pith. the verified trust layer for science. sign in

arxiv: 2605.10111 · v1 · submitted 2026-05-11 · 💻 cs.LG · cs.AI· cs.CV

CFSPMNet: Cross-subject Fourier-guided Spatial-Patch Mamba Network for EEG Motor Imagery Decoding in Stroke Patients

Xiangkai Wang , Yun Zhao , Dongyi He , Qingling Xia , Gen Li , Xinlai Xing , Yuchi Pan , Bin Jiang This is my paper

Pith reviewed 2026-05-12 03:20 UTC · model grok-4.3

classification 💻 cs.LG cs.AIcs.CV
keywords motor imagery EEGstroke rehabilitationcross-subject decodingMamba networkFourier domainbrain-computer interfaceneural state modelingpseudo-label adaptation
0
0 comments X p. Extension

The pith

A network combining Fourier-reorganized Mamba states with shared-private prototype matching improves cross-patient motor imagery EEG decoding in stroke survivors.

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

The paper tries to establish that post-stroke changes in brain dynamics can be modeled as latent neural-state organization to enable reliable MI-EEG decoding across individuals. It proposes CFSPMNet, which reorganizes trial token states in the Fourier domain to guide Mamba propagation and applies shared-private consistency checks to refine pseudo-labels on target patients. A sympathetic reader would care because this could support brain-computer interfaces for rehabilitation that work without collecting large amounts of labeled data from each new patient. Leave-one-subject-out tests on two stroke datasets show gains of 5.63 and 8.25 percentage points over strong baselines, with ablations confirming the contributions of the Fourier and consistency modules.

Core claim

CFSPMNet models post-stroke MI-EEG as latent neural-state organization by combining a Fourier-Reorganized State Mamba Network (FRSM) that represents each trial as a latent physiological token sequence, reorganizes token states in the Fourier domain, and uses Fourier-derived trial context to guide Mamba state-space propagation, together with Shared-Private Prototype Matching (SPPM) that improves target pseudo-label updating by combining semantic confidence with shared-private physiological consistency to filter inconsistent predictions.

What carries the argument

Fourier-Reorganized State Mamba Network (FRSM) combined with Shared-Private Prototype Matching (SPPM), where FRSM handles Fourier-domain token-state reorganization to guide propagation and SPPM enforces physiological consistency for pseudo-label selection.

If this is right

  • The method achieves average accuracies of 68.23 percent on XW-Stroke and 73.33 percent on 2019-Stroke in leave-one-subject-out settings.
  • Gains of 5.63 and 8.25 percentage points over the strongest CNN, Transformer, Mamba, and adaptation baselines.
  • Ablation and sensitivity analyses confirm the contribution of Fourier-domain reorganization and calibrated pseudo-label updating.
  • Neurophysiological visualizations align with the roles of the Fourier and shared-private components in capturing trial-level brain-state context.

Where Pith is reading between the lines

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

  • If the Fourier guidance truly tracks stroke-induced reorganization of neural states, the same token-reorganization step could be tested on other conditions that alter EEG aperiodic activity such as epilepsy or traumatic brain injury.
  • The shared-private consistency filter might reduce the volume of labeled target data needed in clinical BCI pipelines, which could be checked by measuring performance as the number of available target trials is progressively reduced.
  • Extending the prototype-matching step to multi-modal inputs such as EEG plus EMG could produce more robust rehabilitation interfaces when motor imagery signals are weak.
  • Real-time deployment tests would reveal whether the offline accuracy improvements translate to usable control latency and stability in actual BCI loops.

Load-bearing premise

The Fourier-reorganized token states and shared-private physiological consistency checks reliably capture the latent neural-state organization that differs across stroke patients, rather than fitting dataset-specific artifacts.

What would settle it

Apply the full model and its ablated versions (without Fourier reorganization or without consistency filtering) to a fresh stroke MI-EEG dataset collected with different recording hardware or patient demographics and check whether the reported accuracy gains over baselines disappear.

Figures

Figures reproduced from arXiv: 2605.10111 by Bin Jiang, Dongyi He, Gen Li, Qingling Xia, Xiangkai Wang, Xinlai Xing, Yuchi Pan, Yun Zhao.

