Pith. sign in

REVIEW 3 major objections 7 minor 60 references

Cross-Modal Epileptic Signal Harmonization: Frequency Domain Mapping Quantization for Pre-training a Unified Neurophysiological Transformer

T0 review · 3 major / 7 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A single Transformer, pre-trained on scalp and intracranial EEG with a frequency-domain vector quantizer and masked autoencoding, reported the best F1 among compared models across six epilepsy classification tasks.

desk verdict A solid engineering contribution that overstates its cross-modal harmonization: the frequency-domain quantizer is worth a look, but the unified-modality claim is untested because preprocessing hands the model a perfect source tag. read the letter →

arxiv 2506.17068 v1 pith:OYS5U2S3 submitted 2025-06-20 q-bio.NC cs.ETeess.SP

Runkai Zhang , Hua Yu , John Q. Gan , Haixian Wang This is my paper · ORCID
classification q-bio.NCcs.ETeess.SP
keywords EEGpre-trainingintracranialvectorquantizationmaskedautoencoderfrequency-domainmappingepilepsyclassificationTransformerfoundationmodelcross-modaltransfer
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 aims to establish that one Transformer, EpiNT, can serve as a shared backbone for both scalp EEG and intracranial EEG (SEEG and ECoG) across multiple epilepsy-related classification tasks. Its recipe is to treat each recording channel independently, patch it, mask most patches, and pre-train the encoder to predict discrete labels assigned by a frozen, randomly initialized quantizer that maps patches in the frequency domain. Pre-trained on 2,741.1 hours of clinical recordings from 1,199 patients, EpiNT reported the highest F1 among compared models under linear probing on all six downstream tasks and on four of six under full fine-tuning. The reason to care is practical: a single pre-trained model that transfers across montage, amplitude, and frequency differences could reduce the need for task- and patient-specific training in epilepsy diagnostics.

What carries the argument

The load-bearing mechanism is the frequency-domain mapping quantizer: a frozen, randomly initialized set of codebooks whose entries are assigned to patches by circular convolution with a random vector, computed as the inverse DFT of the product of the patch spectrum and the random vector's spectrum, followed by cosine-similarity matching and normalization. Because circular convolution in time is Hadamard multiplication in frequency, this quantizer scores each patch by its spectral content without learned frequency filters. Multiple quantizers (Q) each supply a pseudo-label, and the pre-training loss is the average cross-entropy of predicting those labels for masked patches; the decoder is only a single linear layer. The paper shows this mechanism's necessity by ablation: quantization beats direct L2 reconstruction, and frequency-domain mapping beats time-domain random projection.

What would settle it

Re-run pre-training with matched preprocessing across modalities — for instance, scalp EEG also resampled to 1024 Hz and cut into 3-second epochs — and fine-tune on the same six tasks; if the gap over single-modality pre-training shrinks or disappears, the reported cross-modal advantage is carried by the modality-specific preprocessing cue rather than by the frequency-domain quantizer.

Watch

Extended reading notes

Core claim

EpiNT's central claim is that predicting which entries of a fixed frequency-domain codebook a masked signal patch belongs to is a better pre-training objective for epilepsy neurophysiology than reconstructing the raw waveform, and that doing this on single channels makes the model robust to montage variability. The pre-training corpus mixes scalp EEG (2,548.3 hours) and iEEG (165.5 hours) from 1,199 patients; after fine-tuning, the model achieved the best F1 among the compared methods on tasks that include SOZ localization (TASMC-UCLA, up to 0.925), seizure prediction (CHB-MIT, 0.913), pathology classification (FNUSA 0.886, MAYO 0.965), IED detection, and neonatal seizure detection. Under linear probing, where only a classification head is trained on top of the frozen backbone, EpiNT posted the best F1 on all six tasks; under full fine-tuning it was best on four of six. The paper interprets the consistent linear-probing results as evidence that the frozen backbone, not the fine-tuning head, carries transferable representations.

Load-bearing premise

The load-bearing premise is that the two preprocessing pipelines — scalp EEG downsampled to 256 Hz with 12-second epochs and implanted EEG resampled to 1024 Hz with 3-second epochs — are a neutral harmonization step, not a signature the model can detect and exploit to tell the two recording types apart instead of learning shared neurophysiology.

Editorial extensions

If this is right

  • A frozen EpiNT backbone read out by a linear probe delivers the best F1 of all compared models on all six downstream tasks, so the pre-training itself, not the tuning procedure, is doing the transfer work.
  • Pre-training on a mixture of scalp and intracranial data beats either single-modality corpus, especially for iEEG tasks, so cross-modal exposure contributes beyond total data volume.
  • Quantizing in the frequency domain outperforms both direct raw-signal reconstruction and time-domain vector quantization, pointing to spectral structure as the invariant the model should preserve.
  • High masking ratios destroy clinically relevant sparse events in epilepsy signals; raising the mask ratio drops FNUSA F1 from 0.805 to 0.631, the opposite of what image MAE would predict.
  • Multiple small codebooks outperform a single large codebook, suggesting the quantizer's capacity should be spread across complementary random frequency projections.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the modality-specific preprocessing (256 Hz and 12-second epochs for scalp EEG, 1024 Hz and 3-second epochs for implanted EEG) is the real carrier of cross-modal transfer, then training on a version of the corpus with matched sampling rates and epoch lengths should shrink the reported advantage over single-modality pre-training.
  • Because the quantizer is a fixed random feature map, the codebook need not be learned; one could pre-specify codebooks concentrated on clinically relevant bands such as delta, theta, ripple, and fast ripple, and test whether band-targeted quantization sharpens downstream performance.
  • Channel-independent modeling discards spatial relations between electrodes; a natural extension is to add a small channel-attention head at fine-tuning time and ask whether SOZ localization, which depends on electrode location, improves.
  • The non-monotonic scaling observed on TASMC-UCLA and MAYO suggests data composition matters more than raw hours; a controlled study that varies the scalp-to-iEEG ratio at fixed total hours would clarify when mixing modalities helps.
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, and a circularity audit.

