pith. sign in

arxiv: 1907.08977 · v1 · pith:I6EL74UTnew · submitted 2019-07-21 · 💻 cs.HC · q-bio.NC

Systematic Enhancement of Functional Connectivity in Brain-Computer Interfacing using Common Spatial Patterns and Tangent Space Mapping

Saugat Bhattacharyya , Mitsuhiro Hayashibe This is my paper

Pith reviewed 2026-05-24 18:36 UTC · model grok-4.3

classification 💻 cs.HC q-bio.NC
keywords EEGfunctional connectivitybrain-computer interfacetrial selectiongraph parametersmental taskschannel selection
0
0 comments X

The pith

Filtering unreliable EEG trials via classification improves separability of functional connectivity graph parameters across mental tasks.

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

Low-quality EEG recordings, often caused by lapses in subject concentration, distort the measured interactions between brain regions during cognitive tasks. The paper inserts a classification stage that discards trials the classifier marks as mis-classified or low-probability before the connectivity graphs are built. After this filtering step the remaining trials produce graph parameters that separate more cleanly between different mental tasks. The same cleaned data also yields clearer connectivity maps when only a subset of electrodes is retained instead of using every channel. A reader would care because the method offers a concrete way to reduce the impact of noisy trials on brain-interaction analyses that otherwise rely on averaging across all recordings.

Core claim

By adding a classification step that removes mis-classified or low-probability EEG trials before computing functional connectivity, the separability among graph parameters for different mental tasks increases and the readability of the resulting connectivity maps improves when analysis is restricted to selected channels.

What carries the argument

Common spatial patterns combined with tangent-space mapping, applied to classify and retain only high-probability EEG trials before graph construction.

If this is right

  • Graph parameters for different mental tasks become more separable after unreliable trials are removed.
  • Connectivity maps become easier to interpret when analysis is limited to a selected subset of channels.
  • Functional connectivity better reflects genuine brain-region interactions once low-quality trials are excluded.

Where Pith is reading between the lines

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

  • The same filtering logic could be tested in online BCI pipelines that adaptively drop trials in real time.
  • If the improvement holds, the cleaned graphs might serve as more stable features for downstream task classification.
  • The approach could be checked on other signal modalities where trial quality fluctuates.

Load-bearing premise

Trials rejected by the classifier are exactly the low-quality recordings caused by poor concentration, and discarding them improves connectivity estimates without introducing selection bias.

What would settle it

Compute graph-parameter separability on the full set of trials and again after removing the classifier-rejected trials; if separability does not increase or if the removed trials show no correlation with independent quality markers such as reaction time, the central claim is falsified.

Figures

Figures reproduced from arXiv: 1907.08977 by Mitsuhiro Hayashibe, Saugat Bhattacharyya.

Figure 1
Figure 1. Figure 1: A simplified block diagram of the proposed method for dual purpose of error potential decoding and [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Electrode locations of the 56 channels in (a) Dataset I (b) Dataset II arranged in an extended 10-20 system. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Mean accuracy and standard deviation of the TSLR classifier for 26 subjects after 10-fold cross validation. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Mean accuracy and standard deviation of the TSLR classifier for 5 subjects after 10-fold cross validation. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Average covariances for the six spatially projected channels after CSP of 26 subjects of Dataset I, where (a) [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Average covariances for the six spatially projected channels after CSP of 5 subjects of Dataset II, where (a) [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The average clustering coefficient, local efficiency, participation coefficient and node strength of each [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: The average clustering coefficient, local efficiency, participation coefficient and node strength of each [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: An example of graph visualization for all channels and the selected relevant channels. The error trials from [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
read the original abstract

Functional connectivity of cognitive tasks allows researchers to analyse the interaction mapping occurring between different regions of the brain using electroencephalography (EEG) signals. Standard practice in functional connectivity involve studying the electrode pair interactions across several trials. As the cognitive task always involves the human factor, it is inevitable to have lower quality data from the brain signals influenced by the subject concentration or other mental states which can occur anytime over the whole experimental trials. The connectivity among electrodes are heavily influenced by these low quality EEG. In this paper, we aim at enhancing the functional connectivity of mental tasks by implementing a classification step in the process to remove those incorrect EEG trials from the available set. The classification step removes the trials which were mis-classified or had a low probability of occurrence to extract only reliable EEG trials. Through our approach, we have successfully improved the separability among graph parameters for different mental tasks. We also observe an improvement in the readability of the connectivity by focusing only on a group of selected channels rather than employing all the channels.