Figure 1
Figure 1. Figure 1: Overall framework of CFSPMNet for cross-patient post-stroke MI-EEG decoding. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Architecture of the Fourier-Reorganized State Mamba Network. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Shared-Private Prototype Matching for calibrated target pseudo-label updating. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Hyperparameter sensitivity on XW-Stroke and 2019-Stroke. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: t-SNE feature distributions on XW-Stroke and 2019-Stroke. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Fourier-domain token-state reorganization on XW-Stroke and 2019-Stroke. [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Calibrated target pseudo-label selection on XW-Stroke and 2019-Stroke. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Spatial neurophysiological interpretability on XW-Stroke and 2019-Stroke. [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
read the original abstract

Motor imagery electroencephalography (MI-EEG) decoding offers a non-invasive route for post-stroke rehabilitation, but cross-patient use remains difficult because pathological neural reorganization changes task-related EEG dynamics, aperiodic activity, local excitability, cross-regional coordination, and trial-level brain-state context. This makes source-learned MI representations unreliable for unseen patients. To address this problem, we propose CFSPMNet, a cross-patient adaptation framework that models post-stroke MI-EEG as latent neural-state organization. CFSPMNet combines a Fourier-Reorganized State Mamba Network (FRSM) with Shared-Private Prototype Matching (SPPM). FRSM represents each trial as a latent physiological token sequence, reorganizes token states in the Fourier domain, and uses Fourier-derived trial context to guide Mamba state-space propagation. SPPM improves target pseudo-label updating by combining semantic confidence with shared-private physiological consistency, filtering confident but physiologically inconsistent target predictions. Leave-one-subject-out experiments on two stroke MI-EEG datasets show that CFSPMNet outperforms representative CNN-, Transformer-, Mamba-, and adaptation-based baselines, achieving average accuracies of 68.23% on XW-Stroke and 73.33% on 2019-Stroke, with gains of 5.63 and 8.25 percentage points over the strongest competitors. Ablation, sensitivity, feature-alignment, pseudo-label selection, and neurophysiological visualization analyses further support the roles of Fourier-domain token-state reorganization and calibrated pseudo-label updating. These results suggest that latent neural-state modeling can improve rehabilitation-oriented cross-patient BCI decoding. Code is available at https://github.com/wxk1224/CFSPMNet.

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

2 major / 3 minor

Summary. The manuscript introduces CFSPMNet, a cross-subject adaptation framework for motor imagery EEG decoding in stroke patients. It combines a Fourier-Reorganized State Mamba Network (FRSM) that represents trials as latent physiological token sequences, reorganizes them in the Fourier domain, and uses Fourier-derived context to guide Mamba propagation, with Shared-Private Prototype Matching (SPPM) that filters target pseudo-labels using semantic confidence plus shared-private physiological consistency. Leave-one-subject-out experiments on the XW-Stroke and 2019-Stroke datasets report average accuracies of 68.23% and 73.33%, outperforming representative CNN, Transformer, Mamba, and adaptation baselines by 5.63 and 8.25 percentage points. Supporting analyses include ablations, sensitivity tests, feature alignment, pseudo-label selection, and neurophysiological visualizations. Code is released at https://github.com/wxk1224/CFSPMNet.

Significance. If the results hold under stronger validation, the work could meaningfully advance cross-patient BCI systems for post-stroke rehabilitation by explicitly modeling pathological neural reorganization, aperiodic activity, and cross-regional coordination through latent state organization. Credit is due for releasing reproducible code, evaluating on two real stroke MI-EEG collections, and providing multiple supporting analyses (ablations, visualizations) beyond headline accuracy numbers. The empirical focus on held-out subjects aligns with practical BCI needs.