Referee Report

3 major / 7 minor

Summary. This paper proposes EpiNT, a channel-independent Transformer pre-trained with masked autoencoding and vector quantization. The pre-training uses a 'frequency domain mapping quantizer' that assigns pseudo-labels by cosine similarity to fixed random codebooks in a Fourier-transformed space. EpiNT is pre-trained on 2,741.1 hours of scalp EEG and iEEG from 1,199 patients and then fine-tuned on six epilepsy-related binary classification datasets spanning scalp and intracranial signals. The authors report that EpiNT outperforms randomly initialized models and several pre-trained baselines, and they present ablations on pre-training data composition, codebook size, number of quantizers, mask ratio, and architectural choices, along with interpretability and stability analyses.

Significance. The potential significance is high: a single pre-trained backbone for both scalp and intracranial neurophysiology could benefit epilepsy research, where labeled data are scarce. The paper has notable strengths: a large curated multi-modal pre-training corpus, plans to release code and weights, extensive ablation and comparison experiments, and quantitative stability analysis via CKA. However, the central cross-modal harmonization claim is not yet supported by the experimental design, because the downstream evaluation never crosses modalities and the preprocessing fixes a deterministic modality marker. If the authors add a control for this confound and a genuine cross-modal transfer experiment, the contribution could be valuable to the field.

major comments (3)
  1. [Section 4.8 and Discussion] Section 4.8 and Discussion: The modality-specific preprocessing (scalp EEG downsampled to 256 Hz and cut into 12-s epochs; iEEG resampled to 1024 Hz and cut into 3-s epochs, both truncated to 3072 points) gives every input a deterministic source-modality tag. Because all six downstream datasets are single-modality, the improved F1 scores in Table 3 do not demonstrate that EpiNT learns a shared, modality-invariant neurophysiological representation; the model could simply use the sampling-rate/epoch-length cue as a cheap feature. The Discussion explicitly concedes this 'implicit annotation' but offers no experiment to rule it out. To support the title's and abstract's harmonization claim, the authors should add a cross-modal transfer test (e.g., pre-train on one modality and evaluate on the other modality's tasks) and/or a control with a common preprocessing pipeline for both modalities. This is load-bearing: without it, the results support only within-modality pre-training benefits.
  2. [Abstract and Section 2.5, Table 3] Abstract and Section 2.5, Table 3: The abstract claims EpiNT 'outperformed both randomly initialized models and other pre-trained methods on six downstream classification tasks,' but under full-parameter fine-tuning EpiNT achieves the highest F1 on only four of the six datasets (TASMC-UCLA, CHB-MIT, FNUSA, MAYO), not on CUK-IMHANS or HUH. Section 2.5 itself states 'four out of six evaluation tasks.' The abstract should be corrected or qualified. Furthermore, several top scores are within one standard deviation of the runner-up (e.g., linear-probing MAYO 0.930±0.027 vs VQ-MTM 0.929±0.018), so the use of the term 'outperformed' should be accompanied by explicit significance tests for every comparison claimed.
  3. [Section 2.2 and Figure 3] Section 2.2 and Figure 3: The experiment described as 'Impact of Pre-training Data Modality' does not actually test cross-modal transfer. It pre-trains on different data compositions (scalp only, iEEG only, mixed) and then evaluates on within-modality downstream tasks; the resulting performance differences do not tell us whether a representation learned on one modality transfers to a task on the other. The text's conclusion that 'combining both scalp and iEEG data during pre-training is crucial for optimal model performance and effective transfer of learned features across different types of brain activity recordings' is not supported by this experiment. The authors need a direct cross-modal evaluation or should rephrase the claim.
minor comments (7)
  1. [Abstract and Section 2.1] The abstract and Section 2.1 disagree on the number of downstream tasks: the abstract says 'six downstream classification tasks,' while Section 2.1 says 'five clinically relevant binary classification tasks across six datasets.' Please reconcile.
  2. [Section 4.3] Section 4.3: The default mask ratio m is not specified for the final model; Section 4.9 reports only a range (0.3–0.7). Please provide the exact hyperparameters used for the main results (mask ratio, codebook size, number of quantizers, model dimension, number of layers, etc.) in the main text or a table.
  3. [Table 3] Table 3 has formatting issues: numerical entries are not aligned with their column headers, and the statistical significance asterisks are not accompanied by a description of the test procedure or multiple-comparison correction.
  4. [Discussion] The Discussion states that 'models trained from scratch outperformed pre-trained models on the CHB-MIT dataset,' which appears to contradict Section 2.5 and Table 3, where EpiNT (pre-trained) has the highest F1 on CHB-MIT in all three fine-tuning regimes. Please clarify which comparison is intended.
  5. [Throughout] There are numerous typographical and formatting errors, including 'A PREPRINT - SEPTEMBER13, 2025' (missing space), 'outpus' (§4.6), 'taks' (Figure 1 caption), 'neuronphysio- logical' (§4.11), and 'acrossing' (§4.2).
  6. [Section 4.7] Equation (16) uses inconsistent notation: the normalized codebook vector is called \hat{c}_k in Eq. (15) but \tilde{c}_k in the surrounding text.
  7. [Section 4.7] The paper claims the frequency domain mapping quantizer 'captures crucial frequency features,' but the random vector h_proj is frozen and the codebook is random, so the mechanism is closer to a random projection in the frequency domain. The claim should be tempered or supported by further analysis.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: downstream benchmarks are external labels, the pretraining pseudo-labels are random frozen projections of the input, and the only self-citation is contextual and non-load-bearing.