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 / 0 minor

Summary. The paper proposes enhancing functional connectivity analysis in EEG-based brain-computer interfacing by adding a classification step using Common Spatial Patterns (CSP) and Tangent Space Mapping. This step identifies and removes mis-classified trials or those with low probability, which are presumed to reflect poor subject concentration or low-quality data. The authors claim this yields improved separability among graph parameters for different mental tasks and greater readability of connectivity graphs by restricting analysis to a selected subset of channels rather than all electrodes.

Significance. If the filtering step can be shown to remove noise without introducing selection bias, the method could increase the reliability of graph-theoretic measures of functional connectivity in EEG studies. The approach builds on established techniques (CSP, tangent-space mapping) without introducing new free parameters or ad-hoc entities. However, the manuscript supplies no quantitative metrics, statistical tests, or baseline comparisons to support the claimed improvements in separability or readability.

major comments (3)
  1. [Abstract] Abstract: the central claim that the approach 'successfully improved the separability among graph parameters for different mental tasks' is unsupported by any numerical results, error bars, p-values, or before/after comparisons; without these, the magnitude and reliability of the reported enhancement cannot be assessed.
  2. [Abstract] Abstract (classification step): the premise that trials rejected by the CSP/tangent-space classifier are precisely those corrupted by poor concentration is not independently validated (e.g., via reaction time, eye-tracking, or subject reports). Because the same classifier supplies both the rejection rule and the subsequent 'improved separability' metric, any observed gain risks being an artifact of the selection procedure itself rather than removal of genuine noise.
  3. [Abstract] Abstract: the additional claim of 'improvement in the readability of the connectivity by focusing only on a group of selected channels' likewise lacks any quantitative measure of readability or comparison against full-channel graphs, rendering the statement untestable from the provided text.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below, indicating revisions where the manuscript can be strengthened without misrepresenting our original results.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that the approach 'successfully improved the separability among graph parameters for different mental tasks' is unsupported by any numerical results, error bars, p-values, or before/after comparisons; without these, the magnitude and reliability of the reported enhancement cannot be assessed.

    Authors: We agree that the abstract would benefit from explicit numerical support. The results section of the manuscript includes before-and-after visualizations of graph parameters and connectivity maps demonstrating the effect of the filtering step. To make the central claim more assessable, we will revise the abstract to include a concise summary of the observed changes in graph metrics (e.g., differences in clustering coefficients or path lengths across tasks). revision: yes

  2. Referee: [Abstract] Abstract (classification step): the premise that trials rejected by the CSP/tangent-space classifier are precisely those corrupted by poor concentration is not independently validated (e.g., via reaction time, eye-tracking, or subject reports). Because the same classifier supplies both the rejection rule and the subsequent 'improved separability' metric, any observed gain risks being an artifact of the selection procedure itself rather than removal of genuine noise.

    Authors: The CSP and tangent-space classifier is applied using established BCI methods to retain only trials with high task alignment and probability; this is presented as a data-quality filter rather than a direct measure of concentration. We acknowledge the absence of independent validation (e.g., reaction-time correlates) and the potential for circular evaluation. We will add an explicit limitations paragraph discussing this point and the rationale drawn from prior BCI literature. revision: partial

  3. Referee: [Abstract] Abstract: the additional claim of 'improvement in the readability of the connectivity by focusing only on a group of selected channels' likewise lacks any quantitative measure of readability or comparison against full-channel graphs, rendering the statement untestable from the provided text.

    Authors: We agree that the readability claim would be stronger with a quantitative anchor. The manuscript shows connectivity graphs restricted to selected channels versus the full set. We will revise the abstract and results to include a direct comparison, such as the reduction in edge count or a sparsity metric between the two graph representations. revision: yes

Circularity Check

0 steps flagged

No circularity: standard external methods applied without self-referential reduction or fitted-input prediction.

full rationale

The paper applies established CSP and tangent-space mapping classifiers (external to the connectivity analysis) to reject trials, then computes graph parameters on the retained set. No equations define connectivity measures in terms of the classifier outputs or vice versa; the separability improvement is reported as an empirical observation rather than a constructed identity. No self-citations are invoked as load-bearing uniqueness theorems, and the approach does not rename known results or smuggle ansatzes. The chain remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim depends on the domain assumption that classifier probability reliably flags low-quality trials and that their removal improves connectivity without bias; no free parameters or invented entities are explicitly introduced in the abstract.