major comments (2)
  1. [Experiments / Results tables] Experiments section (results and tables reporting LOSO accuracies): The performance gains over baselines are presented as raw percentage-point improvements without statistical significance testing (e.g., subject-wise paired t-tests, Wilcoxon signed-rank, or permutation tests for the 5.63 pp and 8.25 pp margins). With the small subject counts typical of stroke EEG datasets and known inter-trial variability, this is required to substantiate the central claim of reliable outperformance.
  2. [Ablations and visualizations] Ablation and neurophysiological visualization sections (e.g., §5.3 and associated figures): The visualizations and ablations are offered as evidence that FRSM's Fourier token reorganization and SPPM's consistency checks capture latent physiological organization. However, the manuscript does not report control experiments such as accuracy under phase-scrambled surrogates, Fourier-domain permutation of tokens, or trial-shuffled baselines. Without these, it remains possible that gains arise from dataset-specific artifacts rather than cross-patient neural-state differences, directly affecting the interpretive claim in the abstract and conclusion.
minor comments (3)
  1. [Abstract] Abstract and methods: The strongest baseline for each dataset is not named explicitly when stating the 5.63 pp and 8.25 pp gains; adding the model names (e.g., 'over Mamba-based baseline X') would improve clarity.
  2. [Methods / FRSM] Methods: The precise formulation of the Fourier-derived trial context vector and its injection into the Mamba state update (likely around the FRSM equations) would benefit from an accompanying diagram or expanded pseudocode for readers unfamiliar with state-space models.
  3. [Datasets and preprocessing] Dataset description: Trial rejection criteria, class imbalance handling, and exact preprocessing steps (filtering, artifact removal) are referenced but not tabulated; a supplementary table summarizing these per dataset would aid reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed review. The comments highlight important aspects of validation that we will address in the revision. Below we respond point by point to the major comments.

read point-by-point responses

  1. Referee: Experiments section (results and tables reporting LOSO accuracies): The performance gains over baselines are presented as raw percentage-point improvements without statistical significance testing (e.g., subject-wise paired t-tests, Wilcoxon signed-rank, or permutation tests for the 5.63 pp and 8.25 pp margins). With the small subject counts typical of stroke EEG datasets and known inter-trial variability, this is required to substantiate the central claim of reliable outperformance.

    Authors: We agree that statistical significance testing is necessary to substantiate the reported gains, particularly given the small subject numbers and inter-subject variability in stroke EEG data. In the revised manuscript we will add subject-wise paired t-tests and Wilcoxon signed-rank tests comparing CFSPMNet against each baseline, together with the corresponding p-values, to quantify the reliability of the 5.63 pp and 8.25 pp improvements. revision: yes

  2. Referee: Ablation and neurophysiological visualization sections (e.g., §5.3 and associated figures): The visualizations and ablations are offered as evidence that FRSM's Fourier token reorganization and SPPM's consistency checks capture latent physiological organization. However, the manuscript does not report control experiments such as accuracy under phase-scrambled surrogates, Fourier-domain permutation of tokens, or trial-shuffled baselines. Without these, it remains possible that gains arise from dataset-specific artifacts rather than cross-patient neural-state differences, directly affecting the interpretive claim in the abstract and conclusion.

    Authors: We acknowledge that additional controls would further strengthen the claim that performance gains arise from modeling latent neural-state organization rather than dataset artifacts. Our existing ablations already isolate the contribution of Fourier-domain token reorganization and shared-private prototype matching, and the neurophysiological visualizations provide supporting evidence of physiologically plausible patterns. To directly address the concern, we will add control experiments using phase-scrambled surrogates and trial-shuffled baselines in the revised manuscript and report the resulting accuracies. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical performance on held-out subjects is independent of model internals

full rationale

The paper proposes CFSPMNet (FRSM + SPPM) as an architecture for cross-subject MI-EEG decoding and reports leave-one-subject-out accuracies (68.23% / 73.33%) plus gains over baselines. These are direct empirical measurements on unseen subjects; no equations, fitted parameters, or self-citations are shown that would make the reported numbers equivalent to the model's own inputs by construction. Ablations and visualizations are presented as supporting evidence but do not reduce the central performance claims to tautologies. The derivation chain is therefore self-contained as standard empirical ML evaluation.

Axiom & Free-Parameter Ledger

1 free parameters · 0 axioms · 0 invented entities

The method rests on standard deep-learning training assumptions plus the domain claim that Fourier context and prototype consistency reflect physiologically meaningful cross-patient invariants; no new physical entities are postulated.

free parameters (1)
  • model hyperparameters (learning rate, Mamba state dimension, prototype temperature, etc.)
    Chosen or tuned during training; typical for neural-network papers and required for the reported accuracies.

pith-pipeline@v0.9.0 · 5637 in / 1240 out tokens · 53619 ms · 2026-05-12T03:20:03.725083+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

31 extracted references · 31 canonical work pages

  1. [1]

    Eeg-based predictors of motor recovery during immersive vr-bci rehabilitation.Scientific Reports, 16(1):7870, Feb 2026