full rationale

The derivation chain is not circular. The pretraining target is a set of discrete pseudo-labels produced by a randomly initialized, frozen codebook and a fixed random circular-convolution (frequency-domain) projection (Eqs. 14-17); these labels are not fitted to any downstream label, so the cross-entropy objective in Eq. 3 is a self-supervised pretext task rather than a fitted-input-called-prediction. All downstream F1 scores are measured against external clinical labels (Table 2) under cross-subject splits, so the benchmark claim is not equivalent to the training objective by construction. The only overlapping-author citation (Ref. 12) is a contextual remark about DL for drug-response prediction and is not load-bearing. The paper does not invoke a uniqueness theorem or import a central assumption via self-citation. The one genuine concern is not circularity but confounding: the Discussion explicitly concedes that modality-specific preprocessing (scalp EEG at 256 Hz in 12 s epochs; iEEG at 1024 Hz in 3 s epochs, Section 4.8) 'implicitly differentiated between signal sources,' giving the model a perfect source-modality cue and leaving the unified-harmonization claim untested. That is a limitation in support for the title-level claim, but it does not make any predicted result equal to its inputs by definition. The comparison against other models retrained on the same data further keeps the benchmark claim externally grounded.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim depends on several hand-set design choices (codebook size, quantizer count, mask ratio, preprocessing epochs) and on domain assumptions about channel independence, random pseudo-label quality, and modality preprocessing neutrality. There are no invented physical entities.

free parameters (6)
  • Codebook size K_q = Final value not stated in main text; tested 256, 512, 1024, 2048
    Chosen based on downstream F1 in Section 2.4 and Supplementary Figures 2 and 3, so the reported results depend on a hyperparameter selected on the evaluation tasks.
  • Number of quantizers Q = Final value not stated; tested under codebook sizes 16 and 512
    The multi-quantizer configuration was selected from ablations on the same downstream tasks.
  • Mask ratio m = Final value not stated; tested from 0.3 to 0.7
    Masking ratio strongly affects downstream F1, for example FNUSA drops from 0.805 to 0.631 at high masking, so the chosen value is load-bearing.
  • Epoch length and sampling rate per modality = Scalp: 256 Hz, 12 seconds; iEEG: 1024 Hz, 3 seconds
    These preprocessing choices align input dimensions to 3072 points and encode modality information; they are hand-set and affect all downstream results.
  • Patch length L = 256 time points
    Non-overlapping patch length is a fixed design choice controlling sequence length; it is not swept in the paper.
  • Frequency mapper random vector h_proj = Randomly initialized and non-trainable
    The pseudo-label generator depends on a random filter; different random draws will change pretraining targets, and the paper does not study the variance across draws.
assumptions (5)
  • standard math Circular convolution in the time domain equals the Hadamard product of Fourier transforms (convolution theorem)
    Used in Section 4.7 to justify the frequency-domain mapping quantizer.
  • domain assumption Random, non-trainable codebooks and a random frequency filter provide useful self-supervised targets without any learning of the quantizer
    The pseudo-labels in Eq. (3) are assigned by cosine similarity to random codebooks; the paper offers only empirical ablation to support this choice.
  • domain assumption Channel-independent modeling preserves clinically relevant epileptic information despite discarding all spatial relationships
    Stated in Sections 4.1 and 4.2; montage invariance comes at the cost of ignoring inter-channel context used in clinical reading.
  • domain assumption Modality-specific preprocessing does not teach the model a data-source shortcut
    Section 4.8 fixes sampling rate and epoch length by modality; the Discussion acknowledges this implicit annotation but the paper does not control for it.
  • domain assumption Cross-subject splits and single-channel epochs create independent training and test samples without leakage
    Section 4.1 treats each single-channel epoch as an independent instance, even though epochs from the same patient and recording are correlated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cross-Modal Epileptic Signal Harmonization: Frequency Domain Mapping Quantization for Pre-training a Unified Neurophysiological Transformer." pith.science (2026). https://pith.science/paper/OYS5U2S3

@misc{pith2026250617068,
  author       = {Pith},
  title        = {Pith review of: Cross-Modal Epileptic Signal Harmonization: Frequency Domain Mapping Quantization for Pre-training a Unified Neurophysiological Transformer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OYS5U2S3}},
  note         = {Machine review of arXiv:2506.17068}
}
read the original abstract

