tion results using the topk_mean portfolio-selection heuristic. Rows correspond to Cho2017, PhysionetMotorImagery, and Zhou2016. Left-column panels show the oracle retention ratio as a function of portfolio size K, where K = 1 denotes the fixed best single global pipeline and larger K values de… view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of portfolio-construction strategies. Panel A shows the rank of each strategy within each dataset￾by-band configuration, based on mean oracle-retention ratio averaged across the evaluated portfolio sizes. Cell labels report both rank and mean oracle-retention value. Panel B summarizes the overall mean oracle-retention ratio of each strategy across all dataset-by-band configurations. Panel C show… view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

75 extracted references · 4 canonical work pages · 2 internal anchors

  1. [1]

    Khan, S.et al.Invasive Brain–Computer Interface for Communication: A Scoping Review.Brain Sci.15, 336 (2025)

    2025

  2. [2]

    J.et al.Non-Invasive Brain-Computer Inter- faces: State of the Art and Trends.IEEE Rev

    Edelman, B. J.et al.Non-Invasive Brain-Computer Inter- faces: State of the Art and Trends.IEEE Rev. Biomed. Eng.18, 26–49 (2025)

    2025

  3. [3]

    Saha, S.et al.Progress in Brain Computer Interface: Challenges and Opportunities.Front. Syst. Neurosci.15, 578875 (2021). 19

    2021

  4. [4]

    Chen, J.et al.fNIRS-EEG BCIs for Motor Rehabilita- tion: A Review.Bioengineering10, 1393 (2023)

    2023

  5. [5]

    Stavisky, S. D. Restoring Speech Using Brain–Computer Interfaces.Annu. Rev. Biomed. Eng.27, 29–54 (2025)

    2025

  6. [6]

    V.et al.Sensorimotor ECoG Signal Features for BCI Control: A Comparison Between Peo- ple With Locked-In Syndrome and Able-Bodied Controls

    Freudenburg, Z. V.et al.Sensorimotor ECoG Signal Features for BCI Control: A Comparison Between Peo- ple With Locked-In Syndrome and Able-Bodied Controls. Front. Neurosci.13, 1058 (2019)

    2019

  7. [7]

    J.et al.Invasive Brain Computer Interface for Motor Restoration in Spinal Cord Injury: A Systematic Review.Neuromodulation Technol

    Levett, J. J.et al.Invasive Brain Computer Interface for Motor Restoration in Spinal Cord Injury: A Systematic Review.Neuromodulation Technol. Neural Interface27, 597–603 (2024)

    2024

  8. [8]

    B., Littlejohn, K

    Silva, A. B., Littlejohn, K. T., Liu, J. R., Moses, D. A. & Chang, E. F. The speech neuroprosthesis.Nat. Rev. Neurosci.25, 473–492 (2024)

    2024

  9. [9]

    & Vasques, X

    Oullier, O., Roser, F., Barbaste, P. & Vasques, X. Improving consciousness assessment through neu- roadaptive artificial intelligence and quantum-enhanced brain-computer interfaces.Clin. Neurol. Neurosurg.266, 109396 (2026)

    2026

  10. [10]

    Maslova, O.et al.Non-invasive EEG-based BCI spellers from the beginning to today: a mini-review.Front. Hum. Neurosci.17, 1216648 (2023)

    2023

  11. [11]

    & Kam, T.-E

    Kim, D.-H., Shin, D.-H. & Kam, T.-E. Bridging the BCI illiteracy gap: a subject-to-subject semantic style trans- fer for EEG-based motor imagery classification.Front. Hum. Neurosci.17, 1194751 (2023)

    2023

  12. [12]

    R.et al.Reach and grasp by people with tetraplegia using a neurally controlled robotic arm.Na- ture485, 372–375 (2012)

    Hochberg, L. R.et al.Reach and grasp by people with tetraplegia using a neurally controlled robotic arm.Na- ture485, 372–375 (2012)

    2012

  13. [13]

    Xu, L.et al.Cross-Dataset Variability Problem in EEG Decoding With Deep Learning.Front. Hum. Neurosci. 14, 103 (2020)