axioms (1)
  • domain assumption Mis-classified or low-probability trials correspond exactly to low-quality EEG data caused by subject concentration or mental state fluctuations.
    Invoked when the paper states that the classification step removes incorrect EEG trials to extract only reliable ones.

pith-pipeline@v0.9.0 · 5716 in / 1193 out tokens · 40783 ms · 2026-05-24T18:36:58.481216+00:00 · methodology

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

40 extracted references · 40 canonical work pages · 1 internal anchor

  1. [1]

    Nirs-eeg joint imaging during transcranial direct current stimulation: Online parameter estimation with an autoregressive model

    Mehak Sood, Pierre Besson, Makii Muthalib, Utkarsh Jindal, Stephane Perrey, Anirban Dutta, and Mitsuhiro Hayashibe. Nirs-eeg joint imaging during transcranial direct current stimulation: Online parameter estimation with an autoregressive model. Journal of Neuroscience Methods, 274:71 – 80, 2016

    work page 2016

  2. [2]

    A. Roy, B. Baxter, and B. He. High-definition transcranial direct current stimulation induces both acute and persistent changes in broadband cortical synchronization: A simultaneous tdcs-eeg study. IEEE Transactions on Biomedical Engineering, 61(7):1967–1978, July 2014

    work page 1967

  3. [3]

    Baxter, Bradley J

    Bryan S. Baxter, Bradley J. Edelman, Nicholas Nesbitt, and Bin He. Sensorimotor rhythm {BCI} with simultaneous high definition-transcranial direct current stimulation alters task performance. Brain Stimulation, 9(6):834 – 841, 2016

    work page 2016

  4. [4]

    National academy of sciences of the united states of america

    G Tononi, O Sporns, and G M Edelman. National academy of sciences of the united states of america. volume 91, pages 5033–5037, 1994

    work page 1994

  5. [5]

    Measuring information integration

    Giulio Tononi and Olaf Sporns. Measuring information integration. BMC Neuroscience, 4(1):1–20, 2003

    work page 2003

  6. [6]

    Mahyari and S

    A.G. Mahyari and S. Aviyente. Two-dimensional svd for event detection in dynamic functional brain networks. In Global Conference on Signal and Information Processing (GlobalSIP), 2013 IEEE , pages 37–40, Dec 2013

    work page 2013

  7. [7]

    Stam and J

    C. Stam and J. Reijneveld. Graph theoretical analysis of complex networks in the brain. Nonlinear biomedical physics, 1(1), 2007

    work page 2007

  8. [8]

    Marrelec, J

    G. Marrelec, J. Daunizeau, M. Pelegrini-Issac, J. Doyon, and H. Benali. Conditional correlation as a measure of mediated interactivity in fmri and meg/eeg. Signal Processing, IEEE Transactions on, 53(9):3503–3516, Sept 2005

    work page 2005

  9. [9]

    La Rocca, P

    D. La Rocca, P. Campisi, B. Vegso, P. Cserti, G. Kozmann, F. Babiloni, and F. De Vico Fallani. Human brain distinctiveness based on eeg spectral coherence connectivity. Biomedical Engineering, IEEE Transactions on , 61(9):2406–2412, Sept 2014

    work page 2014

  10. [10]

    Zavaglia, L

    M. Zavaglia, L. Astolfi, F. Babiloni, and M. Ursino. The effect of connectivity on eeg rhythms, power spectral density and coherence among coupled neural populations: Analysis with a neural mass model. Biomedical Engineering, IEEE Transactions on, 55(1):69–77, Jan 2008

    work page 2008

  11. [11]

    A graph- theoretical approach in brain functional networks

    Fabrizio Fallani, Luciano Costa, Francisco Rodriguez, Laura Astolfi, Giovanni Vecchiato, Jlenia Toppi, Gianluca Borghini, Febo Cincotti, Donatella Mattia, Serenella Salinari, Roberto Isabella, and Fabio Babiloni. A graph- theoretical approach in brain functional networks. possible implications in eeg studies. Nonlinear Biomedical Physics, 4(Suppl 1):S8, 2010

    work page 2010

  12. [12]