Scalp electroencephalography (EEG) and intracranial EEG (iEEG) are vital for epilepsy diagnosis and treatment. Their unified analysis offers the potential to harness the complementary strengths of each modality but is challenging due to variations in recording montages, amplitude and signal-to-noise ratio (SNR), and frequency components. To address the aforementioned challenges, this paper introduces EpiNT, a novel Transformer-based pre-trained model for unified EEG and iEEG analysis. EpiNT employs channel-independent modeling with masked autoencoders (MAE) and vector quantization (VQ), along with a frequency domain mapping quantizer to capture crucial frequency features. Pre-trained on over 2,700 hours of multi-modal clinical neurophysiological data from 1,199 patients, EpiNT outperformed both randomly initialized models and other pre-trained methods on six downstream classification tasks, demonstrating robust representation learning capabilities. This work presents a promising approach for unified epilepsy neurophysiology analysis.

Figures

Figures reproduced from arXiv: 2506.17068 by the authors.

Figure 1
Figure 1. Characteristics of different electrophysiological modalities and the proposed pre-trained model. (a) Scalp EEG, electrocorticography (ECoG), and stereo-electroencephalography (SEEG) differ in signal amplitude, temporal and spatial resolution, and typical clinical applications. (b) The proposed EpiNT leverages EEG, ECoG, and SEEG to pre-train, and fine-tuned using downstream taks. 16 [PITH_FULL_IMAGE:figures/full_fi… view at source ↗
Figure 2
Figure 2. Characteristics of pre-training dataset. (a) Patient distribution reveals a substantial proportion of scalp EEG recordings compared to iEEG. (b) Total recording duration demonstrates a significantly higher volume of scalp EEG data (2548.3 hours) relative to iEEG (165.5 hours). (c) Upon partitioning into single-channel data, the disparity in data volume persists, with scalp EEG yielding 48233.8 hours and iEEG 13309.7… view at source ↗
Figure 3
Figure 3. Performance evaluation of models pre-trained on different EEG modalities. The intracranial-only model exhibits limited transferability to scalp EEG downstream tasks. While the scalp-only model shows strong generalization to scalp EEG tasks and comparable performance on iEEG tasks, but underperforms compared to the model pre-trained on a mixture of scalp and iEEG data. Pre-training on a mixed dataset of scalp and iEE… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Visualization of representative electrophysiological signals and their corresponding feature representa￾tions across EpiNT layers. (a) Raw waveforms of epileptic signals from the FNUSA and MAYO datasets, named by the indices in dataset. (b) STFT spectrograms of the raw…
Figure 5
Figure 5. Figure 5: Interpretability analysis of EpiNT. (a) Cosine similarity between EpiNT’s extracted features and conventional EEG features on the FNUSA dataset. (b) CKA matrix illustrating feature consistency across different pre-training runs on the FNUSA dataset. (c) Cosine similari…
Figure 6
Figure 6. Figure 6: Overall architecture of EpiNT. Single-channel electrophysiological signal is processed by instance normal￾ization and patch embedding before being fed into the backbone. During the pre-training stage, the backbone consists of stacked Transformer layers acting as the en…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 47 canonical work pages

  1. [1]

    Amna Rana and Alberto E. Musto. The role of inflammation in the development of epilepsy.Journal of Neuroinflammation, 15(144), 2018

    work page 2018

  2. [2]

    Epilepsy in adults.The Lancet, 393(10172):689–701, 2019

    Roland D Thijs, Rainer Surges, Terence J O’Brien, and Josemir W Sander. Epilepsy in adults.The Lancet, 393(10172):689–701, 2019

    work page 2019

  3. [3]

    The epidemiology of epilepsy.Neuroepidemiology, 54(2):185–191, 2020

    Ettore Beghi. The epidemiology of epilepsy.Neuroepidemiology, 54(2):185–191, 2020

    work page 2020

  4. [4]

    Critical slowing down as a biomarker for seizure susceptibility.Nature Communications, 11(1):2172, 2020

    Matias I Maturana, Christian Meisel, Katrina Dell, Philippa J Karoly, Wendyl D’Souza, David B Grayden, Anthony N Burkitt, Premysl Jiruska, Jan Kudlacek, Jaroslav Hlinka, et al. Critical slowing down as a biomarker for seizure susceptibility.Nature Communications, 11(1):2172, 2020

    work page 2020

  5. [5]

    Noninvasive detection of hippocampal epileptiform activity on scalp electroencephalogram.JAMA Neurology, 79(6):614–622, 2022

    Maurice Abou Jaoude, Claire S Jacobs, Rani A Sarkis, Jin Jing, Kyle R Pellerin, Andrew J Cole, Sydney S Cash, M Brandon Westover, and Alice D Lam. Noninvasive detection of hippocampal epileptiform activity on scalp electroencephalogram.JAMA Neurology, 79(6):614–622, 2022. 12 APREPRINT- SEPTEMBER13, 2025

    work page 2022

  6. [6]

    Source-sink connectivity: A novel interictal EEG marker for seizure localization.Brain, 145(11):3901–3915, 2022

    Kristin M Gunnarsdottir, Adam Li, Rachel J Smith, Joon-Yi Kang, Anna Korzeniewska, Nathan E Crone, Adam G Rouse, Jennifer J Cheng, Michael J Kinsman, Patrick Landazuri, et al. Source-sink connectivity: A novel interictal EEG marker for seizure localization.Brain, 145(11):3901–3915, 2022

    work page 2022

  7. [7]