    Madalena Valente, Diogo Branco, Sergi Bermúdez i Badia, Jean-Claude Fernandes, Patrícia Figueiredo, and Athanasios V ourvopoulos. Eeg-based predictors of motor recovery during immersive vr-bci rehabilitation.Scientific Reports, 16(1):7870, Feb 2026

    work page 2026

  2. [2]

    Attention-adaptive bci-aot system enhances motor recovery and neural engagement after stroke.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 34:686–698, 2026

    Hyunmi Lim, Hyoseon Choi, Bilal Ahmed, Yoonghil Park, and Jeonghun Ku. Attention-adaptive bci-aot system enhances motor recovery and neural engagement after stroke.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 34:686–698, 2026

    work page 2026

  3. [3]

    A comprehensive review of deep learning for motor imagery eeg: From healthy subjects to patients

    Bin Jiang, Xiangkai Wang, Dongyi He, Siyu Cheng, Maoyu Liao, Duoqian Miao, Qingling Xia, Yun Zhao, and Gen Li. A comprehensive review of deep learning for motor imagery eeg: From healthy subjects to patients. Biomedical Signal Processing and Control, 117:109680, 2026

    work page 2026

  4. [4]

    Eegnet: a compact convolutional neural network for eeg-based brain–computer interfaces.Journal of Neural Engineering, 15(5):056013, jul 2018

    Vernon J Lawhern, Amelia J Solon, Nicholas R Waytowich, Stephen M Gordon, Chou P Hung, and Brent J Lance. Eegnet: a compact convolutional neural network for eeg-based brain–computer interfaces.Journal of Neural Engineering, 15(5):056013, jul 2018

    work page 2018

  5. [5]

    Jiaheng Wang, Lin Yao, and Yueming Wang. Ifnet: An interactive frequency convolutional neural network for enhancing motor imagery decoding from eeg.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 31:1900–1911, 2023

    work page 1900

  6. [6]

    Deep learning with convolutional neural networks for eeg decoding and visualization.Human Brain Mapping, 38(11):5391–5420, 2017

    Robin Tibor Schirrmeister, Jost Tobias Springenberg, Lukas Dominique Josef Fiederer, Martin Glasstetter, Katharina Eggensperger, Michael Tangermann, Frank Hutter, Wolfram Burgard, and Tonio Ball. Deep learning with convolutional neural networks for eeg decoding and visualization.Human Brain Mapping, 38(11):5391–5420, 2017

    work page 2017

  7. [7]

    Dbconformer: Dual-branch convolutional transformer for eeg decoding.IEEE Journal of Biomedical and Health Informatics, pages 1–14, 2025

    Ziwei Wang, Hongbin Wang, Tianwang Jia, Xingyi He, Siyang Li, and Dongrui Wu. Dbconformer: Dual-branch convolutional transformer for eeg decoding.IEEE Journal of Biomedical and Health Informatics, pages 1–14, 2025

    work page 2025

  8. [8]

    Eeg conformer: Convolutional transformer for eeg decoding and visualization.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 31:710–719, 2023

    Yonghao Song, Qingqing Zheng, Bingchuan Liu, and Xiaorong Gao. Eeg conformer: Convolutional transformer for eeg decoding and visualization.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 31:710–719, 2023

    work page 2023

  9. [9]

    Multi-scale convolutional transformer network for motor imagery brain-computer interface.Scientific Reports, 15(1):12935, Apr 2025

    Wei Zhao, Baocan Zhang, Haifeng Zhou, Dezhi Wei, Chenxi Huang, and Quan Lan. Multi-scale convolutional transformer network for motor imagery brain-computer interface.Scientific Reports, 15(1):12935, Apr 2025

    work page 2025

  10. [10]

    Slimseiz: Efficient channel-adaptive seizure prediction using a mamba-enhanced network

    Guorui Lu, Jing Peng, Bingyuan Huang, Chang Gao, Todor Stefanov, Yong Hao, and Qinyu Chen. Slimseiz: Efficient channel-adaptive seizure prediction using a mamba-enhanced network. In2025 IEEE International Symposium on Circuits and Systems (ISCAS), pages 1–5, 2025

    work page 2025

  11. [11]

    Unsupervised domain adaptation with synchronized self-training for cross- domain motor imagery recognition