    2020

  14. [16]

    J., Steinberg, F

    Oullier, O., Jantzen, K. J., Steinberg, F. L. & Kelso, J. A. S. Neural Substrates of Real and Imagined Sensorimotor Coordination.Cereb. Cortex15, 975–985 (2005)

    2005

  15. [17]

    J., Joordens, S

    Gibson, E., Lobaugh, N. J., Joordens, S. & McIntosh, A. R. EEG variability: Task-driven or subject-driven signal of interest?NeuroImage252, 119034 (2022)

    2022

  16. [18]

    & Jun, S

    Ahn, M., Cho, H., Ahn, S. & Jun, S. C. High theta and low alpha powers may be indicative of BCI-illiteracy in motor imagery.PloS One8, e80886 (2013)

    2013

  17. [19]

    & Al Dabagh, Y

    Becker, S., Dhindsa, K., Mousapour, L. & Al Dabagh, Y. BCI Illiteracy: It’s Us, Not Them. Optimizing BCIs for Individual Brains. in2022 10th International Winter Conference on Brain-Computer Interface (BCI) 1–3 (IEEE, Gangwon-do, Korea, Republic of, 2022). doi:10.1109/BCI53720.2022.9735007

    arXiv 2022

  18. [22]

    & Ahi, K

    Hosseini, M.-P., Hosseini, A. & Ahi, K. A Review on Machine Learning for EEG Signal Processing in Bioengi- neering.IEEE Rev. Biomed. Eng.14, 204–218 (2021)

    2021

  19. [23]

    & Contreras-Vidal, J

    Craik, A., He, Y. & Contreras-Vidal, J. L. Deep learning for electroencephalogram (EEG) classification tasks: a review.J. Neural Eng.16, 031001 (2019)

    2019

  20. [24]

    Saeidi, M.et al.Neural Decoding of EEG Signals with Machine Learning: A Systematic Review.Brain Sci.11, 1525 (2021)

    2021

  21. [25]

    & Haffen, E

    Joucla, C., Gabriel, D., Ortega, J.-P. & Haffen, E. Three simple steps to improve the interpretability of EEG-SVM studies.J. Neurophysiol.128, 1375–1382 (2022)

    2022

  22. [26]

    & Marchesotti, S

    Wu, S., Bhadra, K., Giraud, A.-L. & Marchesotti, S. Adaptive LDA Classifier Enhances Real-Time Control of an EEG Brain–Computer Interface for Decoding Imag- ined Syllables.Brain Sci.14, 196 (2024)

    2024

  23. [27]

    EEGNet: A Compact Convolutional Network for EEG-based Brain-Computer Interfaces

    Lawhern, V. J.et al.EEGNet: A Compact Con- volutional Network for EEG-based Brain-Computer Interfaces. https://doi.org/10.48550/ARXIV.1611.08024 (2016) doi:10.48550/ARXIV.1611.08024

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1611.08024 2016

  24. [28]

    & Kosslyn, S

    Pearson, J. & Kosslyn, S. M. The heterogeneity of mental representation: Ending the imagery debate.Proc. Natl. Acad. Sci.112, 10089–10092 (2015)

    2015

  25. [30]

    & Jutten, C

    Barachant, A., Bonnet, S., Congedo, M. & Jutten, C. Classification of covariance matrices using a Riemannian- based kernel for BCI applications.Neurocomputing112, 172–178 (2013)

    2013

  26. [31]

    A., Larrondo, H

    Rosso, O. A., Larrondo, H. A., Martin, M. T., Plastino, A. & Fuentes, M. A. Distinguishing Noise from Chaos. Phys. Rev. Lett.99, 154102 (2007)

    2007

  27. [32]

    K., Barthélemy, Q

    Chevallier, S., Kalunga, E. K., Barthélemy, Q. & Mona- celli, E. Review of Riemannian Distances and Diver- gences, Applied to SSVEP-based BCI.Neuroinformatics 19, 93–106 (2021)

    2021

  28. [33]

    Zhong, X.-C.et al.EEG-DG: A Multi-Source Domain Generalization Framework for Motor Imagery EEG Clas- sification.IEEE J. Biomed. Health Inform.29, 2484–2495 (2025)