    Deep learning-based electroencephalography analysis: A systematic review.Journal of Neural Engineering, 16(5):051001, 2019

    Yannick Roy, Hubert Banville, Isabela Albuquerque, Alexandre Gramfort, Tiago H Falk, and Jocelyn Faubert. Deep learning-based electroencephalography analysis: A systematic review.Journal of Neural Engineering, 16(5):051001, 2019

    work page 2019

  8. [8]

    B2-ViT Net: Broad vision transformer network with broad attention for seizure prediction.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2023

    Shuiling Shi and Wenqi Liu. B2-ViT Net: Broad vision transformer network with broad attention for seizure prediction.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2023

    work page 2023

Show all 60 references
  1. [9]

    Yang Li, Yu Liu, Wei-Gang Cui, Yu-Zhu Guo, Hui Huang, and Zhong-Yi Hu. Epileptic seizure detection in EEG signals using a unified temporal-spectral squeeze-and-excitation network.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 28(4):782–794, 2020

    2020

  2. [10]

    vEpiNet: A multimodal interictal epileptiform discharge detection method based on video and electroencephalogram data.Neural Networks, 175:106319, 2024

    Nan Lin, Weifang Gao, Lian Li, Junhui Chen, Zi Liang, Gonglin Yuan, Heyang Sun, Qing Liu, Jianhua Chen, Liri Jin, et al. vEpiNet: A multimodal interictal epileptiform discharge detection method based on video and electroencephalogram data.Neural Networks, 175:106319, 2024

    2024

  3. [11]

    PyHFO: Lightweight deep learning-powered end-to-end high-frequency oscillations analysis application.Journal of Neural Engineering, 21(3):036023, 2024

    Yipeng Zhang, Lawrence Liu, Yuanyi Ding, Xin Chen, Tonmoy Monsoor, Atsuro Daida, Shingo Oana, Shaun Hussain, Raman Sankar, Aria Fallah, et al. PyHFO: Lightweight deep learning-powered end-to-end high-frequency oscillations analysis application.Journal of Neural Engineering, 21...

    2024

  4. [12]

    Runkai Zhang, Rong Rong, Yun Xu, Haixian Wang, and Xiaoyun Wang. OxcarNet: Sinc convolutional network with temporal and channel attention for prediction of oxcarbazepine monotherapy responses in patients with newly diagnosed epilepsy.Journal of Neural Engineering, 21(5):056019, 2024

    2024

  5. [13]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016

    2016

  6. [14]

    Bert: Pre-training of deep bidirectional transformers for language understanding.arXiv preprint arXiv:1810.04805, 2018

    Jacob Devlin. Bert: Pre-training of deep bidirectional transformers for language understanding.arXiv preprint arXiv:1810.04805, 2018

    2018 arXiv

  7. [15]

    Self-supervised representation learning from electroencephalography signals

    Hubert Banville, Isabela Albuquerque, Aapo Hyvärinen, Graeme Moffat, Denis-Alexander Engemann, and Alexandre Gramfort. Self-supervised representation learning from electroencephalography signals. In2019 IEEE 29th International Workshop on Machine Learning for Signal Processing...

    2019

  8. [16]

    Miao Cai and Yu Zeng. MAE-EEG-Transformer: A transformer-based approach combining masked autoencoder and cross-individual data augmentation pre-training for EEG classification.Biomedical Signal Processing and Control, 94:106131, 2024

    2024

  9. [17]

    BrainBERT: Self-supervised representation learning for intracranial recordings.arXiv preprint arXiv:2302.14367, 2023

    Christopher Wang, Vighnesh Subramaniam, Adam Uri Yaari, Gabriel Kreiman, Boris Katz, Ignacio Cases, and Andrei Barbu. BrainBERT: Self-supervised representation learning for intracranial recordings.arXiv preprint arXiv:2302.14367, 2023

    2023 arXiv

  10. [18]

    Large brain model for learning generic representations with tremendous EEG data in BCI.arXiv preprint arXiv:2405.18765, 2024

    Wei-Bang Jiang, Li-Ming Zhao, and Bao-Liang Lu. Large brain model for learning generic representations with tremendous EEG data in BCI.arXiv preprint arXiv:2405.18765, 2024

    2024 arXiv

  11. [19]

    Vector quantization pretraining for EEG time series with random projection and phase alignment

    Haokun Gui, Xiucheng Li, and Xinyang Chen. Vector quantization pretraining for EEG time series with random projection and phase alignment. InInternational Conference on Machine Learning, pages 16731–16750. PMLR, 2024

    2024

  12. [20]

    Learning topology-agnostic EEG representations with geometry-aware modeling.Advances in Neural Information Processing Systems, 36, 2024

    Ke Yi, Yansen Wang, Kan Ren, and Dongsheng Li. Learning topology-agnostic EEG representations with geometry-aware modeling.Advances in Neural Information Processing Systems, 36, 2024

    2024

  13. [21]

    Brant: Foundation model for intracranial neural signal.Advances in Neural Information Processing Systems, 36, 2024

    Daoze Zhang, Zhizhang Yuan, Yang Yang, Junru Chen, Jingjing Wang, and Yafeng Li. Brant: Foundation model for intracranial neural signal.Advances in Neural Information Processing Systems, 36, 2024