    Ghosh, A

    P. Ghosh, A. Mazumder, S. Bhattacharyya, D.N. Tibarewala, and M. Hayashibe. Functional connectivity analysis of motor imagery eeg signal for brain-computer interfacing application. In Neural Engineering (NER), 2015 7th International IEEE/EMBS Conference on, pages 210–213, April 2015. 10 A PREPRINT - JULY 23, 2019

    work page 2015

  13. [13]

    Bolanos, E.M

    M.E. Bolanos, E.M. Bernat, and S. Aviyente. Graph analysis of neuronal interactions for the error-related negativity. In Engineering in Medicine and Biology Society (EMBC), 2010 Annual International Conference of the IEEE, pages 2794–2797, Aug 2010

    work page 2010

  14. [14]

    Bolaños, E.M

    M.E. Bolaños, E.M. Bernat, and S. Aviyente. Identification of small world topologies in neural functional connections quantified by phase synchrony measures. In Engineering in Medicine and Biology Society, 2009. EMBC 2009. Annual International Conference of the IEEE , pages 5308–5311, Sept 2009

    work page 2009

  15. [15]

    Bernat, and Selin Aviyente

    Ali Yener Mutlu, Edward M. Bernat, and Selin Aviyente. A signal-processing-based approach to time-varying graph analysis for dynamic brain network identification. Comp. Math. Methods in Medicine , 2012:451516:1– 451516:10, 2012

    work page 2012

  16. [16]

    Complex brain networks: graph theoretical analysis of structural and functional systems

    Ed Bullmore and Olaf Sporns. Complex brain networks: graph theoretical analysis of structural and functional systems. Nature reviews. Neuroscience, 10(3):186–198, March 2009

    work page 2009

  17. [17]

    Chavarriaga, M.K

    Huaijian Zhang, R. Chavarriaga, M.K. Goel, L. Gheorghe, and J. Del R Millan. Improved recognition of error related potentials through the use of brain connectivity features. In Engineering in Medicine and Biology Society (EMBC), 2012 Annual International Conference of the IEEE , pages 6740–6743, Aug 2012

    work page 2012

  18. [18]

    Pfurtscheller, Ch

    G. Pfurtscheller, Ch. Neuper, D. Flotzinger, and M. Pregenzer. Eeg-based discrimination between imagination of right and left hand movement. Electroencephalography and Clinical Neurophysiology, 103(6):642 – 651, 1997

    work page 1997

  19. [19]

    R. Leeb, F. Lee, C. Keinrath, R. Scherer, H. Bischof, and G. Pfurtscheller. Brain-computer communication: Motivation, aim, and impact of exploring a virtual apartment. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 15(4):473–482, Dec 2007

    work page 2007

  20. [20]

    Common spatial pattern method for channel selelction in motor imagery based brain-computer interface

    Yijun Wang, Shangkai Gao, and Xiaornog Gao. Common spatial pattern method for channel selelction in motor imagery based brain-computer interface. In 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference, pages 5392–5395, Jan 2005

    work page 2005

  21. [21]

    Arvaneh, C

    M. Arvaneh, C. Guan, K. K. Ang, and C. Quek. Optimizing the channel selection and classification accuracy in eeg-based bci. IEEE Transactions on Biomedical Engineering , 58(6):1865–1873, June 2011

    work page 2011

  22. [22]

    Modulation of activity in medial frontal and motor cortices during error observation

    Hein T van Schie, Rogier B Mars, Michael G H Coles, and Harold Bekkering. Modulation of activity in medial frontal and motor cortices during error observation. Nat Neurosci., 7(5):549 – 554, 2004

    work page 2004

  23. [23]

    Errare machinale est: The use of error-related potentials in brain-machine interfaces

    Ricardo Chavarriaga, Aleksander Sobolewski, and José del R Millan. Errare machinale est: The use of error-related potentials in brain-machine interfaces. Frontiers in Neuroscience, 8(208), 2014

    work page 2014

  24. [24]

    Farquhar

    J. Farquhar. A linear feature space for simultaneous learning of spatio-spectral filters in {BCI}. Neural Networks, 22(9):1278 – 1285, 2009. Brain-Machine Interface

    work page 2009

  25. [25]

    A Plug&Play P300 BCI Using Information Geometry

    Alexandre Barachant and Marco Congedo. A plug&play P300 BCI using information geometry. CoRR, abs/1409.0107, 2014