    2025

  29. [34]

    &Krishnan, S

    Singh, A.K. &Krishnan, S. Trends inEEGsignal feature extraction applications.Front. Artif. Intell.5, 1072801 (2023). 20

    2023

  30. [35]

    & Gupta, M

    Gosala, B., Dindayal Kapgate, P., Jain, P., Nath Chaura- sia, R. & Gupta, M. Wavelet transforms for feature engineering in EEG data processing: An application on Schizophrenia.Biomed. Signal Process. Control85, 104811 (2023)

    2023

  31. [36]

    Md., Uddin, M

    Redwan, S. Md., Uddin, M. P., Ulhaq, A., Sharif, M. I. & Krishnamoorthy, G. Power spectral density-based resting-state EEG classification of first-episode psychosis. Sci. Rep.14, 15154 (2024)

    2024

  32. [37]

    Apicella, A.et al.Toward cross-subject and cross- session generalization in EEG-based emotion recog- nition: Systematic review, taxonomy, and meth- ods.https://doi.org/10.48550/ARXIV.2212.08744(2022) doi:10.48550/ARXIV.2212.08744

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2212.08744 2022

  33. [38]

    Hurst, H. E. Long-Term Storage Capacity of Reservoirs. Trans. Am. Soc. Civ. Eng.116, 770–799 (1951)

    1951

  34. [39]

    Approach to an irregular time series on the basis of the fractal theory.Phys

    Higuchi, T. Approach to an irregular time series on the basis of the fractal theory.Phys. Nonlinear Phenom.31, 277–283 (1988)

    1988

  35. [42]

    & Jun, S

    Park, H. & Jun, S. C. Connectivity study on resting- state EEG between motor imagery BCI-literate and BCI- illiterate groups.J. Neural Eng.21, 046042 (2024)

    2024

  36. [43]

    Chen, L.et al.Task-related reconfiguration patterns of frontoparietal network during motor imagery.Neuro- science579, 302–311 (2025)

    2025

  37. [44]

    inBrain Informatics(eds Mahmud, M., Kaiser, M

    Kunjan, S.et al.The Necessity of Leave One Subject Out(LOSO)CrossValidationforEEGDiseaseDiagnosis. inBrain Informatics(eds Mahmud, M., Kaiser, M. S., Vassanelli, S., Dai, Q. & Zhong, N.) vol. 12960 558–567 (Springer International Publishing, Cham, 2021)

    2021

  38. [45]

    P., Torres, E

    Torres, E. P., Torres, E. A., Hernández-Álvarez, M. & Yoo, S. G. EEG-Based BCI Emotion Recognition: A Sur- vey.Sensors20, 5083 (2020)

    2020

  39. [46]

    & Jutten, C

    Barachant, A., Bonnet, S., Congedo, M. & Jutten, C. Multiclass Brain–Computer Interface Classification by Riemannian Geometry.IEEE Trans. Biomed. Eng.59, 920–928 (2012)

    2012

  40. [48]

    & Hou, Y

    Lun, X., Yu, Z., Chen, T., Wang, F. & Hou, Y. A Sim- plified CNN Classification Method for MI-EEG via the Electrode Pairs Signals.Front. Hum. Neurosci.14, 338 (2020)

    2020

  41. [50]

    & Jun, S

    Cho, H., Ahn, M., Ahn, S., Kwon, M. & Jun, S. C. EEG datasets for motor imagery brain–computer interface.Gi- gaScience6, (2017)

    2017

  42. [51]

    Antony, M.J.et al.ClassificationofEEGUsingAdaptive SVM Classifier with CSP and Online Recursive Indepen- dent Component Analysis.Sensors22, 7596 (2022)

    2022

  43. [53]

    & Abderrahim, M

    Omari, S., Omari, A., Abu-Dakka, F. & Abderrahim, M. EEG Motor Imagery Classification: Tangent Space with Gate-Generated Weight Classifier.Biomimetics9, (2024)

    2024

  44. [55]