    2024

  14. [22]

    Towards domain-free transformer for generalized EEG pre-training.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2024

    Sung-Jin Kim, Dae-Hyeok Lee, Heon-Gyu Kwak, and Seong-Whan Lee. Towards domain-free transformer for generalized EEG pre-training.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2024

    2024

  15. [23]

    Reliable evaluation of functional connectivity and graph theory measures in source-level EEG: How many electrodes are enough?Clinical Neurophysiology, 150:1–16, 2023

    Christoffer Hatlestad-Hall, Ricardo Bruña, Mia Liljeström, Hanna Renvall, Kjell Heuser, Erik Taubøll, Fernando Maestú, and Ira H Haraldsen. Reliable evaluation of functional connectivity and graph theory measures in source-level EEG: How many electrodes are enough?Clinical Neu...

    2023

  16. [24]

    Effects of depth electrode montage and single-pulse electrical stimulation sites on neuronal responses and effective connectivity.Clinical Neurophysiology, 131(12):2781–2792, 2020

    Takumi Mitsuhashi, Masaki Sonoda, Hirotaka Iwaki, Aimee F Luat, Sandeep Sood, and Eishi Asano. Effects of depth electrode montage and single-pulse electrical stimulation sites on neuronal responses and effective connectivity.Clinical Neurophysiology, 131(12):2781–2792, 2020

    2020

  17. [25]

    Progress in brain computer interface: Challenges and opportunities.Frontiers in Systems Neuroscience, 15:578875, 2021

    Simanto Saha, Khondaker A Mamun, Khawza Ahmed, Raqibul Mostafa, Ganesh R Naik, Sam Darvishi, Ahsan H Khandoker, and Mathias Baumert. Progress in brain computer interface: Challenges and opportunities.Frontiers in Systems Neuroscience, 15:578875, 2021. 13 APREPRINT- SEPTEMBER13, 2025

    2021

  18. [26]

    Advances in human intracranial electroencephalography research, guidelines and good practices.Neuroimage, 260:119438, 2022

    Manuel R Mercier, Anne-Sophie Dubarry, François Tadel, Pietro Avanzini, Nikolai Axmacher, Dillan Cellier, Maria Del Vecchio, Liberty S Hamilton, Dora Hermes, Michael J Kahana, et al. Advances in human intracranial electroencephalography research, guidelines and good practices....

    2022

  19. [27]

    Heterogeneity of resting-state EEG features in juvenile myoclonic epilepsy and controls.Brain Communications, 4(4):fcac180, 2022

    Amy Shakeshaft, Petroula Laiou, Eugenio Abela, Ioannis Stavropoulos, Mark P Richardson, and Deb K Pal. Heterogeneity of resting-state EEG features in juvenile myoclonic epilepsy and controls.Brain Communications, 4(4):fcac180, 2022

    2022

  20. [28]

    Developmental atlas of phase-amplitude coupling between physiologic high-frequency oscillations and slow waves.Nature Communications, 14(1):6435, 2023

    Kazuki Sakakura, Naoto Kuroda, Masaki Sonoda, Takumi Mitsuhashi, Ethan Firestone, Aimee F Luat, Neena I Marupudi, Sandeep Sood, and Eishi Asano. Developmental atlas of phase-amplitude coupling between physiologic high-frequency oscillations and slow waves.Nature Communications...

    2023

  21. [29]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick. Masked autoencoders are scalable vision learners. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 16000–16009, 2022

    2022

  22. [30]

    Neural fragility as an EEG marker of the seizure onset zone.Nature Neuroscience, 24(10):1465–1474, 2021

    Adam Li, Chester Huynh, Zachary Fitzgerald, Iahn Cajigas, Damian Brusko, Jonathan Jagid, Angel O Claudio, Andres M Kanner, Jennifer Hopp, Stephanie Chen, et al. Neural fragility as an EEG marker of the seizure onset zone.Nature Neuroscience, 24(10):1465–1474, 2021

    2021

  23. [31]

    Resection of high frequency oscillations predicts seizure outcome in the individual patient.Scientific Reports, 7(1):13836, 2017

    Tommaso Fedele, Sergey Burnos, Ece Boran, Niklaus Krayenbühl, Peter Hilfiker, Thomas Grunwald, and Johannes Sarnthein. Resection of high frequency oscillations predicts seizure outcome in the individual patient.Scientific Reports, 7(1):13836, 2017

    2017

  24. [32]

    Dataset of EEG recordings of pediatric patients with epilepsy based on the 10-20 system, 2021

    Dorottya Cserpan, Ece Boran, Richard Rosch, San Pietro Lo Biundo, Georgia Ramantani, and Johannes Sarnthein. Dataset of EEG recordings of pediatric patients with epilepsy based on the 10-20 system, 2021

    2021

  25. [33]

    A practical workflow for organizing clinical intraoperative and long-term iEEG data in BIDS

    Matteo Demuru, Dorien van Blooijs, Willemiek Zweiphenning, Dora Hermes, Frans Leijten, Maeike Zijlmans, and RESPect group. A practical workflow for organizing clinical intraoperative and long-term iEEG data in BIDS. Neuroinformatics, 20(3):727–736, 2022

    2022

  26. [34]

    Epilepsy iEEG interictal multicenterdataset, 2023

    Kristin Gunnarsdottir, Adam Li, Rachel Smith, Joon Kang, Anna Korzeniewska, Nathan Crone, Adam Rouse, Jennifer Cheng, Michael Kinsman, Patrick Landazuri, Utku Uysal, Carol Ulloa, Nathaniel Cameron, Iahn Cajigas, Jonathan Jagid, Andres Kanner, Turki Elarjani, Manuel Bicchi, Sar...