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  26. [26]

    Bci challenge @ ner 2015, 2015

    work page 2015

  27. [27]

    Perrin, E

    M. Perrin, E. Maby, S. Daligault, O. Bertrand, and J. Mattout. Objective and subjective evaluation of online error correction during p300-based spelling. Advances in Human-Computer Interaction , 2012(578295):13, 2012

    work page 2012

  28. [28]

    Dornhege

    G. Dornhege. Toward Brain-computer Interfacing. A Bradford book. MIT Press, 2007

    work page 2007

  29. [29]

    Combaz et al

    A. Combaz et al. Towards the detection of error-related potentials and its integration in the context of a p300 speller brain–computer interface. Neurocomputing, 80(0):73 – 82, 2012. Special Issue on Machine Learning for Signal Processing 2010

    work page 2012

  30. [30]

    P. W. Ferrez and Jose del R. Millan. Simultaneous real-time detection of motor imagery and error-related potentials for improved bci accuracy. In Proceedings of the 4th International Brain-Computer Interface Workshop and Training Course, pages 197–202, 2008

    work page 2008

  31. [31]

    Oppenheim, Ronald W

    Alan V . Oppenheim, Ronald W. Schafer, and John R. Buck.Discrete-time Signal Processing. Prentice-Hall, Inc., Upper Saddle River, NJ, USA, 1999

    work page 1999

  32. [32]

    Bhattacharyya, S

    S. Bhattacharyya, S. Shimoda, and M. Hayashibe. A synergetic brain-machine interfacing paradigm for multi-dof robot control. IEEE Transactions on Systems, Man, and Cybernetics: Systems , 46(7):957–968, July 2016

    work page 2016

  33. [33]

    Ramoser, J

    H. Ramoser, J. Muller-Gerking, and G. Pfurtscheller. Optimal spatial filtering of single trial eeg during imagined hand movement. IEEE Transactions on Rehabilitation Engineering , 8(4):441–446, Dec 2000

    work page 2000

  34. [34]

    Jayasumana, R

    S. Jayasumana, R. Hartley, M. Salzmann, Hongdong Li, and M. Harandi. Kernel methods on the riemannian manifold of symmetric positive definite matrices. In Computer Vision and Pattern Recognition (CVPR), 2013 IEEE Conference on, pages 73–80, June 2013. 11 A PREPRINT - JULY 23, 2019

    work page 2013

  35. [35]

    F. Yger. A review of kernels on covariance matrices for bci applications. In Machine Learning for Signal Processing (MLSP), 2013 IEEE International Workshop on , pages 1–6, Sept 2013

    work page 2013

  36. [36]

    Averaging Covariance Matrices for EEG Signal Classification based on the CSP: an Empirical Study

    Florian Yger, Fabien Lotte, and Masashi Sugiyama. Averaging Covariance Matrices for EEG Signal Classification based on the CSP: an Empirical Study. In EUSIPCO 2015, Nice, France, August 2015

    work page 2015

  37. [37]

    Classification of covariance matrices using a riemannian-based kernel for {BCI} applications

    Alexandre Barachant, Stéphane Bonnet, Marco Congedo, and Christian Jutten. Classification of covariance matrices using a riemannian-based kernel for {BCI} applications. Neurocomputing, 112:172 – 178, 2013. Advances in artificial neural networks, machine learning, and computational intelligenceSelected papers from the 20th European Symposium on Artificial Neu...

    work page 2013

  38. [38]

    Geometric means in a novel vector space structure on symmetric positive-definite matrices

    Vincent Arsigny, Pierre Fillard, Xavier Pennec, and Nicholas Ayache. Geometric means in a novel vector space structure on symmetric positive-definite matrices. SIAM Journal on Matrix Analysis and Applications , 29(1):328–347, 2007

    work page 2007

  39. [39]

    Introduction to Machine Learning (Adaptive Computation and Machine Learning)

    Ethem Alpaydin. Introduction to Machine Learning (Adaptive Computation and Machine Learning) . The MIT Press, 2004

    work page 2004

  40. [40]

    The Elements of Statistical Learning

    Trevor Hastie, Robert Tibshirani, and Jerome Friedman. The Elements of Statistical Learning . Springer Series in Statistics. Springer New York Inc., New York, NY , USA, 2001. 12

    work page 2001