    & Lee, T

    Islam, M. & Lee, T. Functional Connectivity Analysis in Multi-channel EEG for Emotion Detection with Phase Locking Value and 3D CNN. in2023 45th Annual Inter- national Conference of the IEEE Engineering in Medicine & Biology Society (EMBC)1–4(IEEE,Sydney, Australia, 2023). doi:10.1109/EMBC40787.2023.10340922

    arXiv 2023

  45. [56]

    & Bartolomei, F

    Lagarde, S., Bénar, C.-G., Wendling, F. & Bartolomei, F. Interictal Functional Connectivity in Focal Refractory Epilepsies Investigated by Intracranial EEG.Brain Con- nect.12, 850–869 (2022)

    2022

  46. [57]

    EEG analysis based on time domain proper- ties.Electroencephalogr

    Hjorth, B. EEG analysis based on time domain proper- ties.Electroencephalogr. Clin. Neurophysiol.29, 306–310 (1970)

    1970

  47. [58]

    & Jassim, S

    Ghafuri, J. & Jassim, S. Singular-Value-Decomposition- Based Matrix Surgery.Entropy26, 701 (2024)

    2024

  48. [59]

    & Zhu, J

    Gu, X., Fan, Y., Zhou, J. & Zhu, J. Optimized Projection and Fisher Discriminative Dictionary Learning for EEG Emotion Recognition.Front. Psychol.12, 705528 (2021)

    2021

  49. [60]

    & Spasić, S

    Kesić, S. & Spasić, S. Z. Application of Higuchi’s frac- tal dimension from basic to clinical neurophysiology: A review.Comput. Methods Programs Biomed.133, 55–70 (2016)

    2016

  50. [61]

    Preprint at https://doi.org/10.48550/ARXIV

    Chevallier, S.et al.The largest EEG-based BCI repro- ducibility study for open science: the MOABB bench- mark. Preprint at https://doi.org/10.48550/ARXIV. 2404.15319 (2024)

    Pith/arXiv arXiv doi:10.48550/arxiv.2404.15319 2024

  51. [62]

    & Barachant, A

    Jayaram, V. & Barachant, A. MOABB: trustworthy algo- rithm benchmarking for BCIs.J. Neural Eng.15, 066011 (2018)

    2018

  52. [63]

    Zenodo https://doi.org/10.5281/ZENODO.10034223 (2024)

    Aristimunha, B.et al.Mother of all BCI Benchmarks. Zenodo https://doi.org/10.5281/ZENODO.10034223 (2024). 21

    work page doi:10.5281/zenodo.10034223 2024

  53. [64]

    J., Hinterberger, T., Bir- baumer, N

    Schalk, G., McFarland, D. J., Hinterberger, T., Bir- baumer, N. & Wolpaw, J. R. BCI2000: A General- Purpose Brain-Computer Interface (BCI) System.IEEE Trans. Biomed. Eng.51, 1034–1043 (2004)

    2004

  54. [65]

    & Guo, X

    Zhou, B., Wu, X., Lv, Z., Zhang, L. & Guo, X. A Fully Automated Trial Selection Method for Optimization of Motor Imagery Based Brain-Computer Interface.PLOS ONE11, e0162657 (2016)

    2016

  55. [66]

    J., Lazar, M

    Koles, Z. J., Lazar, M. S. & Zhou, S. Z. Spatial patterns underlying population differences in the back- ground EEG.Brain Topogr.2, 275–284 (1990)

    1990

  56. [67]

    Lotte, F.et al.A review of classification algorithms for EEG-based brain–computer interfaces: a 10 year update. J. Neural Eng.15, 031005 (2018)

    2018

  57. [68]

    https://doi.org/10.48550/ARXIV.1201.0490 (2012) doi:10.48550/ARXIV.1201.0490

    Pedregosa, F.et al.Scikit-learn: Machine Learning in Python. https://doi.org/10.48550/ARXIV.1201.0490 (2012) doi:10.48550/ARXIV.1201.0490

    Pith/arXiv arXiv doi:10.48550/arxiv.1201.0490 2012

  58. [69]

    L.et al.PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals.Circulation101, (2000)

    Goldberger, A. L.et al.PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals.Circulation101, (2000)