    Peiyin Chen, Xiaofeng Liu, Chao Ma, He Wang, Xiong Yang, Celso Grebogi, Xiao Gu, and Zhongke Gao. Unsupervised domain adaptation with synchronized self-training for cross- domain motor imagery recognition. IEEE Journal of Biomedical and Health Informatics, 29(5):3664–3677, 2025

    work page 2025

  12. [12]

    Ua-daan: An uncertainty- aware dynamic adversarial adaptation network for eeg-based depression recognition.IEEE Transactions on Affective Computing, 16(3):2130–2141, 2025

    Jian Shen, Lechun You, Yu Ma, Zeguang Zhao, Huajian Liang, Yanan Zhang, and Bin Hu. Ua-daan: An uncertainty- aware dynamic adversarial adaptation network for eeg-based depression recognition.IEEE Transactions on Affective Computing, 16(3):2130–2141, 2025

    work page 2025

  13. [13]

    Ssas: Cross-subject eeg-based emotion recognition through source selection with adversarial strategy.Expert Systems with Applications, 308:130843, 2026

    Yici Liu, Qi Wei Oung, and Hoi Leong Lee. Ssas: Cross-subject eeg-based emotion recognition through source selection with adversarial strategy.Expert Systems with Applications, 308:130843, 2026

    work page 2026

  14. [14]

    A neurophysiological basis for aperiodic eeg and the background spectral trend.Nature Communications, 15(1):1514, Feb 2024

    Niklas Brake, Flavie Duc, Alexander Rokos, Francis Arseneau, Shiva Shahiri, Anmar Khadra, and Gilles Plourde. A neurophysiological basis for aperiodic eeg and the background spectral trend.Nature Communications, 15(1):1514, Feb 2024. 15 arXiv preprint

    work page 2024

  15. [15]

    McDonnell, Akash Umakantha, Ryan C

    Megan E. McDonnell, Akash Umakantha, Ryan C. Williamson, Matthew A. Smith, and Byron M. Yu. Interactions across hemispheres in prefrontal cortex reflect global cognitive processing.Nature Communications, Apr 2026

    work page 2026

  16. [16]

    Arseneau, Shiva Shahiri, A

    Niklas Brake, Flavie Duc, Alexander Rokos, F. Arseneau, Shiva Shahiri, A. Khadra, and G. Plourde. A neurophys- iological basis for aperiodic eeg and the background spectral trend.Nature Communications, 15, 2024

    work page 2024

  17. [17]

    Cross, Samantha M

    Zachariah R. Cross, Samantha M. Gray, Adam J. O. Dede, Yessenia M. Rivera, Q. Yin, Parisa Vahidi, E. M. Rau, Christopher Cyr, A. M. Holubecki, E. Asano, Jack J. Lin, Olivia Kim McManus, Shifteh Sattar, Ignacio Saez, Fady Girgis, David King-Stephens, Peter B. Weber, Stephan U Schuele, Joshua Rosenow, Joyce Y . Wu, Sandi K. Lam, Jeffrey S. Raskin, E. Chang,...

    work page 2025

  18. [18]

    Enhancing real-time cross-subject eeg decoding via pseudo-label tuning and adaptation.Applied Soft Computing, 197:115150, 2026

    Zhanhui Lin, Yonglin Wu, Chenyun Dai, and Yao Guo. Enhancing real-time cross-subject eeg decoding via pseudo-label tuning and adaptation.Applied Soft Computing, 197:115150, 2026

    work page 2026

  19. [19]

    An eeg motor imagery dataset for brain computer interface in acute stroke patients.Scientific Data, 11(1):131, Jan 2024

    Haijie Liu, Penghu Wei, Haochong Wang, Xiaodong Lv, Wei Duan, Meijie Li, Yan Zhao, Qingmei Wang, Xinyuan Chen, Gaige Shi, Bo Han, and Junwei Hao. An eeg motor imagery dataset for brain computer interface in acute stroke patients.Scientific Data, 11(1):131, Jan 2024

    work page 2024

  20. [20]

    Eeg data of motor imagery for stroke, 2 2019

    Tianyu Jia. Eeg data of motor imagery for stroke, 2 2019

    work page 2019

  21. [21]

    Visualizing data using t-sne.Journal of Machine Learning Research, 9(86):2579–2605, 2008