    2023

  27. [35]

    Normative intracranial EEG maps epileptogenic tissues in focal epilepsy.Brain, 145(6):1949–1961, 2022

    John M Bernabei, Nishant Sinha, T Campbell Arnold, Erin Conrad, Ian Ong, Akash R Pattnaik, Joel M Stein, Russell T Shinohara, Timothy H Lucas, Dani S Bassett, et al. Normative intracranial EEG maps epileptogenic tissues in focal epilepsy.Brain, 145(6):1949–1961, 2022

    1949

  28. [36]

    Informa- tion flows from hippocampus to auditory cortex during replay of verbal working memory items.eLife, 11:e78677, 2022

    Vasileios Dimakopoulos, Pierre Mégevand, Lennart H Stieglitz, Lukas Imbach, and Johannes Sarnthein. Informa- tion flows from hippocampus to auditory cortex during replay of verbal working memory items.eLife, 11:e78677, 2022

    2022

  29. [37]

    Refining epileptogenic high-frequency oscillations using deep learning: A reverse engineering approach.Brain Communications, 4(1):fcab267, 2022

    Yipeng Zhang, Qiujing Lu, Tonmoy Monsoor, Shaun A Hussain, Joe X Qiao, Noriko Salamon, Aria Fallah, Myung Shin Sim, Eishi Asano, Raman Sankar, et al. Refining epileptogenic high-frequency oscillations using deep learning: A reverse engineering approach.Brain Communications, 4(...

    2022

  30. [38]

    Big data resources for EEGs: Enabling deep learning research

    L Veloso, J McHugh, E V on Weltin, S Lopez, I Obeid, and J Picone. Big data resources for EEGs: Enabling deep learning research. In2017 IEEE Signal Processing in Medicine and Biology Symposium (SPMB), pages 1–3. IEEE, 2017

    2017

  31. [39]

    The temple university hospital seizure detection corpus.Frontiers in Neuroinformatics, 12:83, 2018

    Vinit Shah, Eva V on Weltin, Silvia Lopez, James Riley McHugh, Lillian Veloso, Meysam Golmohammadi, Iyad Obeid, and Joseph Picone. The temple university hospital seizure detection corpus.Frontiers in Neuroinformatics, 12:83, 2018

    2018

  32. [40]

    EEG synchronization analysis for seizure prediction: A study on data of noninvasive recordings.Processes, 8(7):846, 2020

    Paolo Detti, Giampaolo Vatti, and Garazi Zabalo Manrique de Lara. EEG synchronization analysis for seizure prediction: A study on data of noninvasive recordings.Processes, 8(7):846, 2020

    2020

  33. [41]

    Epileptic EEG dataset.Mendeley Data, 1, 2021

    Wassim Nasreddine. Epileptic EEG dataset.Mendeley Data, 1, 2021

    2021

  34. [42]

    Annotated interictal discharges in intracranial EEG sleep data and related machine learning detection scheme.Scientific Data, 11(1):1354, 2024

    Rotem Falach, Maya Geva-Sagiv, Dawn Eliashiv, Lilach Goldstein, Ofer Budin, Guy Gurevitch, Genela Morris, Ido Strauss, Amir Globerson, Firas Fahoum, et al. Annotated interictal discharges in intracranial EEG sleep data and related machine learning detection scheme.Scientific D...

    2024

  35. [43]

    PhD thesis, Massachusetts Institute of Technology, 2009

    Ali Hossam Shoeb.Application of machine learning to epileptic seizure onset detection and treatment. PhD thesis, Massachusetts Institute of Technology, 2009

    2009

  36. [44]

    Scalp EEG recordings of pediatric epilepsy patients: A dataset for automatic detection of interictal epileptiform discharges from routine EEG.Data in Brief, 39:107680, 2021

    OK Fasil, R Rajesh, and Rajith K Ravindren. Scalp EEG recordings of pediatric epilepsy patients: A dataset for automatic detection of interictal epileptiform discharges from routine EEG.Data in Brief, 39:107680, 2021. 14 APREPRINT- SEPTEMBER13, 2025

    2021

  37. [45]

    A dataset of neonatal EEG recordings with seizure annotations.Scientific Data, 6(1):1–8, 2019

    Nathan J Stevenson, Karoliina Tapani, Leena Lauronen, and Sampsa Vanhatalo. A dataset of neonatal EEG recordings with seizure annotations.Scientific Data, 6(1):1–8, 2019

    2019

  38. [46]

    Multicenter intracranial EEG dataset for classification of graphoelements and artifactual signals.Scientific Data, 7(1):179, 2020

    Petr Nejedly, Vaclav Kremen, Vladimir Sladky, Jan Cimbalnik, Petr Klimes, Filip Plesinger, Filip Mivalt, V ojtech Travnicek, Ivo Viscor, Martin Pail, et al. Multicenter intracranial EEG dataset for classification of graphoelements and artifactual signals.Scientific Data, 7(1):...