    2000

  59. [70]

    Astolfi, L.et al.Comparison of different cortical con- nectivity estimators for high-resolution EEG recordings. Hum. Brain Mapp.28, 143–157 (2007)

    2007

  60. [71]

    & Lotte, F

    Yger, F., Berar, M. & Lotte, F. Riemannian Approaches in Brain-Computer Interfaces: A Review.IEEE Trans. Neural Syst. Rehabil. Eng.25, 1753–1762 (2017)

    2017

  61. [72]

    Neural Eng.16, 051001 (2019)

    Roy, Y.et al.Deep learning-based electroencephalogra- phy analysis: a systematic review.J. Neural Eng.16, 051001 (2019)

    2019

  62. [73]

    & Pfurtscheller, G

    Ramoser, H., Muller-Gerking, J. & Pfurtscheller, G. Op- timal spatial filtering of single trial EEG during imagined hand movement.IEEE Trans. Rehabil. Eng.8, 441–446 (2000)

    2000

  63. [74]

    T.et al.Deep learning with convolu- tional neural networks for EEG decoding and visualiza- tion.Hum

    Schirrmeister, R. T.et al.Deep learning with convolu- tional neural networks for EEG decoding and visualiza- tion.Hum. Brain Mapp.38, 5391–5420 (2017)

    2017

  64. [75]

    & Karabatak, M

    Aksoy, G., Cattan, G., Chakraborty, S. & Karabatak, M. Quantum Machine-Based Decision Support System for the Detection of Schizophrenia from EEG Records.J. Med. Syst.48, (2024)

    2024

  65. [76]

    Fisher, R. A. Theory of Statistical Estimation.Math. Proc. Camb. Philos. Soc.22, 700–725 (1925)

    1925

  66. [77]

    Li, Z.et al.Machine learning-based classification of physiologicalandpathologicalhigh-frequencyoscillations recorded by stereoelectroencephalography.Seizure Eur. J. Epilepsy113, 58–65 (2023)

    2023

  67. [78]

    Seleznov, I.et al.Detrended Fluctuation, Coherence, and Spectral Power Analysis of Activation Rearrangement in EEG Dynamics During Cognitive Workload.Front. Hum. Neurosci.13, 270 (2019)

    2019

  68. [79]

    & Karabatak, M

    Aksoy, G., Cattan, G., Chakraborty, S. & Karabatak, M. Quantum Machine-Based Decision Support System for the Detection of Schizophrenia from EEG Records.J. Med. Syst.48, 29 (2024)

    2024

  69. [80]

    Saha, S.et al.Progress in Brain Computer Interface: Challenges and Opportunities.Front. Syst. Neurosci.15, 578875 (2021)

    2021

  70. [81]

    & Jun, S

    Ahn, M., Cho, H., Ahn, S. & Jun, S. C. High Theta and Low Alpha Powers May Be Indicative of BCI-Illiteracy in Motor Imagery.PLoS ONE8, e80886 (2013)

    2013

  71. [82]

    (Wiley, 2024)

    Vasques, X.Machine Learning Theory and Ap- plications: Hands-on Use Cases with Python on Classical and Quantum Machines. (Wiley, 2024). doi:10.1002/9781394220649

    work page doi:10.1002/9781394220649 2024

  72. [83]

    Grosse-Wentrup, M.&Buss, M.MulticlassCommonSpa- tial Patterns and Information Theoretic Feature Extrac- tion.IEEE Trans. Biomed. Eng.55, 1991–2000 (2008)

    1991

  73. [84]

    Ligthart, S.et al.Minding Rights: Mapping Ethical and Legal Foundations of ’Neurorights’.Camb. Q. Healthc. Ethics32, 461–481 (2023)

    2023

  74. [85]

    Maiseli, B.et al.Brain-computer interface: trend, chal- lenges, and threats.Brain Inform.10, 20 (2023)

    2023

  75. [86]

    & Rossi, F

    Berger, S. & Rossi, F. AI and Neurotechnology: Learn- ing from AI Ethics to Address an Expanded Ethics Land- scape.Commun. ACM66, 58–68 (2023). 22

    2023