    Laurens van der Maaten and Geoffrey Hinton. Visualizing data using t-sne.Journal of Machine Learning Research, 9(86):2579–2605, 2008

    work page 2008

  22. [22]

    Albertson, Eric C

    Asher J. Albertson, Eric C. Landsness, Margaret Eisfelder, Brittany M. Young, Bradley Judge, Matthew R. Brier, Matthew J. Euler, Steven C. Cramer, Jin-Moo Lee, and Keith R. Lohse. Stroke and motor outcomes are associated with regional and age-specific changes in periodic and aperiodic cortical activity.Experimental Physiology, 111(3):1210–1227, 2026

    work page 2026

  23. [23]

    Monchy, J

    N. Monchy, J. Duprez, J-F. Houvenaghel, A. Legros, B. V oytek, and J. Modolo. Functional connectivity is dominated by aperiodic, rather than oscillatory, coupling.Journal of Neuroscience, 45(45), 2025

    work page 2025

  24. [24]

    Kaichao Wu, Beth Jelfs, Qiang Fang, and Leonardo L. Gollo. Criticality and increased intrinsic neural timescales in stroke.npj Systems Biology and Applications, 12(1):4, Dec 2025

    work page 2025

  25. [25]

    Joint spatial feature adaption and confident pseudo-label selection for cross-subject motor imagery eeg signals classification.Expert Systems with Applications, 278:127312, 2025

    Siqi Yang, Zhihua Huang, and Tian jian Luo. Joint spatial feature adaption and confident pseudo-label selection for cross-subject motor imagery eeg signals classification.Expert Systems with Applications, 278:127312, 2025

    work page 2025

  26. [26]

    Yong Peng, Jiangchuan Liu, Honggang Liu, Natasha Padfield, Junhua Li, Wanzeng Kong, Bao-Liang Lu, and Andrzej Cichocki. Prediction consistency and confidence-based proxy domain construction for privacy-preserving in cross-subject eeg classification.IEEE Journal of Biomedical and Health Informatics, 30(2):1115–1127, 2026

    work page 2026

  27. [27]

    Ga-Young Choi, Jeong-Kweon Seo, Kyoung Tae Kim, Won Kee Chang, Sung Whan Yoon, Nam-Jong Paik, Won-Seok Kim, and Han-Jeong Hwang. Clinical feasibility of motor hotspot localization based on electroen- cephalography using convolutional neural networks in stroke.Journal of NeuroEngineering and Rehabilitation, 22(1):193, Sep 2025

    work page 2025

  28. [28]

    Task-state eeg reveals fine hand motor impairment in acute ischemic stroke: A multidimensional analysis of cortical dynamics

    Ke Deng, Jie Zhao, Yuqi Liu, Chune Tian, Xiaofei Lin, Xiaoqing Wei, Rui Li, and Weiping Liu. Task-state eeg reveals fine hand motor impairment in acute ischemic stroke: A multidimensional analysis of cortical dynamics. NeuroRehabilitation, 0(0):10538135261441843, 2026. PMID: 42053233

    work page 2026

  29. [29]

    Dongyi He, Wai Ting Siok, and Nizhuan Wang. Toward robust, reproducible, and widely accessible intracranial language brain-computer interfaces: A comprehensive review of neural mechanisms, hardware, algorithms, evaluation, clinical pathways and future directions, 2026

    work page 2026

  30. [30]

    Yueyang Li, Weiming Zeng, Wenhao Dong, Di Han, Lei Chen, Hongyu Chen, Zijian Kang, Shengyu Gong, Hongjie Yan, Wai Ting Siok, and Nizhuan Wang. A tale of single-channel electroencephalography: Devices, datasets, signal processing, applications, and future directions.IEEE Transactions on Instrumentation and Measurement, 74:1–20, 2025

    work page 2025

  31. [31]

    A somatosensory erp-bci system for post-stroke hand function rehabilitation applications.IEEE Transactions on Instrumentation and Measurement, 75:1–12, 2026

    Junlin Wang, Xiaodong Li, Xingyu Lu, Ningbo Fei, Wei Huang, and Yong Hu. A somatosensory erp-bci system for post-stroke hand function rehabilitation applications.IEEE Transactions on Instrumentation and Measurement, 75:1–12, 2026. 16

    work page 2026