    2020

  39. [47]

    Tripod+ ai statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods.bmj, 385, 2024

    Gary S Collins, Karel GM Moons, Paula Dhiman, Richard D Riley, Andrew L Beam, Ben Van Calster, Marzyeh Ghassemi, Xiaoxuan Liu, Johannes B Reitsma, Maarten Van Smeden, et al. Tripod+ ai statement: updated guidance for reporting clinical prediction models that use regression or ...

    2024

  40. [48]

    Roformer: Enhanced transformer with rotary position embedding.Neurocomputing, 568:127063, 2024

    Jianlin Su, Murtadha Ahmed, Yu Lu, Shengfeng Pan, Wen Bo, and Yunfeng Liu. Roformer: Enhanced transformer with rotary position embedding.Neurocomputing, 568:127063, 2024

    2024

  41. [49]

    Similarity of neural network representations revisited

    Simon Kornblith, Mohammad Norouzi, Honglak Lee, and Geoffrey Hinton. Similarity of neural network representations revisited. InInternational conference on machine learning, pages 3519–3529. PMLR, 2019

    2019

  42. [50]

    MOMENT: A family of open time-series foundation models.arXiv preprint arXiv:2402.03885, 2024

    Mononito Goswami, Konrad Szafer, Arjun Choudhry, Yifu Cai, Shuo Li, and Artur Dubrawski. MOMENT: A family of open time-series foundation models.arXiv preprint arXiv:2402.03885, 2024

    2024 arXiv

  43. [51]

    A time series is worth 64 words: Long-term forecasting with transformers.arXiv preprint arXiv:2211.14730, 2022

    Yuqi Nie, Nam H Nguyen, Phanwadee Sinthong, and Jayant Kalagnanam. A time series is worth 64 words: Long-term forecasting with transformers.arXiv preprint arXiv:2211.14730, 2022

    2022 arXiv

  44. [52]

    Review of the bci competition iv

    Michael Tangermann, Klaus-Robert Müller, Ad Aertsen, Niels Birbaumer, Christoph Braun, Clemens Brunner, Robert Leeb, Carsten Mehring, Kai J Miller, Gernot R Müller-Putz, et al. Review of the bci competition iv. Frontiers in neuroscience, 6:55, 2012

    2012

  45. [53]

    Physiobank, physiotoolkit, and physionet: components of a new research resource for complex physiologic signals.circulation, 101(23):e215–e220, 2000

    Ary L Goldberger, Luis AN Amaral, Leon Glass, Jeffrey M Hausdorff, Plamen Ch Ivanov, Roger G Mark, Joseph E Mietus, George B Moody, Chung-Kang Peng, and H Eugene Stanley. Physiobank, physiotoolkit, and physionet: components of a new research resource for complex physiologic si...

    2000

  46. [54]

    EEG Conformer: Convolutional transformer for EEG decoding and visualization.IEEE Transactions on Neural Systems and Rehabilitation Engineering, 31:710–719, 2022

    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, 2022

    2022

  47. [55]

    CSP-Net: Common spatial pattern empowered neural networks for EEG-based motor imagery classification.Knowledge-Based Systems, 305:112668, 2024

    Xue Jiang, Lubin Meng, Xinru Chen, Yifan Xu, and Dongrui Wu. CSP-Net: Common spatial pattern empowered neural networks for EEG-based motor imagery classification.Knowledge-Based Systems, 305:112668, 2024

    2024

  48. [56]

    Eduardo Santamaria-Vazquez, Victor Martinez-Cagigal, Fernando Vaquerizo-Villar, and Roberto Hornero. EEG- Inception: A novel deep convolutional neural network for assistive ERP-based brain-computer interfaces.IEEE Transactions on Neural Systems and Rehabilitation Engineering, ...

    2020

  49. [57]

    EEGNet: A compact convolutional neural network for EEG-based brain–computer interfaces.Journal of Neural Engineering, 15(5):056013, 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, 2018

    2018

  50. [58]

    TinySleepNet: An efficient deep learning model for sleep stage scoring based on raw single-channel EEG

    Akara Supratak and Yike Guo. TinySleepNet: An efficient deep learning model for sleep stage scoring based on raw single-channel EEG. In2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), pages 641–644. IEEE, 2020

    2020

  51. [59]

    Physics-informed attention temporal convo- lutional network for EEG-based motor imagery classification.IEEE Transactions on Industrial Informatics, 19(2):2249–2258, 2022

    Hamdi Altaheri, Ghulam Muhammad, and Mansour Alsulaiman. Physics-informed attention temporal convo- lutional network for EEG-based motor imagery classification.IEEE Transactions on Industrial Informatics, 19(2):2249–2258, 2022

    2022

  52. [60]

    A DF-SSA analytical framework for revealing variations in multidimensional EEG features of epileptic seizures.Biomedical Signal Processing and Control, 100:107073, 2025

    Guibin Chen, Gang Li, Wanxiu Xu, Hanfan Wu, Suhong Ye, and Bin Zhou. A DF-SSA analytical framework for revealing variations in multidimensional EEG features of epileptic seizures.Biomedical Signal Processing and Control, 100:107073, 2025. 15 APREPRINT- SEPTEMBER13, 2025 Figure...

    2025